Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Benefits of wind power curtailment in a hydro-dominated electric generation system Evans, Joel I. 2009

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_fall_evans_joel.pdf [ 1.86MB ]
Metadata
JSON: 24-1.0063119.json
JSON-LD: 24-1.0063119-ld.json
RDF/XML (Pretty): 24-1.0063119-rdf.xml
RDF/JSON: 24-1.0063119-rdf.json
Turtle: 24-1.0063119-turtle.txt
N-Triples: 24-1.0063119-rdf-ntriples.txt
Original Record: 24-1.0063119-source.json
Full Text
24-1.0063119-fulltext.txt
Citation
24-1.0063119.ris

Full Text

Benefits of wind power curtailment in a hydro-dominated electric generation system  by JOEL I. EVANS B.Sc., The University of Maine, 2004  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF APPLIED SCIENCE in THE FACULTY OF GRADUATE STUDIES (Civil Engineering)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) September 2009 © Joel I. Evans, 2009  Abstract Hydroelectric generation has been well documented as a flexible and fast-responding resource, which can be quite complimentary to variable and uncertain renewable energy resources. However, hydropower is also a variable energy resource that is limited by river inflows and the live storage capacity of its reservoirs. The goal of this thesis is to explore the possible incremental value that curtailment of wind power might contribute, without reducing a wind power producer’s income, in a market context. Therefore, the ultimate use of this research is as a planning study. The modeling in this research was performed at an hourly time resolution and only focused on real (not reactive), available wind power. An hydraulic simulation and electrical energy generation optimization model of a hydropower system, using a linear programming approach, was used. It was a deterministic model, so specific independent variables had to be manually manipulated. These variables were wind power installed capacity, hydrologic regime (wet, average, and dry), import/export transmission capacity, and cost of wind energy. The optimal objective function values of the model at the end of each yearly run were compared to each other. The opportunity for economically beneficial curtailment came from market-based competition for the finite amounts of generation capacity, energy, and reservoir storage in the hydroelectric system. Existing flexibility in the large storage reservoirs was used by the model to shift the wind energy to export periods with the highest market prices. Temporal analysis of wind power curtailment showed that curtailed wind energy was concentrated in the light-load hours (10:00 PM to 6:00 AM) during the spring freshet (snowmelt) period. These were periods of low domestic electrical load, low market energy prices, high energy import levels, and high local river inflows. The total annual amount of curtailed wind energy was between 0.5% and 2%. Based on input parameters, including energy market and generation capacity market prices, intertie capacities, available wind energy, ii  river inflows, and load, an optimal value of wind energy production was determined. Allowing wind power curtailment purely for economic reasons resulted in an incremental increase in this wind energy value between 1% and 4%.  iii  Table of Contents ABSTRACT ....................................................................................................................................................... ii TABLE OF CONTENTS ...................................................................................................................................... iv LIST OF FIGURES ............................................................................................................................................ vii LIST OF ACRONYMS........................................................................................................................................ ix ACKNOWLEDGEMENTS ................................................................................................................................. xii 1  2  3  INTRODUCTION ...................................................................................................................................... 1 1.1  BACKGROUND ...........................................................................................................................................1  1.2  PROBLEM STATEMENT.................................................................................................................................4  1.3  GOALS AND OBJECTIVES...............................................................................................................................5  1.4  THESIS ORGANIZATION ................................................................................................................................7  LITERATURE REVIEW ............................................................................................................................... 8 2.1  OPTIMAL DISPATCH OF WIND AND EXISTING/OTHER GENERATION RESOURCES ........................................................8  2.2  THE MARKET CONTEXT ..............................................................................................................................12  2.3  WIND POWER CURTAILMENT STATE OF THE ART .............................................................................................16  2.4  CONTRACTUAL ARRANGEMENTS ..................................................................................................................20  2.5  CONCLUSIONS .........................................................................................................................................21  MODEL DEVELOPMENT AND MODELING APPROACH ........................................................................... 24 3.1  THE DETERMINISTIC OPTIMIZATION MODEL ...................................................................................................24  3.1.1  Hydraulics .......................................................................................................................................25  3.1.2  Generating stations ........................................................................................................................26  3.1.3  Load-resource balance and objective function ...............................................................................27  3.2  ADDITIONS TO THE EXISTING MODEL ............................................................................................................27  iv  4  3.2.1  Operating reserves..........................................................................................................................28  3.2.2  Wind’s effects to hourly electrical energy prices ............................................................................28  3.2.3  Slack generating capacity valuation ...............................................................................................29  3.2.4  Wind energy as a generation resource ...........................................................................................30  3.2.5  Payment for wind energy ................................................................................................................31  3.3  MODELING APPROACH ..............................................................................................................................31  3.4  SUMMARY ..............................................................................................................................................32  MODEL DATA AND ASSUMPTIONS ....................................................................................................... 34 4.1  FIXED INPUT PARAMETERS .........................................................................................................................34  4.1.1  Load: energy, shapes, and peaks ....................................................................................................34  4.1.2  Fixed-generation energy and shapes ..............................................................................................35  4.1.3  Hydroelectric generating unit outage schedules ............................................................................39  4.1.4  Forebay stage limits and flood control curves ................................................................................39  4.1.5  Energy prices ...................................................................................................................................41  4.1.6  Ancillary services prices ..................................................................................................................43  4.2  MANIPULATED INPUT PARAMETERS .............................................................................................................44  4.2.1  Water years ....................................................................................................................................45  4.2.2  Wind power installed capacity........................................................................................................48  4.2.3  Market depth and intertie capacity ................................................................................................48  4.2.4  Contractual cost per MWh of wind energy .....................................................................................49  4.3  OPTIMIZATION VARIABLES..........................................................................................................................50  4.3.1  Reservoir/forebay stages ................................................................................................................50  4.3.2  Turbine flows ..................................................................................................................................51  4.3.3  Plant spills .......................................................................................................................................51  4.3.4  Energy and capacity market exchanges .........................................................................................52  4.3.5  Integrated wind energy ..................................................................................................................52  v  4.3.6 4.4  6  ASSUMPTIONS: SIMPLIFICATIONS AND ELIMINATED VARIABLES ..........................................................................53  4.4.1  Wind energy data ...........................................................................................................................54  4.4.2  Load and wind reserves ..................................................................................................................57  4.4.3  Load and wind forecasts .................................................................................................................60  4.4.4  Transmission constraints ................................................................................................................60  4.4.5  Slack generating capacity valuation ...............................................................................................61  4.4.6  Available generation resources ......................................................................................................61  4.4.7  Energy price discount percentage during wind events ...................................................................62  4.4.8  Model duration and boundary conditions ......................................................................................62  4.5 5  Allocation of reserve generation capacity ......................................................................................53  SUMMARY ..............................................................................................................................................63  RESULTS AND ANALYSIS ....................................................................................................................... 64 5.1  RESERVOIR/FOREBAY STAGES .....................................................................................................................65  5.2  HYDROELECTRIC SYSTEM EFFICIENCY AND SPILLS .............................................................................................70  5.3  GENERATION CAPACITY RESERVES ................................................................................................................76  5.4  VALUE AND QUANTITY OF CURTAILMENT .......................................................................................................76  5.5  TEMPORAL CHARACTERISTICS OF CURTAILMENT..............................................................................................78  5.6  SUMMARY ..............................................................................................................................................82  DISCUSSION AND CONCLUSIONS .......................................................................................................... 84 6.1  SUMMARY ..............................................................................................................................................84  6.2  ANOTHER APPLICATION .............................................................................................................................87  6.3  FURTHER WORK .......................................................................................................................................88  REFERENCES .................................................................................................................................................. 92  vi  List of Figures FIGURE 1: BRITISH COLUMBIA BULK TRANSMISSION SYSTEM AND GENERATING STATIONS ........................ 25 FIGURE 2: MODELING SCENARIO TREE .......................................................................................................... 32 FIGURE 3: AVERAGE BC HYDRO SYSTEM LOAD .............................................................................................. 35 FIGURE 4: FIXED GENERATION SHAPES ......................................................................................................... 36 FIGURE 5: SMALL HYDRO GENERATION SHAPES ........................................................................................... 37 FIGURE 6: SAMPLE COLUMBIA RIVER PLANTS’ FOREBAY TRAJECTORIES AND CONSTRAINTS ........................ 40 FIGURE 7: MID-COLUMBIA ENERGY MARKET PRICES .................................................................................... 42 FIGURE 8: CAISO ANCILLARY SERVICES PRICES .............................................................................................. 44 FIGURE 9: BC HYDRO INFLOW PATTERNS ...................................................................................................... 46 FIGURE 10: SIMULATED AVAILABLE WIND ENERGY ....................................................................................... 55 FIGURE 11: LOAD AND WIND RESERVE LEVELS .............................................................................................. 58 FIGURE 12: STORAGE RESERVOIR CHANGES – WET WATER YEAR ................................................................. 66 FIGURE 13: STORAGE RESERVOIR CHANGES – AVERAGE WATER YEAR ......................................................... 67 FIGURE 14: STORAGE RESERVOIR CHANGES – DRY WATER YEAR .................................................................. 68 FIGURE 15: SPILL PATTERN CHANGES AT ARD – DRY WATER YEAR ............................................................... 71 FIGURE 16: SPILL PATTERN CHANGES AT ARD – AVERAGE WATER YEAR ....................................................... 72 FIGURE 17: SPILL PATTERN CHANGES AT ARD – WET WATER YEAR ............................................................... 73 FIGURE 18: PLANT EFFICIENCY CHANGES – 2,200 MW INTERTIE CAPACITY ................................................... 74 FIGURE 19: PLANT EFFICIENCY CHANGES – 1,200 MW INTERTIE CAPACITY ................................................... 75 FIGURE 20: VALUE OF CURTAILMENT AND AMOUNT OF ENERGY CURTAILED ............................................... 77  vii  FIGURE 21: DAILY AND SEASONAL PATTERNS OF CURTAILMENT .................................................................. 80 FIGURE 22: WIND ENERGY VALUE VS. WIND ENERGY CONTRACT PRICE ....................................................... 88  viii  List of Acronyms AESO - Alberta Electric System Operator AGC - Automatic Generation Control AIES - Alberta Interconnected Electric System ARD (or KNA) - Arrow Lakes (or Hugh Keenleyside) hydroelectric generating station AWEA - American Wind Energy Association B.C. - British Columbia, Canada BA - Balancing authority BC Hydro - British Columbia Hydro and Power Authority BCTC - British Columbia Transmission Corporation BPA - Bonneville Power Administration CAES - Compressed air energy storage CAISO - California Independent System Operator CFT - Call for Tenders cms - cubic meters per second CO2 - Carbon Dioxide CRT - Columbia River Treaty DNV-GEC - Det Norske Veritas Global Energy Concepts, Inc. EFI - Norwegian Electric Power Research Institute EMPS - EFI’s Multi-area Power Scheduling model EOPS - EFI’s One-area Power Scheduling EPA - Energy Purchase Agreement ERCOT - Electricity Reliability Council of Texas ix  FB - forebay GMS - Gordon M. Shrum hydroelectric generating station GOM - Generalized Optimization Model GW - gigawatt GWEC - Global Wind Energy Council GWh - gigawatt-hour HLH - heavy-load hour HYSIM - Hydrologic Simulation Model IEA - International Energy Association IEEE - Institute of Electrical and Electronics Engineers, Inc. IPP - Independent Power Producer ISO - Independent System Operator LBNL - Lawrence Berkley National Laboratory LLH - light-load hour LTAP - Long-Term Acquisition Plan MCA - Mica hydroelectric generating station Mid-C - Mid-Columbia region in Washington MoEMPR - Ministry of Energy, Mines, and Petroleum Resources MW - megawatt MWh - megawatt-hour NWPP - Northwest Power Pool PCN - Peace Canyon hydroelectric generating station PNW - Pacific Northwest region of the U.S. PTC - Production Tax Credit x  REC - Renewable Energy Credit REV - Revelstoke hydroelectric generating station RPS - Renewable Portfolio Standard SIPREÓLICO - Spanish wind power prediction tool TSO - Transmission System Operator U.S. - United States of America USD - U.S. Dollar USGS - U.S. Geological Survey WECC - Western Electricity Coordinating Council yr - year  xi  Acknowledgements I would like to briefly thank the following people: Tom Siu, Alaa Abdalla, Dan O’Hearn, Doug Smith, Jeff Chun, Magdalena Rucker, Nicholas Miller, and Julija Matevosyan for review and feedback on my assumptions; Yvette Maiangowi for advice and help choosing wind project scenarios; Jian Li for help setting up and running the optimization model; Greg Lawrence for final review of my thesis; BC Hydro for investing in and supporting graduate students at the University of British Columbia; Jimmy Buffett and Radio Margaritaville for facilitating the frame of mind necessary to formulate and assemble my thoughts, and put them in writing, without going insane; and finally Ziad Shawwash for his supervision, mentorship, and tireless enthusiasm for the duration of my research.  