Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The assessment of on-board clean hybrid energy storage systems for railway locomotives and multiple units Hegazi, Mohamed 2016

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

Item Metadata

Download

Media
24-ubc_2017_february_hegazi_mohamed.pdf [ 4.34MB ]
Metadata
JSON: 24-1.0320830.json
JSON-LD: 24-1.0320830-ld.json
RDF/XML (Pretty): 24-1.0320830-rdf.xml
RDF/JSON: 24-1.0320830-rdf.json
Turtle: 24-1.0320830-turtle.txt
N-Triples: 24-1.0320830-rdf-ntriples.txt
Original Record: 24-1.0320830-source.json
Full Text
24-1.0320830-fulltext.txt
Citation
24-1.0320830.ris

Full Text

The Assessment of On-Board CleanHybrid Energy Storage Systems ForRailway Locomotives and MultipleUnitsbyMohamed HegaziBEng, The University of Nottingham, 2013A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF APPLIED SCIENCEinTHE COLLEGE OF GRADUATE STUDIES(Electrical Engineering)THE UNIVERSITY OF BRITISH COLUMBIA(Okanagan)October 2016c© Mohamed Hegazi, 2016The undersigned certify that they have read, and recommend to the College of GraduateStudies for acceptance, a thesis entitled: The Assessment of On-Board Clean HybridEnergy Storage Systems For Railway Locomotives and Multiple Units submittedby Mohamed Hegazi in partial fulfilment of the requirements of the degree of Master ofApplied ScienceLo¨ıc Markley, Applied Science/School of EngineeringSupervisor, Professor (please print name and faculty/school above the line)Gordon Lovegrove, Applied Science/School of EngineeringCo-Supervisor, Professor (please print name and faculty/school above the line)Wilson Eberle, Applied Science/School of EngineeringSupervisory Committee Member, Professor (please print name and faculty/school above the line)Rudolf Seethaler, Applied Science/School of EngineeringUniversity Examiner, Professor (please print name and faculty/school above the line)26 October 2016(Date Submitted to Grad Studies)iiAbstractBatteries, supercapacitors, and hydrogen fuel cells are energy storage devices that have noemissions at the point of use. The idea of powering railway locomotives using these devices is onethat could, theoretically, eliminate emissions from the railway sector. The motivation behindthe research work presented in this thesis is to assess the technical feasibility of employingbatteries, supercapacitors, and hydrogen fuel cells in a railway vehicle. This is meant to serveas reference to future work regarding the cost-benefit analysis, well-to-wheel emissions analysis,and life-cycle assessment of railway vehicles that employ these power sources.In this thesis, the application of on-board clean energy storage systems to railway vehicleswere studied. Simulation models for battery/supercapacitor and hydrogen fuel cell/batteryhybrid powertrains were developed in Simulink. These models were then used to conductsimulations for two train trips. The first trip selected was the 14 km Trehafod to Treherbertroute, on which the British Class 150 diesel motive unit operates. The second trip was the 432km London to Newcastle trip, on which the Intercity 125 train operates. Since no data regardingfreight trains on freight tracks could be obtained, only passenger trains were simulated. Theconclusions made at the end of this thesis could potentially apply to freight trains as well.Based on the case studies considered, it was found out that railway systems are very wellsuited to run on on-board clean energy storage systems from an energy consumption point ofview. Although being slower responding power sources, hydrogen fuel cells proved to be capableof handling dynamic load changes in railway systems to a great extent but still required theassistance of a faster acting power source. Despite having a significantly lower electrochemicalefficiency, employing hydrogen fuel cells resulted in increasing the range of travel without re-fueling/recharging due to the high energy density of hydrogen. Lithium ion batteries provedto be very capable in handling all the required transient power demand. In regeneration,supercapacitors outperformed lithium-ion batteries and reduced the need for frictional brakes.iiPrefaceThe work presented in this thesis was made possible through two consecutive grants byTransport Canada under the umbrella of their Clean Rail Grant program.A technical report containing a portion of Chapter 2 was prepared for Transportation De-velopment Centre of Transport Canada.A version of Chapter 2 was was published in a technical poster at the 2014 AREMA AnnualConference in Chicago, Illinois.The work presented in Section 3.3 was published in a technical poster at the 2015 AREMAAnnual Conference in Minneapolis, Minnesota.The results and associated methods in Subsection 4.2.3 were published in the 2016 CanadianSociety of Mechanical Engineers International Congress (CSME2016).The work presented in Subsection 4.2.4 was published in a technical poster at the 2016AREMA Annual Conference in Orlando, Florida.iiiTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiAcronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xivAcknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xviDedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xviiChapter 1: Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Research Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41.2 Thesis Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5Chapter 2: Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62.1 Diesel-Electric Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62.2 Catenary-Electric Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82.3 Hybrid Energy Storage Technology . . . . . . . . . . . . . . . . . . . . . . . . . 102.4 The Science of Hybridization . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26Chapter 3: Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28ivTABLE OF CONTENTS3.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283.2 Longitudinal Dynamics of Trains . . . . . . . . . . . . . . . . . . . . . . . . . . 303.3 Offline Trajectory Planning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333.3.1 Preprocessing Gradient Data . . . . . . . . . . . . . . . . . . . . . . . . 373.4 Powertrain Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393.4.1 Buck Converter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423.4.2 Boost Converter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 503.4.3 Bidirectional Buck and Boost Converter . . . . . . . . . . . . . . . . . . 563.4.4 Online Deterministic State Machine Control . . . . . . . . . . . . . . . . 573.4.5 Combined Simulink Model . . . . . . . . . . . . . . . . . . . . . . . . . . 593.5 Battery - Supercapacitor Parallel Hybrid Powertrain Description . . . . . . . . 603.6 Fuel cell - Battery Series Hybrid Powertrain Description . . . . . . . . . . . . . 643.7 Numerical Optimization Process . . . . . . . . . . . . . . . . . . . . . . . . . . 66Chapter 4: Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 674.1 British Class 150: Trehafod to Treherbert . . . . . . . . . . . . . . . . . . . . . 684.1.1 Train Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 684.1.2 Route Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 694.1.3 Battery - Supercapacitor Parallel Hybrid . . . . . . . . . . . . . . . . . 704.1.4 Fuel Cell - Battery Series Hybrid . . . . . . . . . . . . . . . . . . . . . . 774.2 Intercity 125: King’s Cross to Newcastle . . . . . . . . . . . . . . . . . . . . . . 904.2.1 Train Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 904.2.2 Route Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 904.2.3 Battery - Supercapacitor Parallel Hybrid . . . . . . . . . . . . . . . . . 924.2.4 Fuel Cell - Battery Series Hybrid . . . . . . . . . . . . . . . . . . . . . . 97Chapter 5: Conclusions and Recommendations . . . . . . . . . . . . . . . . . . . 105Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115Appendix A: Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116vList of TablesTable 2.1 The specifications of the hydrogen tank. . . . . . . . . . . . . . . . . . . 21Table 3.1 Buck converter specifications. . . . . . . . . . . . . . . . . . . . . . . . 44Table 3.2 This table contains the open-loop circuit specifications for the boost con-verter used in this research. . . . . . . . . . . . . . . . . . . . . . . . . . 53Table 4.1 The specifications of British Class 150 DMU. . . . . . . . . . . . . . . . 68Table 4.2 The specifications of the Intercity 125 train as a single rigid body. . . . 90Table A.1 The specifications of Panasonic’s UPF454261 Lithium-ion battery. . . . 116Table A.2 The specifications of Maxwell’s BCAP3400 supercapacitor. . . . . . . . 116Table A.3 The specifications of the 100 kW Honda FCX PEMFC model used in thisstudy as obtained from [33]. . . . . . . . . . . . . . . . . . . . . . . . . 117Table A.4 The specifications of British Class 43 Locomotive. . . . . . . . . . . . . . 117Table A.5 The gradient profile of the Trehafod to Treherbert trip. . . . . . . . . . 118Table A.6 The speed limit profile of the Trehafod to Treherbert trip. . . . . . . . . 119viList of FiguresFigure 1.1 The gravimetric and volumetric energy densities of SC, lithium-ion bat-teries, diesel fuel, and hydrogen gas compressed at 350 bar. . . . . . . . 3Figure 2.1 The main components of a diesel-electric powertrain. . . . . . . . . . . 7Figure 2.2 The operation of continuously electrified railroads. . . . . . . . . . . . . 8Figure 2.3 The operation of discontinuously electrified railroads [24]. . . . . . . . 12Figure 2.4 The charge/discharge curve of a typical battery relating the terminalvoltage of the battery to its SOC [31]. Where Qnom and Qexp are thenominal battery charge and exponential battery charge respectively. . . 16Figure 2.5 Voltage response of a FC stack during a positive step change in load [34]. 19Figure 2.6 The polarization curve of the 100 kW PEMFC presented in [33] with itsparameters detailed in the appendix as generated by the SimPowerSys-tems FC model developed by the authors in [32]. . . . . . . . . . . . . . 20Figure 2.7 Active Series Hybrid Powertrain Architecture. . . . . . . . . . . . . . . 23Figure 2.8 Active Parallel Hybrid Powertrain Architectures. . . . . . . . . . . . . . 23Figure 3.1 This flowchart illustrates the overall process used to conduct a hybridpowertrain simulation for a specific trip. The first phase of the simu-lation combines a forward and a backward velocity profiles producinga fast-as-possible velocity profile. The second phase of the simulationtakes the resultant velocity profile as an input. Fg, and FR refer to thegravitational force and the rolling resistance force respectively. . . . . . 29Figure 3.2 The relationship between a train’s applied power, resultant tractive effortand its velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31Figure 3.3 Forces acting on a train . . . . . . . . . . . . . . . . . . . . . . . . . . . 33viiLIST OF FIGURESFigure 3.4 Railway vehicle trajectory profile, and the maximum allowable velocityprofile are shown. The figure illustrates the three possible modes ofoperation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34Figure 3.5 This flowchart outlines the trajectory planning algorithm used in thefirst phase of simulations in this study. . . . . . . . . . . . . . . . . . . 36Figure 3.6 This figure illustrates the necessary preprocessing steps for the discretegradient data points to be usable in Simulink. . . . . . . . . . . . . . . 38Figure 3.7 The steps required to design, model and control a simulated power elec-tronic converter model in SimPowerSystems. . . . . . . . . . . . . . . . 42Figure 3.8 Circuit layout of a non-isolated buck converter with a purely resistive load. 42Figure 3.9 Output voltage as a function of switching duty cycle for a buck converterwith an input voltage of 1500 V. . . . . . . . . . . . . . . . . . . . . . . 43Figure 3.10 Open-loop output voltage Vout at the operating point shown in Figure3.9 with a ripple amount ∆vout of 1 V. The inset plot highlights theripple component. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45Figure 3.11 Open-loop frequency response of the buck converter given the parame-ters in Table 3.1, and the open-loop frequency response of the linearlyapproximated system. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47Figure 3.12 The open loop output voltage response of the buck converter in compar-ison to that of the linearized model. . . . . . . . . . . . . . . . . . . . . 48Figure 3.13 Closed-loop PWM control of the output voltage in a buck converter usinga PID controller. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49Figure 3.14 Closed-loop response of PID controlled buck converter. . . . . . . . . . 50Figure 3.15 Circuit layout of a non-isolated boost converter with a purely resistiveload. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50Figure 3.16 Output voltage as a function of switching duty cycle. The operatingpoint at 50% duty cycle is highlighted. . . . . . . . . . . . . . . . . . . 51Figure 3.17 The open loop output voltage and input current of a simplified boostconverter given the circuit parameters in Table 3.2. . . . . . . . . . . . 52viiiLIST OF FIGURESFigure 3.18 Open-loop frequency response of the boost converter given the parame-ters in Table 3.2, and the open-loop frequency response of the linearlyapproximated system. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54Figure 3.19 Open loop input current transient response at 50% duty cycle. . . . . . 55Figure 3.20 Closed-loop PWM control of the input current in a boost converter usinga PID controller. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55Figure 3.21 Closed loop input current transient response assuming a 500 A referenceinput current. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56Figure 3.22 The circuitry of the bidirectional buck-boost converter which interfacesthe energy source with the traction machine. . . . . . . . . . . . . . . . 57Figure 3.23 The impact of a two-state state machine deterministic controller on thevehicle velocity as compared to the provided reference velocity profile. . 59Figure 3.24 BSC parallel hybrid powertrain architecture. . . . . . . . . . . . . . . . 61Figure 3.25 The combined battery / SC parallel hybrid powertrain Simulink model. 63Figure 3.26 FCB series hybrid powertrain architecture. . . . . . . . . . . . . . . . . 64Figure 3.27 The combined FCB series hybrid powertrain Simulink model. . . . . . . 65Figure 3.28 A graphical representation of the optimization problem that appearswhen deciding on the hybridization mix. . . . . . . . . . . . . . . . . . 66Figure 4.1 Class 150 train consists of two Class 150 DMU. . . . . . . . . . . . . . 68Figure 4.2 Trehafod to Treherbert altitude profile. . . . . . . . . . . . . . . . . . . 69Figure 4.3 British Class 150 Train Velocity Profile: Trehafod to Treherbert. The in-set plot highlights the perturbation that occurs when the system changesthe mode of operation from Motoring mode to Regenerative braking mode. 70Figure 4.4 Trehafod to Treherbert trip power profile. The battery bank mass ineach DMU was 500 kg, and the SC bank was 1.5 tonnes. The inset plothighlights how the magnitude of the frictional braking power compareswith the magnitude of the regenerated power as absorbed by the SC orthe battery bank. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71ixLIST OF FIGURESFigure 4.5 Trip energy consumption assuming regenerative breaking versus withoutregenerative braking. The battery bank mass in each DMU was 500 kg,and the SC bank was 1.5 tonnes. . . . . . . . . . . . . . . . . . . . . . . 72Figure 4.6 The battery bank’s SOC, and the SC SOC as a percentage change. Thebattery bank mass in each DMU was 500 kg, and the SC bank was 1.5tonnes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73Figure 4.7 Powertrain efficiency excluding the FC stack efficiency. The averagepowertrain efficiency was 91%. The battery bank mass in each DMUwas 500 kg, and the SC bank was 3 tonnes. . . . . . . . . . . . . . . . . 74Figure 4.8 A two dimensional illustration of the ESS sizing optimization problemthat highlights the cases considered for simulation. The ESS in this studyis a BSC hybrid, and the physical constraints are for a British Class 150DMU. The dotted circles represent the ESS cases selected for simulation. 75Figure 4.9 Maximum range of operation in kilometers travelled without rechargingon 70% of the battery bank’s charge as a function of the on-board batterybank mass for different SC masses (legend). . . . . . . . . . . . . . . . . 76Figure 4.10 Energy regenerated as a percentage of net energy consumed as a functionof the on-board ESS mix for a BSC hybrid British Class 150 DMU. . . 76Figure 4.11 Trip power profile. The average FC power was limited to 80 kW in eachDMU and the battery bank mass in each DMU was 3000 kg. . . . . . . 78Figure 4.12 The battery bank’s SOC, and the hydrogen consumption as a percentagechange. The average FC power was limited to 80 kW in each DMU, thebattery bank mass in each DMU was 3000 kg, and the hydrogen storagewas limited to 3 tanks. . . . . . . . . . . . . . . . . . . . . . . . . . . . 79Figure 4.13 Powertrain efficiency excluding the FC stack efficiency. The averagepowertrain efficiency was 84.37%. The average FC power was limited to80 kW in each DMU and the battery bank mass in each DMU was 3000kg. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79xLIST OF FIGURESFigure 4.14 Fuel cell instantaneous and average efficiencies. The average FC powerwas limited to 80 kW in each DMU giving an average efficiency of 64%,the battery bank mass in each DMU was 3000 kg, and the hydrogenstorage was limited to 3 tanks. . . . . . . . . . . . . . . . . . . . . . . . 80Figure 4.15 Hydrogen consumption for a per DMU storage of 1 tank, 3 tanks, 5tanks, 7 tanks and 10 tanks after 15 hours of operation. . . . . . . . . . 81Figure 4.16 Battery SOC for a per DMU battery storage of 100 kg, 500 kg, 1 tonne,1.5 tonnes, 2 tonnes, 2.5 tonnes, 3 tonnes, and 3.5 tonnes for 15 hours ofoperation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82Figure 4.17 A two dimensional illustration of the ESS sizing optimization problemthat highlights the cases considered for simulation. The ESS in this studyis a FCB hybrid, and the physical constraints are for a British Class 150DMU. The dotted circles represent the ESS cases selected for simulation. 