Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Negative sequence impedance measurement for distributed generator islanding detection Wrinch, Michael C. 2008

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2009_spring_wrinch_michael.pdf [ 2.91MB ]
Metadata
JSON: 24-1.0065504.json
JSON-LD: 24-1.0065504-ld.json
RDF/XML (Pretty): 24-1.0065504-rdf.xml
RDF/JSON: 24-1.0065504-rdf.json
Turtle: 24-1.0065504-turtle.txt
N-Triples: 24-1.0065504-rdf-ntriples.txt
Original Record: 24-1.0065504-source.json
Full Text
24-1.0065504-fulltext.txt
Citation
24-1.0065504.ris

Full Text

Negative Sequence Impedance Measurement for Distributed Generator Islanding Detection by Michael C. Wrinch B.Sc., The University of British Columbia, 1995 B.A.Sc., Memorial University of Newfoundland, 2000 M.A.Sc., Memorial University of Newfoundland, 2002 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF Doctor of Philosophy in The Faculty of Graduate Studies (Electrical and Computer Engineering)  The University Of British Columbia (Vancouver) December, 2008 c Michael C. Wrinch 2008  Abstract This thesis presents a method of detecting electrical islands in low voltage distributed generator networks by measuring negative sequence impedance differences between islanded and utility connections. Extensive testing was conducted on a commercial building and 25 kV distributed generator fed network by measuring naturally occurring and artificially injected negative sequence components. Similarly, this technique was tested using the IEEE 399-1990 bus test case using the EMTP software. The practical measurements have been matched to simulations where further system performance characteristics of detecting power system islands has been successfully demonstrated. Measured results indicate that unbalanced load conditions are naturally occurring and readily measurable while deliberately unbalanced loads can increase the accuracy of negative sequence impedance islanding detection. The typically low negative sequence impedance of induction motors was found to have only a small effect in low voltage busses, though large machines can effect the threshold settings. Careful placement of the island detector is required in these situations. The negative sequence impedance measurement method is an improvement on previous impedance measurement techniques for islanding detection due to its accuracy, and distinctly large threshold window which have challenged previous impedance based islanding detection techniques.  ii  Table of Contents Abstract  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  ii  Table of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  iii  List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  vii  List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  viii  Acronyms  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  xi  Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  xiii  Dedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  xiv  1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1 Motivation to Power System Protection for Distributed Generation . . . . . . . . . . .  1 1  1.2  1.1.1  Electrical Energy Supply and Demand  . . . . . . . . . . . . . . . . . . . . .  1.1.2  Distributed Generation as a Viable Alternative  . . . . . . . . . . . . . . . . .  2  1.1.3  Types of Distributed Generation . . . . . . . . . . . . . . . . . . . . . . . . .  3  Technical Challenges Facing Distributed Generation  . . . . . . . . . . . . . . . . . .  1.2.1  Utility Perspective of Distributed Generator Network Islanding  1.2.2  The Problem: Detection of Unplanned Islands  2  3  . . . . . . . .  5  . . . . . . . . . . . . . . . . .  6  1.2.3  Research Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  8  1.2.4  Impedance Measurement for islanding Detection . . . . . . . . . . . . . . . .  8  1.3  Proposed Solution: Negative Sequence Impedance Islanding Detection  . . . . . . . .  10  1.4  Thesis Organization and Contributions  . . . . . . . . . . . . . . . . . . . . . . . . .  11  1.4.1  Thesis Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  11  1.4.2  List of Publications  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  12  2 Review of Islanding Detection Methods . . . . . . . . . . . . . . . . . . . . . . . . . . .  13  2.1  Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  13  2.2  Distributed Generator Related Protection Issues . . . . . . . . . . . . . . . . . . . . .  13  2.2.1  Voltage Issues From Distributed Generator Installations  . . . . . . . . . . . .  14  2.2.2  Short Circuit Current Issues . . . . . . . . . . . . . . . . . . . . . . . . . . .  17  iii  Table of Contents  2.3  2.4  2.5  2.6  2.2.3  Maintenance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  19  2.2.4  Reclosure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  20  2.2.5  Typical Interconnection Protection Schemes  . . . . . . . . . . . . . . . . . .  20  Islanding Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  21  2.3.1  Communication Based Islanding Detection Methods . . . . . . . . . . . . . .  23  2.3.2  Passive Islanding Detection Methods . . . . . . . . . . . . . . . . . . . . . .  25  2.3.3  Active Islanding Detection Methods . . . . . . . . . . . . . . . . . . . . . . .  28  Impedance Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  35  2.4.1  Impulse Response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  36  2.4.2  Islanding Detection using Continuous Injected Noise . . . . . . . . . . . . . .  39  Current and Voltage Measurability . . . . . . . . . . . . . . . . . . . . . . . . . . . .  43  2.5.1  Voltage Measurement  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  43  2.5.2  Current Measurement  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  47  Summary of Research Background  . . . . . . . . . . . . . . . . . . . . . . . . . . .  50  3 Negative Sequence Impedance Islanding Detection . . . . . . . . . . . . . . . . . . . . .  52  3.1  Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  52  3.2  Derivation of System Negative Sequence Impedance Estimation . . . . . . . . . . . .  54  3.2.1  Two Circuit DC System . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  54  3.2.2 3.2.3  Simple Three Phase System . . . . . . . . . . . . . . . . . . . . . . . . . . . Balanced AC Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  55 58  3.2.4  Unbalanced AC Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  60  3.2.5  Negative Sequence Current Flow in Unbalanced Loads . . . . . . . . . . . . .  62  3.2.6  Unbalanced Sources Injection . . . . . . . . . . . . . . . . . . . . . . . . . .  65  3.2.7  Negative Sequence Th´evenin Impedance of the Network . . . . . . . . . . . .  66  3.3  Performance Characteristics of Negative Sequence Impedance Measurement  . . . . .  67  3.3.1  Effect of Changing Per Cent of Unbalanced Load  . . . . . . . . . . . . . . .  69  3.3.2  Effect of Strength of the System vs the Unbalanced Load Power . . . . . . . .  69  3.3.3  Effect of Varying the Power of an Unbalanced Load . . . . . . . . . . . . . .  70  3.3.4  Effect of Changing Source Unbalance . . . . . . . . . . . . . . . . . . . . . .  71  3.3.5  Effect of V2 and I2 Phase Angle on Different Unbalanced Configurations  . . .  72  3.3.6  Multiple Unbalanced Loads in a System  . . . . . . . . . . . . . . . . . . . .  74  Non Fundamental Frequency Negative Sequence Impedance . . . . . . . . . . . . . .  79  3.4.1  . . . . . . . . . . . . .  79  3.5  Sequence Components for Induction and Synchronous Machines . . . . . . . . . . . .  81  3.6  Implementation Strategy for Negative Sequence Islanding Detection . . . . . . . . . .  81  3.6.1  Naturally Occurring Negative Sequence Currents . . . . . . . . . . . . . . . .  82  3.6.2  Injected Negative Sequence Currents . . . . . . . . . . . . . . . . . . . . . .  83  Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  84  3.4  3.7  Sources of Harmonics from Non-Linear Components  iv  Table of Contents 4 Case Studies of Negative Sequence Impedance Islanding Detection . . . . . . . . . . . .  87  4.1  Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  87  4.2  Simulation of Standard IEEE 399-1997 Industrial Bus  . . . . . . . . . . . . . . . . .  87  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  87  4.2.1  System Description  4.2.2  Experimental Results  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  89  4.2.3  Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  94  Practical Example 1: 25 kV Radially Feed Distributed Generator Network . . . . . . .  95  4.3.1  System Description  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  95  4.3.2  Experimental Results  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  98  4.3.3  Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  100  Practical Example 2: 600 V Fed Commercial Building . . . . . . . . . . . . . . . . .  100  4.4.1  System Description  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  100  4.4.2  Experimental Results  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  103  4.4.3  Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  110  4.5  Performance Comparison with other Impedance Based Islanding Detection Methods .  111  4.6  Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  117  5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  119  4.3  4.4  5.1  Summary of Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  119  5.2 5.3  Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Final Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  121 124  Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  125  Appendices A Device Number Reference  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  B Symmetrical Components - The Basics  . . . . . . . . . . . . . . . . . . . . . . . . . . .  C AEMC 3945 Three Phase Power Quality Meter  135 138  . . . . . . . . . . . . . . . . . . . . . .  140  D Simulation Software Tools Used . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  141  D.1 Aspen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  141  D.2 Matlab  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  141  D.3 Microtran Power Systems Simulator . . . . . . . . . . . . . . . . . . . . . . . . . . .  141  D.4 Power World . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  141  D.5 Psim  142  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  D.6 Simulink Power Systems Tool Box D.7 SKM Power Tools  . . . . . . . . . . . . . . . . . . . . . . . . . . .  142  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  142  v  Table of Contents E Simulink Model: 600 V Fed Bus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  143  F Commonly Used Signal Processing Techniques . . . . . . . . . . . . . . . . . . . . . . .  144  vi  List of Tables 1.1  Types of DG and Typical Capacity . . . . . . . . . . . . . . . . . . . . . . . . . . . .  4  1.2 1.3  Technical Challenges for Distributed Generation . . . . . . . . . . . . . . . . . . . . . Technical Challenges Associated with DG Islanding . . . . . . . . . . . . . . . . . . .  5 6  1.4  Factors Influencing Island Detectability . . . . . . . . . . . . . . . . . . . . . . . . .  9  2.1  Communication Islanding Detection Methods . . . . . . . . . . . . . . . . . . . . . .  22  2.2  Passive Islanding Detection Methods . . . . . . . . . . . . . . . . . . . . . . . . . . .  22  2.3  Active Islanding Detection Methods . . . . . . . . . . . . . . . . . . . . . . . . . . .  22  2.4  Data from Wang Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  24  3.1  Base Performance Values for Practical System . . . . . . . . . . . . . . . . . . . . . .  69  3.2  Phase Angle For Different Unbalanced Load Types . . . . . . . . . . . . . . . . . . .  74  3.3  Base Case Values for Figure 3.19 . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  75  3.4  Scenarios and Comments for Multiple Unbalanced Loads (All Values in pu) . . . . . .  77  3.5  Common Power System Non-linear Loads . . . . . . . . . . . . . . . . . . . . . . . .  80  3.6  Sequence Movement For Harmonics [120] . . . . . . . . . . . . . . . . . . . . . . . .  80  4.1  Experiments Modeled using IEEE 399 Standard Bus . . . . . . . . . . . . . . . . . .  89  4.2  Expected Impedances From Perspective of Test Loads (in pu)  . . . . . . . . . . . . .  90  4.3  Experiment 1: IEEE 399 Bus System, Single Phase Unbalance of Load 3 and Load 5 .  91  4.4  Experiment 2a, Load 3 and Load 5 Unbalanced on Phase A . . . . . . . . . . . . . . .  91  4.5  Experiment 2b, Phase A Unbalance on Load 3 and Phase B Unbalance on Load 5 . . .  92  4.6  Using Averaging Equation 3.73 for Load 3 and Load 5 . . . . . . . . . . . . . . . . .  92  4.7  Induction Machine (1.6 MVA) Input Variables for EMTP . . . . . . . . . . . . . . . .  93  4.8  Expected Impedances From Induction Machine . . . . . . . . . . . . . . . . . . . . .  93  4.9  Case 3, IEEE 399 Bus System, Induction Machine on Feeder 2, Bus 2 . . . . . . . . .  93  4.10 Case 4, IEEE 399 Bus System, Induction Machine with Alternate Phase Unbalanced .  94  4.11 Expected Z2 Impedances From Positions ”A”, ”B” and ”C” (in pu) . . . . . . . . . . .  97  4.12 Measured Values for 25 kV System Positive and Negative Sequences . . . . . . . . . .  99  4.13 Experiments Simulated on 25 kV Practical System . . . . . . . . . . . . . . . . . . .  99  4.14 Simulated Values for 25 kV System Positive and Negative Sequences . . . . . . . . . 4.15 Induction Machine Parameters (15 kVA) Input Variables from Simulink Power Library  100 109  4.16 Performance Characteristics to Compare Islanding Detection Techniques . . . . . . . .  112  vii  List of Figures 1.1  2006 United States Projected Summer Generation and Capacity [18] . . . . . . . . . .  3  1.2 1.3  2006 Canadian Projected Winter Generation and Capacity [18] . . . . . . . . . . . . . Power System Islanding Detection Schemes . . . . . . . . . . . . . . . . . . . . . . .  4 7  1.4  Non Detection Zone in Daily Load Profile Illustration . . . . . . . . . . . . . . . . . .  8  1.5  Radially Fed Distributed Generation System . . . . . . . . . . . . . . . . . . . . . . .  9  1.6  IEEE Standard 1547 Resonating Bus Islanding Detection Test Setup . . . . . . . . . .  10  2.1  Distributed Generation and Interconnection on a Radial System . . . . . . . . . . . . .  14  2.2  Ungrounded or Poorly Grounded DG Connections Causing Voltage Rise . . . . . . . .  15  2.3  Voltage Rise From Single Phase Fault on an Ungrounded System . . . . . . . . . . . .  16  2.4  Impedance vs. Frequency for Islanded and Non-Islanded States . . . . . . . . . . . . .  17  2.5  Single Phase Fault Comparison Between Utility only and DG Connected System . . .  18  2.6  Example of Reduced Impedance Relay (21) Reach . . . . . . . . . . . . . . . . . . .  19  2.7  Distributed Generation (left) and DG Interconnection Protection (right) . . . . . . . .  21  2.8  Distributed Generation Multi Power Line Signaling Islanding Detection Issue . . . . .  23  2.9  Distributed Generation Power Line Signaling Islanding Detection . . . . . . . . . . .  25  2.10 Distributed Generation Transfer Trip Islanding Detection . . . . . . . . . . . . . . . .  26  2.11 Frequency Bias Islanding Detection . . . . . . . . . . . . . . . . . . . . . . . . . . .  31  2.12 Voltage Variation Islanding Detection . . . . . . . . . . . . . . . . . . . . . . . . . .  33  2.13 Basic Premise of Using Impedance Measurement for Islanding Detection . . . . . . .  35  2.14 Impedance Measurement Using High Voltage Capacitive Switching (Girgis [36]) . . .  36  2.15 Impedance Measurement Using TRIAC Controlled Capacitive Switching (Hopewell [44]) 38 2.16 US Patent 6,603,290 Drawing 1: Islanding Detection by Signal Injection . . . . . . . .  40  2.17 Negative Sequence Injection Concept and H-Bridge Injector Realization . . . . . . . .  42  2.18 Resistive Divider Voltage Measurement with Isolator . . . . . . . . . . . . . . . . . .  44  2.19 Voltage Transformer Equivalent Circuit . . . . . . . . . . . . . . . . . . . . . . . . .  45  2.20 Coupling Capacitor Voltage Transformer Equivalent Circuit . . . . . . . . . . . . . . .  45  2.21 Pockels Effect Voltage Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.22 Resistive Shunt Current Measurement . . . . . . . . . . . . . . . . . . . . . . . . . .  46 47  2.23 Current Transformer Circuit with Burden . . . . . . . . . . . . . . . . . . . . . . . .  48  3.1  Negative and Positive Sequence Current Flow . . . . . . . . . . . . . . . . . . . . . .  53  3.2  Symmetrical Component Conversion . . . . . . . . . . . . . . . . . . . . . . . . . . .  54  viii  List of Figures 3.3  Two DC Circuits with Different Load Impedances . . . . . . . . . . . . . . . . . . . .  55  3.4  Three Phase Circuit (Single Line) Example with Different ‘Y’ Load Impedances . . . .  56  3.5  Three Phase Circuit Example Expanded with Different ‘Y’ Load Impedances . . . . .  57  3.6  Three Phase Circuit Example Expanded and Logically Grouped . . . . . . . . . . . .  58  3.7  Three Phase Circuit with a Balanced Load . . . . . . . . . . . . . . . . . . . . . . . .  59  3.8  Balanced Symmetrical Component Circuits from Balanced Load . . . . . . . . . . . .  60  3.9  Unbalanced Load and Balanced Source Circuit . . . . . . . . . . . . . . . . . . . . .  60  3.10 Symmetrical Component Concept in an Unbalanced System . . . . . . . . . . . . . .  62  3.11 Symmetrical Components Current Flow in an Unbalanced System Expanded Circuit .  64  3.12 System Performance Test Schematic with System Impedance and Load Impedance . .  68  3.13 Per Cent Unbalanced Load vs. V2 Magnitude and Phase (System Strength = 10 MVA) .  70  3.14 System Strength Vs. V2 with Varying Unbalanced Load from 0.5 MVA to 4.5 MVA . .  71  3.15 Unbalanced Load Size of 0 MVA to 3 MVA vs. V2 , (Load unbalanced by 20%) . . . .  72  3.16 Utility Strength (SCC) vs. V2 (Varying Percent Source Unbalance of 5% to 30%) . . .  73  3.17 V2 Phase For Different Unbalanced Load Types Relative to 0, 240, 120 ABC Phases . .  75  3.18 I2 Phase For Different Unbalanced Load Types Relative to 0, 240, 120 V ABC Phases .  76  3.19 Single Line Diagram of Multiple Unbalanced Equal Load Scenario . . . . . . . . . . .  76  3.20 -10% to + 10% Load Unbalance on Both Phase A, Vs. Calculated V2 (ZLoad = 2000pu )  77  3.21 -10% to + 10% Load Unbalance on Alternating Phases Vs. Calculated V2 . . . . . . .  78  3.22 -10% to + 10% Load Unbalance Vs. Calculated V2 Averaged on A,B,C Alternating Phases 79 3.23 Induction and Synchronous Machine Negative Sequence Impedance . . . . . . . . . . 3.24 Natural Negative Sequence Impedance Islanding Detection Algorithm . . . . . . . . .  81 83  3.25 Negative Sequence Impedance Measurement Islanding Detection Concept . . . . . . .  84  3.26 Injected Components Negative Sequence Impedance Islanding Detection Algorithm . .  86  4.1  IEEE Standard 399-1997 (Brown Book) Reference Bus Case adapted From [65] . . . .  90  4.2  IEEE Standard 399-1997 Reference Bus Case with Induction Machine at Load 3 (Center) 94  4.3  Practical Example 1: 25 kV DG System Fed System Single Line Diagram . . . . . . .  96  4.4  Practical Example 1: 25 kV DG System Fed System Sequence Impedances . . . . . .  98  4.5  Practical Example 2: 600 V Commercial Office DG Fed System Single Line Diagram .  102  4.6  Practical Example 2: 600 V Commercial Office 24 Hour Power Demand . . . . . . . .  103  4.7  Practical Example 2: 600 V Z2 Measured Over 24 Hours . . . . . . . . . . . . . . . .  104  4.8  Practical Example 2: 600 V Combined ZLoad and Z2 Measured Over 24 Hours . . . . .  105  4.9  Practical Example 2: 600 V Z2 At Position 1 During Breaker A and B Opening . . . .  106  4.10 Practical Example 2: 208 V Negative Sequence Injection Experimental Setup . . . . .  107  4.11 Practical Example 2: 208 V Negative Sequence Impedance Phase to Phase Loads . . .  108  4.12 Practical Example 2: 600 V, PV Source Negative Sequence Impedance Transition . . .  109  4.13 Practical Example 2: 600 V Building PV DG Islanding Detection Results . . . . . . .  110  4.14 Practical Example 2: Building Solar DG Islanding Detection with Rotating Machine .  111  4.15 Building Solar DG Islanding Detection with Rotating Machine Voltage ABC . . . . .  112  ix  List of Figures 4.16 Practical Example 2: 208 V Bus Sequence Impedance Diagram from Lab . . . . . . .  113  4.17 IEEE 1547 1 2005 Test Bus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  114  4.18 Impedance Between Islanded and Utility Connected Resonating Bus . . . . . . . . . .  115  B.1 Symmetrical Component Conversion . . . . . . . . . . . . . . . . . . . . . . . . . . .  138  B.2 Radially Fed System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  139  C.1 AEMC 3945 Power Quality Meter . . . . . . . . . . . . . . . . . . . . . . . . . . . .  140  E.1 Simulink Model From Practical Example 2: 600 V Fed Bus . . . . . . . . . . . . . . .  143  F.1  144  Convolution to Fourier Relationship . . . . . . . . . . . . . . . . . . . . . . . . . . .  x  Acronyms CCVT Coupling Capacitor Voltage Transformers CSA Canadian Standards Association CT Current Transformer DG Distributed Generator (Generation in the distribution layer) FJ Frequency Jump IEEE Institute of Electrical and Electronics Engineers IGBT Insulated-gate Bipolar Transistor IPP Independent Power Producer LL Line to Line LN Line to Neutral MVA Mega Apparent Power ( VA =  √ W 2 +VAR2 )  NDS Non Detection Zone OF Over Frequency OV Over Voltage PCC Point of Common Coupling PLL Phase Locked Loop PT Potential Transformer pu Per Unit PV Photo Voltaic Cells (Solar Cells) RL Resistor-Inductor ROCOF Rate of Change of Frequency SCADA Supervisory Control and Data Acquisition SCC Short Circuit Capacity SFS Sandia Frequency Shift  xi  Acronyms SMS Slip Mode Frequency Shift SVS Sandia Voltage Shift THD Total Harmonic Distortion UF Under Frequency UL Underwriters’ Laboratories UV Under Voltage VA Apparent Power ( VA =  √ W 2 +VAR2 )  VAR Reactive Power (Capacitors and Inductors) W Watts  xii  Acknowledgements There are many groups and individuals who helped to make this thesis a success. In this section, the key advisors, experts and supporters of the research will be mentioned. First, I would like to thank the University of British Columbia’s Faculty of Applied Science who provided the opportunity to conduct this work. In particular, my supervisor, Professor Jos´e Mart´ı of UBC, who’s extensive knowledge, vision and expertise played a key roll in the success of this work. I would also like to thank my co-supervisor Dr. Mukesh Nagpal of BC Hydro who consistently offered both technical and professional support throughout my research. I consider myself very fortunate to have had the opportunity to work with these two world class experts. Additional support came from within the Electric Power and Energy Systems lab at the University of British Columbia. The international community of students in the lab offered a diverse background of knowledge, problem solving and expertise that provided the details into various power system concepts for the depth of knowledge that this research required. Some of the more helpful individuals I would like to thank in no particular order are: Amir Rasuli, Liwei Wang, Marcelo Tomim, Michel AlSharidah, Nathan Ozog and Tom De Rybel. The literary style and grammar of this thesis was significantly improved from the expert advice and support of David Greer. The data acquired in this research could not have been attained without the support of several companies and people who gave their time to further this work. The first is UBC Utilities where head electrician, Stan Takenaka, provided the use of specialized three phase power signal monitoring equipment and safe access to hard to reach high voltage areas throughout the campus. David Helliwell, CEO of small energy group, offered additional information and support in this research. Finally I would like to thank BC Hydro, where Dr. Wenpeng Luan and Mike Adams were key links to attaining the network data and providing generalized system information from the British Columbian utility System.  xiii  Dedication For my wife Amy and family.  xiv  Chapter 1  Introduction 1.1 Motivation to Power System Protection for Distributed Generation This thesis presents a novel method of islanding detection for the protection of distributed generator fed systems that has been tested on power distribution busses of 25 kV and less. Recent interest in distributed generator installation into low voltage busses near electrical consumers, has created some new challenges for protection engineers that are different from traditional radially based protection methodologies. Therefore, typical protection configurations need to be re-thought such as re-closures, out-of-step monitoring, impedance relay protection zones with the detection of unplanned islanding of distributed generator systems. The condition of islanding, defined as when a section of the non utility generation system is isolated from the main utility system, is often considered undesirable because of the potential damage to existing equipment, utility liability concerns, reduction of power reliability and power quality. Current islanding detection methods typically monitor over/under voltage and over/under frequency conditions passively and actively; however, each method has an ideal sensitivity operating condition and a non-sensitive operating condition with varying degrees of power quality corruption called the nondetection zone (NDS). The islanding detection method developed in this thesis takes the theoretically accurate concept of impedance measurement and extends it into the symmetrical component impedance domain, using the existence of naturally and artificially produced unbalanced conditions. Specific applications, where this islanding detection method improves beyond existing islanding detection methods, are explored where a generalized solution allows the protection engineer to determine when this method can be used most effectively. The practical concerns of voltage and current measurement accuracy using potential transformers (PTs) and current transformers (CTs) are addressed in this thesis. Through field experimentation, CT and PT bus monitors were evaluated to see if they contain the resolution for measuring negative sequence voltage and current. The practical implementation for this islanding detection scheme comes without significant modifications and capital costs, as the implementation can be realized through choosing the correct existing CT and PT positions and a software addition to digital relays or low voltage inverters. To start, this thesis begins with a brief introduction to power systems in North America and the motivation for the use of distributed generation. Further chapters then detail the background and specifics of this technique.  1  Chapter 1. Introduction  1.1.1 Electrical Energy Supply and Demand Human progress has been linked to the increase of energy consumed per capita [92], [33]. In the last 20 years, electrical consumption has been steadily increasing in North America at a rate of 1.1% for Canada, and 2.0% for the United states [18]; however, the investment into new bulk electric power sources such as hydro dams and nuclear generation plants has become politically, economically and physically limited [75]. For example, transmission investment in the year 2000 was $2.5 billion dollars less than the level of investment in 1975, where over this same period, electricity sales nearly doubled [58]. At the current demand growth, the United States bulk electric power system is estimated to be approximately 5 to 15 years away from the power demand exceeding the generation capacity as seen in Figure 1.1. The United States has historically consumed a median of 7.5 times the power of Canada which can be seen in the Canadian winter demand growth in Figure 1.2. Small localized power sources, commonly known as ”Distributed Generation” (DG), have become a popular alternative to bulk electric power generation [88]. There are many reasons for the growing popularity of DG; however, on top of DG tending to be more renewable, DG can serve as a cost effective alternative to major system upgrades for peak shaving or enhancing load capacity margins. Additionally, if the needed generation facilities could be constructed to meet the growing demand, the entire distribution and transmission system would also require upgrading to handle the additional loading. Therefore, constructing additional power sources and upgrading the transmission system will take significant cost and time, both of which may not be achievable. These trends are not only limited to North America, but worldwide, the demand for electricity is expected to double in the next 20 years [114]. The costs of power outages to a country’s economy can be staggering. The cost associated with power outages to all business sectors in the United States has been determined to be of the order $164 Billion US per year. More specifically, the average cost of a power outage to a medium sized company is $1477 US for one second and $7000 US for one hour. Though the cost of one second of outage is considerable, the cost of one hour, which is a 3600 times longer duration is only 4.7 times of cost increase [70]; hence, initial quick outages are important to avoid significant cost implications to the economy. Distributed Generators can assist in reducing these occurrences by strengthening networks that are near to their stability limit.  1.1.2 Distributed Generation as a Viable Alternative Traditionally, electrical power generation and distribution are purely a state owned utility. However, in order to keep up with the growing demand, many states and provinces in North America are deregulating the electrical energy system. This trend is not without its own challenges. For example, how is an independent power producer (IPP) able to enter the market ? Recent innovations in power electronics such as fast switching, high voltage Insulated Gate Bipolar Transistors (IGBT) and developments in power generation technologies have made DG a considerable alternative to either delaying infrastructure upgrades or as additional cogeneration support [3]. Though the cost per kw-hr is still higher than basic power grid distribution costs, ($0.07/Kw-hr for gas turbines  2  Chapter 1. Introduction 1100  Capacity Growth Projection  Power (1000's of MW)  1000  2.5% 2.0%  900 0.5%  1.5%  800 700  Historical Consumption Consumption Projections  600 500 400 1985  1990  1995  2000  2005  2010  2015  2020  Year Figure 1.1: 2006 United States Projected Summer Generation and Capacity [18]  and as high as $0.5/kw-hr for PV) [66] [59]. The trend to completely deregulate the North American electric power grid along with the increasing trend in the cost of fossil fuels has resulted in the consideration of DG as a viable opportunity. Currently, BC Hydro, Canada’s third largest utility has more than 50 Distributed Generator stations ranging from 0.07 MVA to 34 MVA [79].  1.1.3 Types of Distributed Generation Distributed Generators can be broken into three basic classes: induction, synchronous and asynchronous. Induction generators require external excitation (VARs) and start up much like a regular induction motor. They are less costly than synchronous machines and are typically less than 500 kVA. Induction machines are most commonly used in wind power applications. Alternatively, synchronous generators require a DC excitation field and need to synchronize with the utility before connection. Synchronous machines are most commonly used with internal combustion machines, gas turbines, and small hydro dams. Finally, asynchronous generators are transistor switched systems such as inverters. Asynchronous generators are most commonly used with microturbines, photovoltaic, and fuel cells. A comparison of each type of generation system can be seen in Table 1.1  1.2 Technical Challenges Facing Distributed Generation Distributed Generation (DG) is not without problems. DG faces a series of integration challenges, but one of the more significant overall problems is that the electrical distribution and transmission infrastructure has been designed in a configuration where few high power generation stations that are often distant from their consumers, ”push” electrical power onto the many smaller consumers. DG systems 3  Chapter 1. Introduction 140  Pow wer (1000's of MW)  130  Capacity Growth Projection  120  2.4%  110  Historical Consumption  0.7%  100  1.1%  90  80 -0.6% 70  Consumption  60  Projections 50 1985  1990  1995  2000  2005  2010  2015  2020  Year  Figure 1.2: 2006 Canadian Projected Winter Generation and Capacity [18]  Technology Photovoltaic Wind Geothermal Micro Hydro Reciprocating Engine Combustion Turbine Combined Cycle Microturbines Fuel Cells  Table 1.1: Types of DG and Typical Capacity Typical Capacity Utility Interface 10 VA to 5000 VA 10 VA to 500 kVA 100 VA to several MVA 100 VA to several MVA 1000 VA to several MVA 1000 VA to several MVA 1000 VA to several MVA 10 kVA to several MVA 10 kVA to several MVA  Inverter Induction and Synchronous Generators, Inverters Synchronous Generator Induction or Synchronous Generator Induction or Synchronous Generator Synchronous Generator Synchronous Generator Inverter Inverter  are often smaller systems that are locally integrated into the low voltage distribution system (see Table 1.1) which conflicts with the existing power network design paradigm. An example of a similar radial system is with a large city’s water distribution where one very large pipe of water slowly becomes narrower and narrower until it reaches the customer’s tap at a low flow and low pressure. What would happen if one of the consumers had a water well and started pumping water into the system? Adding DG to the existing electric power distribution system can lead to a reduction of protection reliability, system stability and quality of the power to the customers. More specifically, the technical challenges that the installation of distributed generation face have been reviewed in various studies [9] [24] [63] [118] [116] [89] [32] [65] where the findings of the various studies are listed in Table 1.2. Depending on the amount of DG connected and the strength of the utility power system, the issues outlined in Table 1.2 can become substantial problems [28]. Of the challenges with DG listed in Table  4  Chapter 1. Introduction  Table 1.2: Technical Challenges for Distributed Generation 1. Voltage Regulation and Losses 2. Voltage Flicker 3. DG Shaft Over-Torque During Faults 4. Harmonic Control and Harmonic Injection 5. Increased Short Circuit Levels 6. Grounding and Transformer Interface 7. Transient Stability 8. Sensitivity of Existing Protection Schemes 9. Coordination of Multiple Generators 10. High Penetration Impacts are Unclear 11. Islanding Control 1.2, the problem of protection against unplanned islanding is a significant one. Islanding can be defined as: Islanding: The condition when a portion of the utility system is energized by one or more DG sources and that portion of the system is separated electrically from the rest of the utility system. DG islanding may be inadvertent or intentional [38].  1.2.1 Utility Perspective of Distributed Generator Network Islanding Utilities have a more pragmatic point of view of distributed generation islanding [38]. Their goal is to improve the distribution level (25 kV and below) customer service reliability especially in regions where the reliability is below customer’s needs. It is believed that customer reliability could improve with the addition of DG sources and that the DG may be able to sell electricity back to the utility. However, without complex studies and frequent expensive system upgrades DG islanding is not allowed. Some examples of these studies are: real and reactive power profile and control, planning for islanding, minimum/maximum feeder loading, islanding load profile, minimum/maximum voltage profile, protection sensitivity and DG inertia. One more specific example is how substation auto-reclosers of circuit breakers and main line reclosers may be disabled and other protection devices may need to be removed to allow proper coordination of utility sources and DG sources (covered in further chapters). Maintenance times might also increase as utility workers will not only need to lockout the utility lines but they will need to take additional time to lockout all the installed DG lines. Some of the required installation studies an IPP must complete to be able to island are: 1. inadvertent islanding and planned islanding study, 2. reliability study, 3. power quality study, 4. utility equipment upgrade assessment, 5. safety and 5  Chapter 1. Introduction protection reviews, and 6. commercial benefit study. Clearly the costs of designing a DG to be capable of islanding or to simply be installed into the main utility owned network requires extensive and costly engineering and business reviews which may be outside the financial range of smaller DG suppliers. An Example of Current Utility Islanding Policy: “Nearly all the BC Hydro distribution feeders are radial, resulting in a power outage for all the load customers connected to the feeder during a feeder outage. A DG without planned islanding capability does not provide increased customer service reliability hence the DG must be de-energized during the outage.” [38]  1.2.2 The Problem: Detection of Unplanned Islands As previously stated, Islanding is the condition when a portion of the electrical system is completely disconnected from the rest of the electric utility and left generating electrical power on its own to its local consumers. For unapproved DG systems, the current generalized industry standard is to disconnect all the DG sources from the island as soon as possible. The IEEE society has produced a standard for Distributed Generation IEEE 1547, [54] [56], that highlights the IPP’s distributed generator requirements. One of the requirements for islanding detection states that if an island condition were to occur, the distributed generator should detect and disconnect itself from the network within two seconds of the island state occurring. However, other local utilities also have their own special requirements for IPPs to comply to. One example of this is the Canadian utility, BC Hydro, which has a guideline on islanding detection requirements [38]. Apart from published standards and specialized industry requirements, islanding conditions are technically undesirable. Dr Wilsun Xu’s report [128], highlights many of the challenges and solutions of islanding distributed generation. The reasons for islanding detection are evaluated by [65] [115] [25] [26] [38] and summarized in Table 1.3. As a result of the challenges listed in Table 1.3, islanding detection is essential to the effective integration of distributed generation sources into the existing power network. Table 1.3: Technical Challenges Associated with DG Islanding 1. In general, a distributed generator is a ”weak” supply that does not have the stability and momentum of the typically strong utility system to effectively control transients. 2. A distributed generator’s behavior may be unpredictable if loads are mismatched to the supply characteristics. 3. Upon reclosure from a fault, distributed generators will not be synchronized with the utility system. The result would be potential damage to the distributed generator, the utility, or even the customer. 4. Uncontrolled islands may pose a threat for unaware utility workers. 5. The utility’s liability for the customer’s electricity quality can not be effectively managed with the current mismatch in utility vs. Independent Power Producer’s objectives.  6  Chapter 1. Introduction There exists many islanding detection methods that can be fundamentally split into two basic categories: communication and local where local detection can then be split into two more sub headings of active and passive detection schemes [32] [128] [11] [43] [34]. The families of islanding detection schemes are illustrated in Figure 1.3. Anti-Islanding Schemes Communication Based  Local Detection  Transfer Trip  Passive  Active  Frequency Power Line Signaling  Voltage Power  Impedance Measurement Freq/Phase or Voltage Shift  Harmonics  Figure 1.3: Power System Islanding Detection Schemes  The most reliable and also often the most difficult to implement islanding detection scheme shown in Figure 1.3 is through direct communication between the distributed generators and the utility. Though the trivial case for this method is extremely reliable, the practical implementation of transfer trip or power line signalling can be inflexible, complex and expensive to implement (ie $80,000-$250,000 CAD for a single DG installation) for higher penetration of distributed generators and in other more complex systems. As a result, more cost effective methods of local islanding detection are preferred. Local detection means that the Independent Power Producers are responsible for detecting and disconnecting their generator(s) when an island condition occurs independently and without direct input from the local utility. Local islanding detection, listed in Figure 1.3, can be broken into passive and active techniques [87]. Due to its negligible impact on power quality, a passive islanding detection method is desired. Passive islanding detection monitors the distributed generator terminals for changes in the voltage, current or frequency to estimate the system island state. Unfortunately, passive techniques have sensitivity limitations; hence, active islanding detection methods are being proposed in combination with passive methods [32] [128]. Active detection methods proposed in available literature, consist of a signal or disturbance being injected into the network by the DG or near to the DG and the resulting reaction is then measured and compared to the pre-set threshold. The sensitivity of passive techniques is measured by the Over/Under Voltage (OV) and Over/Under Frequency(OF) variance. Though theoretically OV and OF are trivial to measure, customer loads can vary substantially, and typical +/- 6% variability [12] [13] of OV and +/- 0.5 % OF are allocated in order to prevent unnecessary or “nuisance tripping” of the generators [40]. One particularly difficult state to measure is during near equal generation vs. demand 7  Chapter 1. Introduction by the DG in the islanded area (zero power flow at the point of common coupling). Though originally this was considered to be a rare condition [117] [90], the installation of a large number of DG’s can increase its probability. Figure 1.4 illustrates an example of the non-detection zone of a distributed generation system during a twenty-four hour period.  