xii  1 Introduction Electrical generation systems can be quite extensive and be complicated to operate optimally. Furthermore, within an electrical distribution network generation must always be balanced with demand. The addition of variable electrical generators, wind turbines in this research, can further complicate the operation of existing, conventional generators that can be quickly and accurately manipulated to match generation to demand. Energy resource developers typically need long term power purchase agreements to make their investments profitable, yet short term daily or seasonal market opportunities can still exist which may more than make up for foregone energy production opportunity. One possible way to provide power to consumers in the most economical manner, while still providing a favorable environment for energy development, might be to ensure the purchase of energy from potential resources (i.e. wind developments) while still allowing the integrating utility, independent transmission system operator (ISO/TSO), or balancing authority (BA) autonomy over all generation resources in their area. The utility, ISO/TSO, or BA dispatches and controls all of the electrical generators and perhaps even some of the large electricity consumers. In an arrangement such as this, short term opportunities do not necessarily have to be foregone in order to ensure long term energy goals and requirements.  1.1 Background From 1973 to 2006, worldwide electrical energy production coming from renewable sources including geothermal, solar, wind, combustible renewables and waste increased nearly 12-fold (IEA, 2009). In terms of wind generation capacity, the worldwide capacity of 6.1 GW in 1996 increased to 13.6 GW by 1999 and exploded to 120.8 GW by 2008 (GWEC, 2009). In the United States, wind 1  power capacity increased from 1.7 GW in 1996 to 2.5 GW in 1999 and leapt to 16.8 GW by 2007. By the end of the second quarter in 2009 the total capacity was over 29.4 GW (AWEA, 2009). Although in 2006 this installed wind power capacity only contributed 0.7% of electricity world-wide, and 0.6% in the U.S. (IEA, 2009), the rapidly increasing penetration of wind energy into existing transmission grids is already affecting the ways in which other power generators operate their systems, and how TSOs or BAs operate their transmission networks. BC Hydro is the electrical utility company that serves over 94% of the population in British Columbia, Canada, and owns the hydroelectric plants that are part of this research work. Although there are no active wind farms in B.C., energy purchase agreements (EPAs) were signed in 2006 with independent power producers (IPPs) for 325 MW of wind power capacity due to come online soon, and an additional 19 wind power projects are seeking EPAs in 2009 in response to BC Hydro’s 2008 Clean Power Call. BC Hydro has transmission ties to both the Alberta Interconnected Electric System (AIES) and Bonneville Power Administration (BPA), both of which are experiencing large increases in wind power interconnection. For example, BPA recently released an update in November 2008 on the wind integration issues that it is facing and updated this in a presentation on May 18, 2009 (BPA, 2008). In these documents, they indicate that the 3,000 MW of wind power capacity that they forecasted to be integrated into their balancing area by 2020 will now be generating by late 2010, with over 6,000 total megawatts of installed capacity by 2013. As of May 1, 2009, they had 2,105 MW of wind power capacity in their balancing area, which equates to approximately 20% penetration (installed capacity to peak load). This is already one of the highest penetration levels for a balancing area in North America. The AIES had 498.5 MW of wind power capacity integrated into its electric system (about 5.1% penetration) as of the May 2008, with another 12.2 GW of wind 2  power capacity in the interconnection queue with planned in-service dates in the next 5 years (AESO, 2009). Two contributing factors to the rapid growth of wind power projects in North America are the production tax credit (PTC) and renewable portfolio standards (RPSs) in the U.S., along with similar green energy goals in Canada such as those laid out in the BC Energy Plan of 2007 (MoEMPR, 2007). By the end of 2007, half of the United States (including Washington D.C.) had state-specific RPSs signed into law targeting up to 40% renewable energy with time horizons extending up to 2025. Once fully implemented, these RPSs will cover 46% of electrical load in the U.S. Over 90% of the renewable energy applied to these goals has been wind energy, although solar energy, demand-side management, and energy efficiency initiatives have been gaining support and consideration for classification as renewable energy recently (Wiser, 2008b). Many of these RPS requirements are eligible to be met by the purchase of renewable energy credits (RECs), independent of actual energy generation within the states. REC markets have been established and continue to evolve, and the market prices for RECs in the western U.S. have been steadily increasing from around $3.00 USD/MWh at the end of 2005 to about $12.50 USD/MWh at the end of 2007 (Wiser, 2008a). Along with the planned and recent increases in wind power production both within B.C. and in electricity markets that BC Hydro has access to, comes increased demand for energy storage and electric generation capacity to firm, shift, and shape the variable and uncertain energy coming from wind power producers. Hydroelectric generation has been well documented as a flexible and fast-responding resource, which can be quite complimentary to variable and uncertain renewable energy resources. Therefore, the value of the existing hydroelectric infrastructure and storage reservoirs in British Columbia is becoming increasingly apparent and important in a future that includes an increasing amount of renewable energy resources. 3  1.2 Problem statement Although over 90% of BC Hydro’s current annual energy needs (between 43,000 and 54,000 GWh) are met by its hydroelectric resources, on average it is still about 5,000 GWh/yr short of energy. It can choose to make up this energy deficit by importing energy at lower-cost times because of its reservoir storage capacity and flexibility. Wind power has the potential to offset this energy deficit, but it could also limit the opportunities to import energy from the market when that energy is less expensive than the value of the water in BC Hydro’s storage reservoirs. This situation occurs during times when inexpensive energy is available in the electricity market, but wind power is forecast to be sufficient to meet load. Then, forecasting errors result in not enough wind energy being generated at a time when it is also too late to import the inexpensive energy from the market. The BC Energy Plan, which was unveiled in 2007, is an energy policy document from the British Columbia provincial government. It contains a goal to procure sufficient electrical energy resources, primarily from IPPs, to make up this annual energy deficit during low water years by 2016. And by 2026, B.C. should also be able to provide an extra 3,000 GWh/yr “insurance” energy that represents the annual energy deficit during a critically low water year (MoEMPR, 2007). These goals have led to the ongoing Clean Power Call, which is currently targeting 5,000 GWh/yr of green energy by 2016 (BC Hydro, 2008b). The EPAs for these additional renewable energy resources, with their inherent variability and uncertainty, will place an increased demand on the existing hydroelectric assets that BC Hydro currently operates. During the negotiation of these EPAs, a framework for energy curtailment is typically included for situations where the reliability and security of the electricity transmission system may be negatively impacted. Traditionally, this is the only situation where curtailment of  4  energy delivery is allowed, because discretionary curtailment could affect the financial security of the energy producer whose energy is being curtailed. Research and modeling case studies in Europe and west Texas over the last few years have demonstrated that wind curtailment can help maximize the development of wind energy resources in regions with transmission constraints (Son, 2005; Torre, 2008; Matevosyan, 2006; Korpås, 2006). Transmission upgrades require significant investment, political will, and sufficient time to design, bid, and construct. Often, wind power producers can realize increased profits by building wind farm(s) with generating capacities greater than their available transmission capacity based on the frequency of peak generation in the wind farm(s). These trade-off analyses will be largely ignored in this work because of the lack of wind development that has already taken place in B.C., though they became apparent at a modest level of wind penetration in many situations. Although integrating 100% of wind energy gives maximum income for wind power producers in traditional EPAs, this environment is not necessarily the most profitable for the TSO or BA, other power producers. It also may not even be the most profitable for the wind power producer, depending on contractual terms like penalties for schedule deviations or extreme ramp rates, or choosing to provide reserve capacity instead of some energy. Factors that can influence these situations include market energy prices, EPA-specified prices, REC prices, government incentives, market bidding strategies, power forecast accuracy, generation capacity reserve requirements, reservoir operational flexibility, hydrologic conditions, wind penetration levels, intertie capacities, and others.  1.3 Goals and objectives The particular trade-off that was investigated in this research is that between imported electricity and electricity from wind power projects. In relatively rare though financially significant 5  circumstances, a decision must be made between minimizing the cost of electricity for ratepayers and maximizing the percentage of integrated wind energy. The goal of the research reported by this paper is to provide justification to further investigate curtailment allowances in EPAs under a variety of scenarios. Common large-scale effects to the wind power producers, other power producers, and/or the TSO or BA for the modeled scenarios is presented, including comparisons of their gross incomes. Therefore, the ultimate use of this research is as a planning study to justify accommodating wind power curtailment. And although the model incorporated the generating reserve capacity to operate the electrical system minute-byminute, it made assumptions that must be modified and refined as operational experience with variable generation resources is gained. So the optimal operational regimes in this research do not necessarily accurately represent how the generation system would be operated once these levels of wind power penetration are achieved. The modeling carried out in this study was performed at an hourly time resolution, and focused on the real available wind power. It did not address reactive power, black start, low-voltage ride through, and any other requirements and ancillary services that wind generators could possibly provide, and which are valuable to the TSO, BA, or utility integrating the wind energy. Finally, it is important to note that the physical constraints on the hydroelectric system were inherent in all of the runs (i.e. transmission constraints, water license constraints, ramp rates, etc). These were responsible for the curtailment due to reliability, and these were accounted for and optimized in every model run. The opportunity for economic curtailment came from market-based competition for the finite amounts of generation capacity, energy, and reservoir storage in the hydroelectric system.  6  1.4 Thesis organization This thesis is organized in the following manner: following this introductory section, a concise summary of recent and pertinent work on several aspects of this thesis’s topic is presented in chapter 2. These aspects include optimal operation of electrical generation assets, operation of electrical markets, current wind power control technology, and cooperative agreements between generators of variable electricity and TSOs, BAs, and/or other generators. Next, the simulation and optimization model that was enhanced, and was used in this work, is described in detail in chapter 3. Then the variables, input data, and assumptions that were made during the application of this model are laid out along with the reasoning behind these assumptions in chapter 4. The model was then run and the metrics that were used to evaluate its results are outlined and discussed in chapter 5. Finally, those results are analyzed, summarized, and discussed in chapter 6. Also at the end of this final chapter, some potential applications of the findings are identified and further recommendations on enhancements to this work are outlined.  7  2 Literature review The literature that was reviewed as a part of this research spans several topics related to wind power and power system operation. Many papers have focused simply on integrating wind power into electric generation systems, and a variety of technical factors constraining this process. The first set of these papers reviewed herein used cost-benefit tradeoff analysis, but stopped short of market modeling and income maximization. This is where the next set of research papers takes over. These papers modeled and analyzed generation systems that were physically connected to other balancing authorities and utilities’ generation assets, and who were involved in electricity markets that influenced their optimal operational decisions. Further literature was reviewed to determine the current wind power curtailment capabilities, and how other jurisdictions have implemented wind power control and curtailment strategies in their operating regions. Finally, current contractual constraints that have been put in place were reviewed. This literature provides context for the conclusions that were drawn in this report.  2.1 Optimal dispatch of wind and existing/other generation resources In a report prepared for the Republic of Ireland and Northern Ireland, Gardner, et al. (2003) studied three near-future years, 2005, 2007, and 2010, to determine what level of wind penetration could be integrated into the transmission system. The Irish transmission system was unique in that it was an island system with very little interconnection capacity, namely 450 MW to neighboring Scotland. The 2010 scenario had an installed generation capacity of 7,079 MW, 1,393 MW of which was installed wind capacity. A significant assumption that had to be made was that in order to provide adequate ancillary services and reliability, the conventional generators (primarily thermal) were only allowed to be turned down to minimum generation levels, and not completely shut down, during periods of high wind. This assumption also addressed much of the potential need to carry extra 8  reserves once wind generation was added to the power system, reserves that could have been required by both the variability and uncertainty of the wind resource. This was because the online conventional generation was adequate to meet load in the absence of wind. Therefore, the addition of wind in the simulations simply offset the fuel costs associated with meeting the same amount of load solely with conventional thermal generation. Much work has been done, and many papers have been published, looking at the Nordic power generation system and market. This work has substantial parallels to the BC Hydro system, most notably since the Norwegian generation system has a large proportion of hydropower. Vogstad (2000) performed a case study focused on the hydropower scheduling effects of adding a hypothetical wind farm in central Norway. Like many of the Nordic studies, the Norwegian Electric Power Research Institute’s (EFI) Multi-area Power Scheduling (EMPS) model was used in the analysis. This model included Norway, Sweden, Finland, and Denmark, along with interties to northern Germany. It was run using weekly hydrological data from 1961 through 1990 using a weekly time step. It has two components that calculate the value of water stored in reservoirs using stochastic dynamic programming, and then performs a hydrologic simulation that maximizes the profit from the use of that water. A nested version of this model, called EFI’s One-area Power Scheduling (EOPS) model, was used to perform the actual case study. The hypothetical wind farm’s capacity was increased in 100-MW increments from 100 MW up to 1000 MW. The introduction of wind energy into the hydropower scheduling model decreased the optimal reservoir levels and reduced spillage. The value gained by this reduced spillage, compared with simply valuing wind energy at the prevailing market price for energy, effectively increased the value of the wind energy between 9% for a 100-MW capacity wind farm and 3.7% for a 1000-MW capacity wind farm.  9  Related work was later done by Korpås, et al. (2006). This work focused on constraints that were often experienced in the development of wind power projects because of their remote locations and weak existing transmission networks. Their observation was that the constraints put on wind power development were overly conservative and did not take into consideration the complimentary characteristics that the scheduling of hydropower can have. The case study here established a limit on wind power integration assuming existing operations of the other local hydropower projects, no curtailment of wind power, guaranteed transmission capacity for all generation resources, and a limited grid connection to the load centers and other generation resources. It then incorporated the stochastic variations in wind and hydro into the hydro scheduling algorithm and allowed wind power curtailment. When maximizing wind power penetration, the new limit for economical wind power development was over five times greater than originally thought, with no appreciable loss in market value of exported energy. This added capacity for wind power came from adjustments to the local hydropower production, as well as only providing non-firm transmission capacity for the wind power. This work also emphasized two different control strategies: curtailing wind during times of grid congestion or adjusting storable hydro generation down as much as possible before curtailing wind during times of congestion. (Extra stored water was later used for generation to keep the reservoir storage as close to schedule as possible.) The model domain covered 30 years of hourly data for storable inflow, run-of-river inflow, scheduled hydro generation, market price, electrical consumption, and wind power. The EMPS model was used to generate hydropower schedules, market prices, stochastic inflows, and temperature-dependent electricity consumption. This data was interpolated or shaped to achieve the necessary hourly time step resolution because the EMPS model runs at a weekly time step.  10  Very detailed work was outlined by Korpås (2004) in his doctoral thesis. The scope of this work extended beyond coordination of wind power and traditional hydropower plants, to a variety of energy storage mediums whose purpose was to smooth and firm the variable and uncertain output of wind generators in transmission systems with limited or no external grid connections. Several sections of this thesis were dedicated to the use of dynamic and linear programming algorithms to optimize the use of energy storage devices to maximize the value of wind power in a market context, including scheduling of wind power. Substantial differences in low and high electricity prices were necessary to justify investment in energy storage infrastructure other than existing hydroelectric generators. Finally, in chapter 6 of her doctoral thesis, Matevosyan (2006) developed a planning and evaluation algorithm for the coordinated scheduling of hydro and wind power in areas with limited transmission interconnections. This coordinated operation used the water in the hydro reservoir as a storage medium during congested export situations, and ensured that a maximum amount of available wind energy could be used to meet electrical load. This minimized the need for potentially costly import/export transmission reinforcement. Her research investigated four different coordination cases: 1. “uncoordinated” meaning hydropower was scheduled, and then wind energy was simply added on top of the hydro generation, and curtailed when transmission limits were reached 2. “buy whole wind power production” where the hydropower utility paid for all available wind every hour, and the hydro utility chose whether to curtail wind or spill water in times of transmission congestion 3. “store hydro power and sell later” where hydropower producers shifted excess wind energy that was available during times of transmission congestion to times when there was 11  available transmission capacity to export that excess energy, this was done by beginning with the “uncoordinated” case and then scaling back hydropower production during times of wind energy production in excess of transmission capacity, and later increasing the hydropower production during times when there was extra room on the transmission lines 4. “forced regulation” meaning the hydropower utility had to reduce production during times of transmission congestion to allow all of the wind energy to be exported. These cases reflected a variety of potential ownership and coordination combinations of hydro and wind power generation resources. These cases assumed deterministic load, transmission capacity, wind speed regimes, inflow regime, and electricity prices, similar to the research presented in this study. Matevosyan used only one inflow regime and transmission capacity configuration while integrating an uncertain wind power forecast into the modeling. However, this study’s research does not incorporate forecasts; rather it investigates the effects of separate inflow regimes and varying transmission capacities. She also concluded that the coordinated operation of hydro and wind power producers could be mutually beneficial.  