83Figure 4.18 Maximum number of hours of operation without refueling as a functionof the on-board hydrogen storage mass for a FCB hybrid British Class150 motive unit for different battery bank masses (legend). . . . . . . . 84Figure 4.19 The battery bank’s rate of loss of charge as a function of its mass fordifferent hydrogen tank numbers assuming that each motive unit had a80 kW FC system as the prime mover. . . . . . . . . . . . . . . . . . . 85Figure 4.20 Energy regenerated as a percentage of net energy consumed as a functionof the on-board ESS mix. . . . . . . . . . . . . . . . . . . . . . . . . . . 85Figure 4.21 Battery bank SOC change per trip sensitivity to bank’s mass (legend)and FC system power limit. . . . . . . . . . . . . . . . . . . . . . . . . 86Figure 4.22 The sensitivity of the mean trip FC efficiency to changes in FC powerand battery bank mass (legend). . . . . . . . . . . . . . . . . . . . . . . 88Figure 4.23 The sensitivity of the hourly rate of hydrogen consumption to changesin FC power and battery bank mass (legend). . . . . . . . . . . . . . . 89Figure 4.24 The sensitivity of the regenerated energy as a percentage of the netenergy to changes in FC power and battery bank mass. . . . . . . . . . 89xiLIST OF FIGURESFigure 4.25 Intercity 125 train consist: two Class 43 locomotives hauling 8 Mark 3carriages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90Figure 4.26 Intercity 125 Train Velocity Profile: London’s King’s Cross to NewcastleRound-trip. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91Figure 4.27 Intercity 125 Train Velocity Profile: King’s Cross to Newcastle. Theaverage FC power was limited to 400 kW in each Class 43 locomotiveand the battery bank mass to 5 tonnes. The inset plot highlights theperturbation that occurs when the system changes the mode of operationfrom motoring mode to regenerative braking mode and back to motoringagain. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91Figure 4.28 Intercity 125 Train Velocity Profile: London‘s King’s Cross to NewcastleRound-trip. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92Figure 4.29 Intercity 125 Train Velocity Profile: London‘s King’s Cross to NewcastleRound-trip. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93Figure 4.30 Powertrain efficiency which averaged at 86%. . . . . . . . . . . . . . . . 94Figure 4.31 A two dimensional illustration of the ESS sizing optimization problemthat highlights the cases considered for simulation. The ESS in thisstudy is a BSC hybrid, and the physical constraints are for a BritishClass 43 locomotive. The dotted circles represent the ESS cases selectedfor simulation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95Figure 4.32 Maximum range of operation in kilometers traveled without rechargingon 70% of the battery bank’s charge as a function of the on-board batterybank mass for different SC bank sizes. . . . . . . . . . . . . . . . . . . . 96Figure 4.33 Regenerated energy as a percentage of net energy consumed as a functionof the on-board ESS mix. . . . . . . . . . . . . . . . . . . . . . . . . . . 96Figure 4.34 The power supplied by the FC system, and the battery bank. The av-erage FC power was limited to 800 kW in each Class 43 locomotive andthe battery bank mass to 5 tonnes. . . . . . . . . . . . . . . . . . . . . 97xiiLIST OF FIGURESFigure 4.35 The power supplied by the FC system, and the battery bank for a 10minute period starting at minute 144 to minute 154.. The average FCpower was limited to 800 kW in each Class 43 locomotive and the batterybank mass to 5 tonnes. . . . . . . . . . . . . . . . . . . . . . . . . . . . 98Figure 4.36 Powertrain efficiency excluding the FC stack efficiency which averagedat 86.521%. The average FC power was limited to 800 kW in each Class43 locomotive and the battery bank mass to 5 tonnes. . . . . . . . . . . 99Figure 4.37 Fuel cell instantaneous and average efficiencies. The average FC effi-ciency for the trip was 54.15%. The average FC power was limited to800 kW in each Class 43 locomotive and the battery bank mass to 5tonnes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100Figure 4.38 The battery bank’s SOC and the hydrogen fuel economy as a percentagechange. The average FC power was limited to 800 kW in each Class 43locomotive, the battery bank mass to 5 tonnes, and the hydrogen tanksto 86. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100Figure 4.39 The battery bank’s SOC and the hydrogen fuel consumption as a percent-age change for 3 consecutive trips. The average FC power was limitedto 800 kW in each British Class 43 locomotive, the battery bank massto 5 tonnes, and the hydrogen tanks to 86. . . . . . . . . . . . . . . . . 101Figure 4.40 A two dimensional illustration of the ESS sizing optimization problemthat highlights the cases considered for simulation. The ESS in thisstudy is a FCB hybrid, and the physical constraints are for a BritishClass 43 locomotive. The dotted circles represent the ESS cases selectedfor simulation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102Figure 4.41 Number of hours of continuous operation without refueling as a functionof the on-board hydrogen storage mass. . . . . . . . . . . . . . . . . . . 103Figure 4.42 The battery bank’s rate of loss of charge as a function of its mass. . . . 103Figure 4.43 Regenerated energy as a percentage of net energy consumed as a functionof the on-board ESS mix. . . . . . . . . . . . . . . . . . . . . . . . . . . 104xiiiAcronymsBSC battery/supercapacitor.DMU diesel motive unit.EMI electromagnetic interference.EMS energy management system.ESS energy storage system.FC fuel cell.FCB fuel-cell/battery.GHG greenhouse gas.HEV hybrid electric vehicle.ICE internal combustion engine.LRV light rail vehicle.NiCd nickel-cadmium.NiMH nickel-metal hydride.PEMFC proton exchange membrane fuel cell.PHP power hybridization potential.PID proportional-integral-derivative.PLATHEE platform for energy-efficient and environmentally friendly hybrid trains.PMDC permanent magnet direct current.PWM pulse width modulation.RTRI railway technical research institute.SC supercapacitor.SMPS switch mode power supply.SOC state of Charge.xivAcronymsTTW tank-to-wheel.UDDS urban dynamometer driving schedule.WTW well-to-wheel.xvAcknowledgementsAs a student of science, and an engineer early in his career, the people who aided, supported,and encouraged me during my pursuit of this degree have touched my life in a most substantialway. To those people, I say thank you.First of all, I would like to extend my utmost gratitude to my supervisor, Dr. Lo¨ıc Markley,for his guidance and support over almost three years. His unwavering commitment to highquality research work, and meticulous attention to detail had significant impact on my work.For my co-supervisor, Dr. Gordon Lovegrove, I extend a special thank you. After all, myjourney at UBC began with the first lecture, Social Cost Benefit Analysis by Gordon Lovegrove.His enthusiasm about sustainable transportation inspired me personally. I would also like tothank Dr. Wilson Eberle, and Dr. Rudolf Seethaler for serving on my thesis defense committeeand taking the time to read and comment on my thesis.For all the long hours and days at the Sustainable Transport Safety Research Lab, it wasmy fellow students that kept me focused and motivated. In particular, I would like to thankmy lab partner, Abdul Rahman Masoud, for the interesting political discussions which servedas a nice break between hours of research work. I would also like to thank my friends andcolleagues Adam Lee, and Esraa Jamal, and my wonderful TA, Bara Emran. Special thanksto Victoria You, who took the time to work on a 3D volumetric visualization of a hydrogenpowered locomotive.Finally, I thank all of the friends that supported and encouraged me during the course of thisdegree. To Nasser A. Alqahtani, Fawaz Altamimi, Mohamed Alhashimi, and Aziz Alghamdi Isay thank you. You were my family in Canada.xviTo my mother.xviiChapter 1IntroductionEver since the first steam engine, transportation by rail has enhanced our way of life byreducing the time traditionally required to transport goods by land and sea. A low coefficientof friction between steel wheels and steel rails enabled a very efficient transportation system.Railway propulsion technology has developed tremendously since the introduction of the firststeam engine. The use of liquid fossil fuels in internal combustion engines enabled faster andmore reliable operation. Very powerful trains that could haul hundreds of people and tonnes ofgoods were manufactured. Unaware of their impact on the environment, governments startedcompeting in building an increasing number of complex and innovative railroad networks.Noisy, pollutant emitting, and very hazardous, the diesel train did not survive for long.Its replacement, the diesel-electric train promised better traction, safer operation, higher effi-ciencies and reduced emissions. Unlike diesel trains that have a mechanical coupling betweenthe diesel engine and the wheelsets, diesel-electric trains have an electromechanical coupling.While still having a diesel engine as the prime-mover, diesel-electric trains depended on electricmotors for traction. Mechanical energy produced by the diesel engine is converted to electricalenergy by an on-board generator (alternator), which is then converted to mechanical energyat the wheels by electric traction motors. This arrangement improved the overall reliability oftrains, and drastically reduced the cost of maintenance.The next big step in railway emission reduction was the full electrification of railroads. Thesenew trains did not burn fossil fuels for power generation, and did not store energy on-boardthe train. Fully electrified railway systems were cleaner than diesel-electric trains , although iffossil fuels were used to generate the electricity that powers these trains, then fully electrifiedrailway systems would not have been entirely clean. They were also quieter, faster, safer andmuch more reliable. Having a train directly connected to the electricity grid enabled it access1Chapter 1. Introductionto a practically unlimited power supply, and also improved its acceleration. However, the costof complete electrification remained prohibitively high in a majority of cases [1].Recent developments in energy storage system (ESS) technology prompted research in hy-brid electric powertrains for railway vehicles [2–5]. Electrical ESS such as batteries and super-capacitors can be placed on-board locomotives or motive units, and are able to store brakingenergy that is otherwise lost as heat when using frictional brakes. Contrary to batteries, su-percapacitors have a very high power density but a very low energy density. They are typicallyneeded for two reasons: 1) to provide the needed power for high acceleration rates, and 2) tomore efficiently absorb regenerated energy during braking.Freight and passenger trains are very well suited to run on on-board ESS. This is becausepassenger trains utilize the electrification infrastructure of overhead catenary or under-runningconductor rails, and freight trains are mostly diesel-electric that typically utilize electric motorsfor traction. According to the literature reviewed, electrical ESS have so far been used toreduce the overall energy consumption of fully electrified railways. Research on the viability ofdiscontinuous electrification by employing ESS has so far been theoretical in the majority.Fuel cells, or in particular proton exchange membrane fuel cells (PEMFCs), combine hydro-gen and oxygen to produce electricity with water as waste. The technology has been aroundsince the 1960s, but is only recently showing promise in transportation applications due toimprovements in PEMFC research. Although PEMFCs have efficiencies of levels comparableto that of diesel engines, the high energy density of hydrogen as compared to diesel makesthem a better choice. Since PEMFCs have no moving parts, they are also much quieter andmore reliable than internal combustion engines (ICEs). That being said, due to their relativelyslow dynamic response they must typically be aided by an auxiliary power source. Figure 1.1presents the gravimetric and volumetric energy densities of supercapacitors (SCs), lithium-ionbatteries, diesel fuel, and hydrogen gas compressed at 350 bar.2Chapter 1. Introduction100 101 102 103 104 105Gravimetric Energy Density (Wh/kg)103104Volumetric Energy Density (Wh/L)Diesel FuelLithium-ion BatterySupercapacitorsHydrogen gasat 350 bar pressureFigure 1.1: The gravimetric and volumetric energy densities of SC, lithium-ion batteries, diesel fuel, andhydrogen gas compressed at 350 bar.A vehicle’s duty cycle, or driving cycle, is the cycle of power demand along a specific journey.It depends on many factors, such as driver behavior (driving style), but it primarily dependson the route traveled. Factors like gradient profile (altitude fluctuations), track curves, andspeed limits on certain sections of the tack are important for deciding whether a secondarypower source is needed, and if so, how much it should contribute to the power mix. Studyinga locomotive’s duty cycle on a certain route as a precursor to designing a hybrid powertrainis a well documented practice in the literature [6, 7]. The energy required for propulsionduring the discharge mode and the recoverable braking energy available for charging duringthe regeneration mode must be analyzed to decide on the type and size of the auxiliary powersource [8].Due to the high cost of building railway systems, a theoretical examination of railway dutycycles is required. The first step to generating a theoretical duty cycle is to computationallygenerate a trajectory profile. There are commercially available computational trajectory plan-ning tools that accurately simulate train trips, but none have been developed for clean energyfeasibility studies. Each of these tools has its own objective and motive behind its design, someoptimize trip time and some optimize comfort. The work presented in this thesis explains thedevelopment of a Matlab / Simulink based powertrain simulator to conduct trip analysis andcompare different clean propulsion technologies. It will be demonstrated that the developed31.1. Research Objectivessimulator can be used in optimization, sensitivity, and feasibility studies. One particular fea-sibility study which is the topic of this thesis, is the technical feasibility of the different ESSsizing options on-board a moving railway vehicle or locomotive.A hybrid powertrain is one where more than one type of power source is employed [9].The inherent properties of power sources and the type of load influences the need for hy-bridization, the type of hybrid system, and the degree of hybridization. In particular, loaddynamics and power source energy density are the main factors when it comes to decisionsregarding hybridization. Typically, hybrid systems are employed as an attempt to optimizecertain parameters including, for example : powertrain efficiency, range of travel, acceleration,regenerative braking, and emissions. Like any optimization problem, the optimization of hybridESS is subject to several physical and operational constraints. Physical constraints, such asmass and volume limits, are the only constraints considered in this study.After defining gravimetric and volumetric constraints, a set of the most feasible hybridiza-tion options can be obtained through the simulation of different ESS combinations. Sizingscenarios that obey the physical constrains of the vehicle are chosen as discrete points. Theresults of the powertrain simulation of the chosen sizing scenarios can be then interpolatedto produce functions of the hybridization mix. The more scenarios are simulated, the moreaccurate the interpolation process is, and the more computational power and time is required.This project does not aim to focus on any one particular figure of merit, but to introduce afeasible range of scenarios that obey the physical and longitudinal train dynamics constraints.We do pay special attention to minimum component sizes that are required for continuousoperation. It is up to the end user to decide on the hybridization mix that best suites theiroptimization goal.1.1 Research ObjectivesThe main goal of this research was to compare different clean energy storage options forrailway applications, to find an optimal mix of these options, and to comment on the viability ofthe proposed solutions as a part of a gateway technology in North America from diesel-electricto all-electric locomotives.41.2. Thesis OrganizationThe objectives of the research presented in this thesis were to:• Develop a train trajectory planning algorithm for velocity profile generation.• Develop a battery/supercapacitor parallel hybrid powertrain model and use that modelto simulate a number of railway duty cycles leading to optimization recommendations.• Develop a fuel cell/battery series hybrid powertrain model and use that model to simulatea number of railway duty cycles leading to optimization recommendations.• Simulate a number of trips on railroads of different characteristics in order to find a gener-alized answer as to the hybridization potential and the best candidates for hybridization.1.2 Thesis OrganizationThe thesis is organized as follows:• Chapter 2 includes a review of the literature as well as other background material that isrelevant to the scope of this study.• Chapter 3 explains the work done to develop a two-phase simulation process. The firstphase solves the single-body longitudinal dynamics of the train to generate a target speedprofile. The second phase simulates the actual powertrain.• Chapter 4 presents the results of the conducted simulations, and the interpretation ofthese results. The chapter is devided into two sections, the first one dealing with a short28 km round trip from Trehafod to Treherbert, and the second one dealing with a muchlonger 864 km round trip from London’s King’s Cross Station to Newcastle station.