Network Load  Load  Non Detection Zone (NDZ)  DG Generation Level Islanding is undetectable in these zones  Time of Day  Figure 1.4: Non Detection Zone in Daily Load Profile Illustration  Though there are many islanding detection techniques available, there is no single method that has a non detection zone of zero in all possible scenarios. As a result the power systems engineering community is undecided on what type of islanding detection should be used [87]. For example, IEEE standards 1547-2003 and 929-2000 specify performance characteristics of the islanding detection methods with detailed test circuits that can be used to validate the method. Alternatively, the German standard for islanding detection [102], requires that specific methods be used such as resistive and capacitive load switching impulses to measure changes in the system’s Th´evenin impedance.  1.2.3 Research Motivation An ideal islanding detection system will operate under all system conditions with high security and dependability. Unfortunately, attaining a system with a zero non-detection zone for all situations and that has minimal power quality erosion is difficult. With each islanding detection method, there are factors that can affect sensitivity and quality [128] [98]. Some of these factors are summarized in Table 1.4. As previously mentioned, one of the most difficult states to determine is at zero power flow out of the island to the utility. The objective of this thesis has been to develop an islanding detection technique that can operate well under the zero power flow condition and other operating conditions with a low cost of integration.  1.2.4 Impedance Measurement for islanding Detection Of the many active islanding detection methods and their derivatives [32] [128] [43] [104] [95] [132] [41] [124] [16] [31] [46] [126] [7] [5] , the impedance measurement difference between an island condition and a non-island condition has theoretically a very low non-detection zone for radial systems with strong network connections. Take for example, Figure 1.5 that consists of two 25kV loads, S1 8  Chapter 1. Introduction  Table 1.4: Factors Influencing Island Detectability 1. Penetration density of distributed generators 2. Complex derivatives of RLC loads and resonance decay 3. Harmonic noise 4. Varying and continuously changing loads 5. Predicting and measuring a base thresholds without live experimentation and S2 that are being fed by the utility and DG. Loads S1 and S2 consume a total resistive power of 1 MVA, the source DG has an output power of 5 MVA and the utility has a strength of 100 MVA. The resulting impedances in pu with SBase = 100MVA and VBase = 25kV are ZUtil = 1pu and ZDG = 26pu. The measured Th´evenin impedance at the PCC when Breaker A is open and closed would be nearly 26 times. Such a change in impedance in this system would allow for easy threshold settings for islanding detection.  Figure 1.5: Radially Fed Distributed Generation System There are several released patents that use the impedance measurement technique to detect islanding. These methods are “Signal Injection”[43] [81] and “Variations in the Voltage and Frequency” [126]. Also as published by Asiminoaei in [5] [7], single non-harmonic frequency injection was found to be an effective impedance measurement method. Though the non-harmonic frequency injection method has demonstrated effective lab results, this technique suffers from a difficult and costly interfacing to the power network. The technique used in this thesis is also based on impedance measurements, but as introduced in Section 1.3, it uses signals already present in the power network. Islanding Detection Comparison and Testing Issues Test procedures to validate the performance characteristics of various islanding detection techniques are complex due to the variety of possible types of islands and system characteristics that can exist. Test procedures require trade-offs of practically realized lab experiments and tests that cover as many island 9  Chapter 1. Introduction conditions as possible. This can be a costly and confusing process for standardization organizations which are required to certify each DG for safe operation. One example of a commonly used islanding detection testing platform is the IEEE Resonating bus 1547 seen in Figure 1.6. The LRC system is set to a quality factor of 1 ± 0.05. Islanding Breaker  ZUtil  Util  ZDG  R  L  C  DG  Figure 1.6: IEEE Standard 1547 Resonating Bus Islanding Detection Test Setup A number of methods previously proposed for islanding detection will be briefly discussed in the next chapter with special attention paid to impedance measurement techniques.  1.3 Proposed Solution: Negative Sequence Impedance Islanding Detection In this thesis, the method described in [72] of using negative sequence components to determine power system equivalents for voltage stability prediction is applied and extended to the problem of islanding detection. This method constitutes a novel solution to real time islanding detection for the protection of distributed generators that uses symmetrical components negative sequence impedance measurements. Symmetrical components were developed by Fortescue in 1918 [120] and have long since been used by protection engineers as a tool for power system fault analysis. Previous work using unbalanced voltage for islanding detection appeared in [74] and [108]. However, these references found voltage unbalance techniques to suffer from false tripping and had to be combined with several other techniques such as total harmonic distortion monitoring and active frequency drifting to enhance the accuracy. Other work related to unbalanced distribution systems is presented in [101]. The method operates by unbalanced loads conducting current into the negative and zero sequence symmetrical components networks that can be measured as an associated negative and zero sequence voltage and current. The voltage or current can be calculated using the symmetrical component voltage divider Equation 1.1 and then impedance calculated using Equation 1.2 which can be seen in Figure 1.5. Zsys is the Th´evenin equivalent impedance towards the utility from the PCC (ZUtil + ZLine ), Zload is the positive sequence load impedance from the unbalanced load S1 and E is the source voltage from the  10  Chapter 1. Introduction utility. This Equation has been derived in Section 3.2 of this thesis. [V012−PCC ] = A−1 · [Zload ] · ([Zsys ] + [Zload ])−1 · A · [E012 ]  (1.1)  In general, considering the system in Figure 1.5, the result of measuring the negative sequence voltage, V2−PCC , and negative sequence current, I2−PCC , where the load S1 is unbalanced is the Th´evenin impedance of the network and utility as seen by the PCC bus and given in Equation 1.2. Then applying this for use with DG Islanding detection, Z2−Sys will experience a step increase if the DG suddenly islands from the utility connected system. ZSys ≈ ZUtil + ZLine ≈ −1 ·  V2−PCC I2−PCC  (1.2)  One anticipated challenge to applying this concept was the unbalanced voltage and current measurability. However, field measurements using existing CTs and PTs of naturally occurring unbalanced conditions in distribution systems were found to provide measurable sources of negative sequence current and voltage. These measurements were made at several voltage distribution levels such as: 25 kV, 12 kV, 600 V and 208 V to demonstrate the validity of up stream negative sequence impedance measurement for islanding detection.  1.4 Thesis Organization and Contributions This thesis is divided into five logical Chapters. Chapter 1 serves as the introduction and Chapter 2 covers the background on the state-of-the art of islanding detection, protection of distributed generation and current impedance measurement techniques. Chapter 3 contains a theoretical review and derivation of negative sequence impedance islanding detection. Chapter 4 contains several cases of practical field studies and studies of simulated models to demonstrate the validity of the technique and what typical field values can be expected. Chapter 5 contains the conclusion and future research topics.  1.4.1 Thesis Contributions This thesis introduces a novel solution to islanding detection for distributed generator protection using symmetrical component negative sequence impedance measurements. The key contributions in this thesis are as follows: • Application of the concept of negative sequence impedance measurements to the problem of islanding detection for distributed generator protection.  • Theoretical analysis and performance of negative sequence impedance measurement for islanding detection.  • Field data and network modeling supporting the measurability of naturally occurring of negative sequence voltages and currents.  11  Chapter 1. Introduction • Field data supporting the generation and measurability of injected negative sequence current. • Development, modeling and practical experiments of the novel concept of averaging of threephase sequence injection impedance to enhance impedance measurement accuracy.  • A comprehensive review of all impedance measurement techniques for live systems in the past 20 years.  1.4.2 List of Publications 1. Publication 1: Michael C. Wrinch, Jos´e Mart´ı, Mukesh Nagpal, “Negative Sequence Impedance Island Detection on a Low Voltage Commercial Bus”, Electrical Power & Energy Conference 2008, IEEE Proceedings(accepted), 2008. 2. Publication 2: Woodroffe, Adrian, Wrinch, C. Michael, Pridie, Steven, “Power Delivery to Subsea Cabled Observatories”, Oceans 2008, IEEE Proceedings(accepted), 2008. 3. Publication 3: Wrinch, C. Michael, Tomim, A. Marcelo, Mart´ı,R. Jos´e, “An Analysis of Sub Sea Electric Power Transmission Techniques from DC to AC 60 Hz and Beyond”, Oceans 2007, IEEE Proceedings, 2007.  12  Chapter 2  Review of Islanding Detection Methods 2.1 Introduction Distributed generator protection paradigms have some differences from traditional radial utility systems that pose technical challenges to the protection engineer and safety concerns for the Independent Power Producer and power customers. This chapter contains a review of protection concepts for: distributed generators, how islanding detection fits into the protection mix, a description of typical measurement systems, and a review of previous work on distributed generator islanding detection. The final part of this chapter focuses more closely on localized impedance estimation used over the past 20 years and how it is applied for use with islanding detection.  2.2 Distributed Generator Related Protection Issues Distributed generator protection is an important topic, as it shows how distributed generator protection islanding detection can be practically integrated and the scenarios which islanding detection can be best applied. This section has been developed to cover the topic of distributed generator protection. Distributed generators are low voltage small electrical sources (typically less than 30 MVA) located in or near the customer loads, and like all other generators, they require electrical protection from short circuits and abnormal system conditions. Some of these abnormal conditions are caused by the utility system itself, such as, over voltages, unbalanced currents, abnormal frequency, and breaker reclosures [77] [79] [8] [99] [34]. These conditions can happen very quickly causing generator failure and are of great concern to the owner of the distributed generator. Similarly, the utility is concerned that installations of distributed generators will result in problems on the utility’s distribution equipment or to the customer loads. utility distribution circuits are most commonly configured to supply radial loads where the introduction of distributed generators results in a redistribution of power flow and can increase fault currents as well as possible over voltages. Distributed generator (DG) interconnection protection must address both the concerns of the utility and the distributed generator owner. An example of a typical DG installation can be seen in Figure 2.1 which will be the center of the discussions in this chapter. The system in Figure 2.1 contains the utility system on the left and the DG fed distribution system at the lower right. In between these two electrical power sources are various customer loads. The distance from the utility substation to the distributed generator can be of varying distances resulting in the strength of the utility connections from being very weak(long distance) to very strong(short distance). The utility substation can isolate the DG fed bus or the DG can isolate itself from the system. A hy-  13  Chapter 2. Review of Islanding Detection Methods Fault  Utility Substation  Loads  Loads  A  1 to 100 km  Utility  Loads  Loads  Loads  Loads  DG System  Figure 2.1: Distributed Generation and Interconnection on a Radial System pothetical fault marked with an ’X’ has been inserted for further discussion in this chapter. The first subsections of this section detail the recommended protection standards for DG’s followed by some of the key protection challenges associated with DG’s installed into an existing power network. The key protection challenges associated with installations of distributed generators are: voltage variances, over current, maintenance, utility liability and reclosures. Recommendations, Standards and Guidelines Though each utility will have their own specific guidelines according to the characteristics of each particular region, there are several international standards available that can be used as guidelines. The most important four are as follows [34]: • IEEE C37.95-2003 IEEE Guide For Protective Relaying of Utility-Consumer Interconnections [53]  • IEEE 929-2000 Recommended Practice for Utility Interface of Photovoltaic (PV) Systems [51] • IEEE 242-2001 Recommended Practice for Protection and Coordination of Industrial and Commercial Power Systems (IEEE Buff Book)[52]  • IEEE 1547 Series of Standards for Interconnection of Distributed Resources with Electric Power Systems [54]  2.2.1 Voltage Issues From Distributed Generator Installations There are three ways in which a distributed generator can affect the voltage profile of a distribution system [9] [79]. The first is through the DG interconnecting transformer grounding, the second is through a change in voltage sag from I 2 R profile between the utility and the DG and the third is through floating system resonant effects. Customers expect that the over and under voltages will remain within 14  Chapter 2. Review of Islanding Detection Methods the utility’s specified tolerances which are typically near the ANSI voltage regulation allowance of +/6%, however some distributed generation configurations can potentially push the regulation out of this range. A few specific cases will be examined. Interconnecting Transformer Grounding The first cause of voltage rise in a distributed generation distribution system is through the feeder transformer grounding. An over voltage can occur with the near simultaneous occurrence of three conditions. These conditions are: 1. A single line to ground fault occurs on the line between the distributed generator and the utility substation, leaving the healthy lines untouched (80% of all fault types are single phase [10] ); 2. The substation breaker opens to form an island; 3. A distributed generator has its interconnection transformer connected as a low side WYE or DELTA and high side DELTA, or poorly grounded WYE as seen in Figure 2.2. With these three occurrences, the non-grounded phase to neutral √ voltages on the healthy utility side can rise to as high as 178% ( VLN · 3 ), hence potentially causing  damage to customer loads in only a few cycles. In a more practical system, the transformer would likely saturate, providing a non-inductive path and also the fault impedance of the shorted conductor would not be zero. These factors limit the voltage somewhat to below 178% maximum. The voltage rise has been illustrated in Figure 2.3 where phase B has been grounded and can be seen to pull the ground/neutral to Line ’B’ reference where the other two phases remain at their respective phase voltages. DG  Utility Connection 1  DG  Utility Connection 2  Utility Connection 3  Utility  DG Z  DG Z  Connection 4  Figure 2.2: Ungrounded or Poorly Grounded DG Connections Causing Voltage Rise This over voltage condition can be mitigated in two ways, with a 4 wire WYE strongly grounded system to eliminate the over voltages on the utility side, or with coordinated anti islanding controls that prevent the distributed generator from energizing ungrounded and islanded lines. Unfortunately, both of these options carry additional risks with them. The first option of using a 4 wire WYE system carries the risk of an increased short circuit condition (See Section 2.2.2 for more details) while the coordinated anti-islanding can be complex and requires fast islanding detection times between the Distributed 15  Chapter 2. Review of Islanding Detection Methods Va  Va  Earth Reference  Vb  Earth Reference  Vc  Vb  Vc  Before Single Phase  During Phase B  Ground Fault  Ground Fault  Figure 2.3: Voltage Rise From Single Phase Fault on an Ungrounded System Generator owner and the utility (see Section 2.3 for more specific details). Voltage Sag, Voltage Rise The second cause of over voltages is caused by voltage sag and rise due to changes in the operating DG [34], the utility transformer tap settings and the load conditions. In normal conditions, the voltage sags and rises are designed to be confined between the maximum and minimum allowable voltage tolerances. However, if a DG system is operating in parallel with a utility such as seen in Figure 2.1, the greatest voltage sag occurs under heavy loading but a DG source (particularly a non dispatchable source) poses additional complexity to voltage regulation due to possible reverse power flow. Correction for the voltage drop can be accomplished through capacitive switching, transformer tap changes, or with regulation of the DG output. However, sudden changes in the distributed generation, such as those due to maintenance or for non dispatchable types of DG, such as wind or solar, the regulation can become more difficult. Transformer taps can reach their operational limits and additional intelligent control mechanisms may be required [64]. Network Resonance The third cause of over voltage is by resonant over voltages in an island condition. Resonating transients can occur during natural load switching, through capacitive corrective switching or during faults. Network resonance occurs at the transfer function poles, or in the most simple RLC circuits, when the capacitive and inductive reactance are equal, as seen by Equation 2.1. Here fo is the resonance frequency, C is the capacitance in Farads, and L the inductance in Henrys. However, a distributed generator fed island is, by nature, a weak interconnection system and generally is much weaker and more inductive than a utility connected network. Therefore, as seen in Equation 2.1, as the inductance goes up, the resonant frequency goes down. fo =  1 √ 2π C · L  (2.1)  To demonstrate the difference between an islanded state and a non island state in a system similar to Figure 2.1, the island state (breaker ‘B’ open) and non-islanded state ( breaker ‘B’ closed) impedance vs  16  Chapter 2. Review of Islanding Detection Methods frequency have been analyzed. They are shown together in Figure 2.4. The system resonant characteristics before, and after an island show that the resonance moves down towards the often present 3,5,7,9 harmonics which can cause over voltage problems in the network. Power system harmonic responses have been explored extensively in [4] and [19] where research into injecting harmonic content to evaluate system transfer function has also been explored [131] [5] [110]. Control of system resonances of this type can be managed through island prevention (high impedance busses) and through detailed engineering studies of all operational states to design filters for such states. Impedance as seen from the Distributed Generator 3000 Island State (Breaker "B" Open) Utility Connected (Breaker "B" Closed) 2500  Amplitude (Z)  2000  1500  1000  500  0  0  500  1000 Frequency (Hz)  1500  2000  Figure 2.4: Impedance vs. Frequency for Islanded and Non-Islanded States  2.2.2 Short Circuit Current Issues Single Phase Fault Current Distributed generator interconnection not only causes design challenges for voltage regulation, but it can also increase the short circuit current during faults, while desensitizing utility short circuit protection sensing equipment [79] [82]. Protection systems must take the additional short circuit current capacity into consideration to ensure the system can handle the capacity before installing a distributed generator. Short circuit effects are more clearly seen with symmetrical component reconstruction by Fortescue [120] [29] (Derivation seen in Appendix B). Take for example, a single phase fault in the system in Figure 2.1 where a single phase fault occurs at the cross labeled ”Fault”. The situation is compared with the distributed generator connected to the utility and with the distributed generator disconnected from the utility system, as shown in Figure 2.5. The diagram on the left hand side of Figure 2.5 shows the connection of the three sequence networks for a line to ground fault with only the utility connected. The right side of Figure 2.5 shows the sequence networks with both the utility and DG connected. The 17  Chapter 2. Review of Islanding Detection Methods system with the distributed generator has a delta-connected transformer on the low side, and solidly grounded WYE on the a high side, while using network Th´evenin impedance values similar to those of [79]. With only the utility connected, the short circuit current is 3I0 = 738 A, increasing to 3I0 = 2040 A with the DG and with the utility connected. This also decreases the sensitivity of the ground current detection of the utility relay to 3I0 = 342 A. However, if the interconnecting transformer is ungrounded, the fault current by the DG will flow through the utility ground and will increase the fault current to 3I0 = 888A. The utility or its customers may be required to upgrade their circuit breakers to survive the additional distributed generation increased fault levels [47] [29].  Figure 2.5: Single Phase Fault Comparison Between Utility only and DG Connected System The transformer grounding strategy for the installation of DG’s is a tradeoff between over-voltage, over-current, upgrade costs, and equipment sensitivity. Therefore, the complexity of installation of DG requires proper studies to be conducted to ensure that the system is protected while maintaining economical practicality. Though there is no one general solution for distributed generation current protection, the items listed illustrate the challenges faced by the protection engineer. Reduced Reach of Impedance Relays Distance relays (21) can have a reduced reach with the installation of distributed generators [27]. This is a result of the power infeed from the DG. The relay can reach a maximum fault from an impedance zone that can then trigger the relay over a certain time frame. For example, in Figure 2.6 the distance relay can normally (ie without the DG) estimate the distance of the fault using Equation 2.2. VA = IU · (Z1 + Z2 )  (2.2)  The variable VA is the voltage at the relay, IU is the current towards the fault from the utility, Z1 is 18  Chapter 2. Review of Islanding Detection Methods the impedance to where the DG would be (not connected in this case) and Z2 is the impedance past the DG. However, with the addition of the DG source into the system, the impedance estimate changes to Equation 2.3. VA = IU · (Z1 + Z2 ) + IDG · Z2  (2.3)  The additional IDG is the current from the currently connected DG source. Hence, solving for the impedance seen from the relay with the DG in the system gives the result in Equation 2.4. Therefore, the current ratio of the DG and the utility increases and the accuracy of the relay decreases making it possible for the relay to be inoperable in fault zones. ZRelay =  VA IDG = Z1 + Z2 + Z2 IU IU  (2.4)  Normal Reach of Relay  Reduced Reach of Relay  Fault  VA Z1  Z2  A  Utility  DG System  Figure 2.6: Example of Reduced Impedance Relay (21) Reach  2.2.3 Maintenance Distribution power lines interconnected to distributed generators require periodic maintenance like all other lines. Though this case is more of a practical situation, this detail is worth mentioning. The maintenance procedures are different in regular networks from those in networks with distributed generation. In a radially fed line without distributed generation, only the upstream line is required to be “locked out” for maintenance workers to service the line. However, with the addition of distributed generation, the maintenance worker must know that there is a distributed generator downstream so that they will lock out the lines on both sides of the area they are carrying out maintenance on. This safety concern for each additional distributed generator will come at the cost of documenting, communicating and training for the utility service team. Islanding detection systems help to prevent potential hazards if the lock out procedures are missed or overlooked.  19  Chapter 2. Review of Islanding Detection Methods  2.2.4 Reclosure A large number of overhead line faults are transient in nature and can be cleared if the line is temporarily de-energized. Utility statistics indicate that fewer than 10% of all faults are permanent [29], and it has been discovered that customer service continuity and system stability can be improved by automatically reclosing the breaker. Multiple-shot reclosing breakers are used in areas with tree exposure, however, when there are sources at both sides of the line, high speed reclosing can only safely occur if the system, or if both generators have enough inertia to remain in phase during the dead time. In the case of distributed generators, many of the sources are electronically generated through computer controlled transistor switching, are low inertia generators, or are combinations of the two. These configurations are unable to maintain synchronization after a line is opened from the main utility, forcing protection engineers to re-adjust the reclosure sequence. Due to the effectiveness of reclosures for power system fault clearing, distributed generators require a mechanism to respond to temporary line opening. Islanding detection allows a DG (particularly those at a distance from the interconnection substation) to automatically disconnect themselves and allow the utility system to automatically reclose and clear the fault. After the utility restoration, the distributed generator can locally re-synchronize and re-connect. This process requires that the DG open its breakers shortly after the loss of the utility connection.  2.2.5 Typical Interconnection Protection Schemes As seen in the previous subsections, interconnection of DG requires a mixture of protection considerations to ensure that customer loads and system equipment are safe. The typical protective components of the system illustrated in Figure 2.1, the protection requirements for distributed generators and one bus of the substation can be seen in Figure 2.7 [10] [77] [79]. Both systems include: CT and PT measuring transformers, an AC circuit breaker (52)1 , protective relays for distance (21), over voltage (59) neutral and line, under voltage (27), synchronization (25), directional power (32), phase sequence current (46), phase sequence voltage (47), instantaneous over-current (50) neutral and line, over current (51), directional over current (67), reclosing (79) and frequency (81) over, under and rate of change. Suggested upgrades for distributed generators that may be required if the DG can supply reliable power for an island are follows [79]: • If the DG has a recloser breaker, replace the recloser breaker with a regular circuit breaker. • Add a PT on the load side of feeder to allow dead line check logic • Add three phase PTs on the high voltage side of the DG interconnection to allow for synchronization upon reconnection.  • Use modern multifunction relays equipped with the items seen in Figure 2.7 1 Appendix  A contains ANSI/IEEE Device reference numbers  20  Chapter 2. Review of Islanding Detection Methods PT PT 52  25  G  27  47  PT  59  81  81  U  O  52 U  G  CT  27  59  81  81  U  O  CT  46  32  51  21  79  67  PT 46  32  51  51  50  V  N  N  25  Figure 2.7: Distributed Generation (left) and DG Interconnection Protection (right) Protection of distributed generation installations can be a complex issue involving many optimizations, therefore, care must be taken for each system to maximize the system dependability and security for the customers it serves [40]. Power system dependability is the degree of confidence of correct operation after system trouble, where security is in the degree of confidence that a relay will not operate incorrectly [29]. Islanding detection can be added on top of these basic relay sensors to improve control actions.  2.3 Islanding Detection Islanding detection is an effective tool for distributed generation protection. As introduced in Section 1.2.2 and seen in Figure 1.3, there are a number of islanding detection schemes currently developed. Each class has a limitation and an advantage [11] [128] [34]. It can be difficult to directly compare all the islanding detection methods back to back, as each type will operate more effectively than the other depending on the situation. For example, the change of terminal voltage method may be ideal for rotating machine generators due to their often large reactive component, where as the frequency shift methods work well with inverter based generators that supply more real power. A good performing islanding detection scheme has the ability to securely and dependably detect an island state. IEEE has published standard 1547 [54] which details testing of single phase single source islanding detection methods. Though the testing method is limited to only single inverter systems, it is commonly used by standards agencies such as the Canadian Standards Association (CSA) and Underwriters’ Laboratories (UL), as benchmarks to approve grid tie power converter products for sale in North America. In this section, the state-of-the-art of islanding detection methods will be discussed and reviewed for their particular advantages and disadvantages. This discussion will then lead to a more detailed analysis of impedance measurement techniques in the next section. Islanding detection is broken down into three main classifications: 1. Communication, 2. Passive 21  Chapter 2. Review of Islanding Detection Methods Detection, and 3. Active Detection. For each of these classifications, there are a number of methods which have been listed in Tables 2.1, 2.2 and 2.3 below. These methods have been referenced to several comprehensive reports [11] [128], and [34]. Methods not covered in these reports that have been more recently developed will have references beside them. Each method will be described in relations to how it works, its typical use, its advantages and its disadvantages. The final part of this section will contain a detailed review of the state of the art of impedance measurement techniques and how they are being applied to islanding detection. Table 2.1: Communication Islanding Detection Methods • • •  Power Line Carrier Communications [129] [121] Transfer Trip Supervisory Control and Data Acquisition (SCADA)  Table 2.2: Passive Islanding Detection Methods • • • • • • • • •  Under/Over Voltage (Relay 59, 27) Over/Under Frequency (Relay 81) Voltage Phase Jump / Voltage Vector Shift / Frequency Phase Jump Relay Detection of Voltage and/or Current Harmonics Rate of Change of Frequency Relay (ROCOF) Rate of Change of Voltage Rate of Change of Real/Reactive Power and Power Factor Signal Produced By Disconnect Voltage Unbalance [108] [74]  Table 2.3: Active Islanding Detection Methods • • • • • • • • • •  Slip-mode Frequency Shift Frequency Bias Sandia Frequency Shift Sandia Voltage Shift Frequency Jump ENS or MSD (A device using multiple methods) [102] Varying Terminal Voltage/Voltage Pulse Reactive Error Export Voltage Unbalance [74] Impedance Monitoring (See section 2.4)  22  Chapter 2. Review of Islanding Detection Methods  2.3.1 Communication Based Islanding Detection Methods Power Line Carrier Communications These methods use the power line as a carrier of signals to transmit islanded or non-islanded information on the power lines. The apparatus includes a signal generator at the substation (25+ kV) that is coupled into the network where it continually broadcasts a signal (see Figure 2.9). Due to the low-pass filter nature of a power system (as seen in Figure 2.4), the signals need to be transmitted near or below the fundamental frequency and not interfere with other carrier technologies such as automatic meter reading. Each DG is then equipped with a signal detector to receive this transmitted signal. Under normal operating conditions, the signal is received by the DG and the system remains connected. However, if an island state occurs, the transmitted signal is cut off because of the substation breaker opening and the signal can not be received by the DG, hence indicating an island condition. This method has the advantages of its simplicity of control and its reliability. In a radial system there is only one transmitting generator needed that can continuously relay a message to many DGs in the network. The only times the message is not received is if the interconnecting breaker has been opened, or if there is a line fault that corrupts the transmitted signal. Most recently, Xu [129] has proposed and tested a single phase power line carrier and has demonstrated field data suggesting practical use. Utility Substation  Utility Feeder 1 Signal 1 Signal Generator 1 Signal 2 Towards Radial DG  Utility Feeder 2  System  Signal Generator 2 Signal 3 Utility Feeder 3  Signal Generator 3  Figure 2.8: Distributed Generation Multi Power Line Signaling Islanding Detection Issue There are also several significant disadvantages to this method, the fist being the practical implementation. To connect the device to a substation, a high voltage to low voltage coupling transformer is required. A transformer of this voltage capacity can have prohibitive cost barriers associated with it that may be especially undesirable for the first DG system installed in the local network. Another disadvantage, is if the signalling method is applied in a non radial system, resulting in the use of multiple signal generators. This scenario can be seen in Figure 2.8 where the three feeder busses connect to one island 23  Chapter 2. Review of Islanding Detection Methods bus. The implementation of this system, opposed to a simple radial system, will be up to three times the cost. The following disadvantage is the required transmitting power - enough energy is required to be transmitted into the system at a frequency close to the fundamental, and at a signal-to-noise ratio that exceeds the normal signal to noise level in the system. The non-detection zone of this method occurs when some loads on the system interfere with the transmitted signal. In report [93], it was shown that motor vibrations can cause wide band harmonic voltage fluctuations around the fundamental. These fluctuations may cause potential nuisance trips of this detection method. Though lab experiments have proven to be successful [41], including recent field tests [121], power consumption and system cost are believed to still be challenges. For example, the tests by Wang [121] using intermittent transmission, had an acceptable signal to noise level when the signal amplitude was above 2.5% (approximately 2 kW) and it was not found to interfere with existing automatic meter reading power line carrier devices as previously believed [11]. The lost revenue of a continuous 2 kW signal to this technology would be $1300 CAD to $2570 CAD per year in British Columbia, Canada as seen in the following calculations. Data from Wang [121] and the cost per kW-hr can be seen in Table 2.4. Current costs per kW-hr for small industrial customers has been referenced from the BC Hydro general customer web site 2 . Table 2.4: Data from Wang Experiment Vbase 480 Vin jected 2.5% to 5% Vbase IRMS 168 IPeak 1382 Cost/kWhr $0.0728 Cost calculations for the experiment per year use the following Equation 2.5.  Cost = $/kW hr · hoursyear ·V · I/1000  (2.5)  = 0.00728 · 8760 · 480 · 0.025 · 168/1000 = $1300 = 0.00728 · 8760 · 480 · 0.05 · 168/1000 = $2570 Another problem for power line communication is the complexity of the network and the affected networks. A perfectly radial network with one connecting breaker is a simple example of island signaling; however, more complex systems with multiple utility feeders may find that differentiation between upstream breakers difficult (seen in Figure 2.8). Transfer Trip Transfer trip detection schemes require all circuit breakers which island the DG to be monitored and linked directly to the DG control, or through a central substation SCADA system. When a disconnection is detected at the substation, the transfer trip system determines which areas are islanded and sends the 2 http://www.bchydro.com/policies/rates/rates759.html  24  Chapter 2. Review of Islanding Detection Methods Utility Substation Loads  Signal Generator  Loads  A 1 to 100 km  Utility  Loads  Loads  Loads  Signal Detector Loads  DG System  Figure 2.9: Distributed Generation Power Line Signaling Islanding Detection appropriate signal to the DGs, to either remain in operation, or to discontinue operation. Transfer tip has the distinct advantage similar to Power Line Carrier Signal that it is a very simple concept. With a radial topology that has few DG sources and a limited number of breakers, the system state can be sent to the DG directly from each monitoring point. This is one of the most common schemes used for islanding detection [11]. This system can be seen in Figure 2.10 The weaknesses of the transfer trip system are better related to larger system complexity cost and control. As a system grows in complexity, the transfer trip scheme may also become obsolete, and need relocation or updating. Reconfiguration of this device in the planning stages of DG network is necessary in order to consider if the network is expected to grow or if many DG installations are planned. The other weakness of this system is control. As the substation gains control of the DG, the DG may lose control over power producing capability and special agreements may be necessary with the utility. If the transfer trip method is implemented correctly in a simple network, there are no non-detection zones of operation.  