2.2 The market context In work done at the University of Victoria, British Columbia (Benitez, 2006), a nonlinear mathematical optimization program was developed to investigate the effects of integrating wind power into an existing hydro and thermal power system. The power generation system in Alberta was used as a case study. The program showed that the increased generation variability in the system could be mitigated by either increasing the capacity and operational flexibility of the hydropower plants, or by the addition of peaking gas-fired generators. This work attempted to model all of the economical aspects of wind power integration by minimizing the cost to meet electrical demand using some or all of the following: energy storage, thermal generation, 12  hydropower, and wind power. Outputs of these optimizations were the costs of managing different amounts of wind power variations and the marginal costs of reducing of CO2 emissions. Holttinen, et al. (2001) performed several studies using EMPS, a weekly model of the Nordpool electricity market, to investigate the effects of increased wind power production and reduced CO2 emissions on that market, and to identify regional transmission bottlenecks. This model included Norway, Finland, Sweden, and Denmark, with transmission ties to Central Europe. The Norwegian system, with its significant hydropower and reservoir storage, was modeled in detail. Two system configurations were modeled, the current system and the system 10 years in the future, with wind energy penetrations ranging between 4% and 12% in each case. Wind and inflow inputs to the model came from 30 years of data to represent the stochastic nature of the two resources. This study found that the additional wind energy primarily displaced oil and coal-based generation resources and small amounts of nuclear energy in wet years. The displaced generation was in regions other than where the wind generation was located. The study concluded that in the current system (circa 2001), the market value of wind energy would not be sufficient to stimulate investment in wind resources, but in the 2010 scenario the market value of wind could be high enough to justify wind power development. Angarita and Usaola (2006) developed a mathematical formulation and optimization of hydro and wind generators in a power exchange market context. This work stemmed from the realization that wind power was difficult to forecast (uncertain) and could often be penalized for these comparatively inaccurate forecasts in the existing market context. The hypothesis was that coordinating bids and operations with a hydroelectric generator could reduce these penalties. The study was performed using the following steps. First, optimal bidding and operational strategies for wind and hydro were developed independently. Then synergistic strategies were developed for 13  the same wind and hydro generators under the same conditions and constraints. The conclusion was that cooperation with the hydro generator decreased the imbalances between forecasted and actual wind generation, and therefore the economic penalty to the wind generator for those imbalances was reduced. Day-ahead market energy prices were assumed to be perfectly forecasted in the case study (in Spain). The state-of-the-art Spanish wind forecasting system (SIPREÓLICO) was used to generate wind power forecasts, and several cases were run for one single day with varying forecast accuracy, and either one daily auction for energy or six auctions per day with several corresponding updated forecasts. In Sørensen, et al.’s paper (2008), they modeled the Nordic electrical system with one-hour time steps, while several critical hours were modeled at a five-minute resolution. The most critical hour was defined by a combination of low load, high wind, load up-ramp, and wind down-ramp, and ultimately was found to be a July morning with a wind drop-off during the morning load up-ramp. At their wind energy penetration levels (10% for the high wind scenario), the load ramps overshadowed the wind ramps. Their modeling tool incorporated two of Norway’s three markets: the day-head (spot) market and the regulating (reserve) market. Two wind power scenarios were investigated. The authors found that as wind power was integrated, the accuracy of the two models (hourly and 5-minute) began to diverge. This divergence was due to different levels of grid modeling detail, some stochastic versus deterministic inputs, transmission line loss calculations, and perfect versus realistic forecasts for reserves. More regulating reserves were required for the high wind scenario. A stochastic optimization of wind energy in conjunction with pumped-hydro energy storage resources in a market environment was presented by Garcia-Gonzalez, et al. (2008). The market prices and available wind energy were the stochastic variables in this two-stage optimization. The 14  optimization of day-ahead market bidding strategies was first solved assuming uncoordinated operation of wind and pumped storage resources, and then was subsequently re-solved allowing coordinated operation of these resources. This approach was applied to a realistic case in the Spanish electricity market. The coordinated operation of the wind generator and the pumped storage resource yielded fewer imbalance penalties than uncoordinated operation. The optimal size of the pumped-storage resource could also be determined using a series of these optimizations, studies, or runs. When the pumped storage resource was small compared to the wind generation installed capacity, the incremental value that the pumped storage resource added increased rapidly. However, as the size of the pumped storage resource approached the installed capacity of wind power generation, the incremental decreases in imbalance costs from a larger pumped storage unit became small. Similar to the previous example, but in the United Kingdom market context, Bathurst and Strbac (2003) formulated and compared two optimal bidding and operational strategies for a wind power resource and an energy storage medium. These two strategies represented independent and coordinated operation. The energy storage medium was coordinated with the wind power resource purely through market involvement. The energy storage device balanced benefit from temporal energy arbitrage with ancillary services’ value, which was measured by the avoided costs of wind farms' imbalances. It was also shown that grouping several smaller wind generators into one trading unit could reduce their imbalance penalties in the market. This is commonly referred to as the benefit from diversity of wind generation resources. While providing a high-level overview of some North American electrical systems and their unique attributes, Kirby and Milligan (2008) observed that keys to extracting the maximum value from wind generation resources were: large interconnected balancing areas, conventional generators with fast 15  ramping capabilities, multiple electricity market clearing time horizons (including sub-hourly), and robust ancillary services markets. They also made the following observation: "It is almost always more attractive to curtail fuel-consuming power plants rather than 'spilling' free wind." From a market-specific perspective, Kirby and Milligan observed that by implementing shorter market clearing time steps, such as those some ISOs/TSOs in North America have, load and wind following reserves could be extracted at very low costs when compared with the more traditional hourly market time step. This approach required the operators of traditional or conventional power plants with fast ramp rates to re-dispatch their units more frequently, rather than merely holding reserve capacity to meet variations in load or generation within the operating hour.  2.3 Wind power curtailment state of the art Development of wind power in Texas, along with the increased value potential when coupled with compressed air energy storage (CAES) technology, was investigated by Son (2005). Wind power development has been significant in Texas, and by the early 2003 1,293 MW of wind power capacity was operating. However, the transmission system in the areas with extensive wind power development was not very robust, and wind power curtailment was regularly being exercised at that time. The transmission bottlenecks posed a significant risk to further development of wind power capacity, and the Electricity Reliability Council of Texas (ERCOT) has already made modifications to their market structure to help send price signals to transmission developers. Wind power curtailment not only resulted in lost energy but also in lost production tax credits (PTCs) and/or renewable energy credits (RECs). This research studied the benefits that CAES could provide if coupled with wind power in areas of transmission congestion. Currie, et al. (2008) documented a demonstration of set-point curtailment in the Orkney Islands distribution network in northern Scotland. In this case, curtailment was investigated in order to 16  facilitate increased wind power development, while allowing the electrical system operator the flexibility to continue to ensure system integrity. In this case, the wind power was connected to the electrical system at the distribution level, and the power line connecting the distribution network to the transmission network was the primary location of thermal faults. When this power line became overloaded, set-point curtailment of wind power generators became necessary if more wind power development was to be allowed. In northern Norway, wind resources abound, but the existing transmission system was weak. In order to maintain acceptable voltage levels and maintain grid stability, while still accommodating large-scale development of wind resources, Pálsson, et al. (2003) investigated two types of wind power control: reactive power control and coordinated generation control. The investigations were performed via simulation. The first simulation added a capacitor bank, a static var compensator, and secondary control of each, to control reactive power flow and maintain acceptable voltage levels. The second simulation added Automatic Generation Control (AGC) to a hydropower plant connected to the electrical grid on the same radial transmission line as the wind farm. The AGC control limited total wind-plus-hydro power generation to the thermal limits of that radial electric line. In Spain, regional “special regime” (a.k.a. renewable energy) control centers have been established to maximize the production of these renewable energy resources, while ensuring the security of the electric system. At that time, wind generation represented 60% of the “special regime” generation. These regional control centers must respond within 15 minutes to commands from the central control center, where a quasi-dynamic grid model was run and generation limits were calculated every 20 minutes. The quasi-dynamic grid model consisted of two linear optimization problems, which were solved in series, with the results of one used by the other as constraints. This process 17  typically provided renewable generation limits greater than the limits defined in off-line wind integration studies. The set points determined by these optimizations were sent to the regional control centers and their commands were implemented (Torre, 2008). Miller, et al. (2007) utilized a quasi-steady state time simulation of wind plants to demonstrate active power control capabilities of modern wind turbines. These capabilities helped to facilitate power scheduling, ramp rate limits, and frequency response for wind generators. Minute-by-minute simulations were performed for several three-hour intervals containing challenging grid operation conditions, where active power control of wind farms could be useful. The types of wind power curtailment modeled were: 1. Block curtailment – a set amount of power generation was curtailed to restore downreserve capacity from other generation sources. 2. Load following curtailment – allowed the wind power units to provide some of the downreserve capacity required by the electrical system operator (reduced the amount of foregone wind power). 3. Up ramp rate curtailment – during times of decreasing load and increasing wind generation, the system’s down-reserves may become exhausted and temporary limits on wind energy increases could be required. In the California case study, these conditions were rare, and therefore the lost energy due to the application of these types of curtailment was determined to be small. A survey of electrical grid connection codes that address wind power integration and control was performed by Alegría, et al. (2007). Necessary modifications to grid codes to accommodate wind farm connections were reviewed as well. The key issues identified were voltage and reactive power 18  control, frequency control, and fault ride-through capability. This paper presented a good pictorial summary of the wind power control capabilities that wind turbines could be equipped with, which might be of value to the grid operator in challenging circumstances. These were: 1. Absolute power constraint – a ceiling MW limit of wind power production was specified. 2. Delta production constraint – wind power production was reduced by a specified quantity. 3. Balance regulation – wind farm production was used to help balance generation and load. 4. Power gradient constraint – the increase in power production coming from wind farms was limited to a rate manageable by other conventional or traditional generators connected to the electrical grid. (This was typically in the up-ramping direction, since little can be done in the case where wind speed and corresponding wind power decreased rapidly.) 5. System protection – in a contingency situation, the wind farm was required to rapidly decrease power production until the fault was cleared. The capabilities of the most common wind turbine types (fixed-speed squirrel cage induction generator, variable speed wind turbine with doubly-fed induction generator, and variable speed wind turbine with synchronous generator) were discussed. The latter two turbine types were expected to be required in future wind turbine installations because of their advantages in control capabilities over the fixed-speed squirrel cage induction generator type. A final example of the implementation and effectiveness of curtailment as an operational strategy can be found in chapter 5 of the aforementioned doctoral thesis of Matevosyan (2006). The purpose of this analysis was to develop cost-benefit curves for wind curtailment and transmission line reinforcement. The wind curtailment was estimated in several different ways, depending upon  19  when wind and load data was available. Later in her thesis, Matevosyan went on to investigate other options to reduce the amount of wind energy that might need to be curtailed, and how to further maximize the utilization of the transmission tie capacity.  2.4 Contractual arrangements Wind integration research in Europe has been predominantly looking at the increased integration of wind power if curtailment was allowed for reliability; essentially providing non-firm transmission capacity. Providing wind developers with non-firm transmission capacity would be a diversion from standard practices such, as those in Germany and Greece, where wind power was provided with priority access to transmission capacity. It should be noted that in North America, non-firm transmission capacity was more commonplace. A current example of contractual curtailment arrangements came from Germany and their Renewable Energy Act of 2000, and its revised version of 2004, as summarized by Burges and Twele (2005). The original Renewable Energy Act, from the year 2000, guaranteed transmission grid access and fixed feed-in tariffs to new wind developments. In 2004, the act was revised to provide priority transmission access to new renewable generators, with the caveat that they could be curtailed during times when all of the grid capacity was being used by renewables. This was necessary in the interim because transmission upgrades were taking a number of years to implement, and limited transmission expansion was drastically impeding the development of more renewable energy resources. The uncertainty resulting from possible instances of unpaid curtailment still has a stifling effect on new development of renewable resources. The only wind energy curtailment that has been exercised so far has been in the north of Germany by the grid operator E.On Netz. Their method for curtailment essentially sets limits to wind energy producers of 0%, 30%, or 60% of their nameplate capacity. These limits were only applicable to wind generators who were integrated 20  after the 2004 renewable energy act was established, therefore the installed wind capacity in the region was not curtailable in its entirety. This arrangement has helped resolve some of the renewable energy development bottlenecks in Germany, but the situation was still being studied and addressed further. Greece was another location where accommodations for wind power curtailment were being built into new wind power contracts (Kabouris, 2004). There, maximum wind penetration levels were being sought via the use of interruptible generation contracts. Current legislation required that there be transmission capacity available to accept wind nameplate capacity, plus maximum thermal generation, during minimum load conditions. This had restricted new development of wind resources. For system security and market reasons, existing thermal generators cannot be shut down to allow greater wind power penetration, so the TSO required new wind farm connections to be curtailable. This curtailment was allocated either based on some prioritized list or in proportion to active power generation, and was of the maximum generation curtailment type. An additional investigation was performed to determine a potential wind power penetration level if curtailment was only implemented in cases where violations of grid security were actually happening, rather than curtailing preemptively. This style of transmission system operation was deemed too risky for now, but could become feasible with more sophisticated equipment and was still of interest.  