• Chapter 5 is the final chapter and in it conclusions are made regarding the results of thesimulated models. Recommendations on future research are also mentioned.5Chapter 2Literature Review2.1 Diesel-Electric TechnologyDiesel-electric technology revolutionized the railway industry. Its replacement, the diesel-electric train promised better traction, safer operation, higher efficiencies, and reduced emis-sions. Unlike diesel trains that have a mechanical coupling between the diesel engine and thewheelsets, diesel-electric trains have an electromechanical coupling. While still having a dieselengine as the prime-mover, diesel-electric trains depended on electric motors for traction. Me-chanical energy produced by the diesel engine is converted to electrical energy by an on-boardgenerator (alternator), which is then converted to mechanical energy at the wheels by electrictraction motors. This arrangement improved the overall reliability of trains, and drasticallyreduced the cost of maintenance.Drivetrain Components A vehicle’s powertrain is comprised of the components that areresponsible for power delivery from the main power source to the road surface. In a hybridelectric vehicle, these components are:• Prime Mover: This is typically a diesel engine in diesel-electric trains, and is responsiblefor transforming fuel to useful work as illustrated in Figure 2.1.• Power Electronics Module: Before power electronics were available, traction motors werestarted by varying the resistance of resistor banks that were connected between the powersource and the traction motors. This method, termed hard switching, was the onlyalternative to starting motors directly on-line (DOI).Power electronics are electronic switching devices that can handle high voltages and cur-rents. The main job of the power electronics module is to change the type of electricity62.1. Diesel-Electric Technologyfrom the power source to suit the traction motors. This is established by either changingthe voltage or the frequency of the electricity supplied to the traction motors. Thesemodules allow for soft starting of the traction motors, and allow for a much bigger speedrange.• Propulsion Control: This is the system that controls the distribution of power in a railwayvehicle. A system of “notches” control the power output of the diesel engine and are usedto accelerate the train. Each notch corresponds to a fraction of the maximum availablepower. North American systems employ an eight notch system while in the UK a fivenotch system is used.• Traction Motors: Motors convert electrical energy to mechanical energy. Traction motorsin any land vehicle are used to convert electricity to motion, which is normally transmittedto the wheels through a transmission system.• Braking System: Trains typically employ air actuated frictional brakes. A system ofcompressors, valves and pumps regulates the air that controls the braking effort of thetrain. Currently, newly built systems utilize electronic control of frictional brakes and arecapable of regenerative braking.Figure 2.1: The main components of a diesel-electric powertrain.72.2. Catenary-Electric Technology2.2 Catenary-Electric TechnologyThe benefit of electrifying railway lines using catenary is that the prime mover no longerneeds to be located on-board the locomotive. This allows any train operating on the line toaccess a practically unlimited power supply, and improves its acceleration. Electrified trainsare also safer, as there is less risk of explosion in case of derailment since the train does notcarry any combustible fuel. This comes at a price, however; the price to install continuous-feedelectrification infrastructure can usually only be justified on busy routes. Figure 2.2 illustratesthe operation of catenary-electric technology which is employed by most EU railway systems.Figure 2.2: The operation of continuously electrified railroads.Electrified overhead cables are usually single phase AC power drawn from the main threephase supplied by the local utility company. An electrified route will typically be divided intosections, each supplied from a different substation, and each fed from a different part of theutility network for increased security. Electricity is taken at very high voltage from the grid,transmitted in large cables to the substations, where the voltage is usually stepped down andthen used to electrify a section of the track [10].Single phase transformers inside the feeder stations aim to create a balanced load withrespect to the other phases. Utility companies require balanced loads so as to ensure balancedcurrents and therefore good power quality. Having a train with power levels in the megawattssuddenly show up on a single phase line will inevitably create unbalanced currents within thethree phase utility supply [11].Electrifying an entire route is a complex process, partly due to the reasons mentioned earlierand partly due to other considerations such as safety, security, grounding and electromagneticinterference (EMI). Maintenance personnel who operate close to such high voltages are atconstant risk of electric shock or electrocution. The measures required to ensure the safety of82.2. Catenary-Electric Technologypersonnel add to the cost of any electrification project. Having very high currents in overheadcables or in the rails will create large magnetic fields which could interfere with electronicdevices like pacemakers or other similarly sensitive equipment in the vicinity. There was alsodocumented evidence that EMI causes corrosion in underground pipelines [12, 13].Electrified routes also enable the use of dynamic (regenerative) braking, a mechanism bywhich a train’s kinetic energy was transformed into electrical energy and is fed back through theoverhead cables. The energy returned to the overhead cable can be used for driving other trains[2], or stored on-board the locomotive. If no storage system exists on-board the locomotivesor in the substations, the energy returned can only be consumed if another train is nearby.This creates a need for regeneration and consumption to occur at the same time, otherwise theregenerated energy is wasted [2].Electrifying over large distances requires conductor loss calculations to be taken into con-sideration [14]. The further an electric train is from the substation, the less voltage is availableto it as there is a greater voltage drop across the conductor’ s resistance. This requires allthe power electronic and control systems to be designed to work for a range of voltages whichfurther adds to the complexity of the system.The current collection system, as the name suggests, is the system responsible for collectingthe supply current and delivering it to the on-board electric traction motors. There are manyaccepted designs, but they all fall under two categories: overhead and track-level current col-lection systems. Generally, all overhead running systems utilize a current collector, otherwiseknown as a pantograph that establishes contact with the overhead current carrying cables [15].There are many different configurations and arrangements of the overhead running cables, alsoknown as catenary. These usually depend on the electrification system used, whether AC orDC, and on the level of voltage across the cables.Track level electrification is very popular in metro systems that run in tunnels. It is mucheasier on the eyes than overhead electrification, but carries a significant safety risk factor. Thecurrent collector in track level electrification is typically called a conductor shoe, and it has thejob of establishing contact with the third current carrying rail to close the circuit and allowcurrent to reach the electric traction motors.The powertrain of a fully electric train is the same as a diesel-electric train minus the92.3. Hybrid Energy Storage Technologydiesel generator set. Electric trains will typically employ more sophisticated power electronicsmodules and energy management systems [16].One of the best examples of the early success of electrification in North America is theVirginian Railway. Although it was a short, and a newer system at the time, the companyoperated one of the best engineered railroads in the United States. The Virginian Railroadhauled coal from South Appalachia, and competed with other companies in the region. Itcould only survive the competition because it was financed by the richest man at the time,Henry Rogers. The decision to electrify the Virginian Railroad came in 1922, and after threeyears, the electrification was complete [17]. It operated for 36 years until it was acquired byNorfolk & Western in 1959, and the electrification was shut down in 1961.The Penn Central Transportation Company was another company that operated electrifiedrailroads. It is famous for being the largest bankruptcy in US history at the time, when it wentbankrupt in less than a decade from its inception [18]. One of the main reasons for this was theincredibly high cost of track maintenance, and the high salaries of the workers. Another examplewhere electrification construction and maintenance costs proved too great to profitably operateis the Milwaukee Road Railway, which had capitalization problems that precluded its networkfrom going completely all-electric [19]. Given these historic failures, electrifying freight rail inUSA has remained only economically viable on short isolated routes connecting coal mines withpower plants.Recent developments in ESS technology has sparked a debate on the potential for reintro-duction of electrified railways [20]. This would in turn reduce capital cost requirements, andencourage railway companies to reconsider all-electric locomotives and electrification.2.3 Hybrid Energy Storage TechnologyThis section of the thesis discusses technologies that can replace fossil fuels in railway propul-sion systems. While hybridizing with an ICE is an option that reduces a vehicle’s emissions,it does not meet the objective of this research which is to eliminate all tank-to-wheel (TTW)emissions. TTW emissions, as opposed to well-to-wheel (WTW) emissions, are emissions atthe point-of-use regardless of the power supply chain.102.3. Hybrid Energy Storage TechnologyIn this section, we will explore the state of the art in hybrid propulsion systems appliedto railway vehicles. We will start with an introduction to ICE hybrids, otherwise known as “Green Goats”. This will be followed by a discussion on discontinuous electrification, and thena review of the application of clean power sources as prime movers in railway vehicles.Diesel-Electric Hybrids Green Goats are trains that employ a secondary power source,normally an electrical one. Batteries and SCs have been popular choices in diesel-electrichybrid projects [9, 21–23]. The main goal of hybridization with diesel powered vehicles is theoptimization of the diesel engine efficiency. Diesel engines operate at their maximum efficiencywhen producing rated power, and at their lowest when idling. By reducing the size of theengine to have a rated power equal to the average trip power, we can guarantee that the enginewill operate at its most efficient power levels and hence reduce particulate and greenhouse gas(GHG) emissions and energy waste. A secondary source is therefore needed to supply anyexcess power that cannot be delivered by the engine. The authors in [9] present a review ofthe different types of ESS that can be used in diesel hybrids, and the various diesel-hybridpowertrain architectures.Green goat technology is pioneered by Railpower Technologies Corp, out of Vancouver,British Columbia and Erie, Pennsylvania. This technology however was only applied to switcherlocomotives and not to mainline haul locomotives. Switchers spend about 60% to 80% of thetime idling [21]. This means that the system is highly inefficient for most of the operating timewhich is the motivation behind the development of green goat technology.Discontinuous Electrification Discontinuous electrification aims to reduce the cost of cate-nary technology, as well as extend the range of operation of any train, as a gateway or transitiontechnology from diesel to full electrification or as a means to present a cheaper alternative tofull electrification. On electrified sections, power provided by the overhead catenary is used tosimultaneously propel the vehicle, and charge the on-board ESS. On non-electrified sections aspresented in Figure 2.3, the ESS provides the power required for propulsion. The range of suchtrains could be further enhanced through the use of charging facilities that could be installedat stations along the non-electrified section. Employing ESS in railway systems increases the112.3. Hybrid Energy Storage Technologyutilization of regenerated energy, reduces voltage drop across the conductors, reduces requiredelectrification infrastructure and its maintenance, and reduces the visual impact of overheadcables and in-tunnel electrification. Depending on the capacity of the installed electric storagesystem, it may be possible to increase the regeneration braking force at the high-speed rangeas compared to full electrification [2].Figure 2.3: The operation of discontinuously electrified railroads [24].A research project conducted at the Environmental Engineering Research Laboratory atEast Japan Railway Company [3] studied a discontinuous electrification configuration. Theproject experimented with a typical 600 V line fed from a 1500 V DC electrified line through aDC/DC converter. The electricity fed the traction motors, ESS, and an auxiliary power supplyunit to power a single railcar; no information was given about the route except that it was runin an urban area. The magnitude and direction of the current were controlled by a DC/DCconverter by adjusting the output voltage. The setup relied on 672 lithium-ion battery cells,each operating at 3.6 V with a 30 Ah cell capacity. The cells were arranged in four parallelbranches of 168 series-connected cells [2]. The total capacity of the battery bank was 72 kWhwith the battery state of Charge (SOC) kept between 20%-95% in consideration of battery life.Experimental vehicle operation modes:• Non-electrified sections The battery system provides the entire power required, with the con-verter turned off. Dynamic braking generates electric power that supplies both the batterybank and the auxiliary power unit.• Electrified sections The overhead catenary supplies electricity to the locomotive through the122.3. Hybrid Energy Storage Technologypantograph. This electricity is converted to the 600 V level by the DC/DC converter,which is then used to drive the traction motors. Depending on the SOC of the batterybank, some of the electricity supplied by the overhead catenary could also be used tocharge the battery bank. Electricity regenerated from braking is used to charge the bat-tery bank, or is supplied back to the overhead catenary for use elsewhere in the electricitygrid depending on the battery SOC. This configuration also allows for powering assis-tance mode, in which the storage battery supplies electric power additional to the powerprovided by the catenary in order to overcome steeper grades.In the same project it was shown that in power assistance mode, line voltage fluctuationwas reduced which improved the overall power quality. Having a battery bank to supplementthe power supplied by the catenary reduced fluctuations in the overhead voltage. Hybridizingwith batteries reduced the stress on substations by reducing the peak power required. Thiseliminated the need for substation expansion to accommodate higher power requirements. Suchsetup has reportedly achieved over 30% energy saving in comparison to an inverter-fed regen-erative tram [2]. Traveling at 40 km/h, the length of the non-electrified sections reached 25.8km in total. The result showed that a 1000 A current could charge the on-board battery bankin about 60 seconds storing energy sufficient for 4 km or more [2]. This energy managementstrategy works best with power dense devices such as SCs but will have an adverse impact onthe life time of batteries due to the high charging/discharging frequency.Lithium-ion batteries are not the only option for on-board energy storage. Other options,such as SCs, have demonstrated good results when employed on rail vehicles. The MitracEnergy Saver hybrid light rail vehicle (LRV) built by Bombardier in 2003 used SCs for energystorage and reportedly achieved up to 30% energy saving in comparison to other regenerativeLRVs [14]. Similarly, the Sitras HES hybrid LRV built by Siemens in 2008 promised futureenergy savings of up to 30% [2].Hydrogen Fuel Cell Technology A fuel cell locomotive is more efficient than a diesel-electric locomotive and has emissions at levels comparable to a catenary-electric locomotive. Ithas the low infrastructure cost of a diesel-electric, and the environmental benefits of a catenary-electric. Improved energy security of the rail transport system could be achieved by employing132.3. Hybrid Energy Storage Technologyfuel cell technology in the rail industry. This would also reduce air and noise pollution, vibra-tions, and greenhouse gas (GHG) emissions, and has the possible added advantage of servingas mobile backup energy storage that could feed energy back to the electrical grid for criticalinfrastructure such as hospitals and military installations in emergencies [6].Although there are many types of fuel cells, each one running on a different fuel, themost efficient is the PEMFC which runs on hydrogen. Hydrogen can be produced from manyrenewable energy sources, such as wind and solar energy, which directly translates to lessdependence on fossil fuels. If hydrogen is produced from renewable energy sources, it wouldprovide a zero-emission locomotive, when considering the entire energy cycle.There are many different challenges facing the implementation of fuel cell technology inlocomotives [25]. These challenges include: 1) the lack of hydrogen fuel infrastructure [25], 2)the safety of hydrogen fuel storage on-board a moving vehicle [26], and 3) the fact that dieselengines are still preferred due to their ability to accommodate transient load demands such asacceleration or hill climbing with a fairly constant efficiency