2.3.2 Passive Islanding Detection Methods The following section contains an overview of the most commonly known islanding detection methods using passive means. Passive islanding detection methods measure voltage, current and phase information at the DG’s terminals or Point of Common Coupling (PCC), to determine if an island condition has occurred. Many of these methods are very cost effective, as the relays are already in place for other protective requirements. The biggest challenge with passive detection techniques is setting an appropriate sensor threshold that can identify the difference between islands and natural power system variations. The general consensus is that the currently available passive islanding detection techniques need to be combined with active methods to reduce the non-detection zone, to ensure a higher level of security and dependability. In the following review, many of the methods have been combined into one section as they represent similar phenomena and performance characteristics. Passive islanding detection relia-  25  Chapter 2. Review of Islanding Detection Methods Utility Substation  Transfer Trip Monitor Loads  Loads  A  1 to 100 km Loads  Loads  Loads  Loads  DG System  Figure 2.10: Distributed Generation Transfer Trip Islanding Detection bility is compounded further by the limited accuracy of current and voltage sensors. Accurate analysis of islanding detection techniques must extend beyond computer simulations. Sensor accuracy and the associated practical measurement challenges are discussed in Section 2.5 of this thesis. Despite all the limitations of passive islanding detection, research on this topic continues, as the concept of zero system disturbance for detection is the ideal. Over/Under Frequency, Change of Frequency, Frequency Surge Frequency based passive islanding detection schemes are often used with synchronous and induction machines. If the generated power and the power consuming load are mismatched, the frequency of the system will change. Hence, because the frequency of a power system is generally constant, when a frequency mismatch occurs over a preset threshold, (such as > ±0.5Hz) the system is assumed to be in  an island state, and the DG relay will open the breaker. For example, as seen in the work by Geidl [34], the power unbalance in an island between DG power PDG and the load power PLoad is given by Equation  2.6. By placing ∆P into the Swing Equation 2.7, frequency decline can be seen when the generation capacity is below the load demand [68] [79]. ∆P = ∑ PDG − ∑ PLoad  (2.6)  The variable ∆ f indicates the frequency change from the initial frequency, D indicates the loaddamping factor, or the ratio of percentage load change to percentage frequency change, and H indicates the system’s inertia constant. ∆f =  ∆P (1 − e−Dt/2H ) D  (2.7)  Another method for islanding detection using frequency is called Rate of Change of Frequency 26  Chapter 2. Review of Islanding Detection Methods (ROCOF). This method does not use change of frequency, ∆ f , but ∂∂tf instead. Hence, a DG will trip if the change in frequency exceeds a rate of a specified threshold. Typical threshold values for ROCOF in 60 Hz systems are between 0.1 Hz/s and 1.20 Hz/s [128]. This function needs to be considered during the generator startup frequency change and fault settings to prevent nuisance tripping. It is important to note that frequency, phase and magnitude of the voltage are related to power by Equations 2.8 2.9, and 2.10 where P and Q are the real and reactive power, V1a and V1b are the voltages at two ends of a line, and Z1 is the impedance of the line. Assuming the reactive part of the impedance is dominant, changes in P mostly result in variations in phase and frequency, while changes in Q result in variations in voltage. V1a ·V1b θ Z1  (2.8)  V1a ·V1b · sin(θ) Z1  (2.9)  V1a · (V1b · cos(θ) −V1a ) Z1  (2.10)  P + jQ = Where P can be solved as: P= And Q can be solved as: Q=  The final passive method using frequency is called the Vector Surge, which measures the change of phase angle of the voltage waveform to a known reference waveform. When an island state occurs, there can be an immediate phase shift by the DG to accommodate the change in power requirements. Once again, a threshold is set at the maximum phase jump allowed and if the DG system exceeds that threshold, the relay is triggered. The performance of this method has been previously investigated [128] and it has been found that a generator with a 33% or more power mismatch will result in a minimum detection time of 300ms to operate. The Vector Serge method is not useful for inverter-fed systems as inverters are either self commutated or line commutated. Electronically driven systems do not change frequency when the load changes, because computer control and Phase Locked Loops keep the frequency constant and near the fundamental. For all three passive frequency islanding detection schemes, generator power to consuming power mismatches smaller than 20%, have exponentially increasing trip times, and are prone to nuisance trips if set too sensitive. Of the three, ROCOF experiences the most significant trip time increase below a 20% power mismatch, followed by the Vector Jump and Change of Frequency. The performance of these three frequency based detection method are further complicated by the inertial constant of the generator, multiple generators, voltage and frequency dependency of the feeder loads, and the excitation control of the generator.  27  Chapter 2. Review of Islanding Detection Methods Over/Under Voltage Over and under voltage are also used for passive islanding detection, and often as a complementary device coupled with frequency monitoring. Voltage variations occur as a result of a mismatch of reactive power, Q, mismatch as seen in Equation 2.10. This relay operates on the principle that an excess of reactive power mismatch will drive the voltage up and a deficit of reactive power will drive the voltage down. Once the voltage falls out of the preset thresholds, the relay will open the breaker. Hence, by determining the voltage change or its rate of change, it is possible to detect island states that frequency effects alone cannot. Unfortunately, there is limited experience indicating that the reactive power measurement relay will have higher performance than frequency variations. As real power draw is often much greater than reactive power, a loss of mains is more likely to significantly change the active power than the reactive power. Rate of Change of Voltage Rate of change of voltage has successfully been investigated in [97] where it was found that usual voltage variations are slow in distribution systems, but if one utility system becomes Islanded from the main distribution system, the rate of change of voltage is larger than under regular operation. The nondetection zone of this method is closely coupled with its sensitivity to network disturbances, except in the case of island transitions. Detection of Voltage and/or Current Harmonics This method of islanding detection is generally applied in conjunction with inverter based technologies when system harmonics are likely to be present. In this method, the island detector measures the total harmonic distortion (THD), sets a threshold and then shuts down when the harmonic distortion exceeds that level. If an assumption is made that a utility-connected system is more “stiff” than a DG-only system, the THD will be less for a utility connected system than for a DG-only connected system. There are several factors that can increase the level of harmonics in a network. Examples include switching power supplies, motor drives, and non linear components such as overloaded transformers. The level of harmonics produced by inverters will change between full load and no load conditions. A typical requirement for Inverters is to meet the THD specification of less than 5% under full load conditions. These harmonics are often very small due to the low impedance sink provided by the utility system and the measurability and the threshold setting will exhibit significant issues. This method has found setting thresholds and the ability to accurately measure small harmonics to be very difficult to measure and predict. Similar work has been completed in this field by the injection of harmonics, as discussed seen in Section 2.3.3.  2.3.3 Active Islanding Detection Methods The following section contains an overview of the most common known islanding detection methods for distributed generators using local injection to enhance measurement accuracy. Active islanding de28  Chapter 2. Review of Islanding Detection Methods tection techniques use a mix of advanced control, load switching, and voltage and current variances, followed by measurement at specific times to augment the difference between islanded and non islanded scenarios. Many of the techniques reside inside inverters and switch controlled DGs, using the signal processing capabilities already existing inside these products for power flow and synchronization control. This makes the addition of software modifications for islanding detection very easy to implement. Active islanding detection methods are bound by practical actuation technologies, allowable active disturbance sizes, measurement sensitivity and multiple DG interference that may effect the customer energy quality, reliability or safety of the system. For example, a current source inverter tracks power by using Equation 2.11 to vary either Iinverter , ω or φ to achieve the desired output power. Equation 2.11 can also be varied to make an islanding detection device, as discussed below. iinverter = Iinverter sin(ωt + φ)  (2.11)  Slip-mode Frequency Shift Slip mode frequency shift (SMS) is an inverter based islanding detection scheme that uses a positive feedback control to destabilize the source inverter when an island condition occurs. As seen in Equation 2.11 a current source inverter uses positive feedback of the phase, φ, to slip the frequency out-of-phase hence leading to short term frequency change. SMS is implemented by modifying the phase locked loop (PLL) filter to be naturally out of phase at the fundamental. Under normal operating conditions without SMS, the PLL tracks phase and frequency changes of the network. With SMS, the strength of the utility source keeps the inverter in phase. However, if the frequency during an island is pushed upwards, due to the out of phase filter, the PLL will see a negative phase error and try to shift the frequency away from the fundamental. Due to the positive feedback, the phase shift will be in the wrong direction to correct the phase error. The frequency eventually will fall out of acceptable limits and the frequency relay will open the breaker. SMS also has the advantage of simple implementation as it only requires a modification to existing components in the inverter filter. SMS has been tested to have one of the smallest non-detection zones for islanding detection and is effective in multiple inverter applications [95]. However, the SMS method will fail if the frequency response of an islanded RLC load is greater than the SMS system. Another nondetection zone for SMS is with loads that have high-quality factor, Q, and have resonance frequencies very close to the line frequency [96] [71]. The quality factor is related to the ratio of energy stored over the power loss as seen in Equation 2.12 for a series RLC circuit where R is resistance, L is inductance and C is capacitance. Q=  1 R  L C  (2.12)  Due to the high gain of the positive feedback mechanism, high density installations with SMS may cause power quality problems related to transient response. In smaller density installations, though, the effect on power quality is low. This method has been found to have a non-detection zone in 60 Hz  29  Chapter 2. Review of Islanding Detection Methods systems between 59.3 Hz and 60.5 Hz and for quality factors between 1.5 and higher, though this will vary depending on the specific tuning of the control scheme [71]. Frequency Bias (Active Frequency Drift) Frequency Bias, also known as Active Frequency Drift, is also an inverter and computer based islanding detection technique that distorts the frequency output to create a continuous trend to “drift“ the frequency away from the fundamental. The method works by altering the frequency, f, in Equation 2.11 by slightly increasing the frequency of each 1/2 cycle followed by a “dead time“ where the system waits for the fundamental to catch up to the biased frequency, as seen in Figure 2.11. The current in each cycle can be described by Equation 2.13 where i is the output current, I is the maximum current, fV is the voltage frequency and δ f is the frequency change. i=  √  2Isin[2π( fV + δ f )]  (2.13)  Similarly to Slip-Mode Frequency Shift, when the inverter is connected to the utility-fed network, the strong utility keeps the system frequency stable. However, when the network becomes islanded, the distorted frequency causes the system to seek the system load’s resonance frequency, resulting in the inverter eventually drifting up or down causing the frequency relays to trip. Frequency bias can work with multiple inverters as long as all inverters drift the frequency in the same direction, otherwise they may not drift the frequency fast enough to meet the detection requirements. The Frequency Bias method clearly requires a small amount of output power distortion where the distortion depends on how big the bias is per 1/2 cycle. With Frequency Bias method, the non-detection zone is relatively large compared to other methods and is found that the method is not particularly effective [95]. This method has been found to have a non-detection zone in 60 Hz systems between 58.65 Hz and 60.5 Hz for quality factors of 1.5 to 100 [71]. Sandia Frequency Shift Sandia Frequency Shift (SFS) is an enhancement of the Frequency Bias Islanding detection method which uses positive feedback. The positive feedback takes the error of the “dead zone,” as seen in Figure 2.11, as an error of the line frequency. When this system is connected to the utility, small frequency changes push the inverter out of the range of the line frequency, but the strength of the utility keeps the system stable. The feedback signal can be represented by Equation 2.14, where f is the inverter terminal voltage, f0 is the base frequency (such as 60 Hz), K is the positive feedback gain, and c f0 is the initial chopping fraction [122]. π θ f = (c f0 + K( f − f0 )) 2  (2.14)  When the utility is disconnected, the errors detected in frequency increase and the dead zone increases. This method has been found to be a significant improvement over normal Frequency Bias, except the transient performance degrades with higher density of sources using this system, and large Q 30  Chapter 2. Review of Islanding Detection Methods Frequency Bias 1 Fundamental (60 Hz) 0.8  Biased Frequency (5%)  0.6  Amplitude (pu)  0.4  0.2  0  -0.2  -0.4  Dead Time  -0.6  -0.8  -1 0  0.002  0.004  0.006  0.008  0.01  0.012  0.014  0.016  0.018  Time (t)  Figure 2.11: Frequency Bias Islanding Detection loads also pose a challenge. This method was found to work well with current control inverters, but was found to be inappropriate for constant power inverters [123]. The non-detection zone for this method has been investigated and compared to frequency drift, slip mode frequency shift, and SFS, and by [71] which found the non-detection zone to be between 59.3 Hz and 60.5 Hz for Q greater than 4.8. Sandia Voltage Shift Similarly to Sandia Frequency Shift, Sandia Voltage Shift (SVS) also uses a form of positive feedback to detect islanding. In this case, the inverter decreases its power output and thus its voltage. When the utility is connected, there is little to no change in the output terminal voltage, however when the utility is not connected, the voltage will drop with the reduction of power. The positive feedback control of the voltage reduction is further accelerated downwards until the under voltage protection relay trips. Since SVS is an inverter based scheme, it is easily implemented in software and can couple well with the Sandia Frequency Shift [95]. The drawback of this method is that it creates a reduction of inverter efficiency, and similarly to other positive feedback techniques, may also suffer from poor performance for transient responses. High Q values have negligible effect on this method. Frequency Jump Frequency Jump (FJ) is also known as the Zebra Method and is a close relative of the Frequency Bias method. In the FJ method, “dead zones“ are added similarly to the frequency bias method, but not in every cycle. The frequency is broken into a predefined algorithm, with dead zones added every 31  Chapter 2. Review of Islanding Detection Methods second or third cycle. When connected to the utility, the inverter only sees a modified current and an unmodified utility linked voltage. When in island state, the voltage and current change as per the inverter programmed wave shape. Therefore, the inverter can detect an island by the modified frequency, or by matching the voltage pattern to the inverter’s algorithm. This method is believed to lose effectiveness when used in conjunction with many inverters that use the same algorithm [95]. ENS or MSD (A device using multiple methods) The ENS method is more of a standard than an actual islanding detection technique, but for thoroughness, it is important to describe how it works for this process. ENS (DIN V VDE 0126-1-1 [102]) has been required by law in Germany since 2006 for grid tied devices to use for islanding detection. The standard is different from IEEE performance based standards [54], as it is prescriptive about specific techniques, with particular performances that must be used. It includes a passive measurement component and an active component. The passive component requires that the voltage does not exceed the limits of 80% to 114%, and a frequency of 47.5 Hz to 50.2 Hz (50 Hz is the fundamental frequency in Germany). The DG is to be disconnected within 200 ms (10 cycles) of these limits being exceeded. Along with the instantaneous values, the average voltage measured over ten minutes must also not exceed the limit of 115%. Finally, a resonant LC load is to be switched in parallel with the system to measure the impedance changes on all three phases. Changes outside of the original set range of 5% results in a relay trip within 5 seconds of the occurrence. The performance of this combined method has not been directly measured, but the load switching method is described in the Impedance Measurement Section 2.4 if this chapter. The method has been found to work well but at the cost of power quality degradation. Varying Terminal Voltage/Voltage Pulse Voltage Pulse Perturbation and Correlation are discussed in [34] and [113] where the frequency is kept constant and the voltage output is varied at the inverter terminals. The first method modulates a square pulse on top of the fundamental so that every few cycles the voltage increases for a short duration then decreases. The change of voltage is then measured and the two are compared. This can be seen in Figure 2.12. Due to the usually strong utility connection, the voltage variation will be very small, however, under island conditions the voltage change will be significantly larger. This method suffers from severe limitations when more than one distributed generator is connected to the local network and the threshold values are susceptible to corruption. The method has been further improved by using a random square signal instead of a periodical signal to modulate and then further using a correlation function, φ, as seen in Equation 2.15. φ(τ) =  1 T  T 0  ∆Er (t) · ∆Eg (t + τ)∂t  (2.15)  Here Er is a random signal measured at the point of common coupling and Eg is the signal mea32  Chapter 2. Review of Islanding Detection Methods  Figure 2.12: Voltage Variation Islanding Detection sured at the inverter terminals. With the terminal voltage at the inverter and the voltage at the point of common coupling, the island threshold is no longer a variable of E but also of φ. This method has been demonstrated in simulation to have a better performance, but still suffers from corruption from other DG sources. However, if the number of DG systems are known, the accuracy has been found to increase. This method is, in essence, a form of impedance measurement with an injected signal. Impedance measurement systems are described in the following sections. Voltage Unbalance The voltage unbalance technique is a relatively new form of islanding detection. Distribution level loads are frequently single phased, which results in a system that is impossible to perfectly balance or to have equal power flow down each phase. This unbalance has been found to be significantly more prominent between an islanded and non-islanded state than with typical voltage variations [74]. This method has the drawback of choosing adequate threshold settings when large rotating machines are installed (nearly zero negative sequence voltage drop) also if other intermittent unbalanced loads are present in the system. The method in [108] uses as an indicator, the percent of negative sequence to positive sequence in combination with over/under voltage and total harmonic distortion. Focusing on unbalanced voltage, the technique finds the percent of negative sequence relative to positive sequence as a function of time (Equation 2.16). By assuming the value of the system impedance between an islanded and non-islanded state are significantly different, the change between the two states will result in a significant voltage fluctuation. VUt =  V2,t × 100 V1,t  (2.16)  In Equations 2.17 and 2.18 the average and variance of the voltage fluctuations are determined, 33  Chapter 2. Review of Islanding Detection Methods where VUt (voltage unbalance) is the ratio of negative sequence V2,t to positive sequence V1,t , N is the number of samples in one-cycle, t is the monitoring time, and VUavg,s is the VU reference value initially set for the steady-state normal loading conditions. Then in Equation 2.18, VUavg,s is set for initial conditions and if ∆VUt remains within -100% and +50% for one cycle, the VUavg,s is updated by the VUavg,t to allow for normal load variation. VUavg,t = ∆VUt =  1 N−1 ∑ VUt−i N i=0  VUavg,s −VUavg,t × 100 VUavg,s  (2.17)  (2.18)  The islanding detection decision is made by using the following rule seen in Equation 2.19. RU LE : (∆VUt > 50%)or(∆VUt < −100%)  (2.19)  This technique was found to work very well in simulation, but has not been tested in practical systems. This technique will have a very small non-detection zone if the threshold is set correctly and if the measurements are made accurately. More recently, [74] a simulated voltage unbalance, in combination with active positive feedback frequency drift for islanding detection, small disturbances caused the voltage unbalance technique to falsely trip and the power quality was reduced using the frequency drift method. By combining the two, the false tripping was reduced and the power quality was increased. Current Unbalance Karimi in [60] used negative sequence current injections for islanding detection in simulation. A voltage source inverter controller made the injections. The negative-sequence current injection in the range of 2% to 3% allowed for fast islanding detection. The time of detection was less than 60 ms(3.5 cycles). Active/Reactive Power Control Though active and reactive power control scheme is mentioned here, active and reactive power variations are a result of voltage and phase variations as seen in Equation 2.8. This section has covered variations of voltage, phase and frequency which are implemented passively and actively; hence, employing similar results and similar non-detection zones as if power variation was implemented. One example is Reactive Error Export, discussed in [34]. Here, the generator maintains a specified amount of reactive power. When the island state occurs, the inverter is unable to deliver the reactive power and the phase changes. The error in the ability to export the reactive power is the trigger for an island state.  34  Chapter 2. Review of Islanding Detection Methods  2.4 Impedance Measurement Direct impedance measurement for distributed generation islanding detection has been of significant research interest due to its theoretically negligible non-detection zone. The key challenge associated with impedance measurements is the ability to estimate system Th´evenin impedance from any one point. Significant research has gone into the development of methods for measuring the impedance in live systems for islanding detection. The premise is that the Th´evenin impedance difference between a utility connected system and an islanded system are significant, that is, Zutility << ZDG Loads . If the Th´evenin impedance can be reliably measured from the point of view of the DG, the non-detection zone will be zero for all cases. This premise can be seen in Figure 2.13 where ZLoad is the island impedance which is always much greater than ZUtility . The Th´evenin impedance seen from the DG source is either ZDG + ZLoad for an island condition or ZDG + (ZLoad ||ZUtility ) for a utility connected condition.  Research on islanding detection using impedance measurement was reviewed in [128] and [11] in 2002. This approach has some common limitations which are consistent with those discussed with respect to all methods, e.g. sensitivity, threshold settings, installation cost and practical continuous real time measurement limitations. ZDG  VDG  ZUtility  ZLoad  VUtility  Figure 2.13: Basic Premise of Using Impedance Measurement for Islanding Detection This thesis develops an impedance measuring technique based on the concept of Negative Sequence Th´evenin Equivalents proposed in [72], for application to the islanding detection problem. For this purpose, the most current main impedance measurement techniques are studied in relation to how the proposed method advances the state-of-the-art. This section contains a chronological overview of the last 20 years of techniques used for measuring live power system Th´evenin impedance and its application for DG islanding detection. This review illustrates the difficulty in making accurate measurements of impedance. Many impedance measurements result in either measurement limitations, significant capital costs, power quality reductions, or practical interfacing challenges. Impedance measurements for islanding detection can be broken down into two sub sections of 1. Impulse Response, h(t), and 2. Continuous Injected Signals. In the following Sections, the details and history of these classifications will be reviewed.  35  Chapter 2. Review of Islanding Detection Methods  2.4.1 Impulse Response Impedance Based Islanding Detection using an Impulse, δ(n) One method for determining the island states of a distribution system is to monitor the transfer function or H(e jω ) for changes in the system impedance. Specifically, the transfer function of a system is its frequency response over all frequencies. Some commonly used signal processing techniques used in this section are described in Appendix F. Singular impulse disturbances have been used for impedance measurement as they have the benefit of being short term disturbances with wide bandwidth characteristics. The method requires a disturbance to be created into the power system to create a theoretical impulse of infinite bandwidth, δ(n). As δ(n) contains energy in all bands, V (ω) and I(ω) can be measured to then calculate Z(ω) by means of Ohm’s law as seen in Equation 2.20. Z(ω) =  V (ω) I(ω)  (2.20)  For example, in 1989 Girgis [36] investigated impedance transfer function analysis of a power grid taking historical data using various impulses that were acquired at 12 kV bus with a switched 750 kVAR capacitor (193 ohms or 13µF) to ground as the injected impulse. From the tests, a series of estimated series RLC components were added to match the transfer function poles and zeros of the system as seen in Figure 2.14. The signal processing techniques used to filter the signal were cross correlation and fourier transform. The fourier transform sampling was very course (50 Hz/pt), but the technique was able to demonstrate that the transfer function is not easily predictable like a simple RL circuit, and may contain various resonances. Though the method could also potentially saturate various devices and cause unforseen errors.  Figure 2.14: Impedance Measurement Using High Voltage Capacitive Switching (Girgis [36]) Two methods of measuring grid impedance suggested by DeOliveira [20] in 1991 are to 1. create a disturbance by connecting and disconnecting an industrial load and 2. by connecting a non-linear load and measuring the harmonic content. Laboratory experiments were conducted on both methods 1. by 36  Chapter 2. Review of Islanding Detection Methods connecting various loads 2. using a six pulse bridge as a harmonic source. Both methods demonstrated good results in a lab setting. Measurements of impedance by instantaneous impulses with short sampling time windows of two fundamental cycles, subtracted from each other in a university lab, were made by Staroszczyk[103] in the late 90’s. The main problem discovered was that the system’s state is ”living,” which makes simulation calculations inaccurate or unreliable and experimental measurements challenging. Tests were conducted from a university lab showing consistent expected impedance slopes; however a university lab has the advantage of low voltage coupling and the disadvantage of having high building impedance compared to the utility. Similarly, both Czarnecki [17] and Staroszczyk [103] re-theorized and showed that a switched capacitive impulse on a power grid and advanced signal processing techniques could allow for calculating the transfer function of a power system. The challenge discussed was that the measurement system would have to minimize the disturbance to the customer while maximizing the signal to noise ratio to ensure a signal can be measured similar to Girgis [36]. System impedance was measured in 1998 using switched capacitors but the transformers were found to saturate during the transient, and produced non-linear harmonics up to 1000Hz [94]. Some practical measurements using capacitive switching were made by Nagpal [80] in 1998. Measurements were made of harmonic impedance of three phase distribution feeders with the help of the Canadian utility BC Hydro using continuously switched capacitor banks (five times) up to a usable frequency band of 1.5 kHz. The signal processing used correlation techniques, with statistical accuracy to improve the results. Data acquisition challenges resulted from quantization limitations. Switching noise and fibre optical interconnection was suggested. In 2000, Yao[131] used similar harmonic dv/dt and di/dt changes at the PCC to successfully measure impedance at the PCC. A current injection method for use in systems up to 315 kV was developed by Moreau [76] with Hydro Quebec in 2003. This was accomplished with a capacitively coupled injection made by Isolated Gate Bipolar Transistors (IGBTs) switching, and a spectrum analyzer as the signal source at 120 VAC. The practical results matched the simulations initially made. It should be noted that the noise in the spectral analysis was corrupted at the fundamental harmonics (60·N) and due to other injecting sources, such as arc furnaces and other silicon switching devices. The goal for this paper’s research is to create better models, analyze local sources more thoroughly and to simulate the networks. Impedance measurements have also been attempted using only naturally occurring situations. For example, in 2003, Pedersen [86], estimated short circuit impedance experimentally at 132kV, 230V, and 400kV by waiting for transients to occur, and then measuring the phase, voltage and current differential over many samples to obtain a statistical average over time periods of up to 24 hours. It was shown that different voltage levels require different measurement and disturbance injections to be effective. The drawback of this method is that it waits for transients to occur instead of creating them, hence, the need of statistical values. The naturally occurring δ(t) cannot be used for real-time impedance measurement due to its infrequent occurrence. Finally Xie[130] used transformer energizing inrush current to measure harmonic impedances. It was found that the result of transformer inrush is a 500 Hz band of noise that is injected, and can then  37  Chapter 2. Review of Islanding Detection Methods be used to solve for Equation 2.20. Similarity, disturbances have been accessed using STATCOM, or similar power correction devices in 2002 by Diana[22] who monitored active capacitor banks to obtain injected non-harmonic impedances. In 2002, Xu [127] discussed how capacitor switching harmonic impedance estimation accuracy can be enhanced and calculated in real-time by decomposing capacitive switched transients in the α − β − 0  or Clark transform, instead of symmetrical component transform. Clark’s transform can be seen in Equation 2.21. The subscript “s” is the α − β − 0 and “h” is the harmonic number in Zhs where Z is the impedance. The transform is then processed with the Fourier transform and the harmonic impedances  are calculated as seen in Equation 2.22 where X pre−hs is the voltage or current before the transient, and X post−hs is the voltage or current during the transient.       Va −1 −1 Va √ √ 1        3 −3   Vb  = [T ]  Vb   Vβ  = √  0 √ √ √ 6 V0 2 2 2 Vc Vc Vα      2  Zhs =  Vpre−hs −Vpost−hs Ipre−hs − Ipost−hs  (2.21)  (2.22)  In 1996, DG islanding detection using impulse response was initially published in IEEE by Hopewell [44], who used half fundamental cycle capacitive (20µF ) switching to create disturbances and compared loss of mains (islanded state) to mains connection (utility connected) for distributed generation. As shown in Figure 2.15, multiple inverter coordination and signal corruption was seen as a limiting issue with this technique. Multiple detection devices were discussed, and it was theorized that by out of phase switching between detectors, two DG devices could operate at the same time. Lab and Simulation experiments were run. ZDG  ZUtility  Cswitch VDG  ZLoad  VUtility  Figure 2.15: Impedance Measurement Using TRIAC Controlled Capacitive Switching (Hopewell [44]) Harmonic injector amplifiers were found to be useful for low voltage islanding detection (< 250 VAC), but difficulty arose when synthesizing the injected power requirements. Symmetrical injection can be difficult, but in 2000, Palethorpe [84] used impulse response made from switching transistor bridges using approximated experimental values. Then in Palethorpe’s [85], like his work in [84], with a new approximated network, the power converter bridge was used to make the same measurements to inject signals for impedance measurement with very accurate results with complex linear components up to 2000 Hz. However, when non linear items such as diodes were introduced, the accuracy was at an  38  Chapter 2. Review of Islanding Detection Methods unusable level. Summary of Impulse Response Impedance Measurement Impulse response for impedance measurement and islanding detection has primarily involved the switching of capacitors and power electronic devices to create large enough disturbances that can be measured accurately by CTs and PTs while not consuming real power. The drawbacks of the impulse response method are the practical costs of integration into higher voltage networks, infrequency of naturally occurring changes and power quality reduction making them often impractical for low cost and high penetrations of DG installations. However, it has been shown that if a constant source of noise can be created at a low cost, the impedance measurement technique can be a practical solution for very accurate islanding detection.  2.4.2 Islanding Detection using Continuous Injected Noise Continuously injected signals for impedance based islanding detection addresses the intermittency problem of impulse type injections. They can also be made at a singular frequency and in real-time, hence, reducing the energy requirement. Injection technique inserts a known source, while measuring impedance from dividing voltage and current measurements at a pre-selected frequency. The advantage of using such a technique is that other frequencies near to the dominant fundamental can be made, and if they can be accurately measured, the impedance at the fundamental can be accurately estimated through interpolation. This type of impedance measurement technique is advantageous over the impulse response technique, as it is effective for evaluating the state of a power system continuously in real-time. Where the impulse response is intermittent, it requires an intermittently large transient. A network analyzer was used in 1994 and 1996 by Harris [39] and Rhode [91] respectively, who instrumented grid impedance measurement systems using an HP 3570A network analyzer with a power amplifier injecting non fundamental frequency sweep. The coupling was achieved through a band stop filter in combination with an isolation transformer. The injections were made in the lab (600 VAC or less) through 20 Amp, 30 Amp and 400 Amp feeders. These experiments illustrated the effect of local impedance variations and how this technique was predicably not useful for impedance measurements of fundamental and its harmonics. However, nearby frequencies correlated with the fundamental impedance accurately. In 1997 an injected voltage and a voltage divider to measure the impedance of the power system was used by O’Kane[83], though the power system low pass characteristics and energy requirements were not directly discussed, it can be estimated that to make a measurement around 0.01 pu it can require up to a full load energy related to V 2 . Impedance measurements were then made from continuously injected signals by Di Piazza [21] and Sumner [105] respectively in 2000 and 2001. They modeled and simulated an injected square wave into a power network to obtain impedance values dividing voltage by current to obtain an absolute value of impedance. The technical challenge presented was the amount of energy required to be injected into the system to be able to resolve meaningful measured values. Though simulation tests were made,  39  Chapter 2. Review of Islanding Detection Methods experimental measurements on actual systems were still required. A switching inverter that could be programmed for this algorithm was built by Sumner [106] in 2002, using a similarily controlled current injection square-like wave as with Di Piazza [21] and Sumner [105] to measure the impedance of a system using switching inverter devices as the injection source. The advantage of the system was that the injected signal is computer controlled so it can be executed continuously on demand, and as a current controller, it can be used to limit the size of the current disturbance. Tests were not verified on actual power systems, although the method demonstrates effective signal to noise ratio. Then Sumner [107] (same author as [106] and [105]) used the same system to estimate impedance to actively control shunt loads in a laboratory setting. In 2002, an injection device was used on 77kV and 230 kV HVDC converter sites in Japan by injecting a 0.3% amplitude signal to measure impedance between 30 Hz and 400 Hz [112]. The conclusions found that though the results looked promising, more testing is required to increase the accuracy. Signal processing has been accomplished using: 1. A 60 Hz cancelation filter based on a phase locked loop 2. Clarke’s transformation and Park’s transformation 3. Transfer function computation that reduces the effect of background noise In 2003, a patent filed by Hochgraf [43] for islanding detection by measuring impedance through injecting continuous signals from a three phase bridge inverter. The system is tripped when an impedance changes to indicate an island condition has occurred from a known threshold. The concept of this patent is shown in Figure 2.16, where a three phase bridge can be seen to be computer controlled and monitored for impedance at its interconnection terminals.  System  Va  Vb  Vc  Ia  Ib  Ic  Signal  Impedance  Injector  Estimation  +  +  +  +  +  +  Va*  Vb*  Vc*  Threshold Comparison  Power  Rate of  Generator  Change  Figure 2.16: US Patent 6,603,290 Drawing 1: Islanding Detection by Signal Injection One year later, the injection technique was used with islanding detection standard ENS by Timbus [109] (German abbreviation of: Main Monitoring units with allocated Switching Devices) to detect DG 40  Chapter 2. Review of Islanding Detection Methods islands. A periodic injection of 75 Hz every 1 second is made by an inverter, removing the need for a dedicated islanding detection unit. The injection frequency is ”tuned” or chosen to best fit the network conditions. This technique was also discussed by Xu [128] in 2004. Injection was found to be ideal for islanding detection with 1 kW to 4 kW generators, but other smaller sized inverters may have practical power limitations. Injection amplitudes of 0.02 pu to 0.07 pu were tested and it was determined that 0.05 pu offered the best results while keeping within IEEE power quality standards. Multiple inverters were identified as a special case which would need to be addressed to prevent multi inverter signal corruption. In 2004, the same method was used by Asiminoaei [5] and Timbus [109] where impedance was measured by injecting a single frequency into the network. It was found that the amount of random noise rejection of the transient method is a disadvantage and the data acquisition capabilities need to be high and are often un-achievable. Impedance was measured by injecting a 75Hz signal as a non harmonic for 0.040 s at 1.5 Amps. Asiminoaei [5] considered the impedance at 75 Hz to behave similarly to the fundamental of 50 Hz. It was shown the single frequency injection impedance measuring method can be programmed directly into the inverter’s control algorithm for effective islanding detection. The system consists of voltage injected into 220 VAC single phase utility voltages. In 2005, this topic is further improved with Asiminoaei [7] and [6] as an extension of his work [5] and discusses vector method and FFT method at 75 Hz while discussing THD, power injected, and repetition rate vs. the accuracy of the method. Practical measurements were made using a 10-bit AD converter at 1.3kH sampling rate with high levels of accuracy. The experiment was validated with lab experiments, but practical tests were not completed. In 2006, multiple inverter coordination with single signal injections for islanding detection were addressed by Timbus [111] in the inverter used by the same team as Asiminoaei [5] and Sung-Il [108]. They used multiple inverter coordination to inject the 75 Hz signals with another inverter, by offsetting the injections by 0.05 s from each other. Fourier transforms were used to analyze the results and to detect another inverter; a harmonic compensator was used. In 2006, Timbus [110] discusses the amplitude, frequency and current controller gain technique for impedance based islanding detection. The focus is on the ENS requirement of a 0.5 ohm change measured using an injection amplitude of 0.05 pu. Injection rates made infrequently to lower the total harmonic distortion (THD) is averaged where the overall goal of the work was to assess the value of not having a specialized unit for islanding detection but instead placing it right into the inverter itself. In 2005 and 2006, Katirae [65] and Hernandez-Gonzalez [42] respectively, both explore the idea of micro-grid operation and islanding detection. However, [42] explores impedance measurement using current injection to address island conditions. The injection is through the dq axis (Park’s Transform [67]) using frequencies of 1 Hz to 40 Hz. Results are for a single source only, but claim to be effective in simulations detecting in less than 33.3 ms and most effective at 1 Hz at 1% the fundamental. The non-detection zone occurs in quality factors between 3 and 4 in a resonating bus. Finally, using an unbalanced source to create negative sequence injection, a ship’s impedance network stability was monitored through measurement of the Th´evenin impedance. The impedance was found to be accurately measurable through a multi-variable solution of positive and negative sequence  41  Chapter 2. Review of Islanding Detection Methods a  a  b  b  ibinj ibinj Load  Source  + VDC -  c  icinj  icinj c  Injection Concept  H-Bridge Injection  Figure 2.17: Negative Sequence Injection Concept and H-Bridge Injector Realization injection and conversion, then subsequently to (qd0) park’s transform [45] to solve the d and q axis impedance vectors. The negative sequence component is created by inserting a line to line unbalanced source of varying frequency across two of the three phases as seen in Figure 2.17. The impedance is calculated through the assumption that both the negative, Z2 and positive sequence Z1 impedances are equal by linear independence. Hence, unlike the VIP method in [119] where two measurement at different times are needed to assess impedance, these measurements are taken simultaneously. Therefore, a solution of four variables and four equations are seen in Equations 2.23 results. Various simulation measurements and laboratory measurements have demonstrated that this method is very effective.  