2.5 Conclusions Just like transmission capacity and electrical energy, available generation capacity has value. To reliably serve firm electrical load, each generation technology requires varying levels of available generation capacity related to its predictability, variability, availability, and reliability. For generators with controllable raw resources (i.e. nuclear, natural gas, coal) this available capacity is closely coupled with their energy production. In the case of wind power generators, the raw 21  resource (wind) is uncontrollable, and therefore some amount of controllable generating capacity must be coupled with the wind turbines to maximize their value to the TSO and to the electrical consumer. Hydropower or gas-fired turbine capacity is often used to compliment the energy available from wind power developments. Numerous forms of beneficial wind power curtailment have been demonstrated, although the necessary advanced power controls are not without cost. The perceived need for any wind power control capabilities must be weighed with the additional costs of those appurtenances. These needs will vary depending on numerous factors including the strength of the existing electrical grid, bottlenecks in that grid, the location of generation and load centers, and the type of existing generation assets. In a deregulated electricity environment, price signals should determine the need for advanced wind power controls. This is not always the case though, in part due to a lack of cost associated with pollution, marginally developed markets, and legacy generation and transmission assets from the pre-deregulation era. In nearly all cases, the reality of current generation and transmission conditions is that the optimal level of wind resource development can only be reached by allowing and utilizing wind power curtailment. Whether the constraints are transmission bottlenecks, traditional generation limitations, a lack of sufficient market development and access, or transmission expansion restrictions, wind power curtailment that could mitigate some of these problems should be investigated. The value of wind energy can further be enhanced by coordination with energy storage technologies such as batteries, hydrogen production, or pumped hydro to name a few. Coordination with existing fast-ramping hydropower plants that have storage capacity and operational flexibility can also add to wind energy’s value. In many cases, the optimal coupled value of wind power and 22  hydropower can be greater than the optimal values of each technology on its own, but a contract that appropriately values both resources must be in place to facilitate and motivate this synergy. To investigate the potential benefits of wind power curtailment in the BC Hydro context, the most appropriate tool to use was the current power system optimization and simulation model used in many of their annual planning studies. The model is known as the Generalized Optimization Model (GOM), it has been extensively tested and implemented at BC Hydro, and support and input files were readily available. By using a model that was already well developed, the complexity and realism that could be incorporated into this work, with a reasonable and appropriate level of effort, was greatly increased.  23  3 Model development and modeling approach The model used to perform this research was based on the decision support model developed by Dr. Ziad Shawwash (2000) for BC Hydro. The original model combined an hydraulic simulation with the optimal dispatch scheduling of select generation resources in the BC Hydro system, to maximize the value of inflows and hydropower generation assets in this system. This model was adapted to incorporate wind energy into the optimization process, while continuing to maximize the value of inflows and hydropower generators. The model was run iteratively for various scenarios that were then compared to each other to determine the value that wind energy and wind curtailment capability could add to the existing hydropower system.  3.1 The deterministic optimization model The decision support model that was used is called the Generalized Optimization Model (GOM). It consists of two parts: an hydraulic simulation and a linear optimization model of the generating units within the limits of the simulated river systems. The model was written in AMPL, a mathematical programming language, and the linear programming solving technique is applied using the CPLEX solver. As the name implies, GOM is a generalized model that can be applied to any hydroelectric system. The hydraulic and generation variables, parameters, and constraints within GOM are indexed over time steps that can be defined to represent any study length at any time resolution. This modeling was performed using an one-hour time resolution for a study length of one year.  24  Figure 1: British Columbia bulk transmission system and generating stations  Source: BC Transmission Corporation, 2007  3.1.1 Hydraulics As applied to the BC Hydro case study, GOM simulates two river systems and five hydroelectric generation plants. The river systems that are modeled are the Peace River in northeastern British Columbia and the Columbia River in southeastern B.C. There are two generating stations on the 25  Peace River: the Gordon M. Shrum (GMS) with Williston Lake (40 billion M3 of live storage) as an upstream reservoir and the Peace Canyon (PCN) that is a short distance downstream with a small amount of river storage between it and GMS. On the Columbia River, there are three generating stations: Mica (MCA) with Kinbasket Lake (14.8 billion M3 of live storage) as an upstream reservoir, Revelstoke (REV) a short distance downstream with a small amount of river storage between it and Mica, and Hugh Keenleyside (ARD) further downstream with the Upper and Lower Arrow Lakes (8.8 billion M3 of live storage allowed by the Columbia River Treaty) as its upstream reservoir. For each reach in the river systems, there are flow parameters and variables representing turbine flows and spills from upstream generating stations and local inflows. For each reservoir upstream of each generating station, there are parameters and variables representing optimal, maximum, and minimum reservoir stages (elevations) and their respective storage volumes. There are also parameters defining maximum changes in these stages between each successive time step. Both the Peace and Columbia Rivers have flood control storage and release constraints dictated by flood control curves, and additional constraints on downstream flows in the Columbia River system are imposed according to the Columbia River Treaty (CRT) with the United States.  3.1.2 Generating stations GMS and PCN on the Peace River provide 3,430 MW of installed generating capacity while MCA, REV, and ARD on the Columbia River provide 4,025 MW of installed generating capacity. The combined total hydroelectric generating capacity in the BC Hydro system is about 10,500 MW; therefore, GOM optimizes the operation of about 71% of the generating capacity in the BC Hydro system including the vast majority of its live reservoir storage capacity. Other hydro plants are highly constrained and were not included in these studies for simplicity of analysis.  26  At each generating station, there are parameters and variables representing optimal, maximum, and minimum power generation. There are also parameters defining maximum changes in these optimal values between each successive time step. The relationship between head, discharge, and power generation are approximated by piecewise linear approximations for each generating plant.  3.1.3 Load-resource balance and objective function While obeying all pertinent hydraulic and generation constraints within each time step, GOM seeks to optimize the energy produced by routing of water through the modeled generating units. This energy, combined with other energy produced within B.C. (a parameter) and imported energy from the U.S. and Alberta, must match the aggregate of domestic electrical load, plus exports to the U.S. and Alberta. Exports and/or imports to the U.S. and Alberta are also variables in GOM. The objective function in the optimization problem is the cost or revenue from these energy market exchanges in each time step, coupled with the change in reservoir volume multiplied by the value of water in each reservoir. This sum is maximized in order to extract the most value from the inflows and generation assets in the hydroelectric system for the specified study period, and the optimal routing of water through the hydroelectric system is an output.  3.2 Additions to the existing model The optimization model as just described was only designed to obey hydraulic and generation constraints while optimizing the dispatch of available generating units. It was necessary to expand and augment these capabilities to accurately model and integrate wind power with its unique qualities.  27  3.2.1 Operating reserves Managing the variability of wind power within each time step presents additional challenges that were previously unique to domestic electrical load. The inflows and electricity coming from small hydroelectric power plants and independent power producers can accurately be represented in the model using typical daily and seasonal shapes that do not significantly vary within each hourly time step. Likewise, import or export market transactions generally only vary between each hour and not within the hour. Before the addition of wind power, domestic electrical load was the only parameter that varied within each hour. These intra-hour variations were managed by allocating a specific amount of available generating capacity to respond to the fluctuations in load. Wind power is unique among the modeled generators in that it varies according to wind speed and curtailment instructions, and therefore additional controllable generating capacity had to be allocated to manage its intra-hour variability. These requirements were addressed by explicitly reserving generating capacity within each hour to respond to this variability. This variability in domestic load and wind power was split into two types: following and regulating. The generating capacity allocated to manage the following variability could be provided by any of the optimized hydroelectric generators. However, the regulating variability is on the order of seconds, and the generating capacity that was reserved to deal with this variability had to be equipped with automatic generation control (AGC) equipment. Only the generators at MCA, GMS, and REV currently have this equipment.  3.2.2 Wind’s effects to hourly electrical energy prices Unlike traditional storage hydropower and thermal power sources, the primary energy source of wind power, the wind, is not controlled by the electric system operator. As installed wind power  28  capacity spreads and grows in electrically interconnected areas, downward pressure is exhibited on electricity prices during periods of high wind speed. A similar and perhaps more familiar phenomenon has long been seen in regions with significant hydropower capacity during the spring snow-melt (freshet) period, normally in June and July in the Pacific Northwest (PNW). These trends stem from run-of-river hydropower plants, which also have little or no upstream storage capacity and therefore little control over their primary energy source – local inflows. The downward pressure from this surge in electrical energy availability during the spring freshet has been visible in the electrical energy prices in the PNW. Since the installed wind power capacity is still small in the PNW, similar effects during periods of high wind speed are currently small, and not readily apparent in historically based electricity price shapes. However, since wind power development is increasing quite rapidly, power trading experts widely agree that similar downward pressure to electricity prices is noticeable and relevant, and it will continue to increase. Therefore, an electricity price discount factor was built into GOM according to the ratio of wind power production to wind power installed capacity.  3.2.3 Slack generating capacity valuation The two products that power planning professionals focus on are electrical generating capacity and electrical energy. Both of these demands must be met, and there are corresponding markets for both of these products. In an hydroelectric context, these are represented by the generating capacity of hydro turbine/generator units, measured in megawatts, and electrical energy (or water if viewed as a primary fuel source) measured in megawatt-hours. Wind power, because of its variable primary fuel source, has little generating capacity value in the operating time frame (hours to days). This is compounded by the uncertainty inherent in wind speed forecasts. Because of these realities, it is widely understood that wind power generators are primarily viewed simply as an energy source 29  and not as a source of available generating capacity in the operating time frame. In an hydroelectric context where turbines cannot be run at full capacity all of the time because of variable inflows and finite upstream reservoir capacity, the independent values of generating capacity and energy are quite apparent. With the recent increases in development of wind power as an electrical energy source, it is important that the generating capacity of hydroelectric assets is properly valued. When GOM was initially developed, adequate generating capacity was inherently ensured by the constraint that load and generation must be balanced at every time step. The model modification that was now necessary with the addition of wind power to the generation mix is the valuation of unused and available generating capacity that could be exported via transmission interconnections to adjacent regions. In the optimization model this generating capacity is assigned a market value and the subsequent amount required to manage the variability of load and wind within each hourly time step was incorporated into the objective function as a cost. The remaining available and unused generating capacity in the hydroelectric system in each time step was included in the objective function as a benefit. Because available unused generating capacity is a variable and was now included in the optimization problem’s objective function, it was optimized too.  3.2.4 Wind energy as a generation resource In order to assess the value of the additional wind energy in the optimization problem, wind energy had to be added to the load-generation resource balance first. Then the time series of wind energy production was added as a variable and either fixed as a “must take” resource or allowed to be optimized and therefore varied between zero and the available wind energy for that time step. This  30  approach allowed the objective function values from successive model iterations to be compared and the value of wind energy as a resource derived from this comparison.  3.2.5 Payment for wind energy A final modification was made to the optimization model in order to explore the effects of the following contractual scenarios: 1) payment from the hydroelectric utility to the wind power producer for all available wind energy; or 2) payment for only the portion of available wind energy that is optimal to integrate. Although this comparison is not applicable to the BC Hydro situation, this comparison represented the wind power producer independently and investigated what portion of available wind energy an hydroelectric utility might actually choose to purchase if there was no contract in place. The prices assigned to this wind energy encompassed a range that allowed a curve to be generated that could be used to represent a value of the additional wind energy to the hydroelectric utility. This is an approach that any utility integrating wind could use when deciding to purchase energy from an outside source.  3.3 Modeling approach The goal of this thesis was to explore the possible value that curtailment of wind power, in a market context and under a variety of scenarios, might provide. The previously described simulation and optimization model of a hydroelectric system was the tool that was used to explore this hypothesis. It is a deterministic model, so particular independent variables (input parameters) had to be manually manipulated, and the optimal objective function values of the hydroelectric system at the end of each run were compared to each other. The differences between these values were then attributed to the particular parameter that was varied. Input parameters that could have an effect on the potential value of wind power curtailment are outlined in the scenario tree in Figure 2.  31  Figure 2: Modeling scenario tree  For example, one set of model runs selects one type of water year, one wind energy penetration level, and one set of intertie capacities, and then optimizes the hydroelectric system’s operation for each wind power control scenario and given cost per MWh of wind energy. Then by comparing the objective functions in wind power control scenarios 1 and 2, the value of wind power curtailment for that run was evident. In wind power control scenarios 1 and 2, it was assumed that the hydroelectric utility was contractually obligated to compensate the wind power producer for all available wind energy, and therefore any change in the objective function was only due to the added ability to curtail wind energy. Wind power control scenario 3 could also be compared to the first and second scenarios to represent the case where the hydroelectric utility was not contractually obligated to compensate the wind power producer for any specific portion of available wind energy. Only the wind energy that increased the objective function of the hydroelectric utility was integrated. Again, this part of the investigation was not representative of the BC Hydro case study, but provided insight into the value of wind energy in a competitive market environment.  3.4 Summary The GOM model provided a robust starting point to assess the value of wind power curtailment in an hydroelectric system. In the BC Hydro case, it models most of the flexible generating units and 32  the two river systems with the vast majority of live reservoir storage. The assumptions that account for the remainder of the generating resources are reasonable and consistent with those made in many other planning studies currently carried out by BC Hydro. After deciding that GOM was the most appropriate tool to use for this research, the modifications to the model that were necessary to properly model the addition of wind energy to the hydroelectric system were made in cooperation with Dr. Shawwash, who is the original developer of GOM. This ensured that they worked seamlessly with the existing model. A review of recent research guided the choice of independent variables that were manipulated. These independent variables cover some of the most pertinent aspects of integrating wind energy into an hydroelectric system within a market environment. The results of this thesis will help to understand the effect that these variables could have on the value of wind curtailment in an hydroelectric system and market context, and potentially help guide future system-specific studies.  33  4 Model data and assumptions In this section, the input data necessary to perform the hydraulic simulations and generation optimizations are discussed. Particularly, their sources, simplifications, and assumptions are laid out. The ranges of the independent variables and how they were determined are also explained with supporting references.  4.1 Fixed input parameters When seeking to represent a real world problem with a mathematical model, some level of simplification and assumptions must be made in order to find a timely solution. These simplifications were made by the model designer in consultation with the people most familiar with the problem. In the hydroelectric scheduling model case, the people concerned are the generation system operators, planners, and market traders. Their experience and knowledge of how the variables are inter-related and how they pertain to the ultimate purpose of the modeling, was invaluable. After consultation with these experts, it was determined that the following input data would be kept constant in all of the modeling runs for the BC Hydro case study.  