range.A few proof-of-concept fuel-cell rail projects have been completed in different parts of theworld. The fuel cell mining locomotive developed by Vehicle Projects LLC during 1999-2002 [27]is considered as the first significant application of hydrogen fuel cells to a rail vehicle. During2005-2007, the first fuel cell-battery hybrid switcher locomotive for urban and military-baserail applications [6] was developed by a North American project partnership among VehicleProjects Inc., BNSF Railway Company, and the US Army Corps of Engineers.Research into using hydrogen to power rail vehicles in Japan is spearheaded by two orga-nizations, the railway technical research institute (RTRI), and East Japan Railway Company(JR East). The first tests were conducted by RTRI in 2001, reaching a significant milestonein 2003 when they succeeded in powering one bogie (wheel truck) using hydrogen fuel cells.In 2006, RTRI demonstrated a fully functioning hydrogen powered railcar. It was around thesame time that JR East had transformed a diesel-hybrid railcar to operate using hydrogen,with test runs commencing in 2007 [6].In France, the initiative for cleaner rail technology is headed by the platform for energy-efficient and environmentally friendly hybrid trains (PLATHEE) program [28]. In the UK,similar research is undertaken by the Birmingham Center for Railway Research and Education.142.4. The Science of HybridizationResearchers at the center have developed the UKs first hydrogen powered locomotive, theHydrogen Pioneer [29]. Chinese researchers have also made efforts to integrate fuel cells intorailway systems. The authors in [30] present the work done to build a fuel-cell powered shuntinglocomotive.2.4 The Science of HybridizationThere are a number of factors to be considered when deciding on the need for hybridizationof power sources for any application. The science of hybridization is essentially an assessment ofsupply and demand. The power demand profile must be analyzed to determine peak and meanvalues. The frequency of the power fluctuations, and a statistical description of multiple powerdemand profiles would greatly aid in decisions regarding hybridization. This section discussedthe intrinsic properties of the three different power sources discussed in this thesis, presents anoverview of active hybrid powertrain architectures, discusses the impact of duty cycles on powersource sizing, and lists the different types of energy management systems (EMSs) present inthe literature.A) Power SourcesThe inherent properties of power sources influence the need for hybridization, the type ofhybrid system, and the degree of hybridization. There are three aspects to be considered withany power source: its energy density, its transient response (power density), and its control.This subsection presents an overview of the proposed power sources.Batteries Batteries are a form of energy dense ESS which stores energy in an electrochem-ical form. Different chemistries come at different costs, energy densities, power densities andlife expectancies. Nickel-cadmium (NiCd) batteries are the cheapest, but not the most envi-ronmentally friendly rechargeable battery chemistry on the market. Slightly more expensivechemistries such as nickel-metal hydride (NiMH) and lead-acid batteries were favored to NiCdfor traction applications due to their enhanced energy densities.The state of the art in battery technology are lithium-ion batteries. While being more152.4. The Science of Hybridizationexpensive as compared to NiCd, NiMH and lead-acid batteries, they offer the highest energydensities of over 200 Wh/kg. Lithium-ion batteries have a high charge/discharge efficienciesof 80% to 90% due to the reduced internal resistance of each cell, with the disadvantage ofhaving a short life cycle as compared to other battery chemistries. Lithium based batteries canbe found in most electronic devices, ranging from cell phones and laptops to electric vehicles.Examples of electric vehicles that run on lithium-ion technology include: Tesla Roadster, TeslaModel S, Tesla Model X, Tesla Model 3, Nissan Leaf, and BMW i3.To understand the behavior of any battery, we must first examine its discharge curve.Batteries lose voltage as their charge is depleted. Unlike SCs however, they do not do so ina linear fashion. Figure 2.4 presents the typical charge/discharge curve of a generic battery.The curve can be broken into three segments of interest that define how the battery voltageis related to its SOC: an exponential rise segment, a nominal segment, and an exponentialdecay segment. The authors in [31] briefly discuss the various computer models of lithium-ionbatteries present in the literature. They also propose a generic Simulink based model that relieson a polarization voltage and is freely available in the SimPowerSystems library in Simulink,which was the model used in this study.Figure 2.4: The charge/discharge curve of a typical battery relating the terminal voltage of the batteryto its SOC [31]. Where Qnom and Qexp are the nominal battery charge and exponential battery chargerespectively.Batteries are not meant to be operated in either exponential region. In fact, operating in162.4. The Science of Hybridizationthose regions adversely impacts battery life. Charge controllers or battery management systemsare control circuits that aim to charge batteries while ensuring that they do not operate in theexponential regions. The battery chosen for this study is the UPF454261 lithium-ion 3.7 V cellmanufactured by Panasonic.Supercapacitors Previously termed electric double-layer capacitors, these high-capacity ca-pacitors are now called supercapacitors. Contrary to batteries which discharge at lower powersover longer time periods, supercapacitors have a very high power density but a very low energydensity of typically less than 10 Wh/kg meaning that they discharge at higher powers overshorter time periods. In traction applications, they are typically employed for two reasons:1) to provide the needed power for high acceleration rates, and 2) to absorb more of the re-generated energy during braking. Supercapacitors are rarely used as the sole power source inany application due to their drastically low energy density. The BCAP3400 3400 Farad SCmanufactured by Maxwell Technologies is the SC of choice for all the simulations presented inthis thesis.Hydrogen Fuel Cells PEMFCs combine hydrogen and oxygen to generate electricity. Waterand heat are the only waste products. Like an ICE, PEMFCs do not store energy, they merelyconvert it from one form to another at a certain efficiency. Hydrogen is an excellent energycarrier with a gravimetric energy density of approximately 40 kWh/kg when compressed at 700bar, but a lower volumetric energy density of 1-2 kWh/L at the same pressure. PEMFCs aretypically 50-60% efficient, which is lower than typical battery and supercapacitor efficienciesof over 95%, but this is balanced by the high energy density of hydrogen. The PEMFC modelused in this study was freely available in Simulink as a part of the SimPowerSystems libraryand the details of its development can be found in [32].The fuel cell (FC) stack chosen for this study is manufactured by Honda and commerciallyknown as the Honda FCX family of experimental fuel cell stacks. Table A.3 contains theparameters of the 100 kW PEMFC model used in this study as obtained from [33]. While wedo not aim to discuss the details of the theory of operation of a PEMFC, it is important tohighlight a few key concepts:172.4. The Science of Hybridization• The FC fuel delivery system: Hydrogen and oxygen gas are the two inputs to the FC stack,and must be properly regulated to maintain the electrical output of the stack. Althoughthe stack may be rated at a 100 kW maximum power, it can only deliver such powerwhen the gas flow rates are adequate.Hydrogen is often stored as gas in pressurized vessels typically at 350 bar and 700 bar fortraction applications. The stack itself operates at a 3 bar pressure , which necessitatesthe use of a decompresser and a flow rate controller. Oxygen on the other hand is notstored on-board the vehicle, but is obtained from the air and compressed to the requiredpressure levels. As a result, the hydrogen delivered to the stack is of purity levels up to99.99% while the air is only 21% oxygen.• FC dynamics: Several studies addressed the main challenges associated with implementingFC technology. In [34], the transient power required for acceleration, deceleration andstart-up of a FC vehicle was obtained through simulation. The research team showed thatthe FC response exhibited a delay time as well as an undershoot/overshoot phenomenonwhen exposed to varying operating conditions as shown in Figure 2.5. In a FC vehicle, itis necessary to use a system of pumps and compressors to deliver oxygen at the requiredpressure. The dynamic response of the different subsystems responsible for oxygen deliv-ery may cause oxygen (air) starvation during step load changes, otherwise know as masstransport loss. Several studies [35, 36] placed the maximum power density of PEMFCsat 18 kW/s.The issue of FC durability was investigated in [37], and the results showed that thetransient power demand of the vehicle affected the FC lifespan and performance, whichwas adversely affected with frequent stops. Considering all these factors, it is concludedthat FCs cannot be the sole energy source in an electric vehicle. Instead, they shouldbe coupled with another energy storage device to share the transient load demand andimprove the overall powertrain efficiency.182.4. The Science of HybridizationFigure 2.5: Voltage response of a FC stack during a positive step change in load [34].Both batteries and SCs are viable options for hybridization with FCs for traction applica-tions [38–40]. Unless there is a need for the superior dynamic response of SCs, batteriesoffer more advantages in FC series hybrids due to the higher energy density of batteries.The reason is that the terminal voltage of SCs fluctuates unlike that of batteries. It isimportant that the line voltage of the powertrain be kept constant to not affect the speedof the motor. If a SC bank is to be used as the secondary source in a FC hybrid pow-ertrain, an active parallel architecture would be the better choice. This converter wouldregulate the SC bank’s voltage to result in a relatively fixed DC bus voltage [38, 39].• The polarization curve: Each electric power source has its own unique voltage-current rela-tionship, just like each mechanical power source has its own unique speed-torque rela-tionship. Just like lithium-ion batteries, FC systems have their own unique nonlinearvoltage-current relationship which is often termed as the “polarization curve”.It is important to keep in mind that the FC stack is most efficient when operating in theregion between the rated power and the maximum power. In this case, between 85 kWand 100 kW. It is also important to notice from Figure 2.6 that the current drawn fromthe FC stack decides its terminal voltage and overall efficiency. Although not presentedin the figure, in this particular example, if current of over 350 A is drawn, a sudden dropin voltage and thus power will occur. While there are sophisticated control techniques192.4. The Science of Hybridizationthat reduce FC stack degradation [41, 42], maximize stack efficiency [43], and improve itsdynamics response [44], the approach used in this thesis introduced an upper limit to thecurrent demand from the stack to prevent a sudden drop in FC power .Figure 2.6: The polarization curve of the 100 kW PEMFC presented in [33] with its parameters detailedin the appendix as generated by the SimPowerSystems FC model developed by the authors in [32].• The FC power conditioning system: To prevent overload and fault conditions, a current con-trolled boost converter to control the FC power output. Using this boost converter, wecan ensure that the stack operates continuously and that changes in load (current de-mand) are rate limited. It is important to note that current ripple due to the switchingbehavior of the boost converter can reflect badly on the stack performance and cause itsdegradation. It is also important to mention that the response of the boost convertershould not be faster than that of the FC stack. If the boost converter reacts much fasterthan the FC system, it will attempt to draw current which the FC cannot instantly supplyand mass transport loss may occur. Details regarding the design, modeling and controlof the boost converter will be covered in later chapters.202.4. The Science of HybridizationHydrogen Storage Tanks Hydrogen can be stored in more than one form. Its gaseousform is most popular in traction applications [45]. Hydrogen gas is stored in pressurizedvessels (tanks) typically at 350 bar and 700 bar pressures. The FC stack, which is taskedto combine hydrogen with oxygen gas to produce electricity, operates at approximately3 bar pressure, which necessitates the use of a decompresser and a flow rate controller.Oxygen on the other hand is not stored on-board the vehicle, but is obtained from theatmospheric air and compressed to the required pressure levels. The hydrogen deliveredto the stack is of purity levels up to 99.99% while the air is only 21% oxygen.In this thesis, we assume that hydrogen is stored at a 350 bar pressure. A 350 bar TypeIV hydrogen tank typically weighs around 100 kg, has a volume of approximately 300 L,and can store up to 5.6 kg of usable hydrogen [45]. The exact parameters used in thesimulation are listed in Table 2.1.Table 2.1: The specifications of the hydrogen tank.Physical SpecsVolume 316.4 L (380 L effective)Mass 105 kgUsable Hydrogen 5.6 kg212.4. The Science of HybridizationB) Active Hybrid Powertrain ArchitecturesThere are several ways to categorize the types of hybridization architectures possible inall-electric hybrids that do not rely on an ICE. First they are categorized as either passive oractive hybrids. A passive hybrid architecture does not employ a power electronic converter asan interface between the multiple power sources. Therefore, there is a loss of controllability ofpower flow which is often justified by the reduced powertrain cost. Active hybrids on the otherhand employ power electronic converters to increase the degree of controllability of the powerflow. Only active hybrids were considered in this study.Active Series hybrid powertrains: In series hybrids, the power sources are connected inseries with one source always feeding the other. The secondary power source located betweenthe primary power source and the load, acts as a buffer and is typically employed to handlethe dynamics of the power demand cycle. The main function of the primary power source is toprovide base power demand, and to keep the secondary source charged. Figure Figure 2.7 showsan active series hybrid powertrain architecture. The figure illustrates how each power source isconnected to a DC/DC converter. The DC/DC converter between the primary and secondarypower sources serves to control the voltage/SOC of the secondary power source. In simple terms,it is like a tap that is turned off if the secondary power source is sufficiently charged, and isturned on when the secondary power source drops below required charge levels. This mode ofoperation is often called “charge sustaining control”. The second DC/DC converter betweenthe secondary source and the load controls the motor torque and speed. Both converters canbe bidirectional if regenerative braking is required. If the primary source is a non-rechargeabledevice, like a fuel cell system, the first converter must be unidirectional.222.4. The Science of HybridizationFigure 2.7: Active Series Hybrid Powertrain Architecture.Active Parallel hybrid powertrains: Parallel hybrids utilize power sources connected inparallel to a DC bus of a fixed voltage controlled by DC/DC converters as shown in Figure 2.8.This setup allows more power flow controllability in either direction. Sophisticated controlsystems must be implemented to control how each power source reacts to load change and howmuch of the regenerated energy is fed into each source. Unlike series hybrids, neither of thepower sources acts as a buffer. It is common to keep the DC bus voltage fixed and to use athird DC/DC converter to convert bus voltage according to the desired motor speed.Figure 2.8: Active Parallel Hybrid Powertrain Architectures.D) Duty Cycles and Energy Source SizingA vehicle’s duty cycle, or driving cycle, is the cycle of power demand in a specific journey.It depends on many factors, such as driver behavior (driving style), but it primarily dependson the route traveled. Factors like gradient profile (altitude fluctuations), track curves, speedlimits on certain sections of the tack, must be taken into account when deciding whether a232.4. The Science of Hybridizationsecondary power source is needed, and if it is, how much of the power mix it should contribute.Studying a locomotives duty cycle on a certain route as a precursor to designing a hybrid powerdrive is well documented in the literature [6, 8, 22, 28]. The general procedure is to calculate theaverage power demand on a specific journey as a ratio of the peak power demand resulting inthe power hybridization potential (PHP), which decides if a secondary power source is needed[46].After examining the duty cycle of a locomotive on a certain route, and if the PHP indicatesthat a secondary power source is needed, a decision will have to be taken regarding the sizeof each of the power sources. For the selection of the type of auxiliary power source andthe size of the auxiliary storage system for locomotives, the energy required for propulsionduring the discharge mode and the recoverable braking energy available for charging during theregeneration mode must be analyzed on different duty cycles of the locomotive [22].The authors of [28] report on two approaches to decide on the size of the power sources,namely:1. Standard (Sequential) Design Methodology(a) Decide on the powertrain architecture(b) Size the system components(c) Design an optimal EMS2. Frequency Design Methodology(a) Decide on the powertrain architecture(b) Design an optimal EMS(c) Size