Vq1 = Zqq Iq1 + Zqd Id1  (2.23)  Vd1 = Zdq Iq1 + Zdd Id1 Vq2 = Zqq Iq2 + Zqd Id2 Vd2 = Zdq Iq2 + Zdd Id2 Summary of Continuous Injected Noise Impedance Measurement Continuous injection noise technique is an effective impedance measurement for islanding detection. The technique has evolved to inverter and switch based converter injections due to the ease of software programming of the frequency spectrum that can be injected. Though this method is proven functional, some remaining challenges for this class of impedance measurement are in: the amount of required energy, practical integration into high voltage networks, multi-generator signal corruption and signal measurability. Recently negative sequence injection has been shown to be effective for impedance measurement for stability assessment [45] and can address some of these existing injection challenges. This thesis extends the concept of impedance measurement by continuous negative sequence injection and utilizes it for islanding detection.  42  Chapter 2. Review of Islanding Detection Methods  2.5 Current and Voltage Measurability Of the challenges detailed for the various islanding detection techniques for impedance measurement, voltage and current measurability has been consistently regarded as a significant limiting factor. Therefore, this section contains a discussion of voltage and current sensor technology used today and will cover the existing installed technology, sensor bandwidth, and accuracy at different voltage levels so that simulation measurements can be addressed as either practical or not practical variations. With the negative sequence impedance islanding detection technique developed in this thesis, some cases use naturally occurring signals at distribution voltage levels where existing sensor current transformers and where voltage transformer (CT and PT respectively) devices are already installed. Sensor accuracy is a key item that can limit the ability for some techniques to be effectively used. In this section, both current and voltage measurement techniques will be addressed individually as they are tackled in slightly different ways and require special consideration. Voltage and current transducers in the field today, are a mix of older transducers and more modern computerized devices. However, the islanding detection technique described in this thesis requires complex signal processing from digitized measurements. Therefore, the focus on current and voltage measurability will be on computerized data acquisition, though it is important to keep in mind that the mechanical side of these sensors will not be discussed here. There is significant literature on the specific topic of high voltage measurement. Five key references were chosen: [100] [78] [29] [73] [35] with IEEE standards for CTs and PTs used in power systems as with [55]. More specific references are also mentioned throughout this section.  2.5.1 Voltage Measurement Safe measurement of power system high voltages can not be directly accessed by computerized equipment and must be reduced by often several orders of magnitude before being connected to a computer. Utility distribution voltages range from residential 0.120 kV to line distribution of 50 kV depending on the area and the application. However, for a typical computerized data acquisition analog to digital converter, voltage levels of +/- 5 V or less are required. Therefore, from typical distribution voltage of 25 kV, a voltage drop of nearly five orders of magnitude with minimal error must be achieved. The techniques to achieve this can be broken down into several groups: 1. resistive divider, 2. step down transformer, 3. capacitive voltage divider, 4. optical electrical field measurement and combinations of all four. Each voltage and current method has better optimal performance than the other techniques at different voltage levels. As the technique described in this thesis has been tested at a span from 0.208 kV to 25 kV, the following sub-sections shall review the four techniques listed here for methodology and performance. Resistive Voltage Divider For low voltage applications (below 5000 V), resistive voltage divider electronics are the most common form used to step down signals for accurate measurement. Therefore, resistive divider ratios from 1:100 to 1:1000 are needed. Precision, laser cut, mega Ohm resistor divider pairs can be purchased at a 43  Chapter 2. Review of Islanding Detection Methods relatively low cost and have an accuracy ranging from 1.0 % to 0.01 %. This type of resistor divider is used with low voltage inverters and for precision measurements at higher voltages. Resistive dividers have the advantage of being predicably linear over a wide range of frequencies from DC up to 100’s of kHz. Purely resistive division requires an isolating circuit protector for unforseen transients and ground loops. This concept can be seen in Figure 2.18 where there are two matched resistors, R1 and R2 , which have ratios of up to  R1 R2  = 1000. Following the two resistors, is an isolating amplifier that taps off the  lower voltage resistor R2 . 0 to 5 kV  R1  Isolation Barrier  + R2  Vmeasured -  Figure 2.18: Resistive Divider Voltage Measurement with Isolator  Voltage Transformer and Coupling Capacitor Voltage Transformers Voltage transformers (PTs) and coupling capacitor voltage transformers (CCVTs) are the most common measurement devices for converting higher system voltages to more safe and usable lower voltage levels. CCVTs are less expensive than standard PTs at higher voltage rated voltage levels, but can have inferior transient response due to their ferroresonance effects. A PT consists of a step down transformer with a burden impedance network for mechanical relay or computer based relay inputs [29]. The equivalent circuit of the PT can be seen in Figure 2.19 and has been modeled in EMTP by Fernandes [30]. The circuit has a transformer equivalent circuit and an output burden impedance. The transformer equivalent consists of a primary impedance R1 and jωL1 that are reflected over to the secondary side by dividing each by the square of the turns ratio, n2 . Following the primary impedance, are R2 and jωL2 in series connecting to the burden load where the voltage measurement is made. In between the primary and secondary impedances are the magnetization resistance and inductance, RFe and jωLM . Under normal operating conditions, the magnetization impedance is near 99 pu [23]. Standardized maximum absolute measurement errors for these devices range around 0.3 %, 0.6% and 1.2 % as specified in [55] and [57]. As the current draw for PTs are very low, transformer saturation is not an issue as it can be with current measurement. Conversely, the other common alternative developed for safe high voltage measurement are CCVTs that consist of a combination of a capacitive coupled voltage dividers connected to the step down voltage transformer. CCVTs are more cost effective to PTs at higher voltages (≥ 100kV ) due to a reduction in the isolation transformer size. As seen in the equivalent circuit in Figure 2.20, the voltage division is accomplished through a series of capacitors that are then paralleled by the transformer equivalent 44  Chapter 2. Review of Islanding Detection Methods  R1/n  n:1  j  L1/n  R2  j  Scaled Output  L2  RBurden High RFe  Voltage  j  LM  Connection j  LBurden  Figure 2.19: Voltage Transformer Equivalent Circuit circuit off of the low voltage portion of the capacitive divider chain. Though only two capacitors are shown, these capacitors are often multiple capacitors stacked together, allowing the protection engineer to select specific taps for the voltage level of interest. The other components are the spark gap on the left side of the circuit seen in Figure 2.20, and a resonance suppression circuit on the right. These two additions are added to limit the over voltages during high frequency transients and to limit resonance oscillations during transients. As compared to Figure 2.19, the CCVT’s complexity is higher than the PT with the addition of capacitors. These third order components are most significantly noticed during high speed transients, such as lightning strikes, where the capacitive coupling can act like a short circuit. The shorted high potential signal travels into the measurement system, where the spark gap limits the over voltage into the sensitive circuitry, and the filter reduced the over voltage resonance between the transformer and the voltage divider capacitors. However, under normal operating conditions at the fundamental frequency, CCVTs are as accurate as PTs, but outside of the fundamental frequency, the response is not linear and care must be taken to prevent false readings. Therefore, for steady state voltage measurements used in this thesis, PTs and CCVTs will have comparable accuracy. High Voltage Line     -1/(j  Mc ) 1    Ferro resonance     j  ML  TR  n:1  R1/n  2  j  ML /n 1  2  R2  j  ML  2  suppression Scaled Output     -1/(j    Mc ) 2  Spark Gap     RFe  j  ML  M      Figure 2.20: Coupling Capacitor Voltage Transformer Equivalent Circuit  Optical Electric Field Measurement The final method used for high voltage measurement is with electro-optical phenomena. This phenomena is called the Pockels effect. Though this method is not commonly used due to its recent emergence in the utility market, the optical interface has the distinct advantage of superior electrical isolation char-  45  Chapter 2. Review of Islanding Detection Methods acteristics over PT and CCVT technologies. The Pockels effect (also known as the Linear Electro-optic effect) is used to make Pockels cells, which are voltage controlled transparent crystals that experience change in phase delay at different electric fields. The Pockels cell fundamental can be seen in Figure 2.21. This setup uses the effect that is described by the two Equations 2.24 and 2.25, wherein Equation 2.24, Γ is the induced phase lag between two orthogonal light components, λ is the wavelength of light, ∆n is the induced decomposition of the light into two rays (birefringence), and l is the crystal length. Γ=(  2π ∆nl) λ  (2.24)  Then in Equation 2.25, Vπ is defined as the value of the cell voltage at which Γ reaches π and λ is the light wavelength, n0 is the ordinary refractive index of the crystal, and rxy is electro optic coefficient. In the case of a commonly used Pockels cell, a BGO crystal, rxy = 1.03× 10−12 V/m and n0 = 2.098, where a light wavelength, λ equal to 632.8 nm, Vπ is 33.26 kV. Hence, this sensor is best suited for high voltage measurement [62]. Because of the very high dielectric properties of the Pockels crystal, the cells behave like capacitors which in high voltage applications, the current inrush can pose measurement problems. Industrial grade Pockels sensors are available that have a linear bandwidth from 30 Hz to 3000 Hz and an accuracy ranging between 0.3% to 0.2% for all voltage ranges. These devices are available for voltages above 100 kV [15] [61]. Greater accuracy can be made with larger, but prohibitively expensive crystals, hence, these items are only found in research laboratories at this time. Vπ =  λ  (2.25)  2n30 r41 High Voltage  Laser  Polariser  l/4 Plate  Pockels Cell  Detector  Figure 2.21: Pockels Effect Voltage Measurement  Summary of Voltage Measurement To summarize voltage measurement techniques, there are practical absolute limits for most sensors that are in the range of 1% to 0.1%. Although, measurements beyond these limits are not necessarily useful for the tools that protection engineers currently use, better accuracy can allow for enhanced system state estimates, as the technique discussed in this thesis shall reveal.  46  Chapter 2. Review of Islanding Detection Methods  2.5.2 Current Measurement Similar to voltage measurement, there are a series of different methods for measuring current in power systems. The current measurement techniques used in power systems are an optimization between cost, practicality, target relay and accuracy. The most common methods for current measurement used today are: 1. High and low side resistive shunts, 2. Current transducers, and 3. Opto-magnetic current measurement. Each of these methods have a specific application and advantage over the other where the general accuracy of all three commercial available items are similar. High and Low Side Resistive Shunt High and low side resistive shunts are used in low voltage (less than 1000 V) inverter applications due to their simple installation and high bandwidth. This method consists of a very low value resistor, such as 0.01Ω, in series with the output terminals; with a differential amplifier used to measure the voltage drop across the resistor. Low side measurement places a series resistor along the ground return path, while high side measurement places a series resistor inline with the high voltage output terminals. Low side current measurement carries the disadvantage of inaccuracy if the system has ground loops and can only be used with single phase inverters. For high side current measurement, as the source voltage rises above several hundred volts, or if there is a significant common mode corruption, this method also becomes significantly impractical and unsafe. High side measurement can be seen in Figure 2.22. The accuracy of these sensors range from 1% to 0.1% of the full load current rating. Rshunt Phase A Rshunt 0 to 300~VLN  Phase B  To Load Rshunt  Phase C  +  -  +  -  +  -  Isolation  IA_Measured  IB_Measured  IC_Measured  Figure 2.22: Resistive Shunt Current Measurement  Current Transducers Current transducers are the most common tool used for current measurement in power systems. They have the advantage of protecting both personnel and other equipment from high voltage levels, and 47  Chapter 2. Review of Islanding Detection Methods provide a reduction from high current carrying conductors to the measurement equipment. To measure line currents, a current step down transformer concept is utilized with an iron core around the high voltage current carrying conductor, which has an associated step down coil around the core, as seen in Figure 2.23. The induced voltage in the secondary coil can be very high, as described by Maxwell’s Equation “Faraday’s Law” and seen in Equation 2.26 where Φ(t) is the magnetic flux, and vturn is the volts per turn. In high voltage systems, the magnetic flux is very large and requires the secondary burden impedance be as low as possible so dangerously large voltages are not produced at the terminals. The polarity of the transducer is dependant on the winding characteristics. vturn (t) =  ∂Φ(t) ∂t  (2.26)  The construction of low and high voltage transformers differ in only two ways. The first is the voltage isolation and the second is the ferrite core size. The ferrite core must have the magnetic flux carrying capacity that will not saturate under normal operation and during fault conditions. This requires that the iron core be very large for high power systems. The equivalent circuit for a current transformer is similar to Figure 2.19 with the difference being that the current transformer primary terminals are connected in line (i.e. in series) with the power system, and the voltage transformer is connected line to line. When a current transformer’s core enters saturation, the transformer’s response becomes non linear resulting in inaccurate measurements. Therefore, it is critical for the protection system’s proper operation that measurements can be reliably and accurately made and that the CT does not saturate.  IH  High Voltage Line  RLine  RLine  VB  IH/N  Burden: ZB  Figure 2.23: Current Transformer Circuit with Burden Low voltage and current transformers can be purchased at a relatively low cost with an accuracy of 0.1% or less. However, with larger power system current transformers, accuracy must be carefully  48  Chapter 2. Review of Islanding Detection Methods assessed at the system’s rated current and voltage. The magnetization current of the transformer is not zero, so the assumption of IPrimary = IBurden /N where N is the turns ratio is not accurate and the magnetization current must be taken into account. The accuracy of the measurement at the burden impedance will depend on: the burden impedance, the connecting lead impedance, the current divider ratio, the DC offset and the ferrite residual excitation characteristics. For example, the voltage at the secondary current transformer terminals for a three phase fault VS , is as seen in Equation 2.27 where IL/N is the maximum secondary current, ZB is the connected burden impedance and ZLead is the connection leads and transformer connection impedance [2]. When the current is below the normal transformer operating conditions, the magnetization impedance is very high, however, as the flux density approaches the saturation point of the ferrite core, the excitation impedance drops and the measurement error significantly increases both in amplitude and in phase. The supplier of each current transducer will supply the magnetization curve of the transducer so estimates on the transducer’s accuracy can be properly estimated. VS = IL/N (ZL + ZLead + ZB)  (2.27)  High voltage CT accuracy would normally be difficult to assess due to the large number of CTs out in the field; however, the IEEE standard C57.13-1993(R2003) [55] specifies the standardized characteristic burdens and accuracy classes with the worst case allowable accuracy. For example, power revenue metering CTs have an accuracy of less than 0.3% under full load. This value is critical for proper revenue metering and for protective calculations. When taking field measurements of current transformers, it is important to understand the measurement device so that proper protective decisions can be made for a high level of system security and dependability. Optical Magnetic Field Measurement The final form of current measurement used in power systems is through magneto-optical effects. Unlike the electrical field based Pockels effect, the Faraday effect (or Faraday rotation) is a magneto-optical phenomenon where the interaction between light and a magnetic field change the polarization of the transmitting light. The relationship occurs when the plane of polarization rotates proportionally to the intensity of the magnetic field that is in the direction of the beam of light. The Faraday effect was discovered by Michael Faraday in 1845 which was some of the first experimental evidence of how light and electromagnetism relate. The Faraday effect occurs in many optically transparent and dielectric materials. The relationship between the angle of the optical rotation of the wavelength polarization and the magnetic field is seen in Equation 2.28 where β is the angle of rotation, B is the magnetic flux, d is the length of the path of where the magnetic field interacts with the material, and ν is the Verdet constant for the material. The Verdet constant for most materials is found to be extremely small and wavelength dependent. For example, typical glass-like materials have constants ranging between 0.02 and 0.089 min/Gauss-cm. Measurement errors for these sensors are linear throughout their operating region at  49  Chapter 2. Review of Islanding Detection Methods 0.2% or better [14] [1]. Similarly to Pockels effect, magneto-optical current measurement has superior electrical isolation and no saturation effects over current transducers. However, they are a relatively new commercial product in the utility industry and do not have high installation density. β = νBd  (2.28)  Summary of Current Measurement Current measurement in power systems primarily consists of current transducers. These items have a long product life, hence, when many of them were installed. Over the past 50 years, In North America, there has been little need to replace them. The standardized accuracy depends on their use, but it will range between 1% and 0.1% depending on where and when it was installed. Current measurement in power systems plays a critical role for the protection of power systems. The islanding detection technique described in this thesis has a higher level of performance, depending on the accuracy of the sensors. Therefore, the highest accuracy sensors will offer some of the best results.  2.6 Summary of Research Background This chapter has covered the background topics that are relevant to the current state-of-the art for this thesis’ key contribution of negative sequence impedance measurement for islanding detection. These topics are: distributed generator protection motivation, islanding detection techniques, impedance measurement for islanding detection, and current and voltage measurement techniques and accuracy in power system applications. The installation of distributed generation systems create significant new challenges for the protection engineer. These challenges can be improved with a dependable and secure locally running islanding detection system. Islanding detection can be used to provide additional protection for service workers, assist in limiting voltage surges, reduce short circuit currents, and allow for re-closure functions to operate as they were designed (before the installation of the distributed generator). There are many forms of islanding detection available with some associated standards to support and test them. Of all the forms available, there are various performance characteristics, depending on the application, load types, generator installation density and power system strength. Impedance measurement techniques are the only methods that have a theoretically negligible non-detection zone. Unfortunately, of all the impedance based techniques, they all either have sensitivity, threshold settings, installation cost or practical continuous real time measurement limitations. Additionally, of the sensitivity limitations of impedance measurement techniques, existing power system voltage and current measurement accuracy limits overall effectiveness. There are many techniques commonly used for measuring voltage and current, ranging from parallel/series impedance shunts, step up/down transformers and more recently, electo-optical and magneto-optical effects. However, with all the voltage and current measurement techniques available to the protection engineer, the most common and frequently heavily installed methods at distribution levels are voltage and current 50  Chapter 2. Review of Islanding Detection Methods transformers. Therefore, impedance measurement techniques used today for islanding detection must fall into the physical measurement limitations of a maximum practical accuracy of 0.1%, and a typical accuracy to be expected to match the IEEE accuracy standard of 1% to 0.3 % of the rated load. In the following chapter, a logical increment in the impedance measurement for islanding detection is introduced. This method uses negative sequence impedance measurement, while taking practical and simulations from various locations to demonstrate its effectiveness. The following Chapter 3, first introduces the concept theoretically, followed by Chapter 4 which analyzes some practical case studies.  51  Chapter 3  Negative Sequence Impedance Islanding Detection 3.1 Introduction This chapter contains the derivation and description of negative sequence impedance island detection for distributed generation. Negative sequence impedance islanding detection is a novel form of impedancebased islanding detection that can either use naturally occurring or injected negative sequence currents and voltages. The negative sequence impedance islanding detection technique operates off the principle, that an unbalanced load (or source) has a negative sequence current source at the point of the unbalanced load’s interconnection. Unlike its positive sequence counterpart that are from generators, as seen in Figure 3.1, the negative sequence currents are created at the unbalanced load (if the rest of the system is balanced). Current flows are different between DC systems and AC systems. In a DC system, the direction of current is identified by the sign. In an AC system, the phase angle and amplitude referenced to the voltage dictates the direction of power flow. In Figure 3.1, subscripts 1 and 2 indicate positive and negative sequence components respectively, and subscript UL and BL stand for “Unbalanced Load” and “Balanced Load” respectively. In Figure 3.1, E1 is the positive sequence source, I1 is the positive sequence current, I2a−UL and I2b−UL are the unbalanced load currents from the unbalanced load; UL, V1−UL and V2−UL are the positive and negative sequence voltages at the unbalanced load, I1−BL is the positive sequence current into the balanced load, and V1−BL and V2−BL are the positive and negative sequence voltages at the balanced load. From Figure 3.1, the impedance, Zsource can be expressed as in Equation 3.2 when the negative sequence current is used. This method of impedance measurement allows for the negative sequence components to be used as a real time anti-islanding sensor relay by measuring the impedance away from the unbalanced node and comparing it to a known threshold before and after the island state occurs. The negative sequence impedance islanding detector method requires two assumptions to function correctly: 1. At the distribution level, a power system is rarely balanced, and 2. The utility source is typically much stronger than the distributed generator in the islanded branch as seen in Equation 3.1. The first assumption can be enhanced as discussed in previous impedance measurement techniques [45] by externally creating unbalanced conditions at the cost of the degradation of the system’s power quality. Other factors to consider when using externally injected signal methods are the amount of required energy, practical integration into high voltage networks, multi generator signal corruption, and 52  Chapter 3. Negative Sequence Impedance Islanding Detection  E1  I1  V1-UL  V1-BL  V2-UL  V2-BL  I2a-UL  I2b-UL  I1-BL  Source Unbalanced  Balanced  Load (UL)  Load (BL)  Figure 3.1: Negative and Positive Sequence Current Flow signal measurability. The second assumption is important because the islanding detection triggering threshold can be set without field measurements. In absence of the second assumption, field measurements will be required. Z2−ISLAND ≫ Z2−UT ILIT Y −CONNECT ED ZSource ≈ −1 ·  V2−UL I2a−UL  (3.1)  (3.2)  The negative sequence impedance islanding detection is an improvement upon previous impedance based anti-islanding techniques. As discussed in Section 2.4, the four limitations associated with impedance measurements that negative sequence impedance anti-islanding detection improves upon are: 1. Sensor threshold setting, 2. Real-time measurability, 3. Power quality, and 4. Cost and power requirements for practical implementation. These four improvements are discussed further here in more detail. The first improvement that negative sequence impedance anti-islanding detection offers to impedance based islanding detection techniques is the setting of an appropriate triggering threshold for islanded and un-islanded network impedances. Since the negative sequence impedance of a utility network is typically significantly lower than a distributed generator’s impedance, the threshold detection window is very large. The second improvement to impedance-based islanding detection is the capability of real time continuous monitoring. Most impedance based islanding detection techniques require periodical switching (such as with capacitors) or harmonic injections coupled with transfer function estimations. Negative sequence impedance islanding detection can use the continuous naturally occurring unbalanced conditions to estimate the system impedance. The third improvement is the power quality. Passively measuring negative sequence impedance has no effect on the power quality, however, actively injecting negative sequence currents can cause some corruption of the system’s balance. Unbalanced conditions only effect three phase machines opposed to harmonic injections that effect both single and three phase machines. The fourth improvement of negative sequence impedance anti-islanding detection over the existing impedance measurement techniques is the cost of implementation and power required. Negative sequence impedance islanding detection is a sensor and software solution that can use existing utility CT and PT sensors with an intelligent software addition to a relay system. Negative sequence 53  Chapter 3. Negative Sequence Impedance Islanding Detection impedance islanding detection also does not use expensive large power coupling components such as high VA transformers and capacitors that most other techniques based on injection require. The method can be summarized using Equations 3.1 and Equation 3.2 which are stated at the beginning of this Section. To fully understand the concept of negative sequence impedance islanding detection, this chapter derives the method of negative sequence impedance measurement, demonstrates how the the unbalanced load is a source of negative sequence current, analyzes the performance characteristics over a wide range of scenarios and finally reiterates how the concept expressed in Equations 3.1 and 3.2 can be used as an anti-islanding system for use with distributed generators.  3.2 Derivation of System Negative Sequence Impedance Estimation The negative sequence impedance islanding detection method uses the symmetrical component transform properties to measure the external system impedance. The symmetrical component vector transformation was introduced by Fortescue in 1918 [120] to decouple three phase line interdependencies into three linearly independent systems. The transform requires a matrix operator, A, and the vectorial form of the ABC voltages or currents to convert the unbalanced system into three balanced independent systems called Positive, Negative and Zero sequence, and they are indicated using subscripts 1, 2, 0, respectively, as illustrated in Figure 3.2 and detailed in Appendix B. To introduce the derivation of system negative sequence impedance, a simplified DC system and a three phase system will be presented for comparison.  Figure 3.2: Symmetrical Component Conversion  3.2.1 Two Circuit DC System Consider an example of two nearly identical DC circuits seen below in Figure 3.3 where Eo is the source voltage, Zx is the transmission line impedance, and Zy1 and Zy2 are differing load impedances for each circuit. The two circuits can be described by Equation 3.3 and Equation 3.4 seen below. The Load resistances Zy1 and Zy2 are not the same value in each circuit so the voltages VA = VB and the currents IA = IB . This means that each load is consuming a different power.  54  Chapter 3. Negative Sequence Impedance Islanding Detection  Zx  + Eo  Zx  VA  IA  +  Zy1  Eo  -  VB  IB Zy2  -  Figure 3.3: Two DC Circuits with Different Load Impedances  VA + IA · Zx = Eo  (3.3)  VB + IB · Zx = Eo  (3.4)  If Eo and Zx are not known for either circuit, Zx can be found by measuring VA , IA , VB , IB . Equations 3.3 and 3.4 are combined together as seen in Equation 3.5 and then Zx is solved as seen in Equation 3.6. The solutions for Zy1 and Zx can be compared as seen in Equations 3.7 and 3.6. The solution of Equation 3.6 is only possible when Zy1 = Zy2 . VA + IA · Zx = VB + IB · Zx  (3.5)  VA −VB = Zx IB − IA  (3.6)  VA = Zy1 IA  (3.7)  3.2.2 Simple Three Phase System Extending the same concept seen in the previous Section 3.2.1, Zx solution can also be applied to three phase systems. Consider an example of the system seen in Figure 3.4. All variables are similar to the previous example except the source voltages are sinusoidal: EA = Eo 0, EB = Eo 240, EC = Eo 120. If there is no interphase coupling, the circuit in Figure 3.4 has been expanded into its single phase components as seen in Figure 3.5 and organized so that the three individual circuits can be clearly seen in Figure 3.6. The system of equations for the network can be seen in Equations 3.8, 3.9, 3.10. VA + IA · ZxA = EA  (3.8)  VB + IB · ZxB = EB  (3.9)  VC + IC · ZxC = EC  (3.10)  55  Chapter 3. Negative Sequence Impedance Islanding Detection  [Zx] [IABC] [EABC]  [VABC] [Zy]  Figure 3.4: Three Phase Circuit (Single Line) Example with Different ‘Y’ Load Impedances  If the three source voltages are of the same magnitude and 120 degrees apart, the sum of the three voltage vectors will be equal to zero as seen in Equation 3.11. EA + EB + EC = 0  (3.11)  Then Equation 3.11 can be set to equal the sum of Equations 3.8, 3.9, 3.10 as seen in Equation 3.12. 0 = VA + IA · ZxA +VB + IB · ZxB + EB +VC + IC · ZxC + EC  (3.12)  If ZxA = ZxB = ZxC and ZyA = ZyB = ZyC , Zx , Equation 3.12 can be solved for as seen in Equation 3.13. Normally, if the Zy resistors were all the same, this solution would equate to a non-number, but because of the differing Zy resistances, the solution is non zero. Zx = −1 ·  VA +VB +VC IA + IB + IC  (3.13)  Consider the following more specific example. If the impedances are all considered to be purely resistive then all the voltages and their associated currents are in phase (Figure 3.4). However, when ZyA = ZyB = ZyC the difference of the unbalanced currents is 180 degrees out of phase of the voltage. Take a case for Figure 3.4 when EABC = 1 V , Zx ABC = 20 Ω, ZyA = ZyB = 1000 Ω and ZyC = ZyA + Zunbalance = 1300 Ω where all impedances are resistive. The resulting voltages and currents are described in Equation 3.14 to Equation 3.19.  VA = V 0 = 0.9804 0  (3.14)  VB = V 240 = 0.9804 240  (3.15)  VC = (V + x) 120 = 0.9848 120  (3.16)  56  Chapter 3. Negative Sequence Impedance Islanding Detection  ZxA IA  EA EC  VA  ZyA EB  IB ZxB  ZyB  ZyC  VB  VC  ZxC IC Figure 3.5: Three Phase Circuit Example Expanded with Different ‘Y’ Load Impedances  IA = I 0 = 0.9804 · 10−3 0  (3.17)  IB = I 240 = 0.9804 · 10−3 240  −3  IC = (I − y) 120 = 0.7576 · 10  (3.18) 120  (3.19)  The difference between the voltages and currents are in Equation 3.20 and Equation 3.21. Notice how the phase difference between the two is 180 degrees. VA +VB +VC = 0.0045 120  (3.20)  IA + IB + IC = 2.2282 · 10−4 − 60  (3.21)  Employing Equation 3.13 using the sum of V and I, the result is Zx = 20 Ω as derived. Zx = −1 ·  VA +VB +VC 0.0045 120 = −1 · = 20 Ω IA + IB + IC 2.2282 · 10−4 − 60  (3.22)  These two examples illustrate how similar circuits can be used in combination to solve for resistances in the reverse direction of the current flow. In a real power network, the network is much more complex with mutually coupled components. In the following sections, this concept is be extended to a general solution by decomposing the phases into the negative sequence symmetrical component.  57  Chapter 3. Negative Sequence Impedance Islanding Detection ZxA  VA  IA  EA  ZyA  ZxB  VB  IB  EB  ZyB  ZxC  VC  IC  EC  ZyC  Figure 3.6: Three Phase Circuit Example Expanded and Logically Grouped  3.2.3 Balanced AC Systems A balanced system can be seen in Figure 3.7. The variables EA , EB , EC are the source voltages, Zsys is the impedance of the system up to the load, and ZA , ZB , ZC , ZG are the specific impedances of each phase of the balanced load and the ground impedance. The impedance matrix solution can be seen in Equation 3.23 where [V ], [I], and [Z] represent the vectors of voltage, and current and the matrix of impedances of the system. The resulting impedance matrix of the load, [ZLoad−ABC ], is composed of a combination of ZA , ZB , ZC , and ZG impedances from the original three phase system, as seen in Equation 3.24. [VLine−ABC ] = [ZLoad−ABC ] · [ILoad−ABC ]     [ZABC ] =    ZA + ZG  ZG  ZG  ZG  ZB + ZG  ZG  ZG  ZG  ZC + ZG  (3.23)       (3.24)  For a balanced impedance matrix, ZA = ZB = ZC , and the system has three distinct eigenvalues. This can be seen in Equations 3.25 to 3.30 where the symmetrical impedances, Z012 , are solved for in Equations 3.25 to 3.27 to and the result is shown in Equation 3.30. The variable ‘A’ is the symmetrical  58  Chapter 3. Negative Sequence Impedance Islanding Detection  [ZSys]  V  Line A  V  Line B  V  Line C  ZA EA  EB  EC  ZB  ZC  ZA= ZB= ZC (ZLoad) ZG  Figure 3.7: Three Phase Circuit with a Balanced Load  component operator defined in B.2 in the following appendices. The solution in Equation 3.30 can be illustrated in a conceptual schematic where the three symmetrical components are all independent systems to each other, as seen in Figure 3.8. Since only the positive sequence system has a voltage source, calculations of the voltage and current for this system can be reduced to only one simple circuit where the current flows from the source to the load. Also, it is notable that in many practical systems Z1 is equal to Z2 . Though Z1 is not exactly equal to Z2 such as with rotating machines, for this example, this assumption is made because transmission Z is commonly much greater than the machine Z. More specific details on machine symmetrical impedances will be discussed in further sections.  A  −1  A ·V012 = ZABC · A · I012 · A ·V012 = A Z012 = A  −1 −1  · ZABC · A · I012  (3.26)  · ZABC · A  (3.27)  Z012 = A−1 · ZABC · A  ZA + ZG ZG ZG  ZG ZA + ZG ZG = A−1 ·   ZG ZG ZA + ZG   ZA + 3ZG 0 0   0 ZA 0  =    0 0 ZA  59  (3.25)  (3.28)    ·A   (3.29)  (3.30)  Chapter 3. Negative Sequence Impedance Islanding Detection  E1  Z0 sys  Z2 sys  Z1 sys  Z1 Load  Z2 Load  Positive Sequence  Negative Sequence  Z0 Load Zero Sequence  Figure 3.8: Balanced Symmetrical Component Circuits from Balanced Load  3.2.4 Unbalanced AC Systems In a balanced system as seen in Section 3.2.3, the negative and zero sequence currents and voltages are zero and linearly independent of each other (Figure 3.8); however, in an unbalanced system, this property is no longer true. The result of an unbalanced load shall be further explored with a similar circuit containing the Th´evenin impedance of a balanced grid and an unbalanced load as seen in Figure 3.9.  [ZSys]  V  Line A  V  Line B  V  Line C  ZA EA  EB  EC  ZA  ZB  ZB  ZC  ZC  (ZLoad) ZG  Figure 3.9: Unbalanced Load and Balanced Source Circuit  In Figure 3.9, solving for V012 at VLine between ZSys and ZLoad increases in complexity when the ABC phase impedances of the load in ZLoad are such that ZA = ZB = ZC . When solving for the sequence components at the point VLine , an impedance voltage divider of sequence components can be used. The voltage divider for Figure 3.9 is seen in Equation 3.31 to 3.34 where ZLoad is given by equation 3.35, ZSys is a balanced load given by Equation 3.37, and E012 is a balanced source given by Equation 3.38, where there is only positive sequence voltage generated. Matrix [I3 ] in Equation 3.32 is the three by three identity matrix.  60  Chapter 3. Negative Sequence Impedance Islanding Detection  [VABC ] =  [ZLoad ] · ([ZSys ] + [ZLoad ])−1 · EABC  (3.31)  A−1 · [VABC ] = A−1 · [ZLoad ] · ([ZSys ] + [ZLoad ])−1 [I3 ] · [EABC ]  (3.32)  [V012 ] = A−1 · [ZLoad ] · ([ZSys ] + [ZLoad ])−1 A · A−1 · [EABC ]  (3.33)  [V012 ] = A−1 · [ZLoad ] · ([ZSys ] + [ZLoad ])−1 · A · [E012 ]  (3.34)    ZA + ZG    [ZLoad ] =    ZG ZB + ZG  ZG    ZG       ZG  ZG ZG ZC + ZG  ZG ZG ZA + ZG   [ZLoad (012)] = A−1 ·  ZG ZB + ZG ZG  ZG ZG ZC + ZG   ZSys    [ZSys (012)] = A−1 ·  0  0   0 ZSys 0  0    0      0 ·A  ZSys    E012 =  E1  0  (3.35)      ·A   (3.36)  (3.37)  (3.38)  Consider an example system where E1 = 1 pu, ZSys = 41i pu, and ZA = 500 − 30% pu , ZB =  500 + 30% pu, ZB = 500 pu and ZG = 0 pu. Solving for V012 from Equation 3.34 results in Equation 3.39. With an unbalanced condition not only having a positive sequence voltage, V012 contains negative  and zero sequence components, and the “conceptual” result causing the negative sequence voltage can be seen in Figure 3.10. Unfortunately, the true system is not quite as simple as the one shown in Figure 3.10 as will be discussed with the calculation of current flow from the sequence impedance matrix in Section 3.2.5   −0.0124 − 0.0047i      V012 =  0.9200 + 0.0000i  pu −0.0124 + 0.0047i  61  (3.39)  Chapter 3. Negative Sequence Impedance Islanding Detection  E1  Z2 sys  Z1 sys  Z0 sys  V2  V0  Z0 Load  Z2 Load Z1 Load Positive Sequence  Negative Sequence  Zero Sequence  Figure 3.10: Symmetrical Component Concept in an Unbalanced System  3.2.5 Negative Sequence Current Flow in Unbalanced Loads The circuit in Figure 3.10 has a current source produced at the load for negative and zero sequences. Is the current flowing in to or out of Z2 load and Z0 load ? From Figure 3.10, the current appears to flow out of Z2 load and Z0 load and into ZSys . What is the real schematic where the current connects from the positive sequence circuit to create the current in the negative and zero sequence networks? To evaluate this, Equation 3.36 can be expanded to realize its symmetrical component impedance matrix. The result of this expansion, now referred to as Z012 , is a complex matrix listed in Equations 3.40 to 3.48. If ZA , ZB , and ZC are set unequal to each other, then there are no eigenvalues and the simplification seen in Equation 3.30 will not occur. Therefore, the system needs to be fully expanded into its individual matrix components as in Equations 3.40 to 3.48.  