4.1.1 Load: energy, shapes, and peaks Because BC Hydro is a publicly owned corporation, they are required to publish medium and long term plans and forecasts for their load area and generation assets (BC Hydro, 2008a). These plans include forecasts for annual energy requirements as well as peak annual load. These two values were used to scale historical shapes for typical weeks in each month or season. The result of this preprocessing was an hourly time series of load magnitudes, which was used by the optimization model in the load resource balance constraint, which requires generation plus imports to equal load plus exports. Figure 3 shows the daily and seasonal patterns that electrical load in B.C. exhibits. 34  Figure 3: Average BC Hydro System Load  4.1.2 Fixed-generation energy and shapes The optimization model used in this study only had about 71% of the system’s total hydroelectric generating capacity included in its optimization runs. The remaining capacity that made up the difference between load, imports/exports, and generation came from small hydro, cogeneration, and thermal plants, as well as contracts with other domestic independent power producers (IPPs). The daily and seasonal generation profiles of these resources were largely dependent on local inflows and/or load shapes. Additional preprocessing prior to optimization was performed to generate hourly time series of energy generation from these resources, which were then used by the optimization model as input parameters. In Figure 4, the monthly generation levels of Burrard Thermal (a natural gas fired generator), IPPs, and a portion of Columbia River storage that was not governed by the Columbia River Treaty are 35  shown. Data from the average water year is shown here; however, the differences between dry and wet years were very slight. Burrard thermal generating station’s generation levels only varied an additional 10 or 20 MW during other types of water years and was more influenced by electric load. This can be seen in its operation pattern – relatively constant for all months except the coldest months, when load was at its peak. The majority of IPPs are run-of-river hydroelectric plants, and so the peak amount of electricity from this type of fixed generation was during the freshet in the spring. Daily variations in these plants were small or negligible, so monthly blocks of energy were determined to be sufficient for this modeling. Depending on the type of water year, these monthly energy values varied up to 100 MW. The least variable of these fixed generation sources was the extra energy from non-treaty storage in the Columbia River system. This extra energy was essentially constant year-round and did not appreciably vary based on the type of water year. Figure 4: Fixed Generation Shapes  36  There is more flexibility in the operation of BC Hydro’s small hydro assets. Figure 5 shows the daily and monthly fixed generation shapes for small hydro in B.C. The freshet months still contributed the greatest amount of energy generation because of the limited reservoir storage in these plants; however, because BC Hydro operates these resources and they have some water storage capacity, the daily generation shapes resembled the electrical load shapes. These daily shapes were based on historical optimal operation of these plants, and optimizing them in this study would significantly increase processing time while providing relatively little additional information. Figure 5: Small Hydro Generation Shapes  37  38  4.1.3 Hydroelectric generating unit outage schedules Hydroelectric generating units must occasionally be removed from service for scheduled maintenance. These schedules have been determined by studies that sought to minimize the impact to generation adequacy and market opportunity. Two key components of these schedules were the ability to serve peak load and to meet the goal of minimizing spilled water. The addition of wind energy is likely to have little effect on either of these components, so the existing hydroelectric unit outage schedules were maintained for all of the GOM runs.  4.1.4 Forebay stage limits and flood control curves Other intra-annual hydraulic guidelines and reservoir boundary conditions for the GOM simulation were extracted from the outputs of simulation runs that contain 60 consecutive years of hydrologic data. These simulation runs were performed using a separate high-level model called HYSIM, which operates at a monthly time step (GOM refines these runs one year at a time using shorter time steps). HYSIM runs also approximate the value of water in the two main reservoirs; however, this value was not used in GOM modeling because it would only be applied to water stored at the end of the water year for use in future years. This was not allowed, as the target forebays obtained from HYSIM were fixed in GOM. Otherwise, modeling several consecutive years in series would have been required, greatly increasing the computational time because GOM operates at an hourly time step. Figure 6 shows the monthly forebay targets extracted from HYSIM as vertical lines. For all months except the last month, these targets were relaxed to allow GOM flexibility to make small changes in storage levels. These are typical shapes, with reservoirs being drafted through April in preparation for freshet inflows. By July, the reservoirs are quite full and store water to be used the upcoming winter when electrical loads are high. Mica, at the top of Figure 6, is the most upstream plant on 39  the Columbia River and has the most storage. Revelstoke, in the middle, has very little reservoir storage, and its operation was primarily dictated by local inflows and Mica’s operation. The Arrow Lakes reservoir, at the bottom, has additional constraints placed on its operation by the Columbia River Treaty, and must maintain certain amounts of storage for flood control in the U.S. It typically generated at full capacity all of the time. MCA, REV, GMS, and PCN rarely spilled water. Figure 6: Sample Columbia River Plants’ Forebay Trajectories and Constraints  40  4.1.5 Energy prices For the purposes of this independent academic research, obtaining representative energy market prices can be difficult. Much of the price data that BC Hydro has is proprietary. Its energy and generation capacity in the markets of the PNW put it in a position where its actions can significantly influence the market operations. Some price information is publicly available on a subscription basis at several time horizons, but this data can have inaccuracies because some of it is obtained through a voluntary survey conducted by a third party. Nevertheless, a similar data set was used to develop a typical hourly energy price shape for the mid-Columbia (Mid-C) market in the PNW. This hourly energy price shape came from actual energy prices over the course of a typical year. This was done to avoid the smoothing effects that averaging can have on a data set that contains significant and substantial variability, as well as diurnal, weekly, and seasonal patterns. This shape was then used to prepare the price data set used in this study. The magnitude of energy prices is dependent on the load year being studied. Like some other fixed variables, the annual average price forecasts have been calculated many years into the future by BC Hydro’s long-term plans (BC Hydro, 2008a). The price forecast for the load year in this case study (2010) was applied to the Mid-C energy market price shape. The first graph in Figure 7 shows the median energy market prices, and diurnal as well as monthly trends can be seen. The second graph shows just how volatile these prices can be throughout the year. This graph also shows the different scaling factors applied to each type of water year, with highest overall prices in the dry year and lowest prices in the wet year.  41  Figure 7: Mid-Columbia Energy Market Prices  42  4.1.6 Ancillary services prices The term “ancillary services” usually includes products such as reactive power, black start, lowvoltage ride through, operating reserves, and contingency reserves. These products are necessary in modern electrical systems to ensure their reliable operation. The modeling described in this research only incorporated the operating and contingency reserves, since other services mentioned are more appropriately modeled using power flow simulations, run at very short time steps with complex and detailed transmission models. In North America, ancillary services markets exist in many different evolutionary stages. The incorporation and effect of these markets is highly location specific. At present, the BC Hydro system has synchronized electrical access to the Western Electricity Coordinating Council (WECC), which includes the area extending from western Canada to Mexico. In this region, the ancillary services market administered by the California ISO (CAISO) is robust and mature. Therefore, the CAISO day-ahead market clearing prices for regulating up, regulating down, and spinning reserves for the most recent available water year (2007/08) were used. Their annual average prices are also comparable to contracts that BC Hydro has recently entered into with wind power producers in Montana, and elsewhere, for balancing services. BC Hydro also has contracts with the CAISO to provide a product known as dynamic scheduling, which consists of generating capacity on automatic generation control (AGC), which are directly responsive to electrical grid frequency signals from California. This product is similar to regulating up and down ancillary services. In Figure 8, the CAISO hourly clearing prices for these three types of reserves were plotted. Like energy market prices, there occasionally were large spikes in these ancillary services prices; however, the prices were typically quite constant and lower than energy market prices. 43  Figure 8: CAISO Ancillary Services Prices  4.2 Manipulated input parameters The independent variables and the ranges of their values were carefully chosen based on the goal of the research. The purpose of this thesis was to investigate the viability of allowing wind power curtailment as a valid operational decision. Variables such as water year types and intertie capacities were chosen to represent a range of reasonably expected conditions during the life of an hypothetical wind power project. Other variables, such as wind power installed capacity, market depth, and possible contract values of wind energy were harder to predict and instead were chosen to reflect reasonable conditions in 2010, the studied year, or soon thereafter.  44  4.2.1 Water years In typical generation planning studies in B.C., the series of water years from 1964 to 1973 are used because they have been shown by BC Hydro planners to contain a sample of inflows representative of dry, average, and wet years in the province. To limit the number of study iterations in this research, wet, average, and dry years from these ten water years were chosen to explore the different conditions that could warrant wind curtailment. Classification of wet, average, and dry was based on the inflows to the BC Hydro reservoirs as well as the measured river flows at the USGS hydraulic gauging station at the Dalles Dam near the mouth of the Columbia River. These flow measurements were included because they influence the Mid-C energy prices in the PNW. In Figure 9, the inflows to the five modeled hydropower plants were summed for each hour of the year, and then added cumulatively for the entire water year. It can be seen that the biggest difference between each water year is the timing and quantity of water during the spring freshet. The water years used in this research were 1969/70 representing a dry year, 1968/69 as an average year, and 1973/74 to represent a wet year. The measured flow at the Dalles also influenced this decision because the type of water year in the entire PNW affects energy prices. The water years 1965, 1966, and 1967 were wetter than 1973 in B.C., but they were lower than average water years as measured at the Dalles. The year 1971 was also wetter than 1973 in B.C. and at the Dalles, but its inflows were concentrated in the summer, which sometimes resulted in erratic modeling results. This was because the model year ends in September, and the reservoirs would run out of capacity while they were trying to hit their fixed ending forebay levels.  45  Figure 9: BC Hydro Inflow Patterns  46  47  4.2.2 Wind power installed capacity The selection of the range of wind power installed capacities for this study was driven by the current situation in B.C. Power purchase agreements between BC Hydro and IPPs for 325 MW of installed wind power capacity were signed in 2006 (BC Hydro, 2006). These projects represented about 14% of the total energy targeted in that call. BC Hydro is currently in the process of selecting proposals for another 5,000 GWh of seasonally and/or hourly firm annual energy to be available by 2016 (BC Hydro, 2008b). If half of that came from wind projects, the installed capacity of wind power would total roughly 12% in the province. It was the author’s opinion that a range of installed wind power capacity-to-peak load penetration between 3% and 12% would cover the reasonable possibilities for wind power integration using the existing configuration of the BC Hydro system. Certain assumptions had to be made when selecting individual hypothetical wind projects across the province that would sum to a desired total installed capacity. Section 4.4.1 details the goal and process followed when combining individual wind power projects to deliver an hourly time series of wind energy for each penetration level.  4.2.3 Market depth and intertie capacity The range of transmission intertie capacities was selected to represent two separate but related concepts: the physical constraint that the transmission intertie to the U.S. presents, and the amount of energy that the market in the U.S. can absorb. The physical constraint that the intertie to the U.S. exhibits for exports varies between 2,000 and 2,800 MW in typical modeling at BC Hydro. For imports, this variation is between 1,900 and 2,000 MW. These variations occur throughout the year because of ownership and priority differences for space on those transmission lines. To simplify matters, a single export and import maximum transmission limit value of 2,200 MW was assigned in the modeling performed for this research. It 48  was the author’s opinion, in consultation with others involved in the power trading industry, that physical expansion of these transmission ties was not likely in the near-term. This maximum value was reduced in several steps to represent the second consideration – market depth. If energy production within the region follows growth in demand, then entities will be less likely to need to buy or sell energy from outside of their balancing areas. Even though physical intertie capacity may be available, it could go unused. This market depth was modeled by reducing the maximum and minimum intertie capacity parameters from 2,200 MW down to a minimum of 1,200 MW. At lower levels, load resource imbalance dominated and caused difficulty for the GOM model to converge to a feasible solution. This shrinking access to outside markets had a very real effect on the value of wind power curtailment. At the extreme case, where there was no intertie capacity and/or no access to outside markets, generating capacity and energy surpluses and deficits were not subject to exterior prices. In this case, wind power curtailment would not have any value outside of the balancing area.  4.2.4 Contractual cost per MWh of wind energy The range of costs per MWh of integrated wind energy factored into the optimization problem when the hydroelectric system was allowed to restrict integration of wind energy to economically attractive times, with no penalty for curtailment. The remainder of the scenarios used in this study assumed that the wind power producer received monetary compensation for all available wind energy, regardless of whether it was integrated or not. The range of prices that was represented in the modeling, $15 to $105 per MWh, has two components. The first component is attributed to the value of green RECs that wind energy has because it is typically considered a renewable energy resource. (This definition varies throughout many 49  jurisdictions, but wind energy is almost universally considered renewable and is given RECs for each MWh of energy produced.) The value of these RECs varies widely between markets and is constantly changing, especially as markets for them are established and renewable energy targets are mandated, implemented, and updated by cities, counties, states, and countries. The value of $15 per MWh was a current approximation based on publicly available data contained in BC Hydro’s 2008 Long Term Acquisition Plan in Appendix H (BC Hydro, 2008a). The second component of wind energy value is from the energy itself. There is a quite a range of purported values of wind energy depending on the developer/owner, age and technology of the wind farm, tax credits, and feed-in tariffs. Based on publicly available wind energy contracts with BC Hydro in B.C. in 2006 and with Puget Sound Energy in Washington in 2006, and reports on wind energy purchases from Lawrence Berkeley National Laboratory (LBNL) throughout the western U.S. (Wiser, 2008a; Wiser 2008b), $90 per MWh of wind energy seemed to be at the high end of prices that the market is willing to bear for wind energy purchase contracts. This price is independent of, and additional to, the REC values previously mentioned.  4.3 Optimization variables The following operational decisions were treated as variables in GOM. Their values were optimally determined for each time step throughout the duration of each model run (one year). They, along with the value of the maximized objective function, were the results that form the basis from which conclusions were reached in later chapters.  4.3.1 Reservoir/forebay stages The upstream reservoir or forebay stages for each of the five modeled generation plants were optimized for each time step. This decision was whether to use a unit of water during the current time step to route through the plant’s turbines to generate electricity, to spill it, or to save it for a 50  later time step. These variables were bounded by maximums and minimums that come from either physical constraints or flood control curves. MCA and GMS have large upstream reservoir storage capacities and wide ranges within which their forebays can vary, while REV and PCN are the respective downstream generating plants and are almost run-of-river plants because of their small upstream reservoir storage capacities. ARD has some upstream reservoir storage capacity, but much of it is governed by flood control curves that are guided by the Columbia River Treaty with the U.S. The U.S. is downstream of MCA, REV, and ARD on the Columbia River.  4.3.2 Turbine flows The flow of water through each turbine/generator unit in each generation plant was also optimized, and was a function of flow and net head, including head loss through the plant. These efficiency curves were approximated using piecewise linear functions, which connect the peak efficiency points for each combination of online units and are convex functions that obey the rules of linear optimization problems.  