the system components accordinglyIn [28], it is argued that the second methodology ensures that factors such as system costand volume, battery stress, and atmospheric pollution are taken into account as the intrinsiccharacteristics of the power sources. In [22] a computer program was developed and used forbattery analysis. It took into account the capacity of a battery at different charge rates forcharging and discharging. Charge rates, or C-rates, are multiples of the rated battery capacity242.4. The Science of Hybridizationunder which the battery should operate for one hour. For example, a 1 Ah (Ampere hour) 30 Cbattery could deliver one ampere of current at its rated voltage for one hour, and up to thirtytimes the rated amperage, or 30 A for two minutes.E) Energy Management SystemsEnergy management is needed for systems that allow bidirectional power flow. Systemswith bidirectional power flow ability and more than one source of power require even morecomplex EMSs, each designed with a goal in mind, an optimization objective. Variables to beoptimized include parameters such as: efficiency, cost, energy source life time, emissions, andoverall system volume.Energy management systems can be broadly classified into two categories, rule-based (heuris-tic) EMS and optimal EMS. Rule based EMS obey one or more predefined rules of operation.For example, a charge sustaining battery hybrid will likely be controlled by a heuristic EMSwith the rule being that the battery bank’s SOC be fixed to a certain value. The SOC ofthe battery bank at the beginning of a trip should not be different from the SOC at the endof the trip. Clearly this offers a direct, intuitive and easy to develop system. Although notan optimal approach in itself, rules for heuristic EMS will typically be chosen to optimize acertain parameter. For example, a charge sustaining vehicle may optimize battery lifetime bynot allowing the battery to be overly depleted or overly charged.State machine control is another rule-based control technique that is widely available inthe literature. Although documented to suffer from unwanted oscillations or “chatter” [40]when operating at the threshold between two states, it is perhaps the easiest to simulate andimplement. It was the EMS of choice in the fuel cell/battery/supercapacitor hybrid developedin [40]. Three states (modes) were implemented: a charge state, a discharge state and a recoverystate. In the charge state, the main power source, the fuel cell in this case, supplied powerto the load and the secondary sources. In the discharge state, when the load power was high,both the primary and secondary sources supplied power to the load. In the recovery state, theregenerative braking energy from the load was used to recharge the secondary sources.Optimal EMSs can be classified into two categories: offline or global optimization andonline or real-time optimization. Offline optimization methods rely on the prior knowledge of252.5. Summarythe driving cycle while online optimal EMS attempts to optimize parameters while the systemis running and with no prior knowledge of the driving cycle. Dynamic programming [47–49],linear programming [50], genetic algorithm [7, 51–53] and game theory [54] are examples ofoffline optimization methods.Researchers typically employ a two phase EMS design process, first by using an offlineoptimization method to optimize certain variables given prior knowledge of the driving cycleand then using an online optimization method to improve the results further. In [50] theresearchers combined a two stage offline-online strategy in which linear programming algorithmswere chosen for offline optimization with PID control for the online optimization. The samework was repeated using a dynamic programming algorithm in [47] for the offline optimizationphase. The authors in [55] combined a dynamic programming offline optimization method withan online optimal control based on neural networks and reported a 66% improvement in batterylife when compared to rule-based control applied to a battery/supercapacitor hybrid electricvehicle (HEV).Optimal fuzzy logic is the term coined for control techniques that combine offline optimiza-tion methods with fuzzy logic online control. Fuzzy logic control is a heuristic control methodthat requires accurately defined membership functions. The offline optimization method is usedto decide on the degree of hybridization and the fuzzy controller membership functions. Thetraining sets provided by the offline optimization method replace the experimental calibrationprocess that is often required for fuzzy logic control.Multi-objective optimization using evolutionary algorithms is a rapidly evolving area ofresearch [52]. It is often applied to standardized driving cycles for road vehicles, such as theurban dynamometer driving schedule (UDDS) which is a driving cycle that is developed by theUS Environmental Protection Agency [56].2.5 SummaryIn this chapter, we presented a review of the literature on current propulsion systems em-ployed in the railway sector. Diesel-electric technology was briefly introduced, followed by adiscussion of railway electrification and catenary-electric technology. Regenerative braking,262.5. Summaryimproved acceleration, and the elimination of emissions at the point of use were some of theadvantages of electrification discussed. A brief overview of some of the challenges that arise inelectrified railway systems was presented along with a brief overview of the history of electrifi-cation in North America.The application of hybrid energy storage technology in railway systems was discussed. Dis-continuous electrification and the application of hydrogen fuel cell technology were discussed ingreat detail. Before delving into the technical aspects of both technologies, we first presenteda review of all the successful attempts to hybridize trains that were published in the literature.Later in the chapter we discuss some of the technical aspects of clean hybrid technology. Anoverview of the intrinsic characteristics of lithium-ion batteries, supercapacitors, and hydrogenfuel cells was presented, which was followed by a discussion of hybrid powertrain architectures.The importance of a vehicle’s duty cycle in sizing potential ob-board ESS was discussed, followedby a review of the possible EMS as found in the literature.The following chapters discuss the work done to produce a computer model that simulateshybrid electric powertrains of railway vehicles. Matlab and Simulink were used to conductsuch simulations. Although the goal of this research, stated broadly, is to assess the technicalfeasibility of the proposed on-board clean hybrid electric solutions for railway systems, includ-ing freight locomotives, due to restrictions regarding data availability, we could only simulatepassenger railway vehicles. However, the conclusions drawn could also apply to freight railwayvehicles.27Chapter 3Methodology3.1 OverviewThe work presented in this thesis relied on a two-phase simulation process. The first phaseis a trip simulation that guides the second phase as illustrated in Figure 3.1. This simulationis a high-level equation of motion solver which was developed using Matlab / Simulink as thecomputation platform. The simulator models the train as a single rigid body with a fixedmass and then solves the equation of motion of that mass. It takes inputs that describe thetrack infrastructure, and the vehicle constants and solves the resultant equation of motionto calculate the actual train speed, tractive effort required, power and energy demand. Thecontrol mechanism employed in the first phase relies on a simple two-state logic (on/off) whichgenerates the velocity profile.The second phase of the simulation is the detailed powertrain simulation phase. In thispart of the simulation, detailed models of the prime mover, ESS, traction motors and powerelectronic converters are included in the simulation. Proportional-integral-derivative (PID)controllers in the second phase of the simulation attempt to force the vehicle to follow thevelocity profile generated by the first phase of simulation.283.1. OverviewDriver ModelTEBackward Simulation Forward SimulationTEDistanceVelocityDistanceVelocityDistanceVelocity++ + +v aFRFgA + B × v + C × v21   inertial mass−vaFRFg1   inertial massA + B × v + C × v2−Phase 1: TrajectoryPlanning SimulationEnergy Management SystemDesired speed Actual speed+−Actual speedLoad torque+−FgFRA + B × v + C × v2BatteryBankFuelCellStackPhase 2: PowertrainSimulationVehicle DataInertial massAdhesion coefficientSpeed and acceleration limitsInfrastrucutre DataGradient & curve profileSpeed limit profileminx xnnnnFigure 3.1: This flowchart illustrates the overall process used to conduct a hybrid powertrain simulationfor a specific trip. The first phase of the simulation combines a forward and a backward velocity profilesproducing a fast-as-possible velocity profile. The second phase of the simulation takes the resultantvelocity profile as an input. Fg, and FR refer to the gravitational force and the rolling resistance forcerespectively.293.2. Longitudinal Dynamics of Trains3.2 Longitudinal Dynamics of TrainsLongitudinal train dynamics are the subject of focus in analyzing train trips, and is definedas the motion due to forces acting on rolling stock in the track direction. It includes the motionof the whole train and any relative movements between the railcars. The first motivationbehind the study of longitudinal train dynamics was the need to increase the level of comfortof passengers by reducing longitudinal forces in passenger trains. There was also interest instudying longitudinal train dynamics by the freight rail industry. Auto-coupler tensile failureand fatigue cracking was the primary motivation for research in this case. From this researchan understanding of the different forces and their magnitudes and an awareness of the need tolimit these forces with appropriate driving strategies was developed.Longitudinal forces acting on a train can be divided into two general categories; steadyforces and impact forces. Steady forces arise from the steady application of traction poweror braking, as well as retardation forces such as wheel-rail friction, air resistance, curve andgrade forces. It is common practice to lump curve and grade forces into one combined forcetermed compensated grade. Impact forces are due to changes in locomotive power and brakingsettings, and fluctuations in grade. The steady forces considered are the propulsion forces andthe retardation forces combined. The propulsion force is the tractive effort that is dependenton the power of the prime mover, the speed of the train, and the efficiency of the system. Theretardation force is the sum of the rolling resistance, air resistance and compensated grade.The longitudinal force reaching the wheel-rail contact that is generated by the prime moveron-board the train is known as the tractive effort. To move in a given direction, the tractiveeffort must exceed any retardation forces such as wheel-rail friction, air resistance and elevation.The maximum tractive effort a locomotive can produce to propel a stationary train is thestarting tractive effort (TEstarting). The starting tractive effort is a function of the locomotive’sweight (Wlocomotive) and the wheel-rail adhesion factor (µ) as given by Equation 3.1, and it isindependent of the locomotive’s horsepower and speed. The factor of adhesion depends on thematerial from which the wheels and the rail are made, which is typically 30% for steel wheels303.2. Longitudinal Dynamics of Trainson steel rail. In this mode of operation, the locomotive’s power increases as it gains speed.TEstarting = µ×Wlocomotive (3.1)The horsepower of the locomotive will keep on increasing as the train picks up speed, until itlevels at the rated horsepower as presented in Figure 3.2. This region of operation is termedthe constant power region, and in it the tractive effort drops as the train picks up speed. Therelationship is defined by Equation 3.2. Where P stands for the power of the prime mover inwatts, η stands for the efficiency of the system, V stands for velocity in m/s, FT(N) and FR(N)stand for the tractive force and retardation force respectively.TE(N) = FT(N) + FR(N) =P (Watts)× ηV (m/s)(3.2)Figure 3.2: The relationship between a train’s applied power, resultant tractive effort and its velocityThe retardation forces on a moving train are challenging to calculate, and are often approx-imated using experimental data. The resistance for train movement on straight and level track313.2. Longitudinal Dynamics of Trainscan be determined by the W. J. Davis formula in Equation 3.3:FR = a+ bV + cV2 (3.3)• FR= Train retardation force• a = Rolling resistance component independent of train speed• b & c = Drag coefficients based on the train’s aerodynamics.• V = Train velocity.By inspecting the Davis equation we deduce that the resistance force increases with the speedof the train. Available tractive effort is typically not the controlling criteria for locomotivesoperating at low speeds. Adhesion must be large enough to overcome the train resistance, orelse wheel spin will occur. As speed begins to increase, the available tractive effort also drops.Gravity also plays a very significant role in a train’s motion. Even the slightest elevationscan have significant impact on a train’s performance. This is mainly due to the large mass ofrailway vehicles. Trains that haul loads in the thousands of tonnes require a lot of horsepowerto climb the slightest grades. For this reason, freight railroads in North America are generallylimited to a grade of 2.5%. Typically, the train configuration used to pull a certain load acrossa certain track will depend on the horsepower-per-tonne factor of that particular track. Eachtrack will have a ruling grade, which is the steepest grade along the entire track. Railroadengineers will normally calculate the amount of horsepower required to overcome the rulinggrade for a certain load, the horsepower-per-tonne. A 1% grade will cause a 100 Newtons ofdownward force for each tonne.Descending a grade safely is often a bigger challenge that ascending one. Due to the highmass of trains, even the slightest slope can lead to enormous longitudinal forces. The lowsteel-on-steel coefficient of adhesion which makes rail transportation such an efficient means oftransport also makes it difficult to stop a moving train. Therefore it is of paramount importanceto accurately calculate the expected levels of gravitational force that may act on a train on acertain track before deciding on the train composition. i.e the number of locomotives.323.3. Offline Trajectory PlanningFrTePowered AxleUnpowered AxleFigure 3.3: Forces acting on a trainA train’s stationary mass is different from its effective mass (Meff). This difference is due tothe rotating machines on-board the train. A rotary allowance (RA) of about 5-15% increase intare mass is typically assumed. The higher the number of motored axles in a locomotive, thehigher the rotary allowance. This is due to the added kinetic energy of the rotating machineryinside the vehicle. Equation 3.4 shows how the effective mass is calculated given the rotaryallowance.Meff = Mtare × (1 +RA) +Mpayload (3.4)Curve drag is another retardation force that should be considered. It is a function of the massof the train, the radius of the curve and an experimentally determined constant. Often curvedrag will not be explicitly mentioned, but is often combined with the gradient force to form acompensated grade force. Figure 3.3 illustrates the forces that determine a train’s motion.Tunnel resistance is a factor that is often accounted for when calculating retardation forces,but often not explicitly mentioned since tunnels cover a very small percentage of any giventrack. If the system under study is a metro system, tunnel resistance is continuous and isrepresented by the drag coefficients in the Davis equation, Equation 3.3.3.3 Offline Trajectory PlanningThe topic of this section deals with solving the equation of motion given certain constraintsto generate a trajectory profile for the railway vehicle. In this section, we build on the principlesgoverning a railway vehicle’s motion as discussed in Section 3.2.There are numerous software packages that compute velocity profiles for railway trips.Some are overly simplistic and do not account for train-wagon interactions, while some arevery detailed accounting for the coupling between the railcars, the air-brake system dynamics,environmental conditions and their impact on adhesion, and many other factors [57]. These333.3. Offline Trajectory Planningmore detailed models are usually employed by railway operators and are licensed commercially.Simpler models often approximate the entire train as a single rigid body and can be used forresearch purposes [58].Velocity profiles are determined from the route and train configuration and will typicallybe different under different optimization criteria. For example, the velocity profile generatedusing an algorithm for minimum trip time would typically not be the same as one developedfor minimum energy consumption.Route characteristics determine the maximum allowable speed at every section of the track.Using industry manuals on speed limits, which usually depend on track curvature and grade,a maximum allowable velocity profile can be generated. A railway vehicle’s limited acceler-ation rates introduce additional speed constraints that are further limited by recommendedacceleration/deceleration rates for safe and comfortable operation.The control input in such a system is the engine power, which generates the tractive effortat the wheels. A train can operate in one of three modes, as illustrated in Figure 3.4: inpowering mode where the tractive effort is positive, in coasting mode with zero tractive effort,or in braking mode when the tractive effort is negative (otherwise known as braking effort).The overall driving strategy or optimization target guides the algorithm developed to alternatebetween the possible modes of operation. To minimize trip time, a two mode operating schemeis employed, disregarding the option for coasting.Figure 