Z012 (1, 1) = Z012 (1, 2) =  1 · (ZA + ZB + ZC ) + 3ZG 3 √ √ 3 3 1 1 1 ZA − ZB − ZB i − ZC + ZC i 3 2 2 2 2  (3.40) (3.41)  Z012 (1, 3) = Z012 (1, 2)∗  (3.42)  ∗  (3.43)  Z012 (2, 1) = Z012 (1, 2) 1 Z012 (2, 2) = · (ZA + Zb + ZC ) 3 Z012 (2, 3) = Z012 (1, 2)  (3.45)  Z012 (3, 1) = Z012 (1, 2)  (3.46)  ∗  (3.44)  Z012 (3, 2) = Z012 (1, 2)  (3.47)  Z012 (3, 3) = Z012 (2, 2)  (3.48)  Further substitutions can then be made to simplify the result by inserting the newly defined variables ‘k‘ and ‘y‘ as seen in Equations 3.49 and 3.50. When ‘k‘ and ‘y‘ are placed into Equations 3.40 to 3.48, the result is the simplified form of the system shown in Equation 3.51.  62  Chapter 3. Negative Sequence Impedance Islanding Detection  1 k = − (ZB + ZC ) √6 3 y = (ZC − ZB) 6    [Z012 ] =   1 3 ZA − 2k + ZG 1 3 ZA + k − yi 1 3 ZA + k + yi  (3.49) (3.50)  1 3 ZA + k + yi 1 3 ZA − 2k 1 3 ZA + k − yi  1 3 ZA + k − yi 1 3 ZA + k + yi 1 3 ZA − 2k      (3.51)  This new simplified matrix in Equation 3.51 can be seen to have only real components in the diagonal and imaginary connecting components in the rest of the impedance matrix. Unlike the balanced solution that originally had only diagonal components, the symmetric component transformation results in a filled and complex impedance matrix. This matrix can be expanded and broken into mutual, independent and imaginary matrix components as seen in Equation 3.52. The first two parts are symmetric with the last ’y’ matrix being imaginary and unsymmetrical.  [Z012 ] =    1 1 1      1    (ZA + k) ·  1 1 1  +  3 1 1 1  −3k + ZG  mutual    0 0  0  0  −3k  0  0  independent  −3k        0 y −y   +  −y 0 y ·i y −y 0  (3.52)  imaginary  This can be further realized into a physical representation if the assumption of y << 3k can be made so that only the first two matrices are used and not the ’y’ imaginary matrix seen in Equation 3.52. The advantage of such an approximation is that the system is symmetrical and can be realized as illustrated in Figure 3.11. When does this approximation of y << 3k becomes inaccurate? Set a limiting boundary condition of one order of magnitude for accuracy (10%). Therefore, if  3k y  > 10 the approximation to remove  ’y’ can be assumed valid. This ratio can be expanded as seen in Equation 3.53, and simplified in the solution in Equation 3.54. Therefore, the unsymmetrical ’y’ matrix can be ignored when  ZB ZC  ≤ 0.8908  (and all other perturbations of ZA , ZB , ZC ). Therefore, in a typically unbalanced system, the amount of unbalance may reach up to 100% with a typical value of 10% − 30% and if the impedances differ by less than 10%, it can be seen that the approximation is valid. In other words, this approximation is valid  63  Chapter 3. Negative Sequence Impedance Islanding Detection when Zsys << ZLoad . 3k = y  1 ZB + √6 3 6 ZC −  1 6√ZC 3 6 ZB  √ ZB 3−1 √ > 10 1+10 3 ZC > 0.8908  (3.53)  (3.54)  Using the approximation to the unsymmetrical matrix y, ZLoad (012) can be realized in a schematic with the balanced ZSys (012) impedance as seen in Figure 3.11. What can be noticed in this realization of impedances is that the voltage source E1 remains as the only source; however, the source of the currents are quite easily revealed to be a result of I1 being divided up at node VM . Clearly from the schematic in Figure 3.11, it can be seen that the current from the Z0 and Z2 sequences will indeed be sourced from the E1 , the system load; that is, flowing away from node Vm and not into it.  Figure 3.11: Symmetrical Components Current Flow in an Unbalanced System Expanded Circuit  From the current and voltage derivation in symmetrical components with this configuration, the current flows out of the unbalanced load and into the system allowing the use of ohms law for negative sequence components to be as stated in Equation 3.2.  64  Chapter 3. Negative Sequence Impedance Islanding Detection  3.2.6 Unbalanced Sources Injection Similarly to unbalanced loads, unbalanced sources can also create negative sequence current flow. Unbalanced systems are normally considered undesirable due to the excessive reverse ABC sequence current flow in machines; however, the concept of adding an additional small unbalance into the system may only have a marginal effect on the power quality. There are several key advantages of using an unbalanced source to create its own negative sequence components. The first advantage is an islanding detection scheme is required at the DG source, and the scheme can be directly installed into an inverter or other switched power source. The second advantage of using an unbalanced source for negative sequence current injection, is several “intelligent” measurements can be made at different injection levels hence increasing the accuracy of the impedance measurements. The final advantage of using an unbalanced source is it is possible that the source could actively control the injections based on the already existing unbalance caused from other loads. By controlling the unbalance, the injections could effectively improve the measurement accuracy while reducing the overall system unbalance amplitude. To evaluate this concept, the schematic in Figure 3.7 is re-analyzed with unbalanced sources instead of unbalanced loads. This concept is expressed in Equation 3.55 which is similar to Equation 3.32 for unbalanced loads, with the difference that an unbalanced EABC has been added instead of E012 . In this case, all impedances per phase are balanced so that ZA = ZB = ZC . The sources will be symmetrically unbalanced as seen in Equation 3.58 where the variable ′ PP′ is the per cent of unbalance, such as 1%, 2%, 3%.... Variations of ′ PP′ control the amount of unbalanced current flow. Equation 3.55 is a combination of Equations 3.56 to 3.58. [V012 ] = A−1 · [Zload ] · ([ZSys ] + [Zload ])− 1 · A · [A−1 · [EABC ]]   ZA + ZG  ZG  ZG  ZG  ZB + ZG  ZG  ZG  ZG  ZC + ZG    [ZLoad (012)] = A−1 ·      ZSys    [ZSys (012)] = A−1 ·  0  0   0 ZSys 0  EA  0      0 ·A  ZSys      EABC =  EA · (1 − PP)  EA · (1 + PP)      ·A   (3.55)  (3.56)  (3.57)  (3.58)  Once Equation 3.55 is fully evaluated and expanded (which is too large to include in this work) and then similar variables are collected and canceled, the result is a significant simplification for sequence components V1 and V2 , which can be seen in Equations 3.59 and 3.60. These equations are similar to the impedance division of two series connected loads. The controlling factor for V2 is directly propor65  Chapter 3. Negative Sequence Impedance Islanding Detection tional to the ‘PP‘ per cent value. This ideal case demonstrates the ability to create a more significant negative sequence source through very small changes at the source, opposed to the much larger changes required at a load. To compare the effect of unbalanced sources to an unbalanced load, the example in Section 3.2.4 is used where an unbalanced load of ±30% gave a V2 of approximately 0.013pu. While an unbalanced source of ±5% produces a negative sequence voltage of 0.028pu. Unbalanced sources offer significant performance impact for negligible unbalance. V1 = EA ·  ZA Zsys + ZA  (3.59)  1 ZA V2 = EA · PP · √ · ·i 3 Zsys + ZA  (3.60)  3.2.7 Negative Sequence Th´evenin Impedance of the Network As seen in the previous sections, unbalanced loads and sources create negative sequence voltage sources (or at least the appearance) and an unbalanced current out of them. Therefore, by using Ohm’s law for negative sequence components at the point of the unbalanced load or source; the results in Equation 3.61 became true. The Z2 impedance measured at that point is the system Th´evenin impedance Z2 (Sys) out of that load. Conversely, measurements of positive sequence voltage and current at an unbalanced load will result in the impedance of Z1 (Load). In general, assuming the system in Figure 3.9, the result of measuring the negative sequence voltage and current at the load is the Th´evenin impedance of the system, as seen in Equation 3.61, with the advantage of no positive sequence sources interfering with the result. −1 ·  V2 (Load) ≈ ZSys I2 (Load)  (3.61)  Consider the following example the employs Equation 3.61. Take the system in Figure 3.9 where the following values apply and the Load is ±20% unbalanced:    [ZLoad ] =     [ZSys ] =   1000  0  0  0  1000 ∗ 1.2  0  0    1000 ∗ 0.8  0     10 + 100 j  0  0  0  10 + 100 j  0  0  0  10 + 100 j    1 0      [EABC ] =  1 240  1 120  66  (3.62)      (3.63)  (3.64)  Chapter 3. Negative Sequence Impedance Islanding Detection At steady state, VLine−ABC is:   0.9853 − 5.65    (3.65)    0.0010 − 5.65    (3.66)    [VLine−ABC ] =  0.9884 − 124.72  0.9802 112.96  and ILine−ABC is:    [ILine−ABC ] =  0.0008 − 124.72  0.0012 112.96  The symmetric component of V and I are:    0.012 − 11.08      [VLine−012 ] =  0.984 − 5.80  0.012 156.00   0.0001 84.63  (3.67)      [ILine−012 ] =  0.0010 − 5.96  0.0001 − 108.29  (3.68)  012 The impedance from −1 · VI012 is:      100.50 84    10 + 100 j        [Z012 ] =  973.53 − 179.8  =  −973.5 − 2.4 j  100.50 84 10 + 100 j  (3.69)  The impedance of the positive and negative sequence components are Z1 = 973.53 + 2.4 j pu and Z2 = 10+100 j pu respectively. The positive sequence impedance is near to the 1000 pu base value while the negative sequence impedance is an exact match to the 10 + 100 j line impedance. The angles for the positive sequence voltage and current are nearly identical. The phase angle difference of the negative sequence voltage and the complex conjugate current is 84 degrees indicating reactive components.  3.3 Performance Characteristics of Negative Sequence Impedance Measurement As derived in the previous sections, negative sequence components existing in a power system can be used to measure the impedance away from an unbalanced load or source. The explanation only covered the most basic of systems. How will this method operate over a wide range of system conditions? This section conducts a study of a series of scenarios based on a practical system1 producing qualitatively 1 Practical  System From a Distribution Network in British Columbia, Canada  67  Chapter 3. Negative Sequence Impedance Islanding Detection viewable graphs over a specified range. The performance characteristics are broken down into six components. These components are as follows: 1. Changing Per Cent of Unbalanced Load 2. Strength of System (impedance) vs. Unbalanced Load Power 3. Varying Power of Unbalanced Load 4. Power Factor of Loads 5. Phase Change From Different Unbalanced Load Configurations 6. Multiple Unbalanced Loads in an Island  E  E,I  V,I [ZSys]  [ZLoad]  Figure 3.12: System Performance Test Schematic with System Impedance and Load Impedance  The majority of these studies have been combined with multiple scenarios so trends can be viewed clearly. For example, a varying power factor has been included in all the plots. The performance evaluation comes from the one line diagram in Figure 3.12 where the base system variables are in Table 3.1. In order for the scenarios to be easier to interpret, the impedances ZSys and ZLoad have been converted to power, and referred to in the scenarios as “System Strength” for the transmission short circuit power ZSys and “Load Strength” for the load demand. These values are calculated using Equation 3.70 and the units are in MVA. The base impedances for ZSys and ZLoad in Table 3.1 are similar impedances to one of the case studies run in Chapter 4. The base impedance Zbase is calculated from Equation 3.71. V2 1 · Z 106  (3.70)  V2 (138 · 103 )2 = = 190.0Ω Sbase 100 · 106  (3.71)  Sload =  Zbase =  For example, in the first scenario Zsys is kept constant to the value in Table 3.1 which corresponds to a transmission “System Strength” as seen in Equation 3.72. S=  V2 (138 · 103 )2 = ≈ 10MVA Z pu · Zbase (1.25 + 10 j) · (19)  68  (3.72)  Chapter 3. Negative Sequence Impedance Islanding Detection  Table 3.1: Base Performance Values for Practical System Unit Value Unit VBase 138 kV SBase 100 MVA E(Sys) 1.0 V pu X (Sys) 8 R ZSys 1.25 +10.0j pu ZLoad 100 pu PFLoad 0.95 -  3.3.1 Effect of Changing Per Cent of Unbalanced Load The first performance factor measures the effect an increasing unbalanced load has on V2 . This effect is important to verify that as the unbalance increases, V2 increases, and how the power fact can effect that. The size of the unbalanced load is 1 MVA. The result can be seen in Figure 3.13. As the load becomes increasingly unbalanced, V2 increases. Another way to look at the system is to refer to Figure 3.11 where as the load becomes more unbalanced, the coupling resistor Zm (Load) increases to push V2 higher. Figure 3.13 shows as the load becomes increasingly more unbalanced, V2 increases with a yx type of slope. It also shows that lower power factors increase the effects of the unbalanced load to increase the voltage V2 , but in practical cases of < 30% unbalance, V2 will remain below 0.025 pu and approach 0 as the per cent unbalanced load approaches to zero. A value of 0.025 pu is within the practical measurability limits previously discussed. Before considering using negative sequence impedance measurement islanding detection technique in a network, the system short circuit strength and per cent unbalance will be an indicator of how measurable the negative sequence components will be. Figure 3.13 also shows the phase change throughout the spectrum of unbalanced load conditions. The most significant change in phase comes from the power factor and not from change in per cent unbalance which remains relatively constant.  3.3.2 Effect of Strength of the System vs the Unbalanced Load Power This performance evaluation measures the effect of increasing the system strength relative to the load size has on the amplitude of V2 in the system. The unbalance of the load is 20% and ranges between 0.5 and 4.5 MVA. According to Figure 3.11, when the ratio of system impedance to the unbalanced load impedance grows, the voltage V2 should decrease. In Figure 3.14, the change in system impedance indicates that very estimate with a  1 x  relationship. The initial rise between system strengths of 0 to 10  MVA is a result of the power transfer relationship between impedances. When the transmission system impedance approaches the impedance of the load, an undesirable drop in V2 occurs. Pmax occurs when Zsys ≈ Zload . Near equal system impedance to load impedance is not common and if it occurs, under  voltage relays will likely trip. The phase also has a significant change between a weak system and a strong system varying between 0 and 130 degrees. This is also logical when the system strength reduces 69  Chapter 3. Negative Sequence Impedance Islanding Detection  Figure 3.13: Per Cent Unbalanced Load vs. V2 Magnitude and Phase (System Strength = 10 MVA)  in value; the phase delay of the inductive load becomes more significant, causing phase delay. For a 2 MVA load, the maximum output of V2 occurs when the system strength is between 10 and 30 MVA. The output V2 is from 0.035 pu to 0.02 pu. This evaluation identifies that large unbalanced loads produce the most significant V2 output.  3.3.3 Effect of Varying the Power of an Unbalanced Load This performance evaluation measures the effect of varying the unbalanced load size while keeping a constant system strength of 10 MVA. This evaluation identifies how larger loads contain more pronounced negative sequence voltage and that the voltage is not completely linear with power factors less than 1 (Figure 3.15). Varying the power of the unbalanced load has more of a linear effect. Once again, Figure 3.15 shows how V2 increases nearly linear with an increasing 20% unbalanced load. There is a significantly less phase change per power increase of the load in contrast to Figure 3.14 where the phase makes a full 140 degree shift over the full span of system strength. Figure 3.15 indicates that higher load to system strength ratios have a near linear effect on the available V2 in the system. Some variability in the linearity can be seen which is a result of the reactive component interactions. As the power factor approaches 1 the linearity can be seen to become more pronounced. In this case, the system needs a 20% unbalanced load of near 0.75 MVA at any power factor to have a V2 of 1% and up compared to V1 . This evaluation 70  Chapter 3. Negative Sequence Impedance Islanding Detection  Figure 3.14: System Strength Vs. V2 with Varying Unbalanced Load from 0.5 MVA to 4.5 MVA  has shown that as the load increases, the value of negative sequence voltage increases accordingly.  3.3.4 Effect of Changing Source Unbalance This evaluation identifies how increasing the percent unbalanced input voltage effects the amount of V2 in the system. In this evaluation, the terminal voltages at the source “E” are altered by a per cent on one of the phases. The application of using the source as an injection point for negative sequence current is useful for a DG to directly use negative sequence impedance islanding detection directly from its terminals without having to measure impedance from a nearby unbalanced load. Controlled terminal voltage unbalance can be accomplished with inverter based DG systems through software programming to alter the terminal voltage. This is not possible for rotating generators. Changing unbalanced source has a more prominent effect on creating negative sequence voltages then on unbalanced loads as shown in the contour plot of Figure 3.16. The change in voltage and its measurability can rise proportionately with the per cent unbalance. Unlike the previous performance examples, in which the negative sequence voltage came through the load direct; insertion of negative sequence voltage from the source has the most significant impact and measurability regardless of the system strength. For an unbalanced voltage of 10%, the V2 is approximately 5%. This has the advantage of the injection point being able to control the amount of negative sequence voltage. The voltage can be 71  Chapter 3. Negative Sequence Impedance Islanding Detection  Figure 3.15: Unbalanced Load Size of 0 MVA to 3 MVA vs. V2 , (Load unbalanced by 20%)  increased or decreased depending on the existing system conditions.  3.3.5 Effect of V2 and I2 Phase Angle on Different Unbalanced Configurations The change in negative sequence voltage and current phase angles with different load configurations is evaluated in this section. Some examples of different load configurations are: loading on phase A to neutral only, loading on phase B to neutral only, or loading from phase A to phase B only. The impedance calculated from −1 · VI22 is the same regardless of the unbalanced load configuration, but the the phase angle for both voltage and current change. Consider the circuit in Figure 3.12 where the impedances Zsys = 10 + 100 j pu and Zload = 1000 pu. Zload is connected in WYE or ∆ configuration depending on the scenario. Unbalancing Zload in different configurations changes the phase angle of V2 and I2 . Some examples of different loading configurations are: to increase the impedance in Zload−phase−A by a per cent amount above Zload−phase−B or Zload−phase−C , or by connecting Zload from phase A to phase B while not connecting any other loads to the BC or CA delta. The scenarios and results investigated to demonstrate how the phase angle changes between different loading states are listed in Table 3.2 and illustrated in Figures 3.17 and 3.18 for negative sequence voltage angle and negative sequence current angle respectively. In first scenario listed in Table 3.2, WYE connected Zload−A is increased by 20% where Zload−B 72  Chapter 3. Negative Sequence Impedance Islanding Detection  Figure 3.16: Utility Strength (SCC) vs. V2 (Varying Percent Source Unbalance of 5% to 30%)  and Zload−C remain at the original 1000 pu. This is identified in Figures 3.17 and 3.18 as “V2 A+”. In the second scenario listed in Table 3.2, Zload−A is decreased by 20% where Zload−B and Zload−C remain at the original impedance of 1000 pu. This is identified in Figures 3.17 and 3.17 as “V2 A-”. This process repeats for all the individual remaining phase loading B± and C±. The final three scenarios listed in Table 3.2 are ∆ connected phase to phase unbalanced loads with the other two phases left open. In the first phase to phase loaded case, a load of 1000 pu is placed between phases A and B and no load is connected between phases B and C or phases C and A. This is identified in Table 3.2 as ∆, Zload−AB = 1000 and in Figures 3.17 and 3.18 as “V2 AB”. The negative sequence voltage and current phases are calculated and placed in the table under “Phase Angle of V” for voltage phase and “Phase Angle of I” for current phase. Each of the scenarios show several interesting observations. The V2 phase angle difference is 180o when comparing Zload−A + and Zload−A −. This is similar for the other two single phase scenarios of  Zload−B ± 20% and Zload−C ± 20%. Another observation of interest is the 120o phase angle between the  three items in each “+”, “-” and “phase-to-phase” scenarios. For example, Zload−A+ , Zload−B+ , Zload−C+ , are all 120o out of phase. The 120o angle is clearly seen in Figure 3.17 and Figure 3.18 for voltages and currents respectively. Each “+”, “-” and “phase-to-phase” scenario series presents different phase angles from “0o ”. The difference in the phase angle between the current and the voltage of all scenarios is consistently 95o . The angle Zsys is 85o . The sum of these two angles equates to 180o as 180 − 95 = 85.  73  Chapter 3. Negative Sequence Impedance Islanding Detection  Table 3.2: Phase Angle For Different Unbalanced Load Types Unbalanced Load Phase Angle of V Phase Angle of I o WYE, Zload−A + 20% 73 169o o WYE, Zload−A − 20% −108 −12o WYE, Zload−B + 20% −166o −70o o WYE, Zload−B − 20% 11 107o WYE, Zload−C + 20% −46o 49o o WYE, Zload−C − 20% 131 −132o ∆, Zload−AB = 1000, −41o 54o Zload−CA = Zload−BC = ∞ ∆, Zload−BC = 1000, 78o 174o Zload−CA = Zload−AB = ∞ ∆, Zload−CA = 1000, −161o −65o Zload−BC = Zload−AB = ∞  3.3.6 Multiple Unbalanced Loads in a System In the previous sections, unbalanced impedance calculations were addressed using a single unbalanced load or source in a perfectly balanced system. However in practical systems, other unbalanced loads exist throughout the distribution layer that can corrupt impedance measurements. Not only are the loads simply “unbalanced,” but they are unbalanced on phases A, B, C or in combinations of all three. Therefore, multiple unbalanced components in a radial system pose as a source of corrupting impedance information. Unbalanced loads can be complicated, but the interactions of multiple corrupting loads can be understood using a simple two load radial example as illustrated in Figure 3.19, where the two load impedance matrices, [ZLoad−1 ] and [ZLoad−2 ], are single phase, WYE connected, resistive loads. Of the other components, [E] is a balanced voltage source with the system Th´evenin impedance [ZSys ] and the distribution system impedance is [ZDist ]. The solution for currents and voltages can be expressed in an impedance matrix as [V ] ≈ [E] − [I] · ([Zsys ] + [Zdist ]) if Zload−x are very large where [V012 ] and  [I012 ] for node Y are expressed, as previously, as A−1 [VY ] and A−1 [IY ] respectively. To better understand  the interaction, consider that all impedances are linearly independent connected in a grounded WYE formation. By unbalancing single impedances in each phase, an accurate assessment of the interactions can clearly be seen. The example in Figure 3.19 has the same impedances listed in Table 3.1. These are fictitious values used to demonstrate the concept. In this example, the effect of negative sequence impedance measurements made from the point of [ZLoad−2 ] when both [ZLoad−2 ] and [ZLoad−1 ] vary in levels of unbalance is investigated. Table 3.3 contains the base case information while the four scenarios that will be investigated are listed in Table 3.4. Note that in this table, the result points to the specific reference graph. [ZLoad−2 ] is impedance is found at position “Y” and calculated by using the equation: −1 · VIYY−2 . The source voltage E remains balanced at 1 pu, and both ZSys and ZDist are set to 1 pu. −2  The Loads ZLoad−1 and ZLoad−2 will have their impedances varied by a per cent on single phases. The impedance of the two loads is 2000 pu. During this experiment, the expected impedance measured from position ’Y’ is as follows: the positive sequence impedance of  74  V1 I1  ≈ ZLoad−2 = 2000 pu and the negative  Chapter 3. Negative Sequence Impedance Islanding Detection  Figure 3.17: V2 Phase For Different Unbalanced Load Types Relative to 0, 240, 120 ABC Phases  sequence impedance of  V2 I2  = (ZSys ||ZLoad−1 ) + ZDist ≈ 2 pu.  The purpose of these scenarios is to illustrate how islanding detection threshold settings can be effected when other unbalanced loads exist in the network. The effect that different load configurations have on each other is investigated with a proposed averaging solution to reduce the impedance measurement errors. Table 3.3: Base Case Values for Figure 3.19 Unit Value Unit S(base) 1000 MVA V (base) 138 kV E(Sys) 1 pu ZSys 1 pu ZDist 1 pu ZLoad1 2000 pu ZLoad2 2000 pu PFLoads 1 -  Scenario #1: Both Unbalanced Loads in Phase A Only In this scenario, the load unbalance is made in only phase A of each WYE connected load which can be seen in Figure 3.20. The load unbalance in both identical loads is varied between a range of 0% to 5% on the contours for ZLoad−1 and -10% to +10% for ZLoad−2 values so that the unbalance trends can be seen and how it effects the impedance measurements. Voltage and current values are taken from 75  Chapter 3. Negative Sequence Impedance Islanding Detection  Figure 3.18: I2 Phase For Different Unbalanced Load Types Relative to 0, 240, 120 V ABC Phases  [ZSys]  [E]   X  [VX]  [ZLoad 1]  [IY]  [ZDist]   Y  [VY]   [ZLoad 2]   Figure 3.19: Single Line Diagram of Multiple Unbalanced Equal Load Scenario  point “Y”. The A,B and C Voltages and currents are identified at point “Y” as vectors [VY ] and [IY ] respectively. The trivial result when ZLoad−1 has 0% unbalance and the resulting measured negative sequence source impedance ZLoad−2 at position “Y” is the exact expected value of 2 pu (ZLoad−1 = 2000 is 1000 times larger than Zsys and has little effect on the Th´evenin impedance). However, each consecutive line represents a per cent increase in the unbalance of the impedance of each load. What is shown in this scenario (Figure 3.20) is the measuring impedance ZLoad−2 from position “Y” moves from negative percentage to zero, the measurement accuracy decreases until the “knee” point where the impedance quickly diverges towards 2000 pu. This “knee” point is approximately when the per cent unbalance of each load equals each other. However, when the unbalance of ZLoad−2 grows from zero to a positive per cent unbalance, the performance between 0.1% and 2% diverges to a very low value where the unbalance is low and then quickly becomes more accurate as the unbalance increases. The level and type of unbalance must be considered to avoid inaccurate impedance measurements in the region that 76  Chapter 3. Negative Sequence Impedance Islanding Detection  Table 3.4: Scenarios and Comments for Multiple Unbalanced Loads (All Values in pu) Case ZLoad−1 ZLoad−2 Result 1a 1b 2a 2b 3  2000 + 5% ph A 2000 + 5% ph A 2000 + 5% ph B or C 2000 + 5% ph B or C special averaging case  2000 varying 0% to 10% ph A 2000 varying -10% to 0% ph A 2000 varying 0% to 10% ph A 2000 varying -10% to 0% ph A Z2 swapping to ph A,B, and C averaged  Figure 3.20 Figure 3.20 Figure 3.21 Figure 3.21 Figure 3.22  has poor performance.  Figure 3.20: -10% to + 10% Load Unbalance on Both Phase A, Vs. Calculated V2 (ZLoad = 2000pu )  Scenario #2: Unbalanced Loads in Different Phases In this scenario, the load unbalance consists of the two loads having unbalanced impedances on different phases, which can be seen in Figure 3.21. In previous examples, phase angle changes with different loading conditions and the error generated from multiple loads with different loading conditions can be effected. For example, in this scenario, ZLoad−2 phase A is varied while ZLoad−1 phase B or C is altered. The results differ from Scenario #1, in that the contours do not have the significantly low impedance drop in the positive region. The rest of the curves are nearly a mirror image of Scenario #1 (Figure 3.20). However, similar to Scenario #1, near zero unbalance when the two unbalances nearly equal each other, the impedance accuracy erodes up to the load impedance of 2000 pu. The knee point occurs when 77  Chapter 3. Negative Sequence Impedance Islanding Detection the two unbalance values are equal in size. Negative sequence impedance islanding detection method accuracy benefits when the the source unbalanced load (Zload−2 ) is on a different phase to the unbalanced loads in the network.  Figure 3.21: -10% to + 10% Load Unbalance on Alternating Phases Vs. Calculated V2  Scenario #3: Average Impedance With Unbalance in All Three Phases This Scenario is a combination of of Scenario #1 and #2 by taking the impedance results of phase A, B, C, unbalanced experiments on ZLoad−1 and obtaining the average of the three as seen in Equation 3.73. This Scenario was run by keeping ZLoad−2 unbalance on phase A only, while running an unbalance sweep through ZLoad−1 on phases A, B and C. ZSys ≈  V2−A−unbal V2−B−unbal V2−C−unbal + + I2−A−unbal I2−B−unbal I2−C−unbal  (3.73)  The result of this Scenario can be seen in Figure 3.22. Where both of the previous Scenarios varied differently on the +/- per cent of similar phase load change and on differing phase load change, the averaging creates a symmetric result that converges to the ideal of 2 pu significantly faster than the first two Scenarios. For unbalanced injection techniques, using an active phase unbalance swapping and averaging produces the fastest converging and most accurate results. This multi-phase averaging technique has been validated in a practical system which is described in the following Chapter 4.  78  Chapter 3. Negative Sequence Impedance Islanding Detection  Figure 3.22: -10% to + 10% Load Unbalance Vs. Calculated V2 Averaged on A,B,C Alternating Phases  3.4 Non Fundamental Frequency Negative Sequence Impedance Single frequency impedance measurements are effective for islanding detection, however, the measurement accuracy can be increased with the use of unbalanced harmonics from naturally occurring unbalanced sources and loads to measure island conditions. Although power system impedance above the fundamental are difficult to determine, there are some common trends that are present in distribution systems that have lead to research in this topic. The first is from authors [4] and [19], who mapped power system impedances in many grid configurations showing a predicable LR type impedance trend in frequencies up to several kHz. This Section shall detail the sources of non-linear components and the effect that it can have on negative sequence impedance islanding detection.  3.4.1 Sources of Harmonics from Non-Linear Components Up to now, the systems analyzed have all been unbalanced passive components. Passive components have the advantage of simple mathematical descriptions that can be expressed easily using Ohm’s law. Non-linear loads corrupt the 60 Hz fundamental by injecting a spectrum of harmonics. This nonfundamental energy can activate resonances and cause heating of rotating machines. Some types of non-linear loads are discussed in the IEEE standard 519-1992 [50]. The following list is a summary of the most significant non-linear loads in a power system that can be used for naturally occurring harmonic injections. The most common source of non-linear distortion in DC systems are AC-DC converters that use 79  Chapter 3. Negative Sequence Impedance Islanding Detection  1. 2. 3. 4. 5. 6. 7. 8. 9.  Table 3.5: Common Power System Non-linear Loads Saturated Transformers Diodes and Switching Silicone devices Arc Furnaces Static Var Compensators Inverters and DG sources Electronic Phase Control Switch mode power supplies PWM Drives Voltage dependant resistors Table 3.6: Sequence Movement For Harmonics [120] Harmonic 1 2 3 4 5 6 7 8 9  Sequence Positive 1 2 0 1 2 0 1 2 0  Negative 2 1 0 2 1 0 2 1 0  Zero 0 0 0 0 0 0 0 0 0  diode and SRC bridges to invert the AC waveforms to DC signals and vice-versa. This conversion creates odd harmonics from the fundamental equal to n = x·2+1 (ie 3,5,7,9...) where ‘n‘ is the harmonic order and ‘x‘ is a counter from 0 to ∞. The amplitude of each harmonic decreases by 1/n. Sequence harmonics also appear in power systems, but for each harmonic the sequence number changes due to the transform [120] as shown in Table 3.6. In theory, harmonic sequence impedances are a promising option for increasing the accuracy of impedance estimation. However, the overriding limitation of harmonic negative sequence impedance measurement is the measurability of the small signals. For example, if a typical negative sequence voltage is in the range of 1%, the third and fifth (most prominent) harmonics are typically in the range of 5% of the fundamental. Therefore, 1% of 5% equals 0.05% which exceeds a typical PT’s measurable limit. Although this topic poses significant interest, it is outside of the basic theory and physical limits presented here, and practical experiments were not conducted.  80  Chapter 3. Negative Sequence Impedance Islanding Detection  3.5 Sequence Components for Induction and Synchronous Machines Negative and positive sequence impedances are equal to each other in passive systems. This is not the case for rotating synchronous and induction machines. Rotating machines have a significantly lower negative sequence impedance under steady state conditions. When measuring [ZSys ] from networks that have large machines, the low negative sequence impedance can affect the value of the measured system impedances; however, small rotating machines tend to have small overall significance in comparison to the very low utility system impedance. The V2 and I2 can become too small to measure. In a machine, the magnetic field from the negative sequence stator currents rotates at synchronous speed in the opposite direction to the machine rotor. Viewing the currents from the rotor, the stator currents are double frequency. Hence, currents of two times the rated frequency are then induced through to the rotor circuits. Synchronous machines and induction machines have similar but slightly differing negative sequence impedance. The derivation of the impedances for each device is discussed in [68] and [120]. Figure 3.23 summarizes these impedances, where on the left hand side of Figure 3.23 is the impedance of an induction machine and on the right side of Figure 3.23 is the impedance of a synchronous machine. The induction machine Xl is the stator leakage reactance, Ra is the stator resistance, XM is the magnetizing reactance, Xr is the rotor leakage reactance. Rr is the rotor resistance and s is the slip. For the synchronous machine, Xd” and Xq” are the sub-synchronous d and q axis components from the parks transformation. The variables Ra and Rr are the stator resistance and rotor resistance respectively. For the induction machine, the slip at full load is near 2, resulting in a R2 , in Figure 3.23, approaching − 12 Rr . These low impedances may have an effect on the threshold window for negative sequence impedance  islanding detection. They can create a falsely strong system. Several case studies are investigated in the following chapter with rotating machines to show how various sizes can effect the proposed technique.  Ra  Xl  Rr  Xr  XM  R  Induction Machine  =    s  sR  r  Xd Xq Xd + Xq R = Ra + Rr  X  =          Synchronous Machine  Figure 3.23: Induction and Synchronous Machine Negative Sequence Impedance  3.6 Implementation Strategy for Negative Sequence Islanding Detection This section describes the implementation strategy of negative sequence islanding detection and how it can be applied as a protective relay. In previous sections, negative sequence current is created from  81  Chapter 3. Negative Sequence Impedance Islanding Detection unbalanced loads, resulting in the ability to use negative sequence components to measure the system Th´evenin impedance. What was also revealed is, though corruption can occur from other negative sequence sources, the difference between a high and low impedance connection will show a measurable variance. By using the observable change of impedance between the utility connected system and an islanded system, an island state can be detected with a predictable threshold. The performance experiments conducted, indicate the ideal impedance measurement scenario corresponds to highly unbalanced loads in the local system with respect to the other system loads, with an impedance difference between the two states on either side of the knee point in Figures 3.20 to 3.22. A negative sequence impedance islanding detection relay can be implemented in several ways depending on the network configuration and load characteristics.  3.6.1 Naturally Occurring Negative Sequence Currents Starting with naturally occurring negative sequence currents, take a typical radial system as the one in Figure 3.25, where the utility source is connected to a substation at the left, and the DG source is connected to the distribution system on the lower right. Negative sequence impedance islanding detection relay is used by the DG to detect the opening of circuit breaker ’A’ causing an island. Direct measurement of negative sequence components from the DG terminals (typically balanced) will not give meaningful islanding imformation. A negative sequence impedance islanding detection impedance relay system requires a source of negative sequence current in a location where existing CT and PT components already exist, and where close communication is available to the DG. For example, the CT and PT components at breaker ’B’ can be used by the negative sequence impedance islanding detection relay if ‘Load DG’ is unbalanced. Similarly, Loads 1, 2 and 3 typically have CT and PT connections for revenue metering that could be then linked to the negative sequence impedance islanding detection relay system if they are close to the DG source. Therefore, the positions where CTs and PTs can be used for islanding detection are from revenue meters of the three loads, the DG load or the protection CTs and PTs at the utility substation. These locations have been identified with CT and PT devices shown in Figure 3.25. Of the CT and PT locations identified (protective or revenue type), the ideal location to use negative sequence impedance islanding detection is a practical question of: where the most significant amount of unbalance exists, access to CT and PT data, and the zone of protection required. The algorithm for naturally occurring negative sequence currents can be seen in Figure 3.24. The algorithm starts with setting an initial threshold impedance from either measuring negative sequence impedance directly or with known network information from the local utility. Then the relay can start its islanding detection scanning by measuring the negative sequence impedance continuously. The input voltages and currents locations are the most important considerations when implementing this algorithm. To reduce errors, a count of three positive readings must be made before the islanding detection breaker will open. This initial three positive readings can be any number that allows for the impedance change to be detected with in the current IEEE standard of two seconds.  82  Chapter 3. Negative Sequence Impedance Islanding Detection  Figure 3.24: Natural Negative Sequence Impedance Islanding Detection Algorithm  3.6.2 Injected Negative Sequence Currents If connecting to an unbalanced load in the distribution system is not practical, or the unbalanced loads are not large enough to produce a measurable result, negative sequence currents can also be injected. This case is similar to the original mathematical evaluation of deliberately unbalancing loads to meet the performance requirements as seen in Section 3.3.6. This can be accomplished if the source is inverterbased where it can be programmed internally to deliberately inject unbalanced current into feeder ’C’ in Figure 3.25 where the local CT and PT sensors are available to measure Z2 for the negative sequence impedance islanding detection relay. The specific algorithm can be seen in Figure 3.26. This algorithm operates by deliberately unbalancing each phase individually then measuring Z2 for each of the three unbalanced phases, and then using Equation 3.73 to average out the three calculated impedances to  83  Chapter 3. Negative Sequence Impedance Islanding Detection equate to the Z2 . Phases can be deliberately unbalanced by adjusting the voltage amplitude at the inverter terminals or by briefly connecting highly unbalanced loads onto individual phases of the network.  