4.3.3 Plant spills The total flow from upstream reservoir to downstream reservoir is the sum of turbine flows in a plant and spills at that plant. In an optimal solution, spills were rare at MCA, REV, GMS, and PCN. ARD had more spills because it had to meet minimum flow requirements of the Columbia River Treaty. Occasionally, spills were also necessary at PCN during a set period in the winter when the river forms an ice covering and its operating ranges are more constrained. The reserves required to manage wind power can increase spills in the hydroelectric system, but allowing the curtailment of wind power minimized these additional spills. Although wind energy is wasted during curtailment, water would be conserved that then can be used at more economically opportune times.  51  4.3.4 Energy and capacity market exchanges A quantity of energy, up to the limits of the transmission interties to the U.S., was available to the optimization model in each time step and was subject to the energy price parameters specified for the model. The income or cost attributed to these market exchanges in each time step summed to be part of the maximized objective function in the optimization problem. Another source of value in the objective function of the optimization problem was the export of up and down reserve generation capacity, which was valued at the ancillary services prices that were parameters for the model. The up-reserve capacity quantity was equal to the available hydroelectric generating capacity that was not used to generate electrical energy during that particular time step. The portion of this capacity that was valued was limited to the space left on the export intertie to the U.S. that was not either being used to export energy, or being reserved to deliver dynamic scheduling contracts. The down-reserve capacity quantity was equal to the hydroelectric generating capacity that was currently generating electrical energy, but which could be reduced to absorb excess generation from the U.S. This quantity was also limited to the available space on the U.S. intertie that was not being used to import electrical energy from the U.S.  4.3.5 Integrated wind energy For the model runs that allow wind power curtailment, the variable representing wind energy, whose default value was the available wind energy for that particular time step, could be reduced incrementally all the way down to zero. Allowing this replicated wind power curtailment, and comparing these model runs with equivalent runs where this variable was fixed at its default, showed the effects on the objective function of wind power curtailment capability.  52  4.3.6 Allocation of reserve generation capacity Operating, contingency, and exported generation reserves, which are defined in Section 4.4.2, were optimally supplied by each generation plant in GOM. The division between electrical energy generation and supply of reserve capacity was influenced by the efficiency curves of each plant as well as energy and capacity requirements for that time step. In certain highly constrained circumstances, such as peak load in the winter or minimum load during the spring freshet, the flexibility in the optimized generation plants was insufficient to supply some of the operating reserves. These infeasibilities were discussed with the engineers who operate the hydroelectric system and the following solution was implemented. Two additional variables representing up reserves and load shedding (down reserves) were introduced to the optimization model, with significant associated economic penalty, while an intermediate solution to the optimization problem was found. From this feasible solution, the values for these two variables were fixed for the remainder of the optimization runs to facilitate feasible solutions. These two fixed variables were part of the base optimization case and represented the hydroelectric control capabilities of either supplying some operating reserves from plants that are not included in the optimization model or by shedding domestic electrical load. Both of these operating decisions are credible and reasonable, although they are not commonly used and avoided if possible.  4.4 Assumptions: simplifications and eliminated variables The following set of assumptions were developed with, and reviewed by, subject matter experts within BC Hydro and the greater wind integration community. The goal of these assumptions was to provide a realistic context in which the goals of this research could be met in a timely manner. Conversely, it would have been quite possible to relax and/or vary any number of these 53  assumptions, given adequate time and resources, to investigate their incremental effects on the objective function.  4.4.1 Wind energy data The wind energy data available for the B.C. case study came from BC Hydro’s Wind Data Study (DNV-GEC, 2009). A sub consultant, DNV-GEC, was contracted by BC Hydro to simulate 10 years of 10-minute wind power time series for a great number of physically, environmentally, and economically viable wind farm sites across B.C. The ten years that was simulated included nine Canadian water years: 1998/99 through 2006/07. The simulated wind data from the most recent water year, 2006/07, was used in this research. Wind power development in B.C. is currently performed by IPPs, who then seek power purchase agreements with BC Hydro. Because of this situation, and the lack of existing wind power development, any combination of these hypothetical wind power project time series would be somewhat arbitrary. Therefore, the author used a map of the hypothetical wind farm locations, a list of potential wind farm sizes, and estimates of their development costs, to form four progressively larger one-year, one-hour wind power time series that represented 3%, 6%, 9%, and 12% installed wind power capacities. The size of the selected hypothetical wind farms varied from 28 MW up to 230 MW, and increased diversity in each scenario was ensured so that diversity and smoothing effects would be present in each successively larger wind power scenario. The following figures show the simulated wind power in B.C. from several different perspectives. First the series of one-hour wind power values are displayed, showing the variability. Then they are averaged to show seasonal characteristics. And finally, the monthly averaged diurnal patterns are shown for the 3% and 12% penetration cases.  54  Figure 10: Simulated Available Wind Energy  55  56  4.4.2 Load and wind reserves Regulating reserves for load were approximated in this study as 1% of the actual hourly average load for light-load hours (0:00 to 6:00 and 22:00 to 0:00) and as 2% of actual hourly average load for heavy-load hours (6:00 to 22:00) (BCTC, 2007). Using these percentages, the maximum regulating reserve level for each hour of the day in an entire month was considered the necessary level of up regulating reserves to hold for load during that hour in that month. An equal quantity of down regulating reserves was also held. Wind power’s regulating reserves were approximated simply by plus and minus 3% of installed capacity for all hours in the year. These are slightly lower than the reserve ranges calculated by Hudson, et.al. (2001), but the trend is for fewer reserves necessary as wind farm sizes and wind farm location diversity increase. The wind farms analyzed in Hudson’s paper are small compared to the sizes of wind farms that were considered in this study. Following reserves were approximated by calculating 50% of the step change between each successive hourly average load level (Electrotek, 2003). Then, for each hour of the day, the maximum up and minimum down following reserve levels for that hour in an entire month were considered the necessary levels of following reserves to hold for that hour in that month. This methodology was then applied to the load-minus-wind hourly power time series, and the load-only following reserves were subtracted to yield an incremental amount of following reserves necessary to integrate wind energy. These reserve requirements resulted in blocked generation capacity that otherwise could have been optimized to generate additional energy, or capacity reserves, which could have been imported or exported to neighboring utilities. Generation capacity for contingency reserves was also blocked in the model. Contingency reserves were simply set at 5% of the current level of power generation in the up direction (BCTC, 2007). 57  Unlike regulating and following reserves, which are used to manage normal fluctuations in load and generation, contingency reserves are used to recover from unplanned, forced outages in the power system. The graphs in Figure 11 show the results of the following reserve calculations: the values that were then input into the optimization model. The first graph shows the base reserves for each hour and each month that were attributed to only electric load. The next two graphs show the incremental amounts of following reserve capacity that were required because of the additional wind energy in the power system. It can be seen that the wind following reserves, which were a result of ramping events, were far less diurnally consistent than those from load. Figure 11: Load and Wind Reserve Levels  58  59  It is important to note that in scenarios where wind energy curtailment was allowed, incremental following wind reserves were also reduced by the percentage of wind power curtailment in that time step. Regulating reserves for wind were blocked regardless of curtailment level. Also, the up-generation capacity reserves and load shedding mentioned previously only applied to following reserves and wind contingency reserves, because regulating reserves must be supplied by AGC-equipped generators. BC Hydro’s operators only use the generating plants that were modeled to supply these reserves.  4.4.3 Load and wind forecasts This modeling assumed no load or wind forecast error (i.e. no load or wind forecasts are used in the model). This assumption intuitively results in a realistic lower bound on the value of wind curtailment for zero and consistently negatively correlated load and wind forecast errors. For positively correlated forecast errors, they would tend to cancel each other out and could devalue wind curtailment. The effects of this assumption are somewhat inconsequential because all of the modeled scenarios have this assumption in their optimizations. However, optimal bidding strategies in day-ahead and spot market transactions are interesting and pertinent, and are currently being investigated by others at BC Hydro. These could affect the purported benefits of wind curtailment, especially as wind power penetration levels increase.  4.4.4 Transmission constraints Transmission constraints between regions within B.C. were not included in the optimization model. Only the intertie capacities that provide access to the U.S. market were modeled and varied. Transmission tie upgrades between separate jurisdictions and countries can be the harder to mandate, justify, permit, finance, and/or construct than transmission upgrades within a country or within a single TSO or BA’s jurisdiction. One reason for this is that it is easier to justify transmission 60  upgrades for system stability, redundancy, and/or security rather than simply for economic reasons. The inter-jurisdictional transmission ties are also typically the most rigid and convenient bottlenecks that restrict access to energy and capacity markets. Although there are transmission interties to both Alberta and the United States, the transmission tie capacity used in the model was one single value, and the market prices were shaped to represent those seen in the U.S. market. The transmission tie capacity to Alberta is three to four times smaller than the one to the U.S., and it is not equipped to handle exchanges of generating reserve capacity. This single value for transmission tie capacity was a manipulated variable in this study and was discussed in Section 4.2.  4.4.5 Slack generating capacity valuation Available and unused generation capacity that could provide regulating reserves, following reserves, or dynamic schedules was valued up to the available space on the transmission ties. Limiting the valuation of generating slack capacity to the intertie capacities was a convenient and plausible assumption in this case study. The specifics of generating slack valuation were discussed in Section 4.3.4.  4.4.6 Available generation resources The generation resource stack was represented by forecast predictions for the 2010 load year (BC Hydro, 2008a). Wind energy is simply added to the planned generation mix/stack. Recently IPPs have been developing wind energy projects quickly, and many jurisdictions have a queue of wind projects waiting to be financed and/or built. The development of these projects has been somewhat decoupled from the typical supply/demand pressures induced by well functioning energy and capacity markets because of governments’ financial incentives and renewable energy policy changes, mandates, and commitments. Therefore, the energy from these potential wind power 61  projects would primarily decrease energy requirements on existing generation infrastructure, rather than meet increasing demand for energy or capacity or replace existing generators.  4.4.7 Energy price discount percentage during wind events Regions with large hydroelectric power developments with little storage capacity have a downward influence on prevailing market prices during periods of high inflows. A similar effect is seen in regions with high wind power penetration during wind events. In the Pacific Northwest, these effects are already apparent, and power trading professionals see depressions on the order of about 10% below what the “without wind” expected price might be. This trend was assumed to remain constant, and the energy prices were reduced up to 10% during periods when wind energy was being produced. More specifically, the 10% discount factor was multiplied by the hourly wind penetration level (hourly wind power divided by wind’s installed capacity) and then applied to the expected energy price. A simple example would be a case where 400 MW of wind energy was being generated in an hour, and the installed wind power capacity was 1,000 MW. The expected market price for that hour would be discounted by (400/1,000) * 10%, or 4%. If the wind energy was being produced at a full 1,000 MW in that hour, then the expected energy price would be discounted by (1,000/1,000) * 10%, or the full 10%. More available wind energy pushes prices down, which is reflected by using this discount factor.  4.4.8 Model duration and boundary conditions In this research, model runs covered a period of one year with one-hour time steps. The initial and final reservoir stages were fixed at the optimal values determined by a 60-year HYSIM optimization run. Flood control curves for MCA, REV, and ARD, which comply with Columbia River Treaty requirements, also came from this 60-year optimization run. 62  4.5 Summary These parameters, variables, ranges for manipulation, and assumptions were reviewed by a number of subject experts within BC Hydro and others within the wind power research community. Their feedback on initial ideas was incorporated, and modifications were made as necessary. Finally, since comparisons in this paper are only made between scenarios and model runs from this research, biases that may be introduced by some of these assumptions will be present in all cases, and their overall effects are minimal. With these fixed variables, manipulated variables, optimized variables, and assumptions all realistically tailored to accomplish the goal of this research using the BC Hydro hydroelectric system as a case study, the modeling approach was applied, and model runs were performed. The results of these runs are presented in Section 5.  63  5 Results and analysis The data sets created by the model runs were quite extensive and gave detailed pictures of the optimal operation of the hydroelectric system. Reporting and analyzing these results is done in steps, much like the modeling itself. First, the effects to the optimal operation of the hydroelectric system from the added wind energy are presented. These results came from comparing the fixed (non-curtailable) wind runs with the base case runs that had no wind energy. The differences between these runs are quite noticeable because of the significant increase in available energy and corresponding increase in exported energy. Second, the effects of allowing wind curtailment are presented and analyzed. This analysis compares the runs that allow wind power curtailment to those where the wind energy variable is fixed its full available capacity. These differences are less apparent because the optimal level of curtailment was quite small – between 0.5% and 2.5% of available wind energy in each model run was curtailed. Because the forebay levels were fixed at the start and end of each model run, the annual inflows had to be used to meet domestic load, with the difference made up by imports and exports across the transmission interties to the U.S. When wind energy was added, the optimal forebay stages within the year were changed to maximize the value of the additional energy that could be generated. Because wind generation is non-dispatchable, it had to be used in the hour that it was available, which affected the efficiency of the hydroelectric system. Additional generating reserve capacity also had to be set aside to manage the intra hour variations in the wind energy. The model  64  maximized the value of the additional wind energy by using the flexibility in the storage reservoirs to shift the export of surplus wind energy to times when the market price for energy was the highest.  5.1 Reservoir/forebay stages The two reservoirs with appreciable storage in the BC Hydro system are Kinbasket Lake above Mica Dam and generating station (MCA) and Williston Lake above W.A.C. Bennett Dam and Gordon M. Shrum generating station (GMS). The analysis was limited to these two reservoirs because they are where wind-displaced water could be stored and/or used to shift the wind energy to maximize the value of market energy exchanges. Most trends in the operation of these two reservoirs were consistent within each of the three water years. A few other observed trends were more general and primarily varied as a function of the intertie capacity. In nearly all cases, increasing wind energy penetration simply increased the magnitude of these observed trends. Figures 12, 13, and 14 show the changes in forebay levels as a result of adding wind energy (9% penetration level) and allowing curtailment, as compared to the base (no wind energy) case. The pink and red lines respectively show the changes in the Williston Lake elevations when fixed wind energy was added and when curtailment was allowed. The light green and green lines show the same comparisons for Kinbasket Lake. The light blue and blue lines also show the difference in the cumulative net import/export between the base case and the fixed wind and curtailable wind cases. They constantly increase because the cases with wind have the additional wind energy. Where they end at the upper right corner of each graph is approximately the amount of additional wind energy that was added in that wind case.  65  Figure 12: Storage Reservoir Changes – Wet Water Year  As shown in Figure 12, the changes to the Williston Lake (GMS) elevations were minimal in the wet water year. Optimal reservoir elevations increased slightly over the base case beginning in February and typically remained slightly higher through May, when the excess water was used to generate energy. When curtailment was allowed, the elevations were slightly higher than in the fixed wind case. In Kinbasket Lake (MCA), a significant amount of water was stored in the second half of March when energy prices were at their lowest. This energy was exported throughout the rest of the spring and summer, most notably in mid-April and August. Allowing curtailment decreased the Kinbasket Lake elevation from April through July, which was the same time period where curtailment allowed more water to be stored in Williston Lake.  66  Figure 13: Storage Reservoir Changes – Average Water Year  As shown in Figure 13, Williston Lake (GMS) had the greatest changes in the average water year. The reservoir levels were slightly increased in early winter and then most significantly increased in February. This extra stored energy was exported primarily in May and August. Curtailment had only slight effects on the operation of Williston Lake in this year. Kinbasket Lake showed two unique time periods of shifting energy in the average year. In the winter, excess energy from early December through mid-January was stored and resold through February and early March, before the spring freshet (period of high water flows due to snowmelt). Then during the freshet, energy was shifted by a couple of weeks, from late March to early April and late July. Curtailment had little effect on the reservoir elevations and generally allowed reservoir elevations slightly closer to the base case. 67  Figure 14: Storage Reservoir Changes – Dry Water Year  As shown in Figure 14, there was less water available in the dry year, and therefore smaller changes in lake elevations from wind energy and curtailment in the optimization runs were visible. Like in the average water year, Williston Lake stored water in mid April and gradually released it to generate electricity through the beginning of September. Kinbasket Lake stored water primarily in the fall, and released it for energy before the freshet in late March. During the freshet, additional water was stored which was used in July and September for energy. Some trends were consistent between all of the types of water years. The following list highlights these observations.   