3.4: Railway vehicle trajectory profile, and the maximum allowable velocity profile are shown.The figure illustrates the three possible modes of operation.343.3. Offline Trajectory PlanningOnline trajectory planning is unnecessary in railway applications where the route is pre-defined. Online trajectory planning techniques are often complex to design, simulate andimplement. They are usually employed in robotics applications. Offline trajectory planningtechniques produce very good results without the extra computational power, time and costthat online techniques require. Figure 3.5 presents a more detailed illustration of the algorithmused to obtain the velocity profile in all the case studies presented in this thesis. The algorithmtakes vehicle data and infrastructure data as inputs. Information about the railway vehiclesuch as its Davis equation coefficients (Equation 3.3), inertial mass, adhesion coefficient, speedand acceleration limits are some of the inputs to the algorithm. Other inputs relating to theinfrastructure of the route such as its elevation profile, curve profile and speed limits on eachsection of the route make up the rest of the inputs to the algorithm. A forward and a back-ward velocity profiles are generated and the minimum of the two profiles makes up the demandvelocity profile as shown in Figure 3.5.353.3. Offline Trajectory PlanningVehicleDataInfrastrucutreDataGradient & Curve ProfileSpeed Limit ProfileBackwardSimulationDistanceVelocityDistanceVelocityDistanceVelocityPhase1:TrajectoryPlanningSimulationTE++vaF RFgA+B×v+C×v21   inertial mass−TEmax−TEmaxamax−amaxvmax −∞DriverModelTE++vaF RFgA+B×v+C×v21   inertial mass−TEmax−TEmaxamax−amaxvmax −∞xxnnminForward SimulationInertial MassAdhesion CoefficientSpeed and Acceleration LimitsFigure3.5:Thisflowchartoutlinesthetrajectoryplanningalgorithmusedinthefirstphaseofsimulationsinthisstudy.363.3. Offline Trajectory Planning3.3.1 Preprocessing Gradient DataGradient or elevation data is typically given in discrete steps, which raises the need forinterpolation of the data. The higher the resolution of the data points, the more accuratethe interpolation is. However, it is not out of the ordinary for an elevation dataset to containpoints spaced a few kilometers apart. The straight forward answer to this problem is to linearlyinterpolate the given dataset. However, pure linear interpolation produces sudden changesbetween the different linear segments which is unrealistic and poses a challenge for the PIDcontrollers in charge of vehicle speed control, for these reasons a smoother interpolation methodis needed.As a solution, Matlab was used to interpolate the gradient profile data points. The firststep of the interpolation is to increase the resolution of the data points by generating additionallinearly spaced data points. A trade off exists when deciding on the spacing of the data points.The closer the data points, the better the fit, the more computation time is needed, and thebigger the file size. Data points of 20 meters spacing were generated via Matlab. The nextstep was to use a spline interpolation technique using the newly generated dataset. Figure 3.6illustrates the preprocessing procedure adopted to result in a gradient profile that could workwith Simulink. The result in Figure 3.6(c), is a gradient profile that is linear for the most part,but that has no abrupt changes, which is similar to real world scenarios.It is a given than any attempt to produce a fit line will come at the cost of loss of accuracyas illustrated by Figure 3.6(c). Figure 3.6 presents a section of the Trehafod to Treherbert trip.Although we can observe that in the first 2 km the fit line is not passing through all the discretepoints, this error is acceptable given that the R-square factor was calculated by Matlab to be0.9986, and the Root Mean Square Error to be 0.3603 meters.373.3. Offline Trajectory Planning3 4 5 6 7 8Displacement (km)2468101214Gradient (m/km)(a) Discrete gradient data points.3 4 5 6 7 8Displacement (km)2468101214Gradient (m/km)(b) Linearly interpolated gradient profile.3 4 5 6 7 8Displacement (km)2468101214Gradient (m/km)(c) Spline interpolation of the linearly interpolated gradient profile.Figure 3.6: This figure illustrates the necessary preprocessing steps for the discrete gradient data pointsto be usable in Simulink.383.4. Powertrain Simulation3.4 Powertrain SimulationSwitch mode power supply (SMPS) systems, otherwise known as Power Electronic Con-verters, convert electrical power using semiconductor switching devices. Electrified transportheavily employs medium to high power SMPS systems for power conditioning purposes. Elec-tronic semiconductor switching devices capable of handling high current magnitudes and voltagestresses enable the efficient and fast conversion of electric power.In railway electrification, SMPS systems are often found on-board the motive unit or lo-comotive as well as inside the electrical substations that handle voltage step down from highvoltage transmission to low voltage at the point of use. Depending on the electrification infras-tructure, SMPS systems will convert DC to AC, AC to DC, DC to DC and AC to AC. This isachieved by altering the frequency and magnitude of the supply voltage.All the power sources considered in this study are DC sources. This means that only DCto DC (DC/DC) SMPS will be considered. The simplest DC/DC converters are the buckconverter and the boost converter. The buck converter steps down the voltage at its input toa lower voltage at its output. The boost converter does the opposite. In fact, since the design,modeling and control of SMPS is not the topic of this thesis, all the power electronic modelsused in this study will be ideal and lossless with large inductors and capacitors for low frequencysimulation. This does not impact the overall result of the simulations since the dynamics ofrailway systems are inherently much slower than the dynamics of power electronic converters.SMPS systems are very efficient with efficiency levels in the range of 91%-95%. Therefore,assuming ideal converters does not greatly impact the simulation results.Although electrical isolation of source and load through the use of transformers is commonpractice for safety and protection, it has negligible impact on power ratings and efficiency, andtherefore all the converters considered in this study are non-isolated. That is because isolationis a matter of practical importance which is not the focus in this study.There are two types of hybrid powertrains considered in this study, a fuel-cell/battery (FCB)series hybrid and a battery/supercapacitor (BSC) parallel hybrid. Some parameters were keptconstant throughout the design of both systems so as not to bias the results and conclusions.The same permanent magnet direct current (PMDC) machine was used as a traction motor393.4. Powertrain Simulationin all simulated case studies. The speed of the traction motors in a railway vehicle, and byextension the entire vehicle’s speed, is a function of the voltage applied to the terminals of thetraction motor. In essence, speed control is equivalent to voltage control. The option exists touse a lower voltage power source and a boost converter to give higher voltages at the motorterminals in order for the motor to be able to reach desired high speeds. However, since a boostconverter only steps up voltage from the input side to the output side, it would be impossibleto operate the motor at lower speeds that require voltages at the output of the converter thatare lower than the voltage of the power source connected to its input. This would call forthe use of a buck converter to supplement the boost converter, and a control mechanism toswitch between the two converters depending on the desired speed. Given that the simulatedpowertrain must have bidirectional power flow ability, and if each direction requires a buck anda boost converter, a total of four converters per power source would be needed. For a dualpower source hybrid vehicle, eight converters would be required.As a solution, it was decided to use a higher voltage source and to step-down the voltageusing a buck converter to a lower voltage at the output terminal connected to the tractionmachine. In motoring mode, when power flows from the ESS to the traction machine, voltageis stepped down through a buck converter and stepped up through a boost converter when thedirection of power flow is reversed during regeneration mode. All the models operated at a1500 V DC bus voltage which covered the entire speed range of the traction machine.A fixed gear ratio of 3:1 was kept constant in all simulations, with the faster gear connectedto the traction motor and the slower gear to the wheelset. All references to “battery” in thisstudy refer to lithium-ion batteries only as they are the only battery chemistry considered sincethey are the most energy dense and the most suitable for traction applications.Figure 3.7 is a graphical representation of the main steps taken to design, model and controla power electronic converter using Matlab and Simulink. The figure only highlights the generalprocedure with more details on the specific control of each converter to be presented in thefollowing sections. The first step is the open loop design and simulation. In this step, circuitcomponents are sized according to the converter’s governing equations and open loop specifi-cations. After the components are sized, the converter is then modeled and simulated in theSimulink environment using the SimPowerSystems library. Before proceeding to the next step,403.4. Powertrain Simulationit is important to verify that the stead-state currents and voltages of the simulated convertermatch their corresponding theoretical values.The second and third steps deal with system identification and linear approximation. SMPSare highly nonlinear systems and therefore challenging to model and control. All the modelsproduced in this section were generated computationally and not analytically. While thereare established large signal and small signal models for buck and boost converters, numericalmethods provide faster and more reliable results. Matlab’s Parallel Computing Toolbox wasused in combination with Simulink’s System Identification Toolbox to produce second orderlinear approximations of the proposed converters. The general procedure performed by Matlaband Simulink involves subjecting the open loop converter to an array of test control signalsand storing the output, be it voltage or current. The software then attempts to create a linearsecond order transfer function that relates the test input signal to the output. For example, ifwe wish to control the output voltage of a buck converter, the control signal is the converter’sduty cycle and the system output is the output voltage.The last step in the closed loop control of a SMPS converter is finding the control parameters.PID control is simple, straightforward and easy to simulate. Finding PID parameters can be atedious process, especially with an approximated system. PID controllers were then synthesizedusing Simulink’s PID tuning ability. This greatly reduced the time normally required to tunePID controllers. It was discovered that the simulation results were less sensitive to changes inthe power electronic converters if the PID controller had a very tight error tolerance.413.4. Powertrain SimulationnFigure 3.7: The steps required to design, model and control a simulated power electronic convertermodel in SimPowerSystems.3.4.1 Buck ConverterA buck converter is a DC to DC converter which steps down the input voltage and stepsup the input current. It is used to interface a high voltage source with a lower voltage load.The converter used in this study is an ideal converter with a single ideal switch, ideal diodesand ideal inductor and capacitor. The circuit layout is presented in Figure 3.8.Figure 3.8: Circuit layout of a non-isolated buck converter with a purely resistive load.423.4. Powertrain SimulationOpen-loop design and simulation Buck converters cover a limited range of output volt-ages. The value of the output voltage, Vout, depends on the switching duty cycle, D, and isgoverned by Equation 3.5. A switching duty cycle is defined as the ratio of the time duringwhich a semiconductor switch in a power electronic circuit is turned ON to the time it is turnedOFF. Given Ohm’s Law, the output current, Iout can be found using Equation 3.6.Vout = DVin (3.5)Iout = IL =VoutR(3.6)For component sizing purposes, an operating point has to be carefully chosen to ensureproper operation over a wide range of duty cycles. Figure 3.9 shows the output voltage, Vout,as a function of switching duty cycle as governed by Equation 3.5.0 0.2 0.4 0.6 0.8 1Duty Cycle050010001500Output Voltage (V)1500 x DOperating PointFigure 3.9: Output voltage as a function of switching duty cycle for a buck converter with an inputvoltage of 1500 V.In all the simulations presented in this study, buck converters are used only during motoringmodes to deliver power from the source to the load. The source voltage was kept at 1500 Vin all of the simulations and is then stepped down to lower voltages depending on the desiredmotor speed.The next step in the converter design is to size the inductor and capacitor. Both the433.4. Powertrain Simulationoutput voltage ripple, ∆vout, and the input current ripple, ∆iin, are inversely proportional tothe switching frequency fs as suggested by Equation 3.7 and Equation 3.8 respectively. Theoutput voltage ripple ∆vout is also inversely proportional to the output capacitor size C whilethe output current ripple ∆iout is inversely proportional to the inductor size L.∆vout =Vout(1−D)8LCf2s(3.7)∆iout = ∆iL =Vout(Vin − Vout)VinLfs(3.8)This suggests a trade off between element sizes and switching frequency for any given ripplevalue. For example, a higher switching frequency fs and a small inductor L will give thesame ripple amount as a lower switching frequency fs and a bigger inductor L, as governed byEquation 3.8. The same trade-off exists between the output voltage ripple, capacitor size andswitching frequency fs.Parameters used in the simulation are included in Table 3.1, and the open loop outputvoltage is presented in Figure 3.10. Figure 3.10 shows that the open loop output voltage hasa steady state value of 750 V as governed by Equation 3.5. It also shows that the open loopsettling time of the output voltage is around 1 second, with a ripple component of 1 V asgoverned by Equation 3.7 and calculated in Table 3.1.Table 3.1: Buck converter specifications.Vin 1500 VVout 750 VD 50%R 1 ΩIout 750 Afs 1000 Hz∆vout 1 VL 0.1143 HC 0.2286 F443.4. Powertrain Simulation0 2 4 6 8 10Time (seconds)0100200300400500600700800Output Voltage (V)8.861 8.862 8.863 8.864753.4753.6753.8754754.2Figure 3.10: Open-loop output voltage Vout at the operating point shown in Figure 3.9 with a rippleamount ∆vout of 1 V. The inset plot highlights the ripple component.In practice, the aim is to reduce component sizes and thus reduce cost by operating at ahigher switching frequency fs. For simulation purposes, however, higher switching frequenciestranslate to longer computation time. The Simulink model would have to run at least twicethe switching frequency (Nyquist Frequency) to prevent undersampling and signal distortion.This is both computationally, and memory intensive. Therefore a low switching frequency of1000 Hz was selected.It is important to take note of two facts, the first is that if the inductor size is too small, theconverter will enter a discontinuous conduction mode which produces undesirable results andfurther complicates the control problem. Discontinuous conduction mode is when the inductorcurrent is held at zero for any portion of the sampling time. The critical inductance value atthe boundary between continuous conduction mode and discontinuous conduction mode is Lcritand is governed by Equation 3.9. The second is that if the inductor and capacitor sizes are toolarge, they may store significant energy and distort the results of the simulation. This suggestsa trade off, and that a tuning procedure has to be adopted.Lcrit =(1−D)R2fs(3.9)453.4. Powertrain SimulationLinearized system identification The rotational speed of any electrical motor is a functionof the voltage supply. A closed loop control system can be used to adjust the converter’s outputvoltage depending on the desired speed of the electrical motor connected to the output of thebuck converter. Classical linear control can only be used to control power electronic convertersif the range of operation is bounded to a region where linear approximations of the nonlinearmodel can be made. By examining Figure 3.9, we can see that higher duty cycle values resultin a linear rise in output voltage which simplifies the process of synthesizing a controller forthis converter. An operating point at a 50% duty cycle was selected as shown in Figure 3.9.At this operating point, the output voltage Vout is 750 V.Closed loop control of power electronic converters has been an active research topic eversince their inception. Classical linear control techniques produce reasonable results and aremuch easier to implement when compared to more sophisticated nonlinear control techniques.PID control is the industry standard for all non-sensitive applications. A PID controller actson the error between a desired signal (input signal) and the actual signal (output signal) whichis usually obtained through sensors. It aims to minimize this error to produce an output thatmatches the input. It does so by applying three control parameters to said error signal. Thesethree control parameters are: the proportional parameter (P), the integral parameter (I), andthe derivative parameter (D), hence the controller is termed PID. The proportional parameteracts on the current values of the error signal, while the integral parameter acts on the errorhistory, and the derivative parameter on the future possible values for error.For the purposes of this study, a simple PID controller was synthesized and found to produceaccurate results over a wide range of output and input. The procedure as highlighted inFigure 3.7 treats the plant (system to be