Figure 3.25: Negative Sequence Impedance Measurement Islanding Detection Concept  3.7 Summary This chapter has shown the concept of negative sequence impedance islanding detection for distributed generators. The concept has been demonstrated through detailed derivations, numerical experiments, performance evaluations, and with a suggested implementation strategy. If the systems are unbalanced and available sensors can measure the unbalanced signals, negative sequence impedance measurements can be used for islanding detection. This technique is a logical incremental improvement on the previous challenges of threshold setting, installation cost, power consumption and real time measurability. Thresholds are predicable for negative sequence impedance islanding detection as the impedance between utility connected and island connected states can be significant where the minimum likely differential between an island state and a utility connected system is between 2 and 10 times. Another challenge for previous islanding detection installations were the installation costs and power consumption. Injection or power line carriers require expensive coupling transformers with larger power requirements. However, if there are already existing negative sequence unbalanced conditions, negative sequence impedance islanding detection will only require existing CT and PT access to operate. Finally, the ability to measure for islanding detection continuously is also possible. In this analysis, there have been some potential problems that can only be revealed through actual system measurements. These problems are as follows: Are actual utility systems unbalanced? Is the unbalanced voltage and current truly measurable? How much corruption will exist from other unbalanced 84  Chapter 3. Negative Sequence Impedance Islanding Detection loads? These problems are examined in the following chapter where actual live systems are evaluated for suitability and measurability. What has been revealed is that unbalanced loads that are more than two to three times more unbalanced than the other, can cause accuracy errors that will make threshold setting more difficult. The theoretical analysis of negative sequence impedance islanding detection has been shown to have promising theoretical characteristics. Further studies shall be conducted in the following chapter on practical systems and practical scenarios to further demonstrate the concept’s suitability.  85  Chapter 3. Negative Sequence Impedance Islanding Detection  Figure 3.26: Injected Components Negative Sequence Impedance Islanding Detection Algorithm  86  Chapter 4  Case Studies of Negative Sequence Impedance Islanding Detection 4.1 Introduction In order to properly validate the method of negative sequence impedance islanding detection, three practical case studies were considered. These experiments provided insight into the existence and measurability of unbalanced voltages and currents (challenges identified in Chapter 2). These cases were selected to cover a spectrum of systems where this technique may be useful. Experimentation with live power systems is an essential component for validating any theoretical work (two of these cases contain measurements from actual utility fed systems). Complete shutdowns, deliberate islanding and other potential customer disruptions are outside the allowable scope of experiments on live systems; however, the live islanded state measurements have been simulated in EMTP simulations that closely match the existing network. The practical measurements have been carried out for utility-connected conditions while simulations have been used to evaluate the islanded conditions. The three cases evaluated in this chapter are: 1. IEEE 399-1997 standard industrial bus, 2. A 25 kV distributed-generator fed radial system, and 3. A 600 V distributed-generator fed office building. The first case has been run purely in simulation. Measurements for the second and third cases were taken directly from actual operating utility fed systems.  4.2 Simulation of Standard IEEE 399-1997 Industrial Bus 4.2.1 System Description The first system modeled in this chapter is a non-radial network consisting of multiple branches, two distributed generator sources and a single utility connection. The system is a modification of IEEE 3991997 [48] from the micro grid modeling evaluation by Katiraei [65] (Figure 4.1). This is an industrial bus network with multiple branches used to represent a small “microgrid” system. This case study was conducted purely with simulation in EMTP software. The purpose of this case study was to demonstrate that the change in measured negative sequence impedance between an islanded and utility condition is significant. The size of the load unbalance vs. the measurability will also be evaluated. The unbalanced measurements are taken at the bus feeding two unbalanced loads with several unbalanced loads mixed into the network to provide a practical scenario of corrupting signals.  87  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection There are four main areas in this system (Figure 4.1). From the top, the utility source (1000 MVA) and the connecting transformer and line contain the dominating impedances of this feeding system. The interconnection Breaker “A” will be the point where the islanding will occur (located at near the top of Figure 4.1). The second area is the 5 MVA ‘DG1’ source seen on the bottom right side of Figure 4.1 connected to ‘Bus 1’. This DG source is modeled as an ideal source with an associated impedance. The third area is the 2.5 MVA ‘DG 2’ source on the bottom left side of Figure 4.1 connected to ‘Bus 3’. Both DG1 and DG2 are modeled as ideal sources with impedances. The final point of interest are Loads 1 to 5 which total to 7.7 MVA with an average power factor below 0.9. These loads are distributed along the bottom of Figure 4.1 with 1.5 MVAR power factor correction capacitor bank near Feeder 1 off of the PCC Bus that brings the confined power factor just above 0.9. All the simulated measurements in this test case are in pu with a base voltage of 13.8 kV and a base power of 10 MVA. The system is run at steady state with various scenarios listed in Table 4.1. Breaker “A” connecting the utility bus to the network is opened and closed to demonstrate islanded and utility connected conditions. The scenarios tested consisted of unbalanced load variations between Load 5 and Load 3 linear loads and with a large induction machine. A summary of the particular cases is shown in Table 4.1. Same phase unbalance in Case 2a indicates that the WYE connected loads on Load 3 and Load 5 are both unbalanced on the same phase. For example, the impedance of same phase unbalance on Load 3 is: Z3A = Dx · Zx  (4.1)  Z3B = Zx  (4.2)  Z3C = Zx  (4.3)  Z5A = Dy · Zy  (4.4)  Z5B = Zy  (4.5)  Z5C = Zy  (4.6)  and on Load 5 is:  The variables Dx and Dy are per cent variations of the impedances Zx and Zy . Conversely, the impedance of different phase unbalance on Load 3 is: Z3A = Dx · Zx  (4.7)  Z3B = Zx  (4.8)  Z3C = Zx  (4.9)  88  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection and on Load 5 is: Z5A = Zy  (4.10)  Z5B = Dy · Zy  (4.11)  Z5C = Zy  (4.12)  Dy on Load 5 has moved to Z5B .  Experiment 1a. 1b. 2a. 2b. 3a. 3b. 4.  Table 4.1: Experiments Modeled using IEEE 399 Standard Bus Description Load 5 Unbalanced, Measure Impedance from Load 5 at Bus 3 Load 3 Unbalanced, Measure Impedance from Load 3 at Bus 2 Load 3 and Load 5 Unbalanced, Measure Impedance from Load 3 (Bus 2) and Load 5 (Bus 3) , Same Phase Unbalance Load 3 and Load 5 Unbalanced, Measure Impedance from Load 3 (Bus 2) and Load 5 (Bus 3) , Different Phase Unbalance Load 3 with Induction Machine, Load 3 Unbalanced, Measure Impedance from Load 3 (Bus 2) Load 3 with Induction Machine, Load 5 Unbalanced, Measure Impedance from Load 5 (Bus 3) Load 3 with Induction Machine, Unbalanced Loads 3 and 5 on Alternate Phases, Impedance measured from both at Bus 3 and Bus 2  4.2.2 Experimental Results The computed impedances from the several experiments run on the network are shown in Figure 4.1, listed in Table 4.1 and are detailed in this section. The expected impedances measured from the point of each load are shown in Table 4.2 where ZLoad is the impedance towards the load and Z2 for the Th´evenin impedance towards the utility. The results of the studies are taken from CTs and PTs at Bus 3 connection point for Load 5 and Bus 2 connection point for load 3. Load 3 and Load 5 are unbalanced individually in the fist two experiments (Figure 4.1). The computational analysis values of voltages and currents of each simulation are taken from steady state responses. To allow for a simplified comparison of impedances, the absolute value of each load is shown. The impedance ratio of islanded to utility connected state for Load 3 was 11.7 while the impedance ratio for Load 5 was only 4.4. Under noisy line conditions, threshold settings will be much more flexible with high impedance ratios such as with Load 3. Experiment 1a and 1b - Singular Unbalanced Loads The first experiment in this case study is the trivial case where there is only one unbalanced load in the simulation. The expected outcome should be near to the exact values of the mathematical estimate (Table 4.2). The simulations were run first by setting the power flow towards the utility source to be 89  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection Utility Source  Sbase=10MVA  1000 MVA  Vbase=13.8kV  X/R = 22.2  Line U1-U2 0.515+0.029j % 69.0 kV  A  69/13.8 kV, 15 MVA 0.667+j5.33%  Feeder 3  PCC BUS: 13.8kV  Feeder 2  Feeder 1  Line F1-B2 Line F3-B3  Line F1-B1  3.564+2.66j%  6.065+10.15j %  Bus 2  3.976+5.12j %  0.976km  4.83 km  2.06 km  1.5 MVAR T3 13.8/2.4  Bus 3  Bus 1  3.75MVA 2.44+14.8j%  Line B3-L5  Line B3-L4  Line B1-L2  Line B1-L1  0.732+0.095j%  2.56+0.332j%  0.732+0.095j%  0.104+0.135j%  0.189km  0.362km  0.104km  0.148km  DG2 –  T5 13.8/2.4  T4 13.8/0.48  1.5MVA  1.25MVA  6.48+38.3j%  5.6+48.0j%  Load 3 3.2 MW 1.9 MVAR  2.5 MVA  DG1 –  T2 13.8/2.4  T1 13.8/0.48  2.5MVA  1.0MVA  3.29+2.3j%  8.21+57.5j%  5 MVA  Load 5  Load 4  0.9 MW  0.9 MW 0.6 MVAR  Load 2  Load 1  1.5 MW  0.8 MW  1.0 MVAR  0.47 MVAR  Figure 4.1: IEEE Standard 399-1997 (Brown Book) Reference Bus Case adapted From [65]  Table 4.2: Expected Impedances From Perspective of Test Loads (in pu) State |ZLoad | |Z2 | Z2 Ratio Load 3 Islanded 3.64 1.078 Utility Connected 3.64 0.0916 11.7 Load 5  Islanded Utility Connected  11.1 11.1  0.843 0.193  4.4  near to zero (perfectly zero power flow is impossible with an unbalanced system). Breaker “A” is then opened to simulate an island state. The two resulting impedances from the computational analysis are in separate simulations from two separate unbalanced loads: Load 3 and Load 5. The positive and negative sequence impedances (ZLoad and Z2 ) measured from experiment 1a and 1b are near to the mathematical estimates (Table 4.2) and are listed in Table 4.3. In this experiment, Load 5 is unbalanced by 15% and Load 3 is unbalanced by 10% in separate scenarios. Since Load 5 is deeper in the network and has a smaller value, the negative sequence current created is clearly much smaller than Load 3 which is nearly four times the size of Load 5. The values of the currents and voltages are approaching the measurable limits of 0.1% in some experiments and in others are below the measurable limit. The negative sequence impedances (Table 4.3) show the exact expected answers from the perspective of the test loads as if the CTs and PTs could measure the low voltages and currents presented. This allows the computational analysis of impedances to be compared with the theoretically expected values (Table 4.2).  90  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection  Table 4.3: Experiment 1: IEEE 399 Bus System, Single Phase Unbalance of Load 3 and Load 5 Load State V1 I1 |ZLoad | V2 I2 |Z2 | Z2 Ratio Load 3 Islanded 0.99 0.2720 3.64 0.0105 0.0098 1.078 Utility Connected 0.99 0.2720 3.64 0.0012 0.0132 0.0916 11.7 Load 5 Islanded 0.99 0.0891 11.11 0.0024 0.0028 0.8464 Utility Connected 0.99 0.0891 11.11 5.601e-4 0.0029 0.1960 4.3  Experiment 2a and 2b - Multiple Unbalanced Loads The second experiment is similar to experiment 1, except instead of having only one unbalanced load in the system, there are two unbalanced loads: Load 3 and Load 5. The expected outcome (Table 4.2) of this test is a slightly corrupted impedance measurement. The test was completed twice (2a, 2b) with the unbalanced loads both on Phase A for the first test (same phase unbalance, Table 4.4) and the unbalanced loads on alternate phases for test 2b. For test 2b, the unbalance for Load 5 was placed on Phase A and the unbalance for Load 3 was placed on Phase B (alternating phase unbalance, Table 4.5). Load 5 is, again, the smaller load of the two. What is revealed is that the unbalanced stronger Load 3 has less of an effect on the utility-connected impedance computational analysis while the error for the weaker Load 5 is more severe. The impedance ratio change between the islanded state to the utility connected state on Load 3 in experiments 1 to 2a and 2b increased from 11.7 to 14 and 10. While the ratio between the islanded state to the utility connected state on Load 5 in experiments 1 to 2a and 2b increased from 4.3 to 10.95 and 8.01. Despite the error in the values of the calculated impedances, the ratios of islanded to non-islanded conditions are still significant, thus allowing for reasonable threshold setting. An important consideration is the estimate of the impedance islanding trigger threshold. With Load 3, the utility-connected load is very similar to the first case (Experiment 2a, Table 4.3). However, for Load 5 (lower into the network) the measured impedance value changes to a value 5.2 times smaller between experiment 2a and 2b. Threshold settings and impedance ratios for islanding detection change between locations and initial experimentation may be required on a practical system to asses the particular measured impedances.  Load Load 3 Load 5  Table 4.4: Experiment 2a, Load 3 and Load 5 Unbalanced on Phase A State V1 I1 |ZLoad | V2 I2 |Z2 | Islanded 0.99 0.2720 3.64 0.0072 0.0052 1.393 Utility Connected 0.99 0.2720 3.64 4.45e-4 0.0045 0.0989 Islanded 0.99 0.0891 11.11 0.0079 0.0105 0.7549 Utility Connected 0.99 0.0891 11.11 9.113e-4 0.0132 0.0690  |Z2 | Ratio 14 10.95  When comparing the Z2 impedance values in between experiments 2a and 2b (Table 4.4 and Table 4.5) the measured impedances change along with the impedance ratios between islanded and utility connections. Using Equation 3.73 to apply the impedance averaging method developed in the previous chapter, the impedances improved. Averaging the impedances resulted in an improvement of Load 3 91  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection  Table 4.5: Experiment 2b, Phase A Unbalance on Load 3 and Phase B Unbalance on Load 5 Load State V1 I1 |Zload | V2 I2 |Z2 | |Z2 | Ratio Load 3 Islanded 0.99 0.2720 3.64 0.0109 0.105 1.0316 Utility Connected 0.99 0.2720 3.64 0.0013 0.0133 0.0939 10 Load 5 Islanded 0.99 0.0891 11.11 0.0105 0.0036 2.907 Utility Connected 0.99 0.0891 11.11 0.0016 0.0043 0.3625 8.01  Table 4.6: Using Averaging Equation 3.73 for Load 3 and Load 5 State |Z2 | Z2 Ratio Load 3 Islanded 1.288 Utility Connected 0.0965 13.3 Load 5  Islanded Utility Connected  0.7854 0.0741  10.5  measurements and in a closer ratio to the actual values (Table 4.6). Averaging the Load 5 impedances resulted in the ratio of the original phase A unbalanced condition to improve marginally towards the actual values. In this case, increasing Load 5 unbalance would better counteract the larger unbalance of Load 3 and increase the accuracy for the islanding detection device. Experiment 3a and 3b - A Large Rotating Machine Introduced Into The System In the third experiment, a 1.6 MVA induction motor was added in parallel with Load 3 to explore the effect of the negative sequence impedance of rotating machines on impedance measurement for islanding detection. As shown in Section 3.5, rotating machines can have significantly lower negative sequence impedances due to the reverse rotating slip of negative sequence current. Rotating machines are naturally balanced devices and do not produce negative sequence currents. This experiment was conducted to demonstrate how the machine’s low negative sequence impedance can effect on the proposed islanding detection technique. The single line diagram of this case where the induction machine is seen in parallel with Load 3 is shown in Figure 4.2. The total loading of this bus remains close to 3.5 MVA. In this experiment, the interconnecting lines have a lower impedance than the previous experiments to prevent excessive voltage drop with the induction machine added in the system. The data for the specific values of the induction machine are from Krause [67] and are shown in Table 4.7. Due to the change of Load 3 passive load to an induction machine, the expected negative sequence impedances have changed for both cases. The expected reduced impedances from the perspective of Load 3 and Load 5 are shown in Table 4.8. The new system where the induction machine has been added in parallel with Load 3 in the center of Feeder 2, Bus 2 is shown in Figure 4.2. It is important to note that with a machine of this size, the expected negative sequence impedance on its bus is reduced. The impedance ratio for Load 3 has been reduced by 7.2 times and the impedance ratio for Load 5 has reduced by 5.4 times (Table 4.3).  92  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection  Table 4.7: Induction Machine (1.6 MVA) Input Variables for EMTP Variable S V poles TB f rs Xls XM Xlr′ rr′ J  Value 1.68 MVA 2300 V 2 8.9 · 103 N · m 60 Hz 0.029 Ω 0.226 Ω 13.04 Ω 0.226 Ω 0.022 Ω 63.87 kg · m2  Table 4.8: Expected Impedances From Induction Machine State |Zload | |Z2 | Z2 Ratio Load 3 Islanded 2.64 0.1413 Utility Connected 2.64 0.0324 4.36 Load 5 Islanded 2.9 0.1562 Utility Connected 2.9 0.0787 1.99  The experiment demonstrated how the low negative sequence impedance of the 1.6 MVA induction machine has a significant effect on the impedance and the impedance ratio results from studies from the perspective of Load 3 (Table 4.9). The voltage V2 and the current I2 are smaller than in previous experiments without the large machine (Table 4.9). The amount of unbalance and the load characteristics may have an impact on setting the threshold values to detect islanding conditions. Experiment 4 - A Large Rotating Machine Combined With Multiple Unbalanced Loads In this experiment, the same large rotating machine used in experiment 3 is used in combination with additional unbalanced loads on alternate phases, similar to experiment 2b. Again, the expected impedance ratios are computed from each the point of view of the unbalanced load and are the same expected values  Table 4.9: Case 3, IEEE 399 Bus System, Induction Machine on Feeder 2, Bus 2 Load State V1 I1 |Zload | V2 I2 |Z2 | |Z2 | Ratio Load 3 Islanded 1.0 0.47 2.55 0.0019 0.063 0.143 Utility Connected 0.965 0.45 2.55 4.35 · 10−4 0.016 0.032 4.47 Load 5 Islanded 1.0 0.43 2.79 0.0026 0.020 0.156 Utility Connected 0.97 0.41 2.79 0.0013 0.0192 0.079 1.97  93  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection Utility Source 1000 MVA  Sbase=10MVA  X/R = 22.2  Vbase=13.8kV  Relay ‘A’  A  Line U1-U2 0.515+0.029j % 69.0 kV  69/13.8 kV, 15 MVA 0.667+j5.33%  Feeder 3  PCC BUS: 13.8kV  Feeder 2  Feeder 1  Line F1-B2 Line F3-B3  Line F1-B1  3.564+2.66j% Bus 2  6.065+10.15j %  3.976+5.12j %  0.976km  4.83 km  2.06 km  1.5 MVAR T3 13.8/2.4  Bus 3  3.75MVA  Bus 1  2.44+14.8j%  Line B3-L5  Line B3-L4  Line B1-L2  Line B1-L1  0.423+0.154j%  2.56+0.332j%  0.732+0.095j%  0.104+0.135j%  0.189km  0.362km  0.104km  0.148km  DG2 –  T5 13.8/2.4  T4 13.8/0.48  T2 13.8/2.4  T1 13.8/0.48  1.5MVA  1.25MVA  2.5MVA  1.0MVA  6.48+38.3j%  5.6+48.0j%  3.29+2.3j%  8.21+57.5j%  DG1 –  2.5 MVA  5 MVA  Load 5  Load 4  0.9 MW  0.9 MW  Load 3  Machine 1  2 MW  1.6 MVA 75% Max Load  0.6 MVAR  Load 2  Load 1  1.5 MW  0.8 MW  1.0 MVAR  0.47 MVAR  Figure 4.2: IEEE Standard 399-1997 Reference Bus Case with Induction Machine at Load 3 (Center)  as experiment 3 (Table 4.8). Unbalancing alternate phases resulted in very little additional voltage, V2 , and current, I2 , corruption and the ratios are close to the expected values (Table 4.10). The impedance of ZLoad remains as expected, closely matched to the impedances of the unbalanced loads shown in Table 4.10. There is a small improvement on the impedance ratios and the values of Z2 for Load 3 but not for Load 5. Table 4.10: Case 4, IEEE 399 Bus System, Induction Machine with Alternate Phase Unbalanced Load Load 3 Load 5  State Islanded Utility Connected Islanded Utility Connected  |V1 | 1.00 0.96 1.00 0.98  |I1 | 0.4657 0.4491 0.4201 0.4073  |Zload | 2.55 2.55 2.84 2.84  |V2 | 0.0029 6.44 · 10−4 0.0030 0.0011  |I2 | 0.0167 0.0160 0.0125 0.0127  |Z2 | 0.2037 0.0477 0.2865 0.1037  |Z2 | Ratio 4.27 2.76  4.2.3 Discussion The three experiments in the case study demonstrate the negative sequence impedance detection technique was effective for the experiments proposed. In the ideal balanced case, the computational analysis were very accurate in both cases. However, when another unbalanced load is introduced into the system, 94  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection the results of Z2 impedance accuracy decreases and the impedance ratio increases. The simulated accuracy of Z2 had similar performance when the unbalanced loads were moved between phases resulting in phase A on one load being unbalanced while phase B was unbalanced on the other. The most accurate computational analysis of Z2 came from Load 3, the largest unbalanced load. However, threshold settings for islanding detection are easiest to set if the impedance ratio of Z2 is very large. Load 5 has the highest Z2 impedance ratio with an islanded to non islanded impedance ratio of over 11. In several of the cases, the computational analysis of voltages and currents were below practical measurable limits for CTs and PTs and this method may require additional unbalance added to the system. When determining where to locate the negative sequence impedance islanding detection relay, one must consider the level of unbalance in the system, the utility and load impedances, the threshold setting and sensor accuracy. In this case, the ideal location for islanding detection is either Load 3 or Load 5. The Z2 impedance decreased significantly when the large rotating machine was introduced into the system. The impedance ratio of Load 5 dropped from 10.5 to 1.99 while the impedance ratio of Load 3 remained close to 4.3. In this scenario, the ideal location for the negative sequence islanding detection relay is at Load 3. Care must be taken to choose the right location and to assess the unbalanced system measurability for islanding detection in systems that contain large rotating machines.  4.3 Practical Example 1: 25 kV Radially Feed Distributed Generator Network 4.3.1 System Description The case study in this section corresponds to an actual operating system. The system, located in British Columbia, Canada, is a 25 kV radial network fed by a 3.56 MVA run of the river distributed generator. The purpose of this case study is to demonstrate how naturally occurring negative sequence components can be used to measure the feeding system impedance in a 25 kV distribution network using commonly installed CT and PT equipment. The impedance measurement capability can then be used as a centrallybased islanding detector in the area. The 3.56 MVA distributed generator is capable of supporting an island in the network but requires islanding detection to separate the utility system. The system schematic is in Figure 4.3. The 138 kV feeding utility is located at the top of the diagram, and the DG is connected at the bottom. The points of voltage and current measurement are shown as three circles “A”, “B” and “C” with corresponding CTs and PTs. Point “A” was used to measure the negative sequence current and voltage for the utility. Point “B” was used to used to measure the negative sequence current and voltage for the DG. Point “C” was used to measure the negative sequence current and voltage for the load, but this point was only simulated. The experimental data was taken while the system was running at steady state under typical loading conditions over several days. The utility provided the name plate data and impedances for all of the equipment (lines, transformers, generators, loads). The particular names and locations of the network have been altered to protect the customer’s confidential information. The expected values for Z2 (Th´evenin impedances) measured from each of the three points  95  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection is the impedance towards the utility for “A”, towards the DG for “B” and parallel impedance between the between the utility and the DG for “C”. The impedances can be seen in Table 4.11.  Figure 4.3: Practical Example 1: 25 kV DG System Fed System Single Line Diagram  In previous sections, it was shown that the positive sequence current sources are created from the generators and negative currents come from unbalanced loads. If the loads in the system in Figure 4.3 are unbalanced, the negative sequence impedance can be measured from A and “B”. Negative sequence impedance measured at point A will correspond to the impedance toward the 138kV transmission system and negative sequence impedance measured at point ’B’ will correspond to the impedance towards the DG. The positive sequence impedance measured from either of these locations will not present any meaningful information for islanding detection. Selection of a CT and PT measurement point for the negative sequence impedance islanding detector requires careful consideration. There are three potential locations “A”, “B” and “C” to be considered that are shown in Figure 4.3. Position “A” CTs and PTs are measuring currents from the utility. If 96  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection  Table 4.11: Expected Z2 Impedances From Positions ”A”, ”B” and ”C” (in pu) Position State |Z2 | ”A”  ”B” ”C”  Islanded Utility Connected Islanded Utility Connected Islanded Utility Connected  ∞ 8.52 99.93 99.93 99.93 7.84  the negative sequence current is large enough from Load1 and Load2 , Position “A” negative sequence impedance would equate to the utility impedance. Position “B” CTs and PTs are not ideal because they would measure the negative sequence impedance towards the DG and not detect island conditions. Position “C” at Load1 may be a useful alternative to Position “A”. The CTs at Position “C” are different from Potion “A” because they are likely to be significantly smaller and will have a higher sensitivity to smaller currents. If the CTs at “A” are rated so large that small unbalanced currents are below the large CT’s measurable limits, the CTs at Potion “C” may be used as alternatives. A practical implementation problem may exist with this system for an Independent Power Producer (IPP) who wants to install a negative sequence impedance islanding detector at Potion “A”. Installation of negative sequence impedance islanding detection in many cases is ideally accessed at the utility sub stations. The CTs and PTs at points “A” and “B” are owned by the utility at the substation and the IPP may not have adequate access to them. However, the IPP (or the utility) can install negative sequence impedance islanding detector by accessing privately owned load CTs and PTs on the 25 kV bus at Load 1 or Load 2. This has been demonstrated at circle “C” with simulated experiments of the system. The single line diagram in Figure 4.3 is expressed in a sequence component diagram of impedances shown in Figure 4.4. The top circuit in Figure 4.4 is the positive sequence network, the second level is the negative sequence network and the bottom circuit is the zero sequence network. The utility provided the impedances (at 25 kV) for positive, negative and zero sequence symmetric components. The impedances in Figure 4.4 can be converted from pu in Figure 4.3 by multiplying Z pu by Zbase as seen in Equation 4.13.  Zat  25 kV  = Z pu · Zbase = Z pu ·  (25 · 103 )2 100 · 106  (4.13)  The unbalanced impedance block links the three sequence component circuits to an “unknown schematic”. The connection between symmetrical components could be line-to-line, line-to-neutral, or a combination of the two, but the true circuit is unknown and can only be presented as a “black box”. Chapter 3 describes this problem in more detail. The power demand of the loads on the 25 kV bus vary in time, but they can be assumed to range between 0.5 and 4 MVA (or a 1.25 kΩ to 150 Ω). Refer to 97  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection Figure 4.3 and Figure 4.4 for each of the scenarios described in this section.  Figure 4.4: Practical Example 1: 25 kV DG System Fed System Sequence Impedances  4.3.2 Experimental Results The live measurements at Potion “A” and “B” were taken from the utility during different loading conditions are shown in Table 4.12. The voltages and currents were previously converted to positive and negative symmetric components by the utility acquiring relays. The negative sequence impedances obtained through experimentation in Table 4.12 correlate closely to the actual impedances. The average measured negative sequence impedance at Position “A” is 12 Ω as compared to the expected actual value of 8.52 Ω. The average measured negative sequence impedance measured at Position “B” is 102 Ω as compared to the actual value of 99.93 Ω. Under normal operating conditions, naturally occurring negative sequence components can be used for negative sequence impedance islanding detection at Point “A”. To extend this case for islanding detection for an IPP, the system presented has been simulated for some additional cases (Figure 4.3). The utility can measure unbalanced voltages and currents at “A” and “B” but an IPP may not have access to the utility’s CTs and PTs. An option that was considered in this case was for the IPP to access CTs and PTs from nearby privately owned loads off of the 25 kV 98  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection  Table 4.12: Measured Values for 25 kV System Positive and Negative Sequences Location Positive Sequence Negative Sequence A |V1 | |I1 | |V1 /I1 | |V2 | |I2 | |Z2 |(Ω) Test 1 14900 106.70 140 35.70 2.65 13.47 Test 2 14830 113.90 130 47.28 5.63 8.37 Test 3 14810 63.10 235 44.98 2.95 15.24 Average 14847 95 168 43 4 12 B |V1 | |I1 | |V1 /I1 | |V2 | |I2 | |Z2 | Test 1 15274 25.46 577 215.67 2.62 81.37 Test 2 14567 7.78 1861 54.45 0.33 164 Test 3 15475 149 103 36 0.82 44 Test 4 14910 32 468 121 1.14 106 Test 5 14920 35 423 92 0.81 115 Average: 15029 50 687 104 1 102  Table 4.13: Experiments Simulated on 25 kV Practical System 1. Single unbalanced Load 1 with the utility connected 2. Single unbalanced Load 1 and the system is islanded 3. Both Load 1 and Load 2 are identically unbalanced load magnitudes on phase A with the utility connected. 4. Both Load 1 and Load 2 are identically unbalanced load magnitudes on phase A and the system is islanded 5. Same as case 3 with the unbalanced phase on Load 1 on phase A and Load 2 on phase B 6. Same as case 4 with the unbalanced phase on Load 1 on phase A and Load 2 on phase B Bus at circle “C” in Figure 4.3 (off the 25 kV bus, right hand, middle of the diagram). At this point, the IPP can obtain unbalanced load data. The islanding detection scenarios investigated at point “C” are listed in Table 4.13. The results of each scenario for islanded and non-islanded cases are listed in Table 4.14. The negative sequence impedances for Case #1 at Positions “A” and “B” have been computed to cross-check the expected results from the following cases at Position “C”. The expected negative sequence impedance computed from Position “C” for an islanded state is the impedance of Load 2 in parallel with the impedance of the DG (99.93 Ω). Similarly, the expected negative sequence impedance from Position “C” for a non-islanded state is the parallel impedance of Load 2, the DG, and the utility (7.48 Ω). The negative sequence impedances measured from Positions “A” and “B” confirm the Th´evenin impedance values of the utility and the DG respectively. The negative sequence impedance measured value at “A” for Case #1 is 8.41Ω as compared to the actual value of 8.52Ω. The negative sequence impedance measured at “B” for Case #1 is 97.93Ω as compared to the actual value of 99.93Ω. The negative sequence impedances computed at “C” show the values drifting from the actual expected impedances (Table 4.14). The negative sequence impedance for experiments #2, #4 and #6 are expected to indicate an islanded impedance of ZLoad2 ||ZDG = 96.41Ω. The computed impedances for 99  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection  Table 4.14: Simulated Values for 25 kV System Positive and Negative Sequences Experiment Positive Sequence Negative Sequence Experiment at “A” 1 Experiment at “B” 1 Experiment at “C” 1 2 (island) 3 4 (island) 5 6 (island)  V1 14414 V1 14414 V1 14414 14554 14414 14554 14414 14553  I1 13.3 I1 43 I1 25.64 25.64 25.61 25.88 25.63 25.89  |V1 /I1 | 1080 |V1 /I1 | 334 |V1 /I1 | 562 562 562 562 562 562  V2 11.1 V2 11.1 V2 10.95 117 11.1 116.83 9.80 100.3  I2 1.322 I2 0.113 I2 1.432 1.22 0.72 0.586 0.73 0.69  |Z2 | 8.41 |Z2 | 97.93 |Z2 | 7.64 95.64 15.4 199.26 13.4 145.33  these three experiments are 95.64 Ω, 199.26 Ω and 145.33 Ω respectively. The negative sequence impedances for cases #1, #3 and #5 are expected to equal ZLoad2 ||ZUtil ||ZDG = 7.84 Ω ; however, the  computed negative sequence impedances for the three cases are 7.64 Ω, 15.4 Ω and 13.4 Ω respectively. All the impedance results computed from the six experiments correlate to the estimated impedances.  However, impedance study results for cases 3 to 6 are distorted with unbalanced conditions from the parallel connected unbalanced Loads 1 and 2.  4.3.3 Discussion Field data of naturally occurring unbalanced voltages and currents can be used to measure the Th´evenin impedance towards both the DG and the utility sources. The measured Th´evenin impedance can then be used for islanding detection. Other unbalanced loads near to the impedance measurement system reduces the accuracy of the impedance measurement but not the islanding detection capability. Similar results can be achieved for islanding detection by accessing CTs and PTs from local loads. This case is particularity useful if the utility bus’ CTs are too large to accurately measure small unbalanced currents or if an IPP requires an islanding detection system independent of the utility. This alternative case for the IPP was presented through simulations of unbalanced loads in the network. The most accurate negative sequence impedance measurements and computational analysis came directly from the utility CTs and PTs which did not have the same corruption from parallel unbalanced loads.  4.4 Practical Example 2: 600 V Fed Commercial Building 4.4.1 System Description The case study in this section describes an actual operational office building power system. The system is a 600 V commercial building in British Columbia, Canada with a solar panel installation on the roof. The purpose of this experiment was to demonstrate how negative sequence impedance can be 100  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection used for islanding detection on a low voltage bus. The electrical reach of negative sequence impedance measurement is also investigated in this section. The four-story commercial building features office use on the top three floors and some basic machining rooms and offices on the first floor. The solar panels on the roof currently do not have the capacity to supply the entire building’s energy demand; however, if a large array or other generating technology was installed, this building could potentially island in the local network. The building is a newly constructed and energy efficient ’LEED’ 1 certified building. The experiments conducted in this section use both naturally occurring and injected negative sequence symmetric components to calculate the Th´evenin impedance out of the building. These experiments were conducted at various locations in the building. A single line diagram with an associated impedance diagram from the 69 kV utility distribution system to the 600 V office building is shown in Figure 4.5. The single line diagram is on the left hand side of Figure 4.5 and the associated impedances are shown on the right hand side of Figure 4.5. Details of the distribution system in the single line diagram were supplied by the local, utility provider. The building schematics did not specify the distribution cable and transformer impedances. The impedance values of these components were found by measuring the voltage drop through the items between two different loading conditions using Equation 4.14, Equation 4.15 and Equation 4.16. The variable [E0 ] is a constant voltage source. The variable [Vlow load ] is the three phase voltage vector under low load conditions where, [Ilow load ] is the corresponding three phase current vector out at this low load state, [Vhigh load ] is the three phase voltage vector under load and [Ihigh load ] is the three phase current vector out of the load. By combining Equations 4.14 and 4.15, the result is Equation 4.16. The impedances shown in Figure 4.5 are in pu with a base power of 100 MVA. The base voltages are listed on each bus.  [V0 ] = [Vhigh load ] + [Ihigh load ] · [ZABC ]  (4.14)  [V0 ] = [Vlow load ] + [Ilow load ] · [ZABC ]  (4.15)  ZABC =  [Vlow load ] − [Vhigh load ] [Ihigh load ] − [Ilow load ]  (4.16)  The experiments for negative sequence impedance islanding detection were conducted at the three bold numbers “1”, “2” and “3” in circles seen in Figure 4.5. Position “1” is the main building feeder where natural unbalanced currents and voltages were measured for a 24-hour period. Position “2” is a 208 V feeder on the building’s 3rd floor where unbalanced injection experiments were run. Position 3 is at the solar panel DG inverter system where negative sequence current injection islanding detection computational analysis simulations were run. Limited access to Position “3” prevented direct voltage and current measurements, but detailed simulations were run to demonstrate the performance of the negative sequence islanding detection technique at the DG inverter’s terminals. Realistic simulation conditions were created by matching the loading and unbalanced conditions measured from Position 1 Leadership  in Energy & Environmental Design - Green Building Rating System by the U.S. Green Building Council  101  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection “1” and “2”. The data for all the experiments were acquired using the AEMC 3945 three phase power analyzer and logger. The data was then converted to symmetric components using matlab scripts. The AEMC 3945 is shown in Appendix C.  Figure 4.5: Practical Example 2: 600 V Commercial Office DG Fed System Single Line Diagram  The bold letters ’A’ and ’B’ in squares in Figure 4.5 represent the breakers where DG islanding could occur. Position “A” is the main disconnect breaker for the 12.5kV to the 69 kV network. Position “B” is the main building breaker at the 12kV to 600 V transformer. Testing island and non-island states poses an experimental challenge as the main power switches in this case study could not be simply opened and closed. Therefore, the non-islanded states have been directly measured on the live system and compared to the actual impedance values where the islanded states were validated using the building model built in Matlab Simulink Power Systems Library. The Simulink model was validated by comparing the voltage quality and unbalance at various nodes with actual system measurements. The model is shown in Figure E.1 in Appendix E. The following experiments express impedances in pu. The pu base power for all the experiments is 102  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection 100 MVA and the base voltage will depend on the bus of interest.  4.4.2 Experimental Results Position 1: 600 Volt Main Feeder The purpose of the experiment at Position “1” (Figure 4.5) was to determine if the naturally occurring negative sequence components could be acquired to adequately monitor negative sequence impedance for islanding detection. The measurements were taken from the main 600 V feeder. Voltage and current data were acquired simultaneously over a 24-hour period at a sampling rate of once per minute. The power per phase was plotted (Figure 4.6) and smoothed in 15 minute windows for ease of viewing. The median three phase power over a 24-hour period was 205.4 kVA. The measurements were taken on a weekday in January with the outdoor air temperature ranging from 5 to 10 degrees C. In the twenty-fourhour period, the negative sequence voltage on the 600 V bus ranged from 0.14 V to 0.63 V, while the negative sequence current ranged from 1.0 A to 11.7 A. The primary consuming loads of the building are computers, heaters and lighting with an average per phase steady state Th´evenin impedance of 480 pu and the Th´evenin impedance out of the building towards the utility of 11.6 pu. The 480 pu impedance has been obtained by taking the average daily demand on the building and using Z =  V2 S .  