The modeled water year appeared to be split into two time periods within which wind energy was shifted. 68  o  The first time period for shifting wind energy was before the freshet (October through March). MCA was the primary actor in this time period, while ice conditions in the Peace River (GMS and PCN) system during January limited Williston Lake’s flexibility.  o  The second time period was from the spring freshet through August. Both MCA and GMS stored water during the beginning of this period and released it later to generate energy during economically attractive times.    The optimal time to sell the additional wind energy appeared to be in the early spring, late spring, and late summer. This time could extend into the fall, if the duration of the modeling runs was extended into the next water year.    The reservoir elevations tended to be slightly higher than, and parallel to, the base case runs between May and July, perhaps due to limited system flexibility (high local inflows and generating units out of service for maintenance) during that period.    Although Kinbasket Lake and MCA generating station are on the Columbia River, which is subject to the Columbia River Treaty with the U.S., the operation of Williston Lake and GMS generating station appeared to be less affected by the additional wind energy.    Allowing wind power curtailment affected Kinbasket Lake levels the most.    Adding of wind energy, and to a further degree allowing curtailment, facilitated a “shift” of water/energy between the Peace and Columbia River systems. Often when one reservoir was being drafted the other was storing water, suggesting that the extra wind energy allowed the hydroelectric system greater flexibility to keep water in the storage reservoir where it was most valuable. It should be noted that the values of water in the two river 69  systems are not the same because of differing operational constraints, storage volumes, and hydropower plant efficiency ranges. Incorporating transmission constraints could also affect this difference in water values, but that is beyond the scope of this research.  5.2 Hydroelectric system efficiency and spills A frequency analysis of inter hour generation step change magnitudes was performed for each of the five modeled hydroelectric generation plants, and these all generally decreased with the addition of wind energy. This was interesting because it is generally thought that wind energy, because it is variable and non-dispatchable, would cause existing generating units to fluctuate their output more. What was observed from this data set was that adding wind energy usually only reduced the magnitudes of inter hour generation step changes. The amount of water spilled in all of the model runs was tabulated. Nearly all of the spills were at ARD on the Columbia River, and they are driven by minimum flow requirements dictated by the Columbia River Treaty. When adding wind energy, the total volume of spilled water in dry years only decreased between 0.5% and 1.5% for all wind penetration levels at the 2,200 MW intertie capacity level. The remainder of the dry water year runs, and all of the average and wet water year runs, only changed the total spilled volume by +/- 0.6%. Curtailment only changed total spill volumes by +/- 0.05%, which is negligible. Although the total volume of spilled water was practically insignificant, the model did change the annual spill patterns. A series of plots, similar to the forebay plots, was generated to show the difference between the base (without wind) case and the fixed and curtailable wind cases. These are shown in the following Figures 15, 16, and 17.  70  Figure 15: Spill Pattern Changes at ARD – Dry Water Year  As can be seen in Figure 15, the spills were reduced from the base case runs throughout the fall and winter in dry years, before the onset of the spring snow melt. Then spills were slightly made up in late summer, but remained below base case levels, resulting in the aforementioned spill reductions of 0.5% to 1.5%. The changes to spill volumes were in the dry water year cases. Allowing wind power curtailment did not noticeably affect the spill volumes or patterns in the dry water year cases.  71  Figure 16: Spill Pattern Changes at ARD – Average Water Year  The spill pattern changes in the average water year cases were not as great as they were in the dry water year cases. Spills were not reduced until early winter, and then were not made up until the end of the summer, with the intermediate time exhibiting similar spill patterns to the base case. The net change to the spill volume in the average water year cases was negligible. Curtailment allowed the spill pattern changes to be slightly increased, as can be seen in Figure 16.  72  Figure 17: Spill Pattern Changes at ARD – Wet Water Year  Spill pattern changes were smaller still in the wet water year cases. In wet water years, spills were generally reduced in January and February but quickly made up in March. Spill volume changes in wet water years were negligible, like those in average water years. Curtailment allowed the optimization model to keep the spill patterns closer to the base case patterns. Finally, the generation conversion factor of each plant, which is a proxy for efficiency, was calculated, aggregated, and plotted. The conversion factor is measured in units of kilowatts per cms-hour, and this analysis was carried out at the annual level. Figure 18 shows the efficiency changes that adding wind – “Fixed Wind over Base”, and allowing curtailment – “Curtailable Wind over Fixed”, caused when the intertie limit to the U.S. was +/- 2,200 MW.  73  Figure 18: Plant Efficiency Changes – 2,200 MW Intertie Capacity  These efficiencies were aggregated by river system because, for example, losses at PCN might be more than made up for by gains at GMS. Likewise, the two river systems were aggregated in Figure 18 to show the effects to the entire hydroelectric system when wind was added and when curtailment was allowed.  74  It can be seen that the extra wind energy allowed increases in efficiency at the hydropower plants, especially during the dry year on the Columbia River system. However, this extra efficiency became less notable as wind penetration levels increased. It was also interesting to note that in the dry and average water years, allowing curtailment increased efficiency throughout the system, even as wind penetration levels increased. Figure 19: Plant Efficiency Changes – 1,200 MW Intertie Capacity  75  Figure 19 reports the same data but for the 1,200 MW intertie capacity runs. All of these efficiency changes were smaller than those from the 2,200 MW intertie capacity runs. The increasing efficiency gains with increasing wind penetration were still apparent.  5.3 Generation capacity reserves To analyze the effects that wind energy and curtailment could have on generation reserve capacity, duration curves of all available up and down generation capacity were constructed, as well as similar curves showing just surplus generating capacity that was not being valued on the interties. A variety of factors affect this reserve capacity. Energy almost always has a higher value than generation capacity, and therefore has a higher priority for intertie capacity. Since wind energy was added and the forebays were fixed, that excess energy had to be exported sometime within the year too. The next priority that the optimization model had was to attempt to fill the remainder of the intertie capacity with generation capacity exports. However, this goal was further constrained when wind was added, because wind power’s inherent variability and uncertainty require that additional generation capacity be reserved to manage it. Because of these complexities, it was difficult to draw conclusions from this analysis, although curtailment appeared to have no noticeable incremental effect on generation reserve capacity at this resolution.  5.4 Value and quantity of curtailment For each optimization model run, the objective function was maximized. This objective function is the amount of value derived from market exchanges of energy and generating capacity. An objective function value (1) was determined for just hydropower, then wind energy was added and a new objective function value (2) was returned by the model. Finally, curtailment was allowed and another objective function value (3) was returned. The value of fixed wind energy was the difference between the first two objective functions, (2) – (1), divided by the available wind energy. 76  In other words, the extra value, or income, that resulted from adding wind as an extra source of energy determined the value of the wind energy. Likewise, a value of curtailable wind energy was determined by subtracting objective function values (3) – (1), and dividing by available wind energy. Now, comparing the value of curtailable wind energy, to that of fixed wind energy, yielded what is referred to as the value of wind power curtailment. Figure 20: Value of Curtailment and Amount of Energy Curtailed  77  In Figure 20, the left column labeled “Curtailment’s Effects” shows the incremental value that allowing curtailment added to wind energy. It was always positive and did not show a consistent decreasing trend as wind penetration increased. From the 3% to 9% penetration levels it decreased, but this was probably due to increased diversity rather than the increased amount of wind energy. By the 6% penetration scenario, wind power plants from the four major geographic regions and wind regimes, throughout B.C., were being represented in the wind power time series. The value of curtailment did decrease as intertie capacities decreased. It should be noted that the decreasing intertie capacities we also meant to represent decreasing market depth instead of decreasing physical capacity. A decreasing market depth could be quite plausible in the case where wind energy development outpaces new energy demand in the region, and therefore, utilities would be less likely to need to import electrical energy. In the right column of Figure 20, labeled “Curtailed Wind Energy”, the amount of annual wind energy that was curtailed in each case is shown. For the 3% penetration level, also the least diverse case, the curtailed energy was relatively high. As penetration levels increased, and likewise the diversity increased, a smaller amount of wind energy was curtailed. Again, it should be noted that wind energy was not assigned a value in the model, so curtailing wind energy would not directly decrease expenditures in the model. The wind energy was simply available to be sold in the market to increase the model’s objective function.  5.5 Temporal characteristics of curtailment After determining that curtailment could add value to wind energy, and that wind energy and curtailment affect forebay operations, hydroelectric system efficiency, and spills, the occurrences and magnitudes of wind energy curtailments were analyzed. There are significant diurnal and seasonal shapes in energy market prices, so a more detailed analysis of curtailment by month and 78  time of day was carried out. Each day was split into two parts termed heavy load hours (HLH, the 16 hours between 6:00 AM and 10:00 PM) and light load hours (LLH, the 8 hours between 10:00 PM and 6:00 AM). These are divisions consistent with energy market and hydroelectric system operation practices. The series of bar graphs in Figure 21 were organized with wind penetration increasing left to right, and water years getting wetter top to bottom. Within each graph, there are 12 sets of bars from left to right, beginning in October and ending in September (the Canadian water year). There are two sets of bars for each month. The green and orange bars show the number of either heavy or light load hours within that month that wind power was curtailed. Likewise, the blue and red bars show the amount of energy that was curtailed in heavy and light load hours. The horizontal and vertical scales in all graphs are equivalent. It can be seen from this series of graphs that the number of hours with wind power curtailment (green and orange bars) were greatest when there was little installed wind capacity, which also corresponds to little diversity in wind power development. However, the number of hours when wind power was curtailed was not necessarily related to the amount of energy that was being curtailed. The occurrences of wind power curtailment with little associated loss in wind energy, could have been because the model did not have to hold following reserves when wind power was being curtailed. So if a negligible amount of wind energy was available in a given hour, curtailment could have been enacted to free up following reserve capacity that then could have been sold on the market.  79  Figure 21: Daily and Seasonal Patterns of Curtailment  80  The quantity of wind energy curtailed (blue and red bars) increased as the wind power penetration increased (moving from left to right in the figure) and as the simulated years became wetter (moving from top to bottom in the figure). When viewed as a percentage of total annual wind energy, these energy curtailments were roughly constant throughout wind power penetrations in each representative water year. These ranges were 0.5% to 2.5% (see Figure 20). The annual distribution of wind energy curtailment showed a definite bias towards curtailment during the spring freshet period in late May, June, and July. During the freshet in the Pacific Northwest, there is a relatively large amount of energy available from run-of-river hydropower plants, which depresses electricity prices. This is also a period of the year when the temperatures have risen from their winter lows, and so there is less electrical heat demand. The hydropower facilities in B.C. have a relatively large storage capacity and historically, the optimal operating decision during this period is to store as much water as possible in the reservoirs. Curtailing wind reduced its reserve capacity constraints on hydropower plant production, so they could be run at lower output levels. This allowed import of relatively cheap electricity, and storage of freshet inflows that could be used later in the year during times with higher electricity prices. The diurnal (HLH vs. LLH) division of wind power curtailment also showed a definite tendency to curtail during LLHs. The diurnal power system dynamics during LLH vs. HLH are similar to those observed in the freshet vs. the winter – LLHs are at night when electrical loads and electricity prices are relatively low. When looking at the diurnal distributions, it should be noted that the HLH period was twice as long (16 hours) as the LLH period (8 hours). Even though there were half as many hours in the LLH period, the amount of wind energy curtailed in these hours was nearly always greater than HLH energy curtailment. 81  5.6 Summary Based on the input parameters previously outlined, including energy market and generation capacity market prices, intertie capacities, available wind energy, river inflows, and load, an optimal value of wind energy was determined. Allowing wind power curtailment purely for economic reasons also resulted in an incremental increase in wind energy value. Further analysis of the simulation and optimization model outputs showed potential effects of adding wind power and the further incremental effects of allowing wind power curtailment. In these model runs, existing flexibility in the large storage reservoirs was used by the model to shift the wind energy to more economically attractive time periods. The Columbia River Treaty and ice flow constraints were visible in the forebay outputs, and system inflexibility during the spring freshet typically split the shift of wind energy to either the October through March or March through September time frames. Further analysis of hydroelectric system efficiency, energy generation, and generation capacity reserves yielded limited useful information. Among the trends observed, most were intuitive, such as the decrease in available, surplus generation reserve capacity (because of wind reserves) and the import/export patterns (greatly influenced by the additional wind energy that must be exported). The temporal analysis of wind power curtailment found that it was concentrated in the LLH periods during the spring freshet period. These are periods of low domestic electrical load, low market energy prices, high energy imports, and high local river inflows. The low domestic load and high local inflows reduce the flexibility of the modeled hydroelectric system, and the low market energy prices further reduce the value of wind energy exports during this period. These factors combined to create a situation where the model determined that the optimal operational decision during some hours would be to curtail available wind energy, and this would yield more value to the entire 82  system than what would be lost by the corresponding energy curtailment. This observation supports the hypotheses of hydropower system planners at BC Hydro.  83  6 Discussion and conclusions This section summarizes how the independent variables that were manipulated affected the value of wind energy and curtailment. Then it presents discussion on how the results of this research might be applied to other real world situations. Finally, assumptions that were made in this research, but that might provide a more complete understanding of wind power curtailment if they too were manipulated or incorporated, are outlined.  