controlled) as a black box, and formulates a linearapproximation depending on the test inputs and the corresponding outputs.This process requires iteration with different test input signals, different in magnitude andin frequency. Each time a linear model is found, it is validated through simulation until anacceptable approximation is found. An acceptable approximation is one where the approxi-mated system captures the dynamics of the actual system within 20% error margin. The buckconverter used in this study is a second order system, and therefore it is best to approximateit with a second order linear system.463.4. Powertrain SimulationFigure 3.11 shows the bode plot of the buck converter and its linear equivalent model.While the approximated response is not an identical fit to the actual response, the results areacceptable given that the “goodness of fit” factor, a factor that determines how closely theapproximated system matches the actual system, as calculated by Matlab was 97.17% 1. Thisis also demonstrated by the open loop response in Figure 3.12. The approximated responseis not expected to be an identical replica of the actual response. The most important aspectof comparison is by comparing the transient response of both systems, since any steady stateerror will be compensated by the controller. Equation 3.10 is the transfer function of the secondorder linearized model of the buck converter given the parameters defined in Table 3.1, wheres is the complex variable of the Laplace transform.Vout(s)D(s)=−5350s+ 4.231e7s2 + 5249s+ 3.976e4(3.10)Figure 3.11: Open-loop frequency response of the buck converter given the parameters in Table 3.1, andthe open-loop frequency response of the linearly approximated system.1The “goodness of fit”, also known as the Fit Percent, is the normalized root mean square error fitness valuewhich is calculated using the following equation: gof(%) = 100(1− ||y−yˆ||||y−mean(y)|| ), where y is the validation dataoutput from the actual system, yˆ is the output of approximated system, and || indicates the 2-norm of a vector.473.4. Powertrain Simulation0 2 4 6 8 10Time (s)0100200300400500600700Output Voltage (V) Linear equivalent responseActual responseFigure 3.12: The open loop output voltage response of the buck converter in comparison to that of thelinearized model.Closed loop system and PID controller tuning Figure 3.13 outlines closed-loop controlof the buck converter. The simple buck converter employed in this study is a second ordersystem as it only contains two energy storage elements, the inductor and the capacitor. Ithas two state variables, the inductor current (input current) and the capacitor voltage (outputvoltage). The schematic of the closed loop PID control of the converter’s output voltage ispresented in the Figure 3.13.After producing an acceptable linear approximation of the open-loop system, and validatingthe response of the linear approximation against the actual nonlinear system, the next step isto synthesize a controller. Simulink has a PID tuning tool that can be used with the linearapproximation of the actual system.There are trade offs when tuning a PID controller. Increasing the speed of the response(i.e. reducing the settling time) will most likely lead to an increase in overshoot and oscillationsresulting in an under-damped system. The opposite is also true, an over-damped system willhave a significant settling time but will not overshoot or oscillate. A settling time of less than 5seconds is permissible given that railway systems are inherently slow. Once the PID controllerhas been tuned given the linearly approximated second order plant, it is then applied to the483.4. Powertrain Simulationactual non-linear buck converter circuit. The closed loop response of both systems are thencompared.The output of the PID controller is an analog signal that corresponds to the duty cycle asillustrated in Figure 3.13. A pulse width modulation (PWM) block must be used to convertthat duty cycle into switching signals that trigger the switch, S1, ON and OFF. There aremany techniques that perform Pule Width Modulations. The simple technique used in thisstudy is one where the duty cycle analog signal is compared to a sawtooth waveform of theswitching frequency, fs, calculated earlier. Both signals must be amplitude restricted between0 and 1, corresponding to a 0% duty cycle (always ON) and a 100% duty cycle (always OFF)respectively. Whenever the duty cycle signal is higher than the sawtooth waveform, the switchis turned OFF, and ON when the opposite is true. The result is a train of pulses of fixedfrequency, fs, but of varying duty cycle.nFigure 3.13: Closed-loop PWM control of the output voltage in a buck converter using a PID controller.493.4. Powertrain Simulation0 0.5 1 1.5 2 2.5 3Time (s)0100200300400500Output Voltage (V) Reference voltageClosed loop response of approximated systemClosed loop response of actual systemFigure 3.14: Closed-loop response of PID controlled buck converter.3.4.2 Boost ConverterA boost converter is a DC to DC converter which steps up the input voltage and steps downthe input current. It is used to interface a low voltage source with a higher voltage DC bus.The converter used in this study is an ideal converter with a single ideal switch, ideal diodeand ideal inductor and capacitor. The circuit layout is presented in Figure 3.15.Vin S1LRLIin = ILCVL+ −Vout+−IoutFigure 3.15: Circuit layout of a non-isolated boost converter with a purely resistive load.503.4. Powertrain SimulationOpen-loop design and simulation Boost converters cover a wide range of output voltageas Figure 3.16 illustrates. The value of the output voltage, Vout, depends on the duty cycle,D, and is governed by Equation 3.11. For component sizing purposes, an operating point hasto be carefully chosen to ensure proper operation even as parameters slightly change.Vout =Vin(1−D) (3.11)0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1Duty Cycle00.20.40.60.811.21.41.61.82Output Voltage (V)# 10 4400 / (1-D)Operating PointFigure 3.16: Output voltage as a function of switching duty cycle. The operating point at 50% dutycycle is highlighted.In the FCB hybrid powertrain presented in Subsection 4.2.4 and Subsection 4.1.4, a boostconverter is used to interface between a 400 V FC system and a 1500 V battery bank. Inpractice, both these values will change throughout a trip. Depending on the current draw fromthe FC, its voltage may drop. Although fuel flow rate control systems should exist to keep theFC’s voltage constant, the boost converter must be able to adapt to input voltage fluctuations.Similarly, the battery bank’s voltage will not be constant throughout the trip. The battery’scharge will deplete as it provides transient power to the traction system, and will recharge ifenergy is available from the FC or from regenerative braking.For this reason, the use of closed loop control systems was necessary. As mentioned earlier,513.4. Powertrain Simulationclassical linear control can only be used to control power electronic converters if the rangeof operation is bounded to a region where linear approximations of the nonlinear model canbe made. By examining Figure 3.16, we can see that higher duty cycle values result in anincreasing rise in output voltage. The figure also highlights the selected operating point at 50%duty cycle. At this operating point, the output voltage Vout is 800 V.The next step in the converter design is to size the inductor and capacitor. Both theoutput voltage ripple ∆vout and the input current ripple ∆iin are inversely proportional tothe switching frequency fs as suggested by Equation 3.12 and Equation 3.13 respectively. Theoutput voltage ripple ∆vout is also inversely proportional to the output capacitor size C whilethe input current ripple ∆iin is inversely proportional to the input inductor size L.∆vout =VoutR× DCfs(3.12)∆iin = ∆IL =VinDLfs(3.13)0 1 2 3 4 5 6 7 8 9 10Time (seconds)0100200300400500600700800900Output Voltage (V)9.50029.50049.50069.5008799.8799.9800800.1800.2(a) Open-loop output voltage Vout at 50% dutycycle with a ripple amount ∆vout of 0.5 V0 1 2 3 4 5 6 7 8 9 10Time (seconds)020040060080010001200140016001800Input Current (A)9.50029.50049.50069.50081599.81599.916001600.11600.2(b) Open-loop input current Iin at 50% dutycycle with a ripple amount ∆iin of 0.5 AFigure 3.17: The open loop output voltage and input current of a simplified boost converter given thecircuit parameters in Table 3.2.523.4. Powertrain SimulationThis suggests a trade off between element sizes and switching frequency for any given ripplevalue. For example, a higher switching frequency fs and a small inductor L will give the sameinput current ripple magnitude as a lower switching frequency fs and a bigger inductor L,as governed by Equation 3.13. The same trade off exists between the output voltage ripple,capacitor size and switching frequency fs. The minimum inductor size (Lcrit) for continuousconduction is governed by Equation 3.14.Lcrit =VinD(1−D)2Ioutfs(3.14)In this study, boost converters have one function regardless of their location in the pow-ertrain. They are either used to boost the FC voltage to charge the battery in a FCB serieshybrid, or to boost the regenerated voltage during regenerative braking. Parameters used inthe simulation are included in Table 3.2.Table 3.2: This table contains the open-loop circuit specifications for the boost converter used in thisresearch.Vin 400 VVout 800 VD 50%R 1 ΩIin 1600 AIout 800 Afs 3500 Hz∆iin 0.5 A∆vout 0.5 VL 0.1143 HC 0.2286 FLinearized system identification Following the same process explained in the previous sec-tion, a linear approximation of the boost converter was synthesized and its frequency responseis compared to the frequency response of the actual non-linear system as shown in Figure 3.18.According to Matlab, the “goodness of fit” was 83.7%. In a FC series hybrid, it is importantto control the boost converter’s input current which is also the FC’s output current. Equa-tion 3.15 is the transfer function of the second order linearized model of the boost convertergiven parameters defined in Table 3.2.533.4. Powertrain SimulationIin(s)D(s)=3222s+ 8.338e4s2 + 6.826s+ 24.67(3.15)Figure 3.18: Open-loop frequency response of the boost converter given the parameters in Table 3.2,and the open-loop frequency response of the linearly approximated system.Figure 3.19 shows how the open loop input current of the actual system and the approx-imated system compare. As mentioned earlier, the approximated response is not expected tobe an identical replica of the actual response. The most important aspect of comparison is bycomparing the transient response of both systems. Any steady state error will be compensatedby the controller.543.4. Powertrain Simulation0 2 4 6 8 10Time (s)0200400600800100012001400160018002000Input Current (A)Linear equivalent responseActual responseFigure 3.19: Open loop input current transient response at 50% duty cycle.Closed loop system and PID controller tuning Figure 3.20 outlines closed-loop controlof the input current in a boost converter. After producing an acceptable linear approximationof the open-loop system, and validating the response of the linear approximation against theactual nonlinear system, the next step is to synthesize a controller. Simulink has a PID tuningtool that can be used with the linear approximation of the actual system.Figure 3.20: Closed-loop PWM control of the input current in a boost converter using a PID controller.553.4. Powertrain Simulation0 2 4 6 8 10Time (s)0100200300400500600700Input Current (A)Reference currentClosed loop response of approximated systemClosed loop response of actual systemFigure 3.21: Closed loop input current transient response assuming a 500 A reference input current.3.4.3 Bidirectional Buck and Boost ConverterThe previous sections discussed the design, modeling and closed-loop control of buck andboost converters. The load interface converter used in this study is a bidirectional converterthat is a combination of the buck and the boost converters. To be bidirectional, a convertermuch allow power flow in both directions. Through the use of switching devices and diodes,the direction of the electric current can be controlled. A bidirectional converter is needed toallow for energy regeneration during braking.To simplify circuitry and therefore control, the power source has to be of sufficiently highvoltage as to cover the entire rotational speed range of the traction machine without the needfor a voltage boost. This would necessitate the use of a buck converter when power flow is fromsource to load. Operating at higher voltages is better than operating at higher currents for thesame power levels to minimize the energy lost in the conductors.This creates the need to use a boost converter when power is to flow from load to source,for example during regeneration. The boost converter would reverse the current direction inthe circuit, and boost the voltage across the traction machine terminals to suitable levels forcharging the ESS. Figure 3.22 illustrates the modes of operation of the converter employed inthis study.563.4. Powertrain SimulationMotoring Mode Regenrative ModeBidirectional Buck and BoostLS1S2S1S2LFigure 3.22: The circuitry of the bidirectional buck-boost converter which interfaces the energy sourcewith the traction machine.3.4.4 Online Deterministic State Machine ControlTo simplify the design and simulation of the hybrid systems discussed in this study, a simpletwo state heuristic controller was implemented in Simulink. Since we do not aim to study orcomment on the impact of control systems on energy consumption, it would be unproductiveto spend time designing and simulating a more complex control system.The two states that determine the operation of the power electronic converter and thedirection of energy flow are: a 1) Motoring State, and a 2) Regenerative Braking State. As thenames suggests, the motoring state is the state in which the energy flow is from storage deviceto traction machine, and the power electronic converter interfacing the source and the load isoperating in buck mode. In this state, the source voltage of 1500 V is stepped down througha buck converter, and the duty cycle is decided by the PID controller. The opposite of thatis the regenerative braking state, in which the energy flow is from traction machine to storagedevice, and the power electronic converter interfacing the source and the load is operating inboost mode.As mentioned in section 2.4, state machine control is documented to suffer from unwantedoscillations or “chatter” when operating at the boundary condition between any two states [40].573.4. Powertrain SimulationIt takes time for the PID controller to adjust the switching duty cycle of the converter to producea zero velocity error between the actual velocity and the reference velocity profile. This time istermed the “settling time”, referred to as ts in Figure 3.23. This highlights the importance ofchoosing a critically damped (no overshoot) PID controller, otherwise the state machine controlmechanism will alter the state of the controller unnecessarily producing undesirable results oroscillations as presented in Figure 3.23.The boundary condition in this case is the zero error condition, if the speed error is positiveindicating that the actual speed is less than the reference speed, the motoring state is invoked,otherwise the regenerative braking state is invoked. If the error oscillates around zero however,this will result in an uncontrollable system behavior. One solution to this problem is to intro-duce a delay in the state transition mechanism. This delay should be larger than the settlingtime to filter out oscillations due to any overshoot produced by the PID controller, but sensitiveenough to realize when a change of state is needed. Another solution would be to introduce aboundary gap between the two states. It is therefore obvious that this control technique is nota robust one, but its main advantage is its simplicity. Figure 3.23 explains the operation of thecontrol system.Figure 3.23 illustrates how the chosen control mechanism functions. In it we see how theacutual vehicle speed attempts to match the reference speed. The figure is of a vehicle thatstarts in regenerative braking mode and transitions to motoring mode when the actual speedbecomes lower than the reference speed. After the state of the control system changes, ittakes some time, ts, for the power source to deliver enough power for the vehicle to accelerate,matching the actual speed to the reference speed. Towards the end of the trip, the vehicletransitions from motoring mode to regenerative braking mode. In motoring mode, the vehiclewill attempt to follow the reference speed profile by adjusting the magnitude of power deliveredto the traction motors. If the vehicle continues to accelerate and no power is being supplied,then the direction of power flow must be reversed to absorb energy from the system. The figurealso illustrates the oscillations that may occur at boundary conditions as mentioned earlier.583.4. Powertrain SimulationFigure 3.23: The impact of a two-state state machine deterministic controller on the vehicle velocity ascompared to the provided reference velocity profile.3.4.5 Combined Simulink ModelProducing a locomotive powertrain model and simulating an actual train trip is a complextask. So far, all the subsections of this chapter discussed all the different components of theSimulink model. Section 3.2 presented an overview of the longitudinal dynamics of railwayvehicles which was later combined with concepts mentioned in Section 3.3 to produce a velocitytrajectory profile given a certain railway-consist (train) and a defined route. In Section 3.4 wediscussed the need for power electronic converters on-board the railway vehicle, and presentedan overview of the design, modelling and control of a buck and a boost converter. We alsomentioned how the two converters were combined into one simplified converter that allowedbidirectional power flow. Other powertrain components such as the traction machine, lithium-ion battery bank, supercapacitor bank, and FC system models were not explicitly