The 11.6 pu  impedance has been obtained by calculating the Th´evenin impedance from the impedance diagram in Figure 4.5 from the 600 volt bus at Point 1 to the 69kV utility feeder. Building Power Demand Over 24 Hours, Winter 100 Phase A Phase B Phase C  Apparent Power, kVA  90  80  70  60  50  40  5  10 15 Time − 1PM to 1PM (hours)  20  Figure 4.6: Practical Example 2: 600 V Commercial Office 24 Hour Power Demand  The positive and negative sequence impedances at this feeder vs. time are shown in Figure 4.8 for direct comparison of the two. The negative sequence is shown on its own in Figure 4.7. While the solar panel is not connected, the positive sequence impedance is Z1 =  V1 I1 ,  meaning that Z1 equates  to the building’s load demand on one phase. The negative sequence impedance Z2 = 103  V2 I2  means that  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection Z2 equates to the Th´evenin impedance out of the building. The average negative sequence impedance measured was 27.5 pu and the average positive sequence impedance measured was 480 pu. The actual Th´evenin impedance towards the utility was 11.6 pu and the average Th´evenin impedance towards the building was 480 pu. The experimentally measured negative sequence impedance correlates to the actual Th´evenin impedance away from the building averaging at 27.5 pu. The negative sequence impedance varied randomly by ±17.5 pu (Figure 4.7) and was not affected by the building load profile throughout the day as the positive sequence impedance highlights (Figure 4.8).  Figure 4.7: Practical Example 2: 600 V Z2 Measured Over 24 Hours  The opening of Breakers “A” and “B” were simulated using the equivalent Simulink model as shown in Figure 4.9. The model was created using the schematic given by the local utility and by modeling live unbalanced conditions previously measured in the Simulink model. As Breaker “A” and “B” were opened, the negative sequence impedance quickly increases. When Breaker “A” was opened, the negative sequence increased from 27 pu to 150 pu in less than 0.02 s. When Breaker “B” was opened, the negative sequence increased from 150 pu to 86,000 pu in a similar amount of time. 86,000 pu is very large meaningless number and it can be considered to be infinite. The building DG cannot support the power requirements inside the network contained within Breaker “A” and so the voltage collapses when Breaker “A” is opened. Some inaccuracy of the negative sequence impedance accuracy is expected due to the other unbalanced loads in the system. However, when Breaker “B” was opened, the change in negative sequence impedance was far more significant rising to near infinite levels.  104  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection  Figure 4.8: Practical Example 2: 600 V Combined ZLoad and Z2 Measured Over 24 Hours  Position 2: Negative Sequence Injection on the 208 Volt Bus This experiment at Position “2” in Figure 4.5 demonstrated how negative sequence components can be injected when there is not enough naturally occurring unbalance of the phase voltages and currents, or there is another significantly stronger negative sequence source that corrupts the measurement. The experiment was run from the office building third floor of at a 208 V feed. Negative sequence currents are injected and the negative sequence impedance is measured from it. The experiment uses the three phase injection averaging method derived in Section 3.3.6 to improve the accuracy: ZSys ≈  V2−A−unbal V2−B−unbal V2−C−unbal + + I2−A−unbal I2−B−unbal I2−C−unbal  (4.17)  In Equation 4.17 the variable V2−x is the negative sequence voltages on each phase and I2−x is the negative sequence currents on each phase, with the ‘x’ depicting either phase A, B or C injections. Injections of negative sequence currents are created by connecting a variable load from Phase A to Phase B, leaving Phase C open on the 208 V bus. The schematic for the negative sequence injection is shown in Figure 4.10. Practical access limitations and safety concerns led to the experiment being run on the higher impedance 208 V bus at Position “2”, instead of on the lower impedance 600 V bus. Position “2” is shown in Figure 4.5 on the single line diagram in the left and the impedance diagram on the right. The Th´evenin impedance of the system measured from the third floor 208 V feeder at Position “2” was 253.6 pu. The 253.6 pu impedance can be broken down into several components: from the experimental setup (Position “2”) to the first breaker, the impedance was 120 pu (the 208 Bus in Figure 105  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection  Figure 4.9: Practical Example 2: 600 V Z2 At Position 1 During Breaker A and B Opening  4.5), the impedance from the 208 V bus to the 208:600 volt transformer was 92 pu and the transformer’s impedance is 35 pu (8%). The Th´evenin impedance from the transformer towards the feeding system was 11.6 pu (Position “1”) as shown in Figure 4.5. Experiments of the non-islanded state were run on the live system while islanded states were conducted in simulation. The impedance changes are measured and computed when Breaker “B” was opened (Figure 4.5). The measured impedance from injecting negative sequence components is shown in Figure 4.11, impedance vs. power. The delta connected phase combinations listed in the legend of AB, BC and CA represent the negative sequence impedance vs. injected power. For example, Phase AB represents a load being placed on phases AB only while leaving phase BC and CA open. This is repeated for phases BC and CA. The measured impedance measured from this experiment gave different negative sequence impedances ranging from 500 pu to 100 pu on different phases. Additionally, the impedance measurements do not stabilize until the injected power is over 1000 watts. The measurements achieved an average of 277 pu by using the averaging technique of impedances (Figure 4.11 - “Smoothed Averaged 3 Phases”). When the averaging technique was employed with smoothing, more accurate impedance readings were made with only 100 watts of injected power. The system was islanded by opening Breaker “B” in simulation. The changes in impedance are shown in Figure 4.12. After the initial stabilization of the islanding, there is a small change in impedance between an islanded condition and a non-islanded condition. The change between an islanded and non islanded state is only 20 pu is due to the high input impedance from the 208 V bus. The small change of 20 pu relative to the 200 pu steady state impedance is not adequate for accurate islanding detection. The actual impedance of 253 pu is much higher than the small change of 11.6 pu that occurs from opening 106  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection  Figure 4.10: Practical Example 2: 208 V Negative Sequence Injection Experimental Setup  Breaker “B”. This may make threshold settings for an islanding detection relay to detect Breaker “B” opening very difficult at this Position because of the small 11.6 pu change relative to the 253 pu total impedance. Position 3: 600 Volt Solar Panel Coupling The experiment at Position “3” as shown in Figure 4.5 is run at the point of electrical coupling of the roof top solar DG system. This system is currently integrated into the building directly through an inverter that is then connected to a 208:600 transformer on the main bus of the building. The purpose of this case is to show how negative sequence islanding detection can be used directly by the inverter. The inverter injects negative sequence components into the network and computes the negative sequence impedance. Islanded and non-islanded conditions have been tested by simulating the network and monitoring the negative sequence components. The negative sequence injection technique is made by deliberately lowering the voltage on one phase for several cycles and then switching to subsequent phases. As seen in case study 2, individually unbalancing each phase and subsequently measuring the negative sequence impedance allows the more accurate impedance averaging technique to be employed. Steady state conditions for the experimental simulations are set so that there is near zero power flow from the utility into the building. Zero power flow is the most difficult case for many islanding detection techniques. The level of output unbalanced voltage at the inverter’s terminals was set to 2.5%. The resulting two computed impedances ZLoad and Z2 of this experiment both quickly increase when Breaker “B” was opened to island the system (Figure 4.13). The Th´evenin impedance of the system under non-islanded conditions from the solar DG is 37 pu. The negative sequence impedance  107  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection  Figure 4.11: Practical Example 2: 208 V Negative Sequence Impedance Phase to Phase Loads  computed was 39.2 pu. When Breaker “B” was opened, the negative sequence impedance increases to 567.2 pu in under 0.015 s. The islanding detection relay threshold can be set in between these two values. The building load impedance for this case is 525 pu. Normally, the impedance ZLoad would not contain meaningful information. However, the power flow out of the building is set to nearly zero so the positive sequence impedance computed from the inverter is approximately equal to the building impedance, ZLoad (Figure 4.13). Effect of Rotating Machines In the same case study examined in Section 4.4.2, a large 15 kVA HVAC Fan (about 20 HP) was added to the simulations. The machine was placed on the 3rd Floor 208 V Bus. The purpose of this experiment was to compare the impedance change when a single large rotating machine was added into the system. As shown in Section 3.5, rotating machines can have significantly lower negative sequence impedances due to the reverse rotating field created by the negative sequence currents. When rotating machines are known to be in the system, the islanding detection threshold settings may require more precision. A typical office building contains many small ventilation fans ranging from 0.500kW to 2.5kW. The rotor resistance of these machines dominates their negative sequence impedance component. A single large machine has a lower overall impedance than many smaller machines making it a worst case scenario for this case study. Data acquisition for this case study starts when the machine reaches synchronous speed and the network is stable. After the start of the case, the phase voltages and currents are monitored. Breaker “B” is opened at 0.1 s and the machine is disconnected from the network after 0.2 s. During this  108  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection  Figure 4.12: Practical Example 2: 600 V, PV Source Negative Sequence Impedance Transition  time, the solar panel DG is unbalanced by 5 % on one phase and the negative and positive sequence impedances are computed in pu similar to the previous section. The simulations are made using standard linear components from the Simulink Power Library, and in particular, the machine power, voltage and impedance parameters which are shown Table 4.15. All the machine values are in pu. Table 4.15: Induction Machine Parameters (15 kVA) Input Variables from Simulink Power Library Variable Value S 15 kVA V 208 V poles 2 TB 0.1 pu f 60 Hz rs 0.01965 pu Xls 14 pu XM 510 pu Xlr′ 14 pu rr′ 0.01965 pu  As shown in Figure 4.14, the impedance changed significantly between islanded and non-islanded, and when the machine was disconnected from the network. When the system was in a non-islanded state, the negative sequence impedance computed was 42 pu. This impedance was slightly higher than the expected system Th´evenin impedance of 37 pu. Opening Breaker “B” increased the impedance to an average of 316 pu in 25 ms. Disconnecting the 15 kW machine increased the impedance to an average 109  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection  Figure 4.13: Practical Example 2: 600 V Building PV DG Islanding Detection Results  of 573 pu after 25 ms of instability. The ratio between islanded and non-islanded states with the rotating machine is 7.5 opposed to the ratio of 14.3 for the same system without the rotating machine. Clearly, setting the threshold for islanding detection in this scenario would require additional care to take the machine negative sequence impedance into account. The output of positive sequence impedance ZLoad and the negative sequence impedance Z2 is shown in Figure 4.14 and the raw voltage output sinusoidal wave form for each phase is shown in Figure 4.15.  4.4.3 Discussion Islanding detection using negative sequence impedance measurement on a 600 V bus has been demonstrated. The method functions quickly (under a few cycles) and the impedance threshold window is wide. The natural load unbalance of the building served as an effective means for passively measuring negative sequence impedance at the 600 V feeder. Injection methods at the 208 V busses are dependant on the impedance of the connection to the 600 V bus, and the amount of corrupting unbalance in the system. In all cases, the measurement accuracy of negative sequence impedance is dependant on the amount of unbalance in the rest of the feeder. Additional improvements in measurement accuracy the amount of required injected negative sequence current was the result of employing the three phase impedance averaging method developed in this thesis. In all cases, the threshold windows were very wide. The 208 V bus experiment highlighted the errors caused from other unbalanced loads in the system. This became more pronounced when different phase-to-phase injections were made and the impedances were all different. This error is understood more clearly by looking at a symmetric component sequence 110  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection  Figure 4.14: Practical Example 2: Building Solar DG Islanding Detection with Rotating Machine  impedance diagram of the corrupting impedances illustrated in Figure 4.16. The single phase building loads, Z208 , inject single phase unbalanced symmetric currents through the “unknown” symmetric component network box. The phase to phase lab experiment ZLab injects the negative sequence current into the system.The two unbalanced symmetrical component currents from “unknown” symmetric component network and ZUnbal1 combine to make a complex circuit representation. The complex combination of unbalanced loads in the system results in Th´evenin impedance measurement errors made at ZLab . In the third experiment, the 2.5% unbalanced voltage solar panel source accurately computed Th´evenin impedances of the system. The changes in islanding transitions were well pronounced. The solar inverter had no other equipment tied to the bus to corrupt the negative sequence impedance computations; and the impedance of the 208:600 V transformer was 5%. A low Th´evenin impedance of 37 pu made a significant difference in islanding detection capability as compared to the previous experiment where the Th´evenin was in the hundreds of pu. The addition of a 15 kVA rotating machine to the network reduced the measured Th´evenin impedance by close to 50%, but a clear threshold for islanding detection is available.  4.5 Performance Comparison with other Impedance Based Islanding Detection Methods Negative sequence impedance islanding detection performs well under the presented cases. But, how does this compare with other impedance based islanding detection methods? In this section, the negative sequence impedance islanding detection is compared with two other impedance based islanding 111  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection Solar Panel, Machine, and Unbalanced Voltage vs. Time  800  V V  600 V  A  B  C  400  Volts (V)  200  0  -200  -400  -600  -800  0  0.01  0.02  0.03  0.04  0.05  0.06  0.07  0.08  0.09  0.1  Time (s)  Figure 4.15: Building Solar DG Islanding Detection with Rotating Machine Voltage ABC  detection methods. These methods are: 1. Non Harmonic Injection [7], and 2. Negative Sequence Voltage [74]. These methods will be compared to passive and active negative sequence impedance islanding detection and to these two presented methods. Quantitative performance characteristics are difficult to directly compare, due to the multitude of different configurations of networks. Nevertheless, there are some key characteristics employed in this section to provide insight into how each method performs. The characteristics these methods will be compared to were chosen to address the weaknesses of impedance based islanding detection (Table 4.16). Table 4.16: Performance Characteristics to Compare Islanding Detection Techniques 1. Non Detection Zone 2. Threshold Setting 3. Power Efficiency 4. Power Quality  Non-Detection Zone and Threshold Setting The non-detection zone and threshold settings are the two most significant characteristics that determine the value of an islanding detection system. The non-detection zone is the state in which the islanding detection technique loses its sensitivity. The threshold setting is the point that triggers the detection 112  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection  Figure 4.16: Practical Example 2: 208 V Bus Sequence Impedance Diagram from Lab  relay. Though the non-detection zone and the threshold setting of islanding detectors are different in definition, they overlap when compared to islanding detection. Therefore, these two characteristics are covered together in this section. Positive sequence injection impedance measurement techniques are limited because the Th´evenin impedance is not measurable at the fundamental or at any of its harmonics. This is because the assumption of Equation 4.18 does not apply when there are multiple sources at the same frequency in the network. For example, injecting 60 Hz into a live 60 Hz system would not reveal any meaningful impedance information about the network if Equation 4.18 is employed. ZT hevenin = ´  Vin ject Iin ject  (4.18)  Therefore, impedance measurements with injection techniques can only be made at frequencies near to the fundamental. If it is assumed the system is similar to a resistor-inductor (RL) type circuit, the impedance at the fundamental frequency is easily obtained through interpolation. However, resonance characteristics (below 2kHz) in a practical system from power factor correcting capacitors and nonlinear devices make estimating the impedance much more difficult and in some cases, impossible. If interpolation to the fundamental is not used, a non fundamental frequency impedance can then be used. However, this carries additional threshold setting problems. Consider the IEEE standard 1547.1  113  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection resonance bus [56] for testing DG islanding detection as shown in Figure 4.17. In this scenario, there is a single DG source, a load, an LC resonator and a Utility source. The Quality Factor, QF, is set to 1 and the inductor, L, and the capacitor, C, are set to resonate at 59.5 Hz. The QF can be found using Equation 4.19. To understand how the threshold settings are problematic, refer to Figure 4.18. Graphs ‘a’ and ‘b’ both contain plots of an islanded state and a non islanded state where graph ‘a’ is a high impedance utility connection and graph ‘a’ is a low impedance utility connection. If 265 Hz is chosen for the injected signal in the high impedance utility connection (Figure 4.18, graph a). The impedance difference between and islanded state and a non-islanded states is zero (the intersection of the two curves). Setting a threshold for islanding detection at this point is not possible. Similarly, with a low impedance utility connection, the same condition occurs at 451 Hz. When the system impedance is already known, setting the measuring frequency and threshold settings are relatively simple; however, in the system shown here, the specifics of the system are not known and thresholds will need to be calibrated each time the network is changed. Therefore, the non-detection zone of the injection techniques are at frequencies in which the threshold settings are near to zero. QF = R ·  C L  (4.19)  Figure 4.17: IEEE 1547 1 2005 Test Bus For the negative sequence voltage method [74], the non detection zone problem is due to different factors from those affecting the harmonic injections method. The negative sequence voltage method relies on a consistent and non changing negative sequence voltage source in order to prevent the island detector from false tripping. This is because after the threshold voltage is initially set, re-setting the threshold can be difficult and problematic. Unfortunately, the starting and stopping machines and unbalanced loads can cause significant voltage fluctuations and the threshold would require a continual updating algorithm for the voltage. Consider the negative sequence voltage method for the case study in Section 4.2 of this chap114  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection  Figure 4.18: Impedance Between Islanded and Utility Connected Resonating Bus  ter, which includes an induction machine (Figure 4.2). When Load 3 is connected in parallel with Machine 1, the negative sequence voltage varies significantly. The 2 MW load with the rotating machine connected in parallel has an islanded negative sequence voltage of 0.0019 pu, whereas with the 3.2 MW+1.9 MVAR load connected without the rotating machine, has a negative sequence voltage of 0.015. This is The change is 7.8 times larger. This large difference is particularly prevalent when the Th´evenin impedance ratio between islanded and non-islanded states is small. Similarly, in the case with different unbalanced loads, Load 5 (0.9 MW + 0.6 MVAR) is 15% unbalanced and outputs a negative sequence voltage of 0.0024 pu whereas Load 3 (3.2 MW+1.9 MVAR) is unbalanced by 10% and has a negative sequence output voltage of 0.0105 pu. The voltage ratio between the two loads is 4.3. The negative sequence voltage change in these two examples illustrates how the unbalanced voltage can vary depending on the loading. Although there is no real non-detection zone for the negative sequence voltage technique, the threshold setting poses a significant challenge in areas with large machines and constantly changing loads. The negative sequence impedance islanding detection method proposed in this thesis does not have the threshold or frequency selection problems that the two other islanding detection techniques that were presented pose. The most significant non-detection zone in the negative sequence impedance technique occurs when large rotating machines exist in the network. These machines can reduce the impedance ratio significantly and can result in difficult threshold settings. In the example shown in Section 4.2, the addition of a large rotating machine changed the impedance ratio from 11.7 down to 4.3. Setting the threshold becomes increasingly difficult when large rotating machines are present in the system, and when there are other large unbalanced loads corrupting the unbalanced voltage and current 115  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection measurements. Power Efficiency and Power Quality Efficiency and power quality are both important design considerations when using any active islanding detection technique. Power efficiency is a measurement of the amount of power the islanding detection scheme consumes. Power quality is measured by how much the islanding detection scheme alters or disrupts regular 60 Hz systems. Harmonic injection islanding detection adds to the total harmonic distortion(THD) of the system. For example, if a harmonic injection is made at an amplitude of 1% of the fundamental in the commercial building example, an increase of 1% of THD may place the system outside of the allowable 5% limits. Continuous injected harmonic power is extremely inefficient. Consider the previous example in the office building discussed in Section 4.4. Injection method used near to the fundamental on the 600 bus at the test building would have an impedance of approximately 11.6 pu. The base impedance is found using Equation 4.20. The total power for a continuously injected signal of 1% of 600 V is given by Equation 4.21. The continuous power required to run the islanding detection scheme can add up to 1% of the total power produced by the DG. Zbase = V ( f )2in j Z( f )thev  2 Vbase 6002 = = 0.0036 Sbase 100 · 106  (4.20)  (600 ∗ 0.01)2 = 862 VA 11.6 ∗ Zbase  (4.21)  =  Consider using the injection technique on the 25 kV bus test case in Section 4.3 seen in Figure 4.3 using an injection amplitude of 1% of the fundamental. By placing the injection system on the 25 kV bus, the Th´evenin impedance near the fundamental is 13.8 pu. The resulting injected power at this point is 7.2 kVA (Equation 4.22). The impedance used in these examples is at the fundamental, but the power required will vary depending on the frequency in which the injection occurs. An alternative to continuous harmonic injection to lower the power wasted is using periodical injection. V ( f )2in j (25 · 103 ∗ 0.01)2 = = 7.2 kVA 3 )2 Z( f )thev 13.8 ∗ (25·10 6 100·10  (4.22)  Harmonic injection and unbalanced voltage (caused by negative sequence current injection) affect the power quality in similar ways. Both will cause “non-torque” current to flow in rotating machines wasted as heat. Voltage unbalance has little effect on single phase loads, but can cause excessive heating on three phase rotating machines. This excessive heating reduces the machine’s lifetime and performance over time. In the past Sections, voltage unbalance as low as 0.3% is measurable. For example, in the lab setting in Practical Example 2: 600 V Commercial Building, the unbalanced load was injected into the system between 0.3% and 1.2%. At 0.6% (approximately 1000 W), the impedance could be resolved (seen in Figure 4.11).  116  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection  4.6 Summary In this chapter negative sequence impedance islanding detection has been applied to three test cases, and then compared to two other islanding detection methods. The three test cases highlight how negative sequence impedance islanding detection can function in different network configurations such as radial networks, non radial networks, medium voltage buses and low voltage buses. Voltage and current measurements from all the physical networks had naturally occurring unbalanced symmetric components. The first case, based off of the IEEE standard industrial bus, was evaluated with multiple DG sources, multiple unbalanced loads and a large rotating machine. The impedances measured were near exact values of the anticipated outcome when there was only one single unbalanced load in the system. However, multiple unbalanced loads were found to affect the impedance measurement accuracy and the ratio between the islanded and utility connected states. The size and amount of unbalanced loading on the system produced errors in the measurements. The total impedance ratio increased when the two loads were on the same phase, and the impedance did not change when the unbalanced conditions for the two loads were on opposite phases. The introduction of a large rotating machine resulted in the biggest reduction of the islanded to non-islanded impedance ratio and very small unbalanced voltages. The machine used in this case was very large in size, but it showed how significantly rotating machines can affect the negative sequence impedance measurement capability. The unbalanced voltages and currents were below the measurable limit of typical PT and CT sensors and implementing negative sequence impedance islanding detection in this system using current CT and PT technology may be difficult. In the second case study, negative sequence islanding detection was tested on a 25 kV utility bus. Unbalanced conditions were found to be measurable from the utility CT and PT sensors and the negative sequence impedance measured closely correlated to the system’s impedance. Further computer simulations indicated similar results during island conditions. Sensor placement was investigated using computer simulation. Nearby unbalanced loads were found to be ideal sources of naturally occurring negative sequence voltage and current, which the IPP could use as a method of independently conducting negative sequence impedance islanding detection. The third case study evaluated negative sequence islanding detection in a single office building with a solar DG installation. The first experiment at the building’s main feeder measured naturally occurring unbalanced conditions. The resulting negative sequence impedance closely correlated to the Th´evenin impedance away from the building. The second experiment was conducted off a 208 V feeder from the third floor of the building. Negative sequence current injection from the lab demonstrated how current injections can be used to measure the Th´evenin impedance away from the lab. Additionally, averaging three phase negative sequence impedance measurements was found to improve the accuracy while reducing the amount of needed injected current. The last experiment of this case study used generated negative sequence components from the solar DG for islanding detection. The inverter unbalance accurately measured the difference between island and utility connected states up to the building main breaker. The addition of a rotating machine decreased the impedance ratio between an islanded and non-islanded condition, although islanding detection threshold could be set. 117  Chapter 4. Case Studies of Negative Sequence Impedance Islanding Detection The last part of this chapter compared negative sequence impedance islanding detection with two common impedance based islanding detection schemes. The effects of harmonic injection and voltage unbalance on power systems were addressed. Power quality, non-detection zone, efficiency, and threshold settings were qualitatively compared with negative sequence impedance islanding. Negative sequence impedance islanding detection is an improvement upon these two impedance based islanding detection schemes in all the characteristics reviewed.  118  Chapter 5  Conclusion Distributed generator installations into busses near electrical consumers have created new challenges for protection engineers. The typical protection configurations such as unplanned islanding, reclosures, out of step monitoring, impedance relay protection zones and of distributed generator systems need to be reevaluated. This chapter summarizes the main contributions presented in this work with suggested future research on the method of negative sequence impedance islanding detection. This thesis has presented a novel method of islanding detection for the protection of distributed generator fed systems. The proposed islanding detection method measures the negative sequence impedance and compares it to a known impedance threshold that signals an island condition when the impedance increases past it. Negative sequence impedance islanding detection is a unique addition to the existing islanding detection techniques. Negative sequence impedance islanding detection addresses many of the existing limitations of islanding detection.  5.1 Summary of Contributions This thesis has made the following contributions to the field of islanding detection for distributed generation. 1. Development of a novel solution for islanding detection of distributed generators using negative sequence impedance This thesis presents a new concept of islanding detection based on the concept of negative sequence Th´evenin impedance measurements [72]. The change in impedance between island and utility connected systems is the threshold trigger. The proposed concept provides a more effective impedance based islanding detection scheme that can be used with naturally occurring unbalanced loads or with injected unbalanced currents. The advantages of negative sequence impedance islanding detection are the wide impedance threshold range setting, minimal integration costs, accuracy using naturally occurring unbalanced loads and a fast response time of 25 ms compared to the 2 s IEEE 1547 standard. The minimal integration costs are a result of only requiring existing system CT and PT sensors without having to install expensive highly sensitive PTs and CTs to access voltage and current signals. In the cases investigated, the existing naturally occurring unbalanced loads offer the necessary amount of unbalanced conditions to support reliable impedance measurements for islanding detection. 119  Chapter 5. Conclusion 2. Theoretical analysis and performance of negative sequence impedance measurement for islanding detection The concept of negative sequence impedance islanding detection has been proven algebraically and through simulations. The proof identified the correlation of the negative sequence impedance to the Th´evenin impedance away from the unbalanced load. This thesis models how each symmetrical component is created, and demonstrates the negative sequence performance characteristics in terms of load sizes, power factor, connecting utility strength and the effects of multiple unbalanced loads. The advantage of this analysis is that the performance characteristics are clearly expressed and can be used to further determine the suitability of negative sequence impedance islanding detection in a large number of scenarios. 3. Field data and network modeling supporting the measurability of naturally occurring negative sequence voltage and current Field data was collected and the associated systems were modeled after to test the ability to acquire real live data in the field for use with the negative sequence impedance islanding detection method. The field data supports the measurability of naturally occurring negative sequence voltages and currents on 25 kV, 600 V, and 208 V networks that lead to accurate impedance measurements with clear threshold settings for islanding detection. Actual CT and PT sensors were used to make measurements on live power systems proving that current technology can be used. Three test cases were addressed to cover a range of scenarios. The test cases covered were 1. an IEEE standard bus system modeled similarly seen in [65], 2. a 25 kV radial distributed generator fed bus and 3. a 600 V office building. Live measurements and matching computer simulations were made. Though the negative sequence voltages in scenario #1 were below typical CT and PT measurement limits, the concept capability is still effectively demonstrated in simulation. The advantage of these field experiments and simulations is two of the three practical scenarios explicitly demonstrated that negative sequence impedance islanding detection can be used in the field; and the currently installed equipment has the sensitivity for accurate measurements. 4. Field data supporting the creation and measurability of injected negative sequence current Simulations, numerical analysis and lab experiments were conducted to create negative sequence injection for measuring negative sequence impedances. The method of injection can be applied using a simple load or may be used in connection with equipment such as inverters, which can deliberately inject negative sequence currents. Live experiments and simulations run on a 208 V bus in an office building demonstrated that this method can be practically used for measuring impedances for distributed generator islanding detection. The advantage of this work occurs in areas where there is no access to other unbalanced loads or where the local unbalance from other loads is too great. This method can be used to offset  120  Chapter 5. Conclusion the other unbalanced loads or simply as a negative sequence source in an area that does not have naturally occurring unbalanced loads. 5. Design, modeling and practical experiments of a novel concept of three phase sequence injection impedance measurement averaging to enhance impedance measurement accuracy The concept of individual three phase sequence injection impedance measurement averaging with injected negative sequence components was verified in an algebraic analysis, simulation and with practical experiments on a 208 V bus in this thesis. Unbalanced systems were found to be consistently unbalanced on one particular phase, however, by injecting negative sequence current into the unbalanced system by deliberately loading phases AB, BC, then CA different impedance answers were revealed. The average of the three impedances was found to equate to a more precise Th´evenin impedance. This concept has the advantage of increasing the accuracy of negative sequence impedance measurements while reducing the required power of the current injections for measuring Th´evenin impedance of the system. 6. A comprehensive review of all impedance measurement techniques for live systems in the past 20 years In this thesis, a comprehensive review of all impedance measurement techniques for live power systems over the past 20 years was conducted. The review was organized by style of impedance method and chronologically sorted. This comprehensive review is not available in any past works and is necessary to fully understand the value of the concept presented in this thesis. The advantage of this comprehensive review is that progression and limitations of current impedance measurement techniques are revealed.  5.2 Future Work Negative sequence impedance islanding detection technique has been demonstrated to operate in many ideal network conditions; however, there are some future research topics that may further enhance the concept. There are several significant areas to be explored in this work: 1. Techniques to improve or extend existing PT and CT performance Through all the practical measurements and simulations, the unbalance voltage is often very small, which limits the accuracy of the measurements. Many of the unbalanced voltages are in the range of 1 % to 0.1 % which is near the absolute accuracy of some CTs and PTs. Though the accuracy limitation of the CTs and PTs is around the typical measurements, the relative noise platform may be much lower. This means the performance of these CTs and PTs may be extended beyond the original expectations. Further investigation into PT and CT noise platforms and the typical accuracy beyond the nameplate would assist in furthering this research. 121  Chapter 5. Conclusion Measuring beyond the CT and PT noise platform has the potential to lead to nuisance tripping. Higher accuracy system measurements can improve the reliability of measurements. Therefore, further research into methods of improving existing installed CT and PT accuracy could be beneficial. This would involve taking several typical CT and PT systems into a lab setting to evaluate whether simple modifications could be made to further improve or estimate actual measurements. Several enhancements to high voltage measurement techniques have been made in the past decade, such as, using high accuracy laser cut resistor dividers, and optical current and voltage sensors. Although they have not garnered wide spread industry acceptance. Investigations on more cost effective replacements of existing old CTs and PTs would accelerate this; allowing for more wide spread application of negative sequence islanding detection to be introduced. This technique has not yet been applied to harmonic impedance measurements due to sensitivity limitations. Network harmonics are often less than 5% of the fundamental, whereas unbalanced harmonics would be only a fraction of that per cent limiting the measurability of these components. Enhancements in voltage and current measurability will further enhance the ability to address this concept beyond the theoretical analysis. 2. Sensor Reach Through Weak Connections and Heavy Unbalanced Systems As seen in previous investigations, complex and multiple heavily unbalanced high system impedance networks limit the ability to accurately measure island states, and limit the negative sequence impedance sensor’s reach to the islanding breaker. Further research into these conditions may extend the performance. The research could investigate alternatives with injection types with the combination of phase and amplitude measurements. As this work has dealt with absolute impedance measurements, phase errors and system amplitude trend signatures may also present potential for further research into the reach of negative sequence measurements. 3. Enhanced Understanding of Practical Unbalanced Conditions This work consisted of a study of several sites with actual practical measurements on three scenarios, one being simulation. However, a more comprehensive study of natural system unbalance is needed. To date, there has been little research completed on unbalanced power systems whereas most studies and approximations are made assuming perfectly balanced network states. This assumption has lead to very little research and published data on unbalanced networks and their overall effect. Practical published cases are rare. The research community and the concept presented in this thesis would benefit from a detailed review of typical conditions for utilities around the world. Practical measurements are extremely difficult to attain due to national security requirements of utilities. This leads to a reluctance of utilities to release any information indicating network performance or locations of measurements. However, a specialized study with a partnership between academia and industry with only Th´evenin impedances of the specific locations may allow for a study such as this to be completed. Understanding of the unbalanced conditions that exist at vari122  Chapter 5. Conclusion ous voltage levels from 500 kV to 4 kV will benefit further implementation and other applications of using negative sequence impedance to measure the Th´evenin impedance. 