6.1 Summary As was shown in the analysis of the reservoir/forebay stages, integrating wind energy into the modeled hydroelectric system allowed the system greater flexibility to shift the integrated wind energy to the most opportune export time periods. This flexibility is important for wind power producers within the study area that might sign power purchase agreements with the hydroelectric utility, as well as to customers outside of the study area that are accessible via the transmission interties. At the river and system levels, the efficiencies of the hydropower plants were consistently affected by the shifts in generation patterns that wind power integration caused and allowed. Practical operating experience would be necessary to confirm these findings. The value of wind energy, when normalized to the average annual market energy prices, did not appear to have a noticeable correlation to the type of water year. However, the value of wind power curtailment did appear to increase in wetter years. The value of wind energy consistently declined as more wind energy was integrated (increased wind energy penetration). This decline ranged from 5% to 10% when comparing the value of wind energy at 3% to 12% penetration, respectively. The value of wind power curtailment was highest at the 3% wind penetration level and initially declined by about 1% when the penetration level was increased 84  to 6%. However, as the wind energy penetration levels were further increased to 9% and 12%, the value of wind power curtailment ceased to decrease and remained approximately constant. The diversity of wind farms producing the wind energy increased in each successive penetration level, but the initial added diversity between the 3% and 6% penetration level appeared to be the most important increment when looking at the value of wind power curtailment. As the intertie capacities were reduced, the value of wind energy appeared to vary based on whether the hydroelectric utility was a net importer or exporter of energy. For the dry and average years (when the hydroelectric utility was a net exporter of energy), the value of wind energy decreased as the intertie capacities were reduced. For the wet year (when the hydroelectric utility was a net importer of energy because it was refilling its reservoirs, even with the additional wind energy), reduced intertie capacities increased the value of wind energy. The value of wind power curtailment consistently decreased as the intertie capacities were reduced, presumably because access to market-priced energy was also reduced. Because reducing the intertie capacities also represents a regional market depth decrease, the trend of reduced curtailment value could possibly become evident as wind power development in the electrically connected region increases and other utilities need to import less energy. The other independent variable that was manipulated was the contractual price for wind energy. As was mentioned in the methodology section, this situation explored the effects when the hydroelectric utility could choose to purchase, or not purchase, any portion of the available wind energy. In many situations, and particularly the BC Hydro case study, this is purely hypothetical. However, this approach could be useful in other contexts, including providing justification for a negotiated contract price for wind energy, or to explore the financial viability of a utility investing in its own wind power generation facilities. 85  When the hydroelectric utility was required to integrate all available wind energy, this contractual wind energy price simply shifted the additional income from these energy sales between itself and the wind power producer. When the hydroelectric utility was not required to purchase all available wind energy, the variations in contractual price plotted with curtailment yielded a curve approximating what portion of available wind energy was profitable at a given wind energy price (see Figure 22 in Section 6.2). For all of the modeling runs, the vast majority of curtailed energy was in the freshet period, particularly in late May, June, and July. The percentage of annual wind energy that was curtailed was largest at the 3% wind penetration level during average and wet years, and was never greater than 2.5% of total annual wind energy. This percentage decreased at the 6% penetration level when diversity was greater. However, the percentage of total annual wind energy remained steady or increased slightly as penetration levels continued to increase up to 12% penetration. As intertie capacity/market depth decreased, so did the percentage of curtailed annual wind energy. Therefore, the greatest percentage of curtailed wind energy was at the lowest penetration and diversity levels, during wetter years, with highest intertie capacities. The cases with the least percentage of wind energy curtailed were dry years, with greater wind energy penetration levels coming from more diverse portfolios of wind projects, and at lower intertie capacities. The value added to the available wind energy, by wind power curtailment, followed the same trends. When there were larger amounts of wind energy curtailed, the value that curtailment added was greatest, and when smaller quantities of wind energy were curtailed, the added value of wind power curtailment was smaller. Like the percentage of curtailed wind energy, the value of wind power curtailment appeared to level off after the initial added diversity and energy between the 3% and 6% penetration cases, and even slightly increased when the wind penetration reached 12%. 86  6.2 Another application In order to allow the purchaser of the wind energy to curtail it, the appurtenances required to enact curtailment, as well as the contractual flexibility to allow it, must be incorporated early in both the negotiation and design phases of a project. This puts regions with young wind power development markets in the best position to more closely investigate the opportunities that wind power curtailment might present. Some areas, such as northern Germany, build curtailment allowances into their new contracts, but have not always done so. As a result, they experience the added difficulty of how to distribute curtailment instructions between eligible wind power developments, as well as the lack of these provisions in older wind power contracts, before it was widely recognized that curtailment would be necessary. Figure 22 depicts the cases where the optimization model only requires the hydroelectric utility to pay for the wind energy that it integrates. The contractual price that the hydroelectric utility could pay for each unit of wind energy is plotted across the x-axis, while the extra income per unit of total available wind energy is on the left y-axis (red lines). The right y-axis (purple lines) shows what percent of available wind energy the utility would curtail, or choose not to purchase, at the specified contract price. If this analysis was extended to include more contractual energy prices, the downward bend in the red curves would indicate the highest contract price that the hydroelectric utility should be willing to pay in energy purchase agreements. It can be seen that very little energy is curtailed until the wind energy contract price (x-axis) equals the wind energy value (y-axis). Once the contract price exceeds the value, the optimization model began to curtail wind energy that had less value than the contract price, and curtailment increased sharply.  87  Figure 22: Wind Energy Value vs. Wind Energy Contract Price  From the perspective of a wind power developer, they could look at the red line that represents the current wind penetration level in their region to see what wind energy might be worth to a purchasing entity. As the penetration level increases, the value of the extra energy decreases, as would the price that a utility should be willing to pay for it.  6.3 Further work There are almost endless ways to extend the analysis presented in this research and to tailor it to meet specific needs. An approach has been laid out herein, and a more robust investigation of water years, wind energy penetration levels, wind power diversity scenarios, intertie limits, or contractual arrangements and energy prices could quickly be incorporated into the modeling the way it is currently designed. Likewise, other parameters in the model such as market energy prices, ancillary service market prices (specifically generation capacity reserves), reserve generation capacity levels, electrical load,  88  other generation sources, generator outages schedules, and forebay constraints could quite easily be varied because they are simply time series data sets in the model runs. Beyond these sensitivity analyses, the modeling methodology could be expanded in several ways to more robustly model other aspects of power and energy trading, as well as the planning and operation of the hydroelectric system. For example, electrical load and wind power forecasts for the next few hours and days are important in both energy trading and system planning. This modeling did not incorporate forecasts and assumed perfect foreknowledge. Realistically, the sale of energy and hydroelectric system planning happens in the days and hours leading up to the current operating hour, and therefore the optimal dispatch plans and energy exchange schedules that this modeling optimizes are typically unachievable. Models that incorporate forecasts can have one or more stochastic variables representing these load, price, and/or wind energy forecasts, and can facilitate trading and planning horizons specific to the markets the modeled utility uses. All of these forecasts have errors associated with them, which are reduced leading up to the operating hour, and that could be stochastically modeled as well. On a related track, a simplification made in this research that could be further investigated is the effect that available wind energy, and new clean, green, and renewable sources of energy in general, can have on energy market prices. In this modeling, an energy price depression factor was incorporated based on the existing experience of market traders. This depression factor should presumably increase as the installed wind capacity in the region increases, similar to the effects seen in regions with large proportions of run-of-river hydroelectric generation capacity during the spring freshet. It is also possible that the value of ancillary services, specifically reserve generation capacity, could be affected by increasing variable generation resources in the region, as well as demand side 89  management or smart grids. Hydroelectric generating capacity is particularly valuable because of its fast response times, ramping abilities, and relatively broad efficient operating ranges. This supports the need to properly value these commodities in existing and new market structures. Another market modification that could be investigated is shortening the time step below one hour. This research used an one-hour time step in accordance with the market in the Pacific Northwest, however several other markets in North America, and some in Europe, operate at a time step as short as ten minutes. This could not only affect the energy and capacity market transactions, but also the quantity of reserve generation capacity that needs to be set aside to handle intra-time step variations in load and variable generation, such as wind. There are numerous ways to calculate reserve generation capacity requirements, and these have a definite effect on the wind power curtailment illustrated in this research. An assumption laid out earlier was that the “following” portion of the reserve capacity necessary to manage wind’s variability was reduced as wind energy was curtailed. In practice, there are several ways in which a wind power plant can be curtailed, each with different effects to its electrical generation variability. Some of the latest developments in wind turbine technology allow them to so precisely control their output that they might even be able to provide some of these ancillary services when curtailed. An addition to the simulation and optimization model that could have significant impacts on wind power curtailment would be modeling the electrical transmission system. In this research, the only transmission limits that were modeled were those between British Columbia and the U.S. market. However, most transmission systems have many bottlenecks, also known as pinch points. As wind power resources far from load centers are developed, they could end up competing with other energy resources for transmission capacity, and limit the operational flexibility of the rest of the generation system. 90  Finally, after analyzing the results of this research, it became evident that the ability of the modeled hydroelectric system to shift surplus energy to the most opportune times was important. Traditionally, this simulation and optimization model has been run for individual water years (October through September in Canada) with boundary conditions coming from longer-term hydraulic and energy balance models. However, when additional wind energy is added in successive runs, these longer-term models have not considered this, and the surplus energy must be exported to the market. By only modeling one water year at a time, any extra water that wind energy has allowed to be stored, must be used to generate energy exports by the end of the study in September. Realistically, this surplus stored water could be carried over further into the fall and winter months, or even longer in dry years, before the reservoir levels must be reduced to make room for spring snowmelt and runoff. Likewise, surplus energy from wet years could be used to keep reservoir levels higher so that more energy would be available during later years that may be drier, with less available hydroelectric energy. Therefore, modeling runs with durations of several years, or with carry-over storage capacity, could provide a more complete assessment of the capabilities and values of available reservoir storage flexibility and volume.  91  References Alberta Electric System Operator (AESO), “Wind Power Integration,” Alberta, 2009. Available on the internet at: http://www.aeso.ca/gridoperations/13902.html. Alegria, I.M., J. Andreu, J.L. Martin, P. Ibanez, J.L. Villate and H. Camblong, “Connection requirements for wind farms: A survey on technical requirements and regulation,” Renewable and Sustainable Energy Reviews, 11, 1858-1872, 2007. American Wind Energy Association (AWEA), "AWEA 2nd Quarter 2009 Market Report," 2009. Available on the internet at: http://www.awea.org/publications/reports/3Q08.pdf. Angarita, J.M. and J.G. Usaola, “Combining hydro-generation and wind energy Biddings and operation on electricity spot markets,” Electric Power Systems Research, 77, 393-400, 2007. Bathurst G.N. and G. Strbac, “Value of combining energy storage and wind in short-term energy and balancing markets,” Electric Power Systems Research, 67, 1-8, 2003. BC Hydro, “F2006 Open Call for Power Report on the CFT process conducted by BC Hydro,” British Columbia, Canada, 2006. Available on the internet at: http://www.bchydro.com/planning_regulatory/acquiring_power/open_call_for_power/cft_ results.html. BC Hydro, “2008 Long-Term Acquisition Plan,” British Columbia, Canada, 2008a. Available on the internet at: http://www.bchydro.com/planning_regulatory/integrated_electricity_planning/2008_ltap.h tml. BC Hydro, “Clean Power Call,” British Columbia, Canada, 2008b. Available on the internet at: http://www.bchydro.com/planning_regulatory/acquiring_power/clean_power_call.html.  92  BC Transmission Corporation (BCTC), “System Operating Order 1T-61 BCTC Operating Reserve and NWPP Reserve Sharing Procedures,” British Columbia, Canada, 2007. Available on the internet at: http://www.bctc.com/transmission_system/system_operating_orders/. Benitez, P.C., L.E. Dragulescu and G.C. van Kooten, “The Economics of Wind Power with Energy Storage,” University of Victoria Working Paper 2006-2, Victoria, British Columbia, 2006. Bonneville Power Administration (BPA), “Balancing act: BPA grid responds to huge influx of wind power,” Portland, Oregon, 2008. Burges, K. and J. Twele, “Power systems operation with high penetration of renewable energy - the German case,” Ecofys, Berlin, Germany, 2005. Currie, R.A.F., G.W. Ault, R.W. Fordyce, D.F. MacLeman, M. Smith and J.R. McDonald, “Actively managing Wind Farm Power Output,” IEEE Transactions on Power Systems, 23(3), 15231524, 2008. DNV Global Energy Concepts, Inc. (DNV-GEC), “BC Hydro Wind Data Study,” Washington, 2009. Available on the internet at: http://www.bchydro.com/planning_regulatory/energy_technologies/wind_energy/wind_da ta_study.html. Electrotek Concepts, “Characterizing the Impacts of Significant Wind Generation Facilities on Bulk Power System Operations Planning,” Virginia, 2003. Garcia-Gonzalez, J., R.M. Ruiz de la Muela, L.M. Santos and A.M. Gonzalez, “Stochastic Joint Optimization of Wind Generation and Pumped-Storage Units in an Electricity Market,” IEEE Transactions on Power Systems, 23(2),460-468, 2008. Gardner, P., H. Snodin, A. Higgins and S. McGoldrick, “The Impacts of Increased Levels of Wind Penetration on the Electricity Systems of the Republic of Ireland and Northern Ireland: Final Report,” Garrad Hassan and Partners Ltd., Glasgow, Scotland, 2003. 93  Global Wind Energy Council (GWEC), “Global Wind Report 2008,” second edition, Brussels, Belgium, 2009. Available on the internet at: http://www.gwec.net/. Holttinen, H., K. Vogstad, A. Botterud and R. Hirvonen, “Effects of Large-Scale Wind Production on the Nordic Electricity Market,” Proceedings of European Wind Energy Conference, Copenhagen, Denmark, July 2-6, 2001. Hudson, R., B. Kirby and Y. Wan, “Regulation Requirements for Wind Generation Facilities,” American Wind Energy Association, 2001. International Energy Association (IEA), “Key World Energy Statistics,” 2009. Available on the internet at: http://www.iea.org/Textbase/stats/index.asp. Kabouris, J. and C. Vournas, “Application of Interruptible Contracts to Increase Wind-Power Penetration in Congested Areas,” IEEE Transactions on Power Systems, 19(3), 1642-1649, 2004. Kirby, B. and M. Milligan, “Facilitating Wind Development: The Importance of Electric Industry Structure,” Electricity Journal, 21(3), 40-54, 2008. Korpås, M., “Distributed Energy Systems with Wind Power and Energy Storage,” doctoral thesis, Trondheim, Norway, 2004. Korpås, M., J.O.G. Tande, K. Uhlen, E.S. Huse and T. Gjengedal, “Planning and operation of large wind farms in areas with limited power transfer capacity,” Proceedings of 2006 European Wind Energy Conference, 2006. Matevosyan, J., “Wind power integration in power systems with transmission bottlenecks,” doctoral thesis, Stockholm, Sweden, 2006. Miller, N.W., K. Clark and G.A. Jordan, “Planning and Operating Power Systems with a High Percentage of Wind Generation: Taking Advantage of the Latest Wind Plant Controls,” WindPower 2007 Conference & Exhibition, Los Angeles, California, June 2007. 94  Ministry of Energy, Mines and Petroleum Resources (MoEMPR), “The BC Energy Plan: A Vision for Clean Energy Leadership,” Victoria, British Columbia, 2007. Available on the internet at: www.energyplan.gov.bc.ca. Pálsson, M., T. Toftevaag, K. Uhlen and J.O. Tande, “Control Concepts to Enable Increased Wind Power Penetration,” Proceedings of 2003 IEEE Power Engineering Society General Meeting, 3, 1984-1990, 2003. Shawwash, Z., “A Decision Support System for Real-time Hydropower Scheduling in a Competitive Power Market Environment,” doctoral thesis, Vancouver, University of British Columbia, 2000. Son, Y.S., “Dispatchable Wind Power Valuation in Texas,” Proceedings of 2005 IEEE Power Engineering Society General Meeting, 1944-1947, 2005. Sørensen, P., I. Norheim, P. Meibom and K. Uhlen, “Simulations of wind power integration with complementary power system planning tools,” Electric Power Systems Research, 78, 10691079, 2008. Torre, M., T. Domínguez, G. Juberías, E. Prieto and O. Alonso, “Operation of a Power System with large integration of renewable energies,” 7th International Workshop on Large-Scale Integration of Wind Power into Power Systems, Madrid, Spain, May 26-27, 2008. Vogstad; K., “Utilising the complementary characteristics of wind power and hydropower through coordinated hydro production scheduling using the EMPS model,” Norwegian Wind Power Conference, Trondheim, Norway, 2000. Wiser, R. and M. Bolinger, “Annual Report on US Wind Power Installation, Cost, and Performance Trends: 2007,” Lawrence Berkeley National Laboratory, Berkeley, California, 2008a.  95  Wiser, R. and G. Barbose, “Renewables Portfolio Standards in the United States: A Status Report with Data Through 2007,” Lawrence Berkeley National Laboratory, Berkeley, California, 2008b.  96  

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.24.1-0063119/manifest

Comment

Related Items