4. Transient and Disturbance Performance Although negative sequence impedance estimation for islanding detection has been found to function well under steady-state conditions, transients and other disturbances create errors that may lead to nuisance tripping. Performance details during transients require further study to improve this limitation. For example, network transients can cause difficulty in sequence conversion with very large errors. There are opportunities for modern filtering and other signal processing techniques to enhance the ability to process the negative sequence impedance measurement beyond the steady-state studies investigated in this work. 5. Integration challenges with grid tie inverters and relays Stand alone devices and natural measurements have been studied in this work, however, integration strategies into inverter devices and power system relays requires further understanding. For example, the active unbalanced sources using the averaging impedance method is limited by the effect on the system and power quality degradation. Active negative sequence current injection requires study in minimizing the disturbance to the customer. A relay prototype or an inverter for negative sequence impedance measurement and injection algorithms would be beneficial to further understand the integration challenges. 6. Rotating Machine Negative Sequence Impedance Rotating machines pose limitations to the measurement ability of negative sequence impedance islanding detection. In smaller buildings where small motors and small HVAC systems exist, the high impedance of these items is still much higher than low impedance utility connection. Larger systems such as industrial facilities with rotating machines dominating the load can have impedances that significantly lower the ability to detect an island. The position of the islanding detection sensor is critical to its sensitivity. Further study into how to reduce the effect of rotating machines on this technique would greatly extend the value of this research. This evaluation extends beyond motors to rotating induction and synchronous generators. The injected unbalance in an inverter was investigated in this work, but rotating machine injections were not. This limitation requires a detailed study which can benefit the community two-fold. The first is to extend the use of negative sequence impedance islanding detection for machines as well as investigating the life of machines with deliberate unbalanced output. 7. Negative Sequence Impedance Phase Response This thesis covered absolute values for impedance measurements. This was because of the phase errors encountered during signal processing and Fourier transform of the data. Very small errors or sampling inaccuracies cause significant phase errors. Further investigation into the phase response of negative sequence impedance may enhance the accuracy further. 123  Chapter 5. Conclusion  5.3 Final Remarks The novel concept of negative sequence impedance islanding detection has been introduced in this thesis. The practical cases and mathematical evaluations highlight some practical applications where this islanding detection scheme may be used. Negative sequence impedance islanding detection has improved upon the threshold settings, non-detection zone, and installation cost limitations of other impedance based methods. As has been shown, negative sequence impedance islanding detection is a viable method for islanding detection applications.  124  Bibliography [1] ABB. Abb focs fiber-optic current sensors datasheet. ABB, 1(1-2):2, 2005. [2] ABB. Outdoor instrument transformers application guide. ABB, 1(2.1):1–124, 2005. [3] Spencer Abraham. National transmission grid study. Technical report, US Department of Energy, may 2002. [4] J. Arrillaga. Power system harmonic analysis. Wiley, Chichester; New York, 1997. Jos Arrillaga. [et al.].ill.; 25 cm. [5] L. Asiminoaei, R. Teodorescu, F. Blaabjerg, and U. Borup. A new method of on-line grid impedance estimation for pv inverter. In Applied Power Electronics Conference and Exposition, volume 3, pages 1527–1533. APEC ’04. Nineteenth Annual IEEE, IEEE, 2004. [6] L. Asiminoaei, R. Teodorescu, F. Blaabjerg, and U. Borup. A digital controlled pv-inverter with grid impedance estimation for ens detection. Power Electronics, IEEE Transactions on, 20(6):1480 – 1490, 2005. [7] L. Asiminoaei, R. Teodorescu, F. Blaabjerg, and U. Borup. Implementation and test of an online embedded grid impedance estimation technique for pv inverters. Industrial Electronics, IEEE Transactions on, 52(4):1136, 2005. [8] P. Barker. Overvoltage considerations in applying distributed resources on power systems. In Power Engineering Society Summer Meeting, 2002 IEEE, volume 1, page 109, 2002. [9] P. P. Barker and R. W. De Mello. Determining the impact of distributed generation on power systems. i. radial distribution systems. In Power Engineering Society Summer Meeting, 2000., volume 3, page 1645. IEEE, 2000. [10] J. Lewis Blackburn. Protective relaying: principles and applications. Power engineering; 5. M. Dekker, New York, 2nd edition, 1998. J. Lewis Blackburn.ill.; 24 cm. [11] Ward Bower and Michael Ropp. Evaluation of islanding detection methods for photovoltaic utility-interactive power systems. Report IEA PVPS Task 5 IEA PVPS T5-09: 2002, Sandia National Laboratories Photovoltaic Systems Research and Development, March 2002. [12] CSA Standard CAN3-C235-83. Preferred voltage levels for AC systems, 0 to 50 000 V. Canadian Standards Association, 1983. 125  Bibliography [13] C.Malone. Distribution system voltage and power quality, hydro ontario. Hydro Ontario, 2006. [14] NXTPhase Corporation. Nxct optical current sensor datasheet. NXTPhase Corporation, 1(1-2):8, 2006. [15] NXTPhase Corporation. Nxvt optical voltage sensor datasheet. NXTPhase Corporation, 1(1-2):2, 2006. [16] Neil Cullen, Jim Thornycroft, and Alan Collinson. Risk analysis of islanding of photovoltaic power systems within low voltage distribution networks. Technical report, International Energy Agency, 2002. [17] L. S. Czarnecki and Z. Staroszczyk. On-line measurement of equivalent parameters of distribution system and its load for harmonic frequencies. Power Delivery, IEEE Transactions on, 45(2):467 – 472, 1996. [18] Kevin J. Dasso. The reliability of bulk electric systems in north america. Technical report, North American Electric Reliability Council, September 2006. [19] Francisco De la Rosa. Harmonics and power systems. Electric power engineering series. CRC Taylor Francis, Boca Raton, FL, 2006. Francisco C. De La Rosa. ill.; 25 cm. [20] A. de Oliveira, J. C. de Oliveira, J. W. Resende, and M. S. Miskulin. Practical approaches for ac system harmonic impedance measurements. Power Delivery, IEEE Transactions on, 6(4):1721 – 1726, 1991. [21] M. C. Di Piazza, P. Zanchetta, M. Sumner, and D. W. P. Thomas. Estimation of load impedance in a power system. In Harmonics and Quality of Power, 2000. Proceedings. Ninth International Conference on, volume 2, pages 520 – 525. IEEE, 2000. [22] V. Diana, M. Sumner, P. Zanchetta, and M. Marinelli. Non-invasive power system impedance monitoring for improved power quality. In Power Electronics, Machines and Drives, 2004. (PEMD 2004). Second International Conference on, volume 1, page 265. IEEE, 2004. [23] Hermann W. Dommel. EMTP Theory Book, volume 6. Microtran Power System Analysis Corp., Vancouver, 2 edition, 1996. [24] M. K. Donnelly, J. E. Dagle, D. J. Trudnowski, and G. J. Rogers. Impacts of the distributed utility on transmission system stability. Power Systems, IEEE Transactions on, 11(2):741–746, 1996. [25] R. C. Dugan. Distributed resources and reliability of distribution systems. In Power Engineering Society Summer Meeting, 2002 IEEE, volume 1, page 106. IEEE, 2002. [26] R. C. Dugan and T. E. McDermott. Operating conflicts for distributed generation on distribution systems. In Rural Electric Power Conference, 2001. IEEE, 2001.  126  Bibliography [27] R. C. Dugan and T. E. McDermott. Distributed generation. Industry Applications Magazine, IEEE, 8(2):1–19, 2002. [28] Ismail Mohamed Elmarkabi. Control and Protection of Distribution Networks with Distributed Generators. dissertation, North Carolina State University, 2004. [29] Walter A. Elmore. Protective relaying theory and applications. Marcel Dekker, New York, 2nd edition, 2004. [30] D. Fernandes, W. Neves, and J. Vasconcelos. A coupling capacitor voltage transformer representation for electromagnetic transient studies. In International Conference on Power Systems Transients, pages 1–6, New Orleans, 2003. [31] Pai Fu-Sheng and Huang Shyh-Jier. A detection algorithm for islanding-prevention of dispersed consumer-owned storage and generating units. Energy Conversion, IEEE Transactions on, 16(4):346–351, 2001. [32] T. Funabashi, K. Koyanagi, and R. Yokoyama. A review of islanding detection methods for distributed resources. Power Tech Conference Proceedings, 2003 IEEE Bologna, 2:1– 6, June 2003. [33] Steve Gehl. Electricity technology roadmap: 2003 summary and synthesis. Technical report, Electric Power Research Institute, November 2003. [34] Martin Geidl. Protection of power systems with distributed generation: State of the art. Report, Power Systems Laboratory of the Swiss Federal Institute of Technology, July 20 2005. [35] F. Ghassemi, P. F. Gale, B. Clegg, T. Cumming, and C. Coutts. Method to measure cvt transfer function. Power Delivery, IEEE Transactions on, 17(4):915 – 920, 2002. [36] A. A. Girgis and R. B. McManis. Frequency domain techniques for modeling distribution or transmission networks using capacitor switching induced transients. Power Delivery, IEEE Transactions on, 4(3):1882 – 1890, 1989. [37] John J. Grainger, William D. Stevenson, and William D. Stevenson. Power system analysis. McGraw-Hill series in electrical and computer engineering. Power and energy. McGraw-Hill, New York, 1994. [38] BC Hydro Systems Group. Distribution power generator islanding guidelines. BC Hydro, 2006. [39] M. B. Harris, A. W. Kelley, J. P. Rhode, and M. E. Baran. Instrumentation for measurement of line impedance. In Applied Power Electronics Conference and Exposition, 1994. APEC ’94., page 887. IEEE, 1994. [40] W. G. Hartmann. How to nuisance trip distributed generation. In Rural Electric Power Conference, 2003, pages C5–1. IEEE, 2003. 127  Bibliography [41] Jatinder Singh Hayer. Signal Detection technique for Islanding Detection of Distributed Generators Using Power Line Signaling. PhD thesis, University of Alberta, 2004. [42] R. Hernandez-Gonzalez, G. Iravani.  Current injection for active islanding detection of  electronically-interfaced distributed resources.  Power Delivery, IEEE Transactions on,  21(3):1698–1705, 2006. [43] Clark Glenn Hochgraf. Anti-islanding detection scheme for distributed power generation. Visteon Global Technologies, Inc., August 5 2003. [44] P. D. Hopewell, N. Jenkins, and A. D. Cross. Loss-of-mains detection for small generators. Electric Power Applications, IEE Proceedings-, 143(3):225 – 230, 1996. [45] J. Huang and K. A. Corzine. Ac impedance measurement by line-to-line injected current. In Industry Applications Conference, 2006. 41st IAS Annual Meeting, volume 1, page 300. IEEE, 2006. [46] R. M. Hudson, T. Thorne, F. Mekanik, M. R. Behnke, S. Gonzalez, and J. Ginn. Implementation and testing of anti-islanding algorithms for ieee 929-2000 compliance of single phase photovoltaic inverters. In Photovoltaic Specialists Conference, 2002. Conference Record of the Twenty-Ninth IEEE, page 1414. IEEE, 2002. [47] BC Hydro. 35 kv and below interconnection requirements for power generators. Technical report, BC Hydro, February 2007. [48] IEEE-SA-Standards-Board. Ieee recommended practice for industrial and commercial power systems analysis. IEEE Std 399-1990, pages 1–495, 1990. [49] IEEE-SA-Standards-Board. Ieee standard electrical power system device function numbers. IEEE Std 519-1992, 1991. [50] IEEE-SA-Standards-Board. Ieee recommended practices and requirements for harmonic control in electrical power systems. IEEE Std 519-1992, pages 1–101, 1993. [51] IEEE-SA-Standards-Board. Ieee recommended practice for utility interface of photovoltaic (pv) systems. IEEE Std 929-2000, pages 1–32, 2000. [52] IEEE-SA-Standards-Board. Ieee recommended practice for protection and coordination of industrial and commercial power systems. ANSI/IEEE Std 242-2001, pages 1–710, 2001. [53] IEEE-SA-Standards-Board. Ieee guide for protective relaying of utility-consumer interconnections. IEEE Std C37.95-2002 (Revision of IEEE Std C37.95-1989), pages 1–49, 2003. [54] IEEE-SA-Standards-Board. Ieee standard for interconnecting distributed resources with electric power systems. IEEE Std 1547 2003, pages 1–27, 2003.  128  Bibliography [55] IEEE-SA-Standards-Board. Ieee standard requirements for instrument transformers. IEEE Std C57.13-1993(R2003) (Revision of IEEE Std C57.13-1978, pages 1–73, 2003. [56] IEEE-SA-Standards-Board. Ieee standard conformance test procedures for equipment interconnecting distributed resources with electric power systems. IEEE Std 1547.1-2005, pages 1–63, 2005. [57] IEEE-SA-Standards-Board. Ieee standard for high-accuracy instrument transformers. IEEE Std C57.13.6-2005, pages 1–6, 2005. [58] Edison Electric Institute. Eei survey of transmission investment. Report, May 2005. [59] Electrical Power Research Institute. 1999 summary and synthesis. Technical Report CI-112677V1, Electrical Power Research Institute, July 1999. [60] H. Karimi ; A. Yazdani; R. Iravani. Negative-sequence current injection for fast islanding detection of a distributed resource unit. Power Electronics, IEEE Transactions on, 23(1):298–307, Jan. 2008. [61] N. A. F. Jaeger and F. Rahmatian. Integrated optics pockels cell high-voltage sensor. Power Delivery, IEEE Transactions on, 10(1):127–134, 1995. [62] Santos JC, Taplamacioglu MC, and Hidaka K. Optical high voltage measurement using pockels microsingle crystal. Review of Scientific Instruments, 70(8):3271–3276, 1999. [63] M. A. Kashem and Gerard Ledwich. Multiple distributed generators for distribution feeder voltage support. IEEE Transactions On Energy Conversion, 20:676–684, 2005. [64] F. Katiraei, C. Abbey, and R. Bahry. Analysis of voltage regulation problem for a 25 kv distribution network with distributed generation. In Power Engineering Society General Meeting, 2006. IEEE, page 8, 2006. [65] F. Katiraei, M. R. Iravani, and P. W. Lehn. Micro-grid autonomous operation during and subsequent to islanding process. Power Delivery, IEEE Transactions on, 20(1):248 – 257, 2005. [66] S. Knoke, J. Majeski, K. Smit, and K. Tarp. Installation, operation, and maintenance costs for distributed generation technologies. Technical Report 1007675, Electric Power Research Institute, February 2003. [67] Paul C. Krause, Oleg Wasynczuk, Scott D. Sudhoff, and Paul C. Krause. Analysis of electric machinery and drive systems. IEEE Press power engineering series. IEEE Press; Wiley-Interscience, Piscataway, NJ New York, 2nd edition, 2002. Paul C. Krause, Oleg Wasynczuk, Scott D. Sudhoff.ill.; 25 cm. Rev. ed. of: Analysis of electric machinery. c1995. [68] Prabha Kundur. Power System Stability and Control. Power System Planning and Operations Program. McGraw-Hill, Inc., Palo Alto, 1 edition, 1994. 129  Bibliography [69] B.P Lathi. Modern Digital and Analog Communication Systems, chapter 2.7, pages 40–45. Oxford University Press, 189 Madison Avenue, New York, New York, 10016, 1998. [70] David Lineweber and Shawn McNulty. The cost of power disturbances to industrial and digital economy companies. Technical report, Primen, June 2001. [71] L. A. C. Lopes and Sun Huili. Performance assessment of active frequency drifting islanding detection methods. Energy Conversion, IEEE Transaction on, 21(1):171–180, 2006. [72] Jos´e R. Mart´ı. Provisional patent application: Method and apparatus for finding th´evenin equivalents in a power system and detecting voltage stability conditions. 61/006,056(056), 2007. [73] Jos´e R. Mart´ı, L. R. Linares, and H. W. Dommel. Current transformers and coupling-capacitor voltage transformers in real-time simulations. Power Delivery, IEEE Transactions on, 12(1):164– 168, 1997. [74] V. Menon and M. H. Nehrir. A hybrid islanding detection technique using voltage unbalance and frequency set point. Power Systems, IEEE Transactions on, 22(1):442–448, 2007. [75] John Micklethwait. Atomic renaissance. The Economist, 384(8545):102–104, September 2007. [76] G. Moreau, H. H. Le, G. Croteau, G. Beaulieu, and E. Portales. Measurement system for harmonic impedance of the network and validation steps. In Quality and Security of Electric Power Delivery Systems, pages 69 – 73. CIGRE/PES 2003. CIGRE/IEEE PES International Symposium, 2003. [77] Chuck Mozina. Distributed generation interconnection protection course. Beckwith Electric, Course Handouts, 2007. [78] M. Nagpal and C. Henville. Protection accessories course notes. Technical report, BC Hydro, January 2007. [79] M. Nagpal, F. Plumptre, R. Fulton, and T. G. Martinich. Dispersed generation interconnectionutility perspective. Industry Applications, IEEE Transactions on, 42(3):864–872, 2006. [80] M. Nagpal, W. Xu, and J. Sawada. Harmonic impedance measurement using three-phase transients. Power Delivery, IEEE Transactions on, 13(1):272 – 277, 1998. [81] Zhihong Ye; Pengwei Du; John Keith Nelson; Wright Miller Nicholas. Anti-islanding protection systems for synchronous machine based distributed generators. General Electric Company, September 18 2007. [82] H. Nikkhajoei and R. H. Lasseter. Microgrid protection. In Power Engineering Society General Meeting, 2007. IEEE, page 1, 2007.  130  Bibliography [83] P. O’Kane and B. Fox. Loss of mains detection for embedded generation by system impedance monitoring. In Developments in Power System Protection, Sixth International Conference on, page 95. IEEE, 1997. [84] B. Palethorpe, M. Sumner, and D. W. P. Thomas. Power system impedance measurement using a power electronic converter. In Harmonics and Quality of Power, 2000. Proceedings. Ninth International Conference on, volume 1, page 208. IEEE, 2000. [85] B. Palethorpe, M. Sumner, and D. W. P. Thomas. System impedance measurement for use with active filter control. In Power Electronics and Variable Speed Drives, 2000. Eighth International Conference on, page 24. IEEE, 2000. [86] K. O. H. Pedersen, A. H. Nielsen, and N. K. Poulsen. Short-circuit impedance measurement. Generation, Transmission and Distribution, IEE Proceedings-, 150(2):169 – 174, 2003. [87] Daniel Persson. Islanding detection in power electronic converter based distributed generation. Master thesis, Industrial Electrical Engineering and Automation, February 2007. [88] Manisa Pipattanasomporn, Michael Willingham, and Saifur Rahma. Implications of on-site distributed generation for commercial/industrial facilities. IEEE Transactions of Power Systems, 20(4):206–212, 2005. [89] Arne Faaborg Povlsen. Impacts of power penetration from photovoltaic power systems in distribution networks. Technical Report Report IEA PVPS T5-10: 2002, International Energy Agency, February 2002. [90] Satish J. Ranade, Deepak R. Sagi, Ravindra Mulpuri, Raghu Surabhi, and Joydeep Mitra. Likelihood of islanding in distribution feeders with photovoltaic generation. In Power Engineering Society General Meeting, 2007. IEEE, page 1, 2007. [91] J. P. Rhode, A. W. Kelley, and M. E. Baran. Complete characterization of utilization-voltage power system impedance using wideband measurement. Industry Applications, IEEE Transactions on, 33(6):1472 – 1479, 1997. [92] Jeremy Rifkin. The Hydrogen Economy: The Creation of the Worldwide Energy Web and the Redistribution of Power on Earth, volume 1. Penguin Putnam Inc., New York, NY, 1 edition, 2002. [93] C. M. Riley, B. K. Lin, T. G. Habetler, and G. B. Kliman. Stator current harmonics and their causal vibrations: a preliminary investigation of sensorless vibration monitoring applications. Industry Applications, IEEE Transactions on, 35(1):94–99, 1999. [94] A. Robert, T. Deflandre, E. Gunther, R. Bergeron, A. Emanuel, A. Ferrante, G. S. Finlay, R. Gretsch, A. Guarini, J. L. Gutierrez Iglesias, D. Hartmann, M. Lahtinen, R. Marshall, K. Oonishi, C. Pincella, S. Poulsen, P. Ribeiro, M. Samotyj, K. Sand, J. Smid, P. Wright, and Y. S. 131  Bibliography Zhelesko. Guide for assessing the network harmonic impedance. In Electricity Distribution. Part 1: Contributions. CIRED. 14th International Conference and Exhibition on, volume 1. IEEE, 1997. [95] M. E. Ropp, M. Begovic, and A. Rohatgi. Prevention of islanding in grid-connected photovoltaic systems. Progress in Photovoltaics: Research and Applications, 7(1):39 – 59, 1999. [96] Michael Eugene Ropp. Design issues for grid-connected photovoltaic systems. Dissertation, Georgia Institute of Technology, 1998. [97] S. K. Salman, D. J. King, and G. Weller. New loss of mains detection algorithm for embedded generation using rate of change of voltage and changes in power factors. In Developments in Power System Protection, 2001, Seventh International Conference on (IEE), page 82, 2001. [98] O. Samuelsson and N. Strath. Islanding detection and connection requirements. In Power Engineering Society General Meeting, 2007. IEEE, page 1, 2007. [99] Tony Seegers. A report to the line protection subcommittee of the power system relay committee of the ieee power engineering society. Technical report, IEEE, August 2004. [100] Ken Sonnenberg. Current and voltage transformer basics. GE Multilin - Instrument Transformers, Inc., March 2007. [101] N. L. Soultanis, S. A. Papathanasiou, and N. D. Hatziargyriou. A stability algorithm for the dynamic analysis of inverter dominated unbalanced lv microgrids. Power Systems, IEEE Transactions on, 22(1):294–304, 2007. [102] German Standard.  Din vde 0126-1-1.  Technical report, Selbstttige Schaltstelle zwischen  einer netzparallelen Eigenerzeugungsanlage und dem ffentlichen Niederspannungsnetz,, February 2006. [103] Z. Staroszczyk. Problems in real-time wide band identification of power systems. In Instrumentation and Measurement Technology Conference, 1998. IMTC/98. Conference Proceedings., volume 2, page 779. IEEE, 1998. [104] John Stevens, Russell Bonn, Jerry Ginn, and Sigifredo Gonzalez. Development and testing of an approach to anti-islanding in utility-interconnected photovoltaic systems. Technical report, Sandia National Laboratories, August 2000. [105] M. Sumner, B. Palethorpe, D. Thomas, P. Zanchetta, and M. C. Di Piazza. Estimation of power supply harmonic impedance using a controlled voltage disturbance. In Power Electronics Specialists Conference, 2001. PESC. 2001 IEEE 32nd Annual, volume 2, page 522. IEEE, 2001. [106] M. Sumner, B. Palethorpe, D. W. P. Thomas, P. Zanchetta, and M. C. Di Piazza. A technique for power supply harmonic impedance estimation using a controlled voltage disturbance. Power Electronics, IEEE Transactions on, 17(2):207 – 215, 2002. 132  Bibliography [107] M. Sumner, B. Palethorpe, P. Zanchetta, and D. W. P. Thomas. Experimental evaluation of active filter control incorporating on-line impedance measurement. In Harmonics and Quality of Power, 2002. 10th International Conference on, volume 2, page 501. IEEE, 2002. [108] Jang Sung-Il and Kim Kwang-Ho. An islanding detection method for distributed generations using voltage unbalance and total harmonic distortion of current. Power Delivery, IEEE Transactions on, 19(2):745 – 752, 2004. [109] A. V. Timbus, R. Teodorescu, F. Blaabjerg, and U. Borup. Online grid measurement and ens detection for pv inverter running on highly inductive grid. Power Electronics Letters, IEEE, 2(3):77 – 82, 2004. [110] A. V. Timbus, R. Teodorescu, F. Blaabjerg, and U. Borup. Ens detection algorithm and its implementation for pv inverters. Electric Power Applications, IEE Proceedings-, 153(2):206 – 212, 2006. [111] A. V. Timbus, R. Teodorescu, F. Blaabjerg, and U. Borup. Online grid impedance measurement suitable for multiple pv inverters running in parallel. In Applied Power Electronics Conference and Exposition, 2006. APEC apos;06. Twenty-First Annual, page 5. IEEE, 2006. [112] K. Tomiyama, S. Ihara, R. J. Piwko, E. R. Pratico, and L. Crane. Impedance monitor performance tests. In Power Engineering Society Winter Meeting, 2002., volume 1, page 538. IEEE, 2002. [113] O. Tsukamoto, T. Okayasu, and K. Yamagishi. Study on islanding of dispersed photovoltaic power systems connected to a utility power grid. Solar Energy, 70(6):505–511, 2001. [114] Mossadiq Umedaly. A vision for growing a world-class power technology cluster in a smart, sustainable british columbia. British Columbia Reports and Publications, 2005. [115] G. J. Vachtsevanos and H. Kang. Simulation studies of islanded behavior of grid-connected photovoltaic systems. Energy Conversion, IEEE Transactions on, 4(2):177–183, 1989. [116] Bas Verhoeven. Utility aspects of grid connected photovoltaic power systems. Technical Report IEA PVPS T5-01:1998, International Energy Agency, December 1998. [117] Bas Verhoeven. Probability of islanding in utility networks due to grid connected photovoltaic power systems. Technical report, International Energy Agency, september 2002. [118] P. L. Villenueve. Concerns generated by islanding [electric power generation]. Power and Energy Magazine, IEEE, 2(3):49– 53, 2004. [119] K. Vu, M. M. Begovic, D. Novosel, and M. M. Saha. Use of local measurements to estimate voltage-stability margin. Power Systems, IEEE Transactions on, 14(3):1029–1035, 1999. [120] C. F. Wagner and Robert David Evans. Symmetrical components as applied to the analysis of unbalanced electrical circuits. McGraw-Hill, New York, London,, 1933. 133  Bibliography [121] W. Wang, J. Kliber, G. Zhang, W. Xu, B. Howell, and T. Palladino. A power line signaling based scheme for anti islanding protection of distributed generators part ii field test results. Power Delivery, IEEE Transactions on, 22(3):1767–1772, 2007. [122] Xiaoyu Wang, Walmir Freitas, Wilsun Xu, and Venkata Dinavahi. Impact of dg interface controls on the sandia frequency shift antiislanding method. IEEE Transactions on Energy Conversion, 22(3):792–794, 2007. [123] Xiaoyu Wang, Walmir Freitas, Wilsun Xu, and Venkata Dinavahi. Impact of interface controls on the steady-state stability of inverter-based distributed generators. In Power Engineering Society General Meeting, 2007. IEEE, page 1, 2007. [124] Yuguang Wang. Utility Interactive Power Conditioning Systems Islanding Problem Solutions. PhD thesis, Arizona State University, 1990. [125] P. Welch. The use of fast fourier transform for the estimation of power spectra: A method based on time averaging over short, modified periodograms. Audio and Electroacoustics, IEEE Transactions on, 15(2):70 – 73, 1967. [126] Robert Wills. Anti-islanding method and apparatus for distributed power generation. Plug Power Inc., 2001. [127] Xu Wilsun, E. E. Ahmed, Zhang Xiqin, and Liu Xian. Measurement of network harmonic impedances: practical implementation issues and their solutions. Power Delivery, IEEE Transactions on, 17(1):210–216, 2002. [128] Xu Wilsun, K Mauch, and S. Martel. An assessment of distributed generation islanding detection methods and issues for canada. Natural Resources Canada, 2004. [129] J. Wilsun Xu; Guibin Zhang; Chun Li; Wencong Wang; Guangzhu Wang; Kliber. A power line signaling based technique for anti islanding protection of distributed generators part i: Scheme and analysis. Power Delivery, IEEE Transactions on, 22(3):1758–1772, 2007. [130] C. Xie, S. B. Tennakoon, R. Langella, D. Gallo, A. Testa, and A. Wixon. Harmonic impedance measurement of 25 kv single phase ac supply systems. In Harmonics and Quality of Power, 2000. Proceedings. Ninth International Conference on, volume 1, page 214. IEEE, 2000. [131] Xiao Yao, J. C. Maun, H. B. Mahmoud, T. Detroz, and Do Stephane. Harmonic impedance measurement using voltage and current increments from disturbing loads. In Harmonics and Quality of Power, 2000. Proceedings. Ninth International Conference on, volume 1, page 220. IEEE, 2000. [132] Jung Youngseok, Sol Junghun, Yu Gwonjong, and Choj Jaeho. Modelling and analysis of active islanding detection methods for photovoltaic power conditioning systems. In Electrical and Computer Engineering, 2004. Canadian Conference on, volume 2, page 979. IEEE, 2004.  134  Appendix A  Device Number Reference This information comes from IEEE Standard Electrical Power System Device Function Numbers [49]. 1 - Master Element 2 - Time Delay Starting or Closing Relay 3 - Checking or Interlocking Relay 4 - Master Contactor 5 - Stopping Device 6 - Starting Circuit Breaker 7 - Anode Circuit Breaker 8 - Control Power Disconnecting Device 9 - Reversing Device 10 - Unit Sequence Switch 11 - Reserved for future application 12 - Overspeed Device 13 - Synchronous-speed Device 14 - Underspeed Device 15 - Speed - or Frequency-Matching Device 16 - Reserved for future application 17 - Shunting or Discharge Switch 18 - Accelerating or Decelerating Device 19 - Starting to Running Transition Contactor 20 - Elect. operated valve (solenoid valve) 21 - Distance Relay 22 - Equalizer Circuit Breaker 23 - Temperature Control Device 24 - Over-Excitation Relay 25 - Synchronizing or Synchronism-Check Device 26 - Apparatus Thermal Device 27 - Undervoltage Relay 28 - Reserved for future application 29 - Isolating Contactor 30 - Annunciator Relay 31 - Separate Excitation Device 32 - Directional Power Relay 33 - Position Switch 34 - Motor-Operated Sequence Switch 35 - Brush-Operating or Slip-Ring Short-Circuiting Device  135  Appendix A. Device Number Reference 36 - Polarity or Polarizing Voltage Devices 37 - Undercurrent or Underpower Relay 38 - Bearing Protective Device 39 - Mechanical Conduction Monitor 40 - Field Relay 41 - Field Circuit Breaker 42 - Running Circuit Breaker 43 - Manual Transfer or Selector Device 44 - Unit Sequence Starting Relay 45 - Reserved for future application 46 - Reverse-phase or Phase-Balance Current Relay 47 - Phase-Sequence Voltage Relay 48 - Incomplete-Sequence Relay 49 - Machine or Transformer Thermal Relay 50 - Instantaneous Overcurrent 51 - AC Time Overcurrent Relay 52 - AC Circuit Breaker 53 - Exciter or DC Generator Relay 54 - High-Speed DC Circuit Breaker 55 - Power Factor Relay 56 - Field Application Relay 57 - Short-Circuiting or Grounding Device 58 - Power Rectifier Misfire Relay 59 - Overvoltage Relay 60 - Voltage or Current Balance Relay 61 - Machine Split Phase Current Balance 62 - Time-Delay Stopping or Opening Relay 63 - Pressure Switch 64 - Ground Detector Relay 65 - Governor 66 - Starts per Hour 67 - AC Directional Overcurrent Relay 68 - Blocking Relay 69 - Permissive Control Device 70 - Electrically Operated Rheostat 71 - Level Switch 72 - DC Circuit Breaker 73 - Load-Resistor Contactor 74 - Alarm Relay 75 - Position Changing Mechanism 76 - DC Overcurrent Relay 77 - Pulse Transmitter 78 - Phase-Angle Measuring or Out-of-Step Protective Relay 79 - AC-Reclosing Relay 80 - Reserved for future application  136  Appendix A. Device Number Reference 81 - Frequency Relay 82 - DC-Reclosing Relay 83 - Automatic Selective Control or Transfer Relay 84 - Operating Mechanism 85 - Carrier or Pilot-Wire Receiver Relay 86 - Lockout Relay 87 - Differential Protective Relay 88 - Auxiliary Motor or Motor Generator 89 - Line Switch 90 - Regulating Device 91 - Voltage Directional Relay 92 - Voltage and Power Directional Relay 93 - Field Changing Contactor 94 - Tripping or Trip-Free Relay 95 - Reluctance Torque Synchrocheck 96 - Autoloading Relay 97 - For specific applications where other numbers are not suitable 98 - For specific applications where other numbers are not suitable 99 - For specific applications where other numbers are not suitable Note: A suffix letter may be used with the device number; for example, suffix N is used if the device is connected to a Neutral wire (example: 59N in Siemens Relay is used for protection against Neutral Displacement); and suffixes X,Y,Z are used for auxiliary devices. Similarly, the ”G” suffix denotes a ”ground”, hence a ”51G” being an time over-current ground relay  137  Appendix B  Symmetrical Components - The Basics Symmetrical component vector transformation was introduced by Fortescue in 1918, [120] [37] to decouple three phase line interdependencies into three linearly independent lines. The transform requires a matrix operator and the vectorial form of the ABC voltages or currents to convert the unbalanced system into three balanced independent systems called Positive, Negative and Zero sequence as illustrated in Figure B.1.  Figure B.1: Symmetrical Component Conversion  Symmetrical components are calculated through the transformational operator ’A’ as seen in equations B.1 and B.2. The result is a sum of the vectors from VABC and IABC in reference to the ′ A′ transformational matrix frame (Equation B.2). The transformation of phasors VABC and IABC is accomplished by multiplying the phasors by ’A’. The result are what is called Zero, Positive and Negative sequences, or denoted as subscripts of: 0, 1, 2 for Zero, Positive and Negative sequences respectively. The sequence conversion is as follows in Equation B.3. a = 1 120o   1  1  1    A =  1 a2 a  1 a a2 V012 = A−1 ·VABC  (B.1)       (B.2)  (B.3)  For example, a balanced system seen in Figure B.2 has an impedance matrix solution for [V ] as seen in Equation B.4 where [V ], [I], and [Z] all represent matrix compositions of voltage, current and 138  Appendix B. Symmetrical Components - The Basics impedance of the system. The resulting impedance matrix, [ZABC ], is composed of a combination of ZA , ZB , ZC , and ZG (ground) impedances from the original three phase system as seen in Equation B.5. E  E,I  V,I [Z Sys ]  [Z Load ]  Figure B.2: Radially Fed System  [VLoad ] = [ZLoad ] · [ILoad ]     [ZABC ] =    (B.4)   ZA + ZG  ZG  ZG  ZG  ZB + ZG  ZG  ZG  ZG  ZC + ZG       (B.5)  As the impedance is matrix is balanced, ZA = ZB = ZC , the system then has distinctive eigenvalues. Or in Equations B.6 to B.9, Z012 is solved and the result is seen in Equation B.9. As only one of the three systems has a voltage source, calculations of the voltage and current is reduced to only one simple circuit where the current flows from the source to the load. Also, note that Z1 is equal to Z2 (in most practical systems, this assumption is acceptable but not always true)  A ·V012 = ZABC · A · I012  (B.6)  A−1 · A ·V012 = A−1 · ZABC · A · I012  [Z012 ] = A−1 · ZABC · A    ZA + 3ZG 0 0 Z0   Z1  =  0 ZA 0  Z2 0 0 ZA  139  (B.7)  (B.8)      (B.9)  Appendix C  AEMC 3945 Three Phase Power Quality Meter  Figure C.1: AEMC 3945 Power Quality Meter  140  Appendix D  Simulation Software Tools Used Various analysis tools have been used thoughout out this thesis. This appendix contains a very brief description of each of these tools so the reader may refer to use them to further their research.  D.1 Aspen ASPEN OneLiner is a PC-based short circuit and relay coordination program for relay engineers. The program OneLiner is a type of productivity tool. The engineer can change the relay settings and network configuration and see the results of the change immediately such as: breaker curves, one-line diagrams, and short circuits.  D.2 Matlab Matlab is an interactive system and programming language for general scientific and technical computation and visualization. The basic Matlab data element is a matrix. Matlab commands are expressed in a form very similar to that used in mathematics and engineering. There are two basic versions of the software, the professional version and the student edition. The student edition is distributed by Prentice-Hall, the professional version is distributed by The MathWorks, Inc.  D.3 Microtran Power Systems Simulator MicroTran is the UBC version of the EMTP, and is an advanced and reliable software tool for the simulation of electromagnetic transients in power systems. Microtran was founded in 1987 by Hermann W. Dommel, Jose R. Marti and Luis Marti. Professor Dommel is the father of the EMTP which he originally developed in the late 1960s.  D.4 Power World PowerWorld is a power systems analysis software, for real-time steady-state power system visualizations and analysis. The interface is a single-line diagram. PowerWorld is most effective for complex power flow solutions in detailed networks allowing matching of phase and power flows through the addition of different loads and busses at different voltages.  141  Appendix D. Simulation Software Tools Used  D.5 Psim Psim is a EMTP type simulation tool and design software for power electronics, power systems, motor drives, and switching for dynamic systems. It has a fast simulation time and an easy-to-use graphic interface. PSIM tool has an advantage of reducing common switching numeric oscillations.  D.6 Simulink Power Systems Tool Box Simulink is an interactive system for the nonlinear simulation of dynamic systems in a state space modeling atmosphere. It comes with basic signal processing and control blocks, but with the Power Systems Tool Box, it becomes a module for the simulation of power systems. Its primary interface is a graphic block diagramming tool and a customizable set of block libraries. It can handle linear, nonlinear, continuous-time and discrete-time power systems. Simulink is closely integrated with Matlab and the Simulink Power Systems Toolbox where it forms an effective electrical systems design and analysis tool. Though this simulation tool is much slower and difficult to stabilize than EMTP type (non-state space) simulators, the graphical interface and direct link to Matlab’s superior analysis capability makes it a worthwhile investment.  D.7 SKM Power Tools SKM Power Tools is a set of modules and integrated tools that allow the user to simulate three-phase power systems for design and analysis in steady-state. Some tools include load flow, voltage drop calculations, motor starting, fault analysis with feeder, raceway and transformer sizing.  142  Appendix E  Simulink Model: 600 V Fed Bus  Figure E.1: Simulink Model From Practical Example 2: 600 V Fed Bus  143  Appendix F  Commonly Used Signal Processing Techniques Impulse response often takes significant signal processing to fully extract the wave forms and to calculate the transfer function. A quick introduction on some of the most common signal processing techniques used has been inserted here to facilitate the subsequent description of the techniques. The most basic signal manipulation process can be seen in Figure F.1, which shows the paths for calculating the transfer function of a system and breaking a signal into its individual frequencies. The frequency domain transfer function, H(e jω ) can be found from the input x[n] and the output y[n] through convolution of the Fourier Transform. The most simple route is to take the Fourier Transform of x[n] and y[n] then divide the two as seen in equation F.2. The discrete Fourier Transform is used to transform a time domain sampled signal to its frequency domain representation. The transform is accomplished by calculating the sum of all the products of a function at point “n” with the cosine wave and sine wave at point “n” with respect to a specific frequency. All frequencies are set to the normalized interval between 0 to 2π where 2π is equal to the sampling frequency. The results for each frequency are real and imaginary values (Figure F.1). The squared sum of the two values yields the magnitude of the specific frequencies. X (e jω ) =  ∞  ∑  x[n]e jωn  (F.1)  n=−∞  convolution  −−−−−−→  x[n]   F   h(n)  multiplication  y[n]   F   X [e− jω ] −−−−−−−→ Y [e− jω ] H(e− jω )  Figure F.1: Convolution to Fourier Relationship H(e jω ) =  Y (e jω ) F (y[n]) = X (e jω ) F (x[n])  (F.2)  ∞  y[n] =  ∑  k=−∞  x[k]h[n − k] = x[n] ∗ h[n] 144  (F.3)  Appendix F. Commonly Used Signal Processing Techniques Equation F.1 is for continuous infinite and requires an infinite time window. A sampled signal in a finite window contains undesirable high frequency artifacts at the boundaries of the window. This phenomenon is called the Gibbs Effect. To reduce the Gibbs Effect, a windowing function such as Hamming and Hanning type windows can be multiplied onto the signal. These windows slowly taper a signal’s amplitude at each end to zero, thereby diminishing the discontinuity and reducing the Gibbs Effect. While all the windows listed effectively reduce the Gibbs Effect, one of the most commonly used is the Hamming window due to its consistently good results. Applying the window to a discretised finite length sample results in Equation F.4, L  X (e jω ) =  ∑ x[n]e jωn · w[n]  (F.4)  n=0  Power system frequency windowing is slightly different because the fundamental and its harmonics are often significantly larger than all other non-harmonic signals being measured. The Gibbs effect can be virually eliminated by carefully using a window the size of t =  n f  where n is an integer value between  1 and inf, t is time and f is the frequency of the fundamental. This ensures that the sampling time window exactly fits an integer value of the fundamental. Therefore to maximize signal integrity using Fourier Transform, windowing combined with specifically chosen time windows is critical. In some cases a square window may be suitable. Another common signal processing technique for improving signal-to-noise ratio and often used in impedance measurement is cross-correlation [69]. The advantage of cross-correlation is that it matches signal shapes regardless of the signal amplitude. For example, if two decaying sine waves of the same frequency and different starting amplitudes are cross-correlated together, the strongest correlation occurs when the start of the first wave matches the start of the second wave. The formula for cross-correlation is given in Equation F.5, where g and z are signals of the same length. The resultant output length is 2*M-1 where M is the length of the original two functions. ψgz (τ) ≡  ∞ −∞  g(t)z(t + τ)∂t  (F.5)  Finally, an improved approach to using traditional Fast Fourier Transform (FFT) of voltage and current is to use the MATLAB embedded function “transfer function estimation“ which uses Welch’s averaged periodogram method [125]. Welch’s method auto-correlates and cross-correlates before using an FFT to reduce the signal to noise ratio.  145  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0065504/manifest

Comment

Related Items