defined inthe body of this thesis since the study relied on models present in the literature and readilyavailable in the SimPowerSystems library in Simulink. Section 3.5 and Section 3.6 discuss howall the different subsystems are integrated in a combined Simulink model.593.5. Battery - Supercapacitor Parallel Hybrid Powertrain Description3.5 Battery - Supercapacitor Parallel Hybrid PowertrainDescriptionBSC Parallel Hybrid Powertrain The parallel hybrid architecture chosen for this studyallows both power sources to react independently to changes in load. Essentially, it allows forcomplete control for the power split between the two sources. If a series hybrid architecturewas chosen, one source will always be feeding the other which would not allow the independentevaluation of each source.For an unbiased transient response, Lbat and Lsc have the same value. This would allowus to estimate how each source reacts to changes in load. As illustrated in Figure 3.24, eachpower source is connected to a separate bidirectional buck-boost converter. Control signals toswitches S1 and S3 are synchronized so that both switches turn on and off at the same time.The same is applied to switches S2 and S4. This guarantees that both sources are either inmotoring mode or in regenerative braking mode without the option for one source to be in onemode and the other in another mode. This prevents one source from feeding the other, whichallows to judge the ability of each source to respond to transient load changes based solely onits intrinsic characteristics, and simplifies the control system design.603.5. Battery - Supercapacitor Parallel Hybrid Powertrain DescriptionLscLbatS1S2S3S4BuckSwitchesBoostSwitchesBatteryBankSupercapBankCinCinCoutPMDCMotorMotoring Mode Regenerative ModePowertrain LayoutLscLbatS1S2S3S4LscLbatS1S2S3S4Figure 3.24: BSC parallel hybrid powertrain architecture.The combined model All of the different subsystems discussed earlier have to be combinedinto a single Simulink model. Figure 3.25 presents an illustration of the combined model. Thetraction machine’s rotational speed and torque are stepped up or down according to a predefinedgear ratio. Rotational speed is then converted to translational velocity through the wheelsetattached to the gearbox. This translational velocity is then used to calculate the retardationforce FR using the Davis equation described in Equation 3.3.The distance traveled (location) can be calculated by integrating the translational velocity.The location is used to find the corresponding velocity limit and elevation magnitude providedin distance coded lookup tables. The output of the elevation lookup table is the gravitationalforce FG which is added to the other retardation forces resulting in a net retardation force Fnet.The net retardation force is converted to load torque on the traction machine.613.5. Battery - Supercapacitor Parallel Hybrid Powertrain DescriptionThe output of the velocity profile lookup table is the desired velocity at each section ascalculated using methods discussed in Section 3.3. This output is then compared to the actualspeed of the vehicle and an error signal is produced. Two PID controllers act on that errorsignal and produce a PWM signal of a duty ratio corresponding to the desired motor speed.The error signal is also used to determine the direction of power flow, either motoring mode orregenerative braking mode. The power electronic converters will then operate to deliver powerto the traction machine, resulting in an increase in the vehicle’s speed. It is then integrated toproduce displacement that is then fed to the lookup tables to find the elevation and velocitylimits. The simulation is terminated once the train reaches its destination.623.5. Battery - Supercapacitor Parallel Hybrid Powertrain DescriptionDiferror>0iferror<0+ −VelrefVelDerrorDDωpi×rwheel30m/smm/sωFRFGFnetτ loadτ loadDesterministicStateMachineEMSGearRatioGearRatioDistanceVelocityDistanceGradientPIDPID∑∫∑1 0 1 0PWM PWM1 01 0-1LscLbatS1S2S3S4BuckSwitchesBoostSwitchesBatteryBankCinCinCoutPMDCMotorMotoringModeRegenerativeModeLscLbatS1S2S3S4LscLbatS1S2S3S4UCBankr wheelFR=a+b×V+c×V2Figure 3.25: The combined battery / SC parallel hybrid powertrain Simulink model.633.6. Fuel cell - Battery Series Hybrid Powertrain Description3.6 Fuel cell - Battery Series Hybrid Powertrain DescriptionThe FCB series hybrid powertrain As explained earlier in Subsection 2.4, there areadvantages and disadvantages for every type of powertrain architecture. In series hybrids oneof the power sources acts as a buffer between the prime-mover and the load. This seriesarchitecture does a better job of maintaining the battery SOC whilst meeting all the requiredpower transients [59].FCB series hybrids have been studied in a wide array of traction applications, ranging fromrailway [60] and collection trucks [61] to lighter duty experimental setups [62, 63].PMDCMotorMotoring Mode Regenerative ModeLbatS1S2LbatS1S2BatteryBankFuelCellStackBoost Converter Bidirectional Buck Boost Converter400 V 1500 VFigure 3.26: FCB series hybrid powertrain architecture.The combined model As mentioned in Section 3.5, all of the various subsystems have tobe combined into one Simulink model that represents a FCB series hybrid train. Figure 3.27illustrates how all the different subsystems were interconnected into one Simulink model in thisstudy.643.6. Fuel cell - Battery Series Hybrid Powertrain DescriptionBatteryBankFuelCellStackDRegenerativeModeMotoringModeiferror>0iferror<0+ −Vel refVelDerrorDDLS1S2LS1S2ωpi×rwheel30m/smm/sωFRFGFnetτ loadτ loadDesterministicStateMachineEMSGearRatioGearRatioDistanceVelocityDistanceGradientPIDPID∑∫∑1 0 1 0PWMPWM1 01 0+−I refI inD1 0PIDPWMPSaturation∑∑RateLimiter1 0SOCSOCref-1BoostConverterBidirectionalBuck/BoostConverterH2FlowRateRegulatorAirFlowRateRegulatorr wheelFR=a+b×V+c×V2Figure 3.27: The combined FCB series hybrid powertrain Simulink model.653.7. Numerical Optimization Process3.7 Numerical Optimization ProcessAs mentioned in Section 1.1, one of the main goals of this thesis is to make an attemptat optimizing the sizing mix of the on-board ESS. Normally, we would need to identify theexact parameters that are to be optimized before attempting to optimize the on-board ESSmix. Instead, this thesis will present how the mix of the on-board ESS impacts certain keyparameters. These parameters include figures of merit such as fuel economy and battery SOCdepletion.The first step in any optimization problem is to identify limits or boundary conditions. Inthis study, two physical boundary conditions exist: a mass boundary condition and a volumeboundary condition. All possible solutions must satisfy both conditions, meaning that theymust fall within the unshaded region. Note that the optimal solution for a given applicationmust lie within this region.Since this problem is too complex to be solved analytically, computational means wereemployed. Iterating the simulation at different ESS sizing ratios and logging the results of eachsimulation at a sufficiently high resolution can help in identifying trends and optimal regions.The results can be interpolated to find the estimated fuel economy at any point.Figure 3.28: A graphical representation of the optimization problem that appears when deciding on thehybridization mix.66Chapter 4Results and DiscussionThis chapter of the thesis presents the results obtained from the simulations conducted onMatlab / Simulink. The chapter is divided into two sections that present the results for each ofthe two case studies considered. The first case study is a 28 kilometer round trip from Trehafodto Treherbert in the UK. The trainset is made up of two British Class 150 diesel motive units(DMUs). The second case study is a much longer 864 kilometer round trip from London’sKing’s Cross Station to Newcastle. The trainset in the second case study is the Intercity 125.It is made up of two British Class 43 locomotives hauling 6-8 passenger carriages that can carryapproximately 600 seated passengers.Each section is further divided into four subsections that 1) describe the train, 2) describethe route, 3) discuss the results for the battery/ultracapacitor hybrid study, and 4) discuss theresults for the fuel cell / battery hybrid study.Since there exists an infinite number of ESS sizing combinations, each with its own setof results, only a selected set is presented. A detailed set of single trip results for a specificsizing combination as a representative example of all the simulations conducted is presentedand discussed. This is then followed by a discussion of results obtained if the train were torepeat the trip continuously over a 15-16 hour operational day. Finally, the results of the overalloptimization study are presented and discussed.674.1. British Class 150: Trehafod to Treherbert4.1 British Class 150: Trehafod to Treherbert4.1.1 Train DescriptionBritish Class 150 is a DMU that runs on a 200 kW diesel engine. Typically, two or three ofthese units will form a trainset. Each DMU has one powered bogie and one trailing bogie asshown in Figure 4.1. The case considered in this study is one where a trainset is formed of twoDMUs that can carry 200 - 250 passengers. Table 4.1 contains the DMU specifications.The space available for hybrid ESS installation was assumed to be 3.5 tonnes at 4000 Liters.This assumption was based on the combined mass and volume of the diesel engine and fuelcompartment. Whenever mass and volume are discussed in this chapter, they refer to per-wagon mass and volume. For example, the mass constraint mentioned earlier is for a singleDMU, and not for the entire trainset which is comprised of two DMUs. Also, when ESS massis discussed, it refers to mass per DMU or locomotive and not for the entire trainset.Table 4.1: The specifications of British Class 150 DMU.Physical SpecsLength 20.06 mWidth 2.816 mHeight 3.774 mWheel Diameter 0.834 mMass 35.8 tonnesFuel Capacity 1500 LDavis Equation (kN) 2.089 + 0.0098v(m/s) + 0.0065v2(m/s)Performance SpecsMaximum Speed 120 km/hMaximum Engine Power 213 kWFigure 4.1: Class 150 train consists of two Class 150 DMU.684.1. British Class 150: Trehafod to Treherbert4.1.2 Route DescriptionThe route chosen for this study is the 14 kilometer route from Trehafod to Treherbert inthe UK. It is a standard short route that is typically found in urban areas. It has an uneventfulelevation profile. As the train leaves the Trehafod station, it almost steadily gains elevationtill it reaches the destination at Treherbert. The elevation change in 14 kilometers is 100meters (see Figure 4.2), or about 7.14 meters altitude gain for every kilometer traveled. Thiscorresponds to a gradient of less than 1%, which is acceptable for railway operation.5 10 15 20 25Distance (km)020406080100120Altitude(m)Figure 4.2: Trehafod to Treherbert altitude profile.Figure 4.3 shows the velocity profile for the trip. The Simulink simulator works by solvingsystem equations with time as the independent variable. Therefore, if we allow the velocityprofile to reach zero, meaning that the vehicle comes to a full stop at designated stations,the simulator would freeze. As a solution, we generate a velocity profile that does not goto zero, but instead to 1 m/s for a period of time that would correspond to the equivalentdelay of a complete stop. By examining the figure, we can see that the train makes 9 stopsbetween Trehafod and Treherbert resulting in a one way trip time of approximately 23 minutes.Although the train is capable of a 120 km/h top speed, it barely reaches 80 km/h assuming afast-as-possible trajectory as computed using methods discussed in Section 3.3.694.1. British Class 150: Trehafod to Treherbert0020406080100Velocity(km/h)Actual Speed Desired Speed5 10 15Distance (km)20 25Station StopsFigure 4.3: British Class 150 Train Velocity Profile: Trehafod to Treherbert. The inset plot highlightsthe perturbation that occurs when the system changes the mode of operation from Motoring mode toRegenerative braking mode.4.1.3 Battery - Supercapacitor Parallel HybridIn this section, the results of the BSC power train simulation for the given case studyare presented and discussed. The results of this section are highly dependent on how muchpower each source can contribute. Since neither on-board power sources are restricted to acertain maximum power output, they can contribute up to their maximum potential, makingthe mass of each source the deciding factor. Assuming a 3500 kg 4000 L space, a variety of ESScombinations can exist. For illustration purposes, we focus on a case where the battery bankis 500 kg in mass and the SC bank is 1.5 tonnes. 18,630 UPF454261 lithium-ion batteries arearranged in 46 parallel branches of 405 series-connected cells each giving a terminal voltage of1500 volts. The SC bank is made up of 5,786 BCAP3400 SCs arranged in 11 parallel branchesof 526 SCs each.Figure 4.4 presents the power demand profile during the 45 minute round trip from Trehafodto Treherbert on the British Class 150 DMU assuming a 500 kg battery bank and a 1.5 tonne SCbank in each DMU. Motoring mode is defined as the mode of operation when the power drawnfrom the power sources is positive; by contrast regenerative mode is the mode of operationwhen it is negative. The first time the system switches from motoring to regenerating is atapproximately minute 23. The first thing to be noticed is that the battery power is capable704.1. British Class 150: Trehafod to Treherbertof handling almost all of the transient power requirements when the vehicle is moving up hillin motoring mode. As the battery loses charge and thus voltage, power from the overheadcatenary is used to assist in propulsion. Power is not drawn from the SC bank since its voltageis linearly related to its SOC. Due to the very small energy density of SCs, if power is drawnfrom the SC bank, its terminal voltage would drop drastically affecting the vehicle’s speed.Although the SC bank loses charge slightly in the motoring mode, it appears that its mainfunction is to absorb the regenerated power.With the presence of a SC bank, the need for frictional braking was drastically reduced.Figure 4.4 shows the times during which a mechanical brake was required to prevent the vehicle’sspeed from overshooting its desired value. We can see that the braking power needed is smallwhen compared to the amount of regeneration power that is absorbed by the SC bank or thebattery bank.0 5 10 15 20 25 30 35 40 45Time (minutes)-1000100200300Power(kW)Bat tery Power S C Power Catenary Power Frictional Brake PowerFigure 4.4: Trehafod to Treherbert trip power profile. The battery bank mass in each DMU was 500 kg,and the SC bank was 1.5 tonnes. The inset plot highlights how the magnitude of the frictional brakingpower compares with the magnitude of the regenerated power as absorbed by the SC or the batterybank.It is important to examine how the SOC of the on-board ESS changes as the trip progresses.Figure 4.6 presents SOC percentage change of on-board ESS as a function of trip time for thechosen case study. The battery bank’s SOC was 70% depleted in approximately 18 minutes.The battery bank’s SOC depletion was limited to 70% because excessive depletion of lithium-ionbatteries has an adverse impact on their lifetime. The energy regenerated was approximately714.1. British Class 150: Trehafod to Treherbert5.5 kWh as demonstrated by Figure 4.5, which is about 12.2% of the net energy consumed.0 5 10 15 20 25 30 35 40 45Time (minutes)01020304050Energy (kWh)Assuming regenerationNot assuming regenerationFigure 4.5: Trip energy consumption assuming regenerative breaking versus without regenerative brak-ing. The battery bank mass in each DMU was 500 kg, and the SC bank was 1.5 tonnes.The 1.5 tonne SC bank (7 Wh/kg) has the potential to hold 10.5 kWh of energy, makingthe 5.5 kWh regenerated energy approximately 52.3% of full SC charge. However, the SC bankonly gained 45% reaching its maximum charge. The difference between the two percentages isattributed to powertrain inefficiencies including SC self discharge phenomenon.If no SC bank was present, and we assume that the battery bank could absorb power atthe required rate, then a 5.5 kWh energy returned to the battery bank would constitute anapproximate SOC gain of 5.5% assuming a fixed battery bank of 500 kg at an energy densityof 200 Wh/kg. It is important to note that the energy stored in the SC bank can be used torecharge the on-board battery bank.By comparing Figure 4.6 with Figure 4.4, we observe that the more depleted the SC bankis, the less the need for frictional braking. Before the end of the trip, when the SC bank is at80% charge, we observe an exponential increase in frictional brake power.724.1. British Class 150: Trehafod to Treherbert0 5 10 15 20 25 30 35 40 45Time (minutes)20406080100Consumption(%)Bat tery SOC S C SOCFigure 4.6: The battery bank’s SOC, and the SC SOC as a percentage change. The battery bank massin each DMU was 500 kg, and the SC bank was 1.5 tonnes.The mean trip powertrain efficiency for this simulation was 91%. The powertrain efficiency isthe ratio between the electrical power supplied by the source to the mechanical power suppliedby the traction motor. This is an expected value since it represents the efficiency for thePMDC traction machine, the power electronics, and other powertrain inefficiencies. Figure 4.7illustrates how the powertrain efficiency changes with trip time. By inspecting Figure 4.7 alongwith Figure 4.4 we can deduce that the powertrain is more efficient when operating in motoringmode than when operating in regenerative braking mode.734.1. British Class 150: Trehafod to Treherbert0 5 10 15 20 25 30 35 40 45Time (minutes)020406080100120(%)Instantaneous Average Moving Average EfficiencyEfficiencyEfficiencyEfficiencyFigure 4.7: Powertrain efficiency excluding the FC stack efficiency. The average powertrain efficiencywas 91%. The battery bank mass in each DMU was 500 kg, and the SC bank was 3 tonnes.ESS Optimization study In this section we demonstrate how the energy storage mix on-board the locomotive impacts certain key parameters such as the maximum regenerated energy,net SOC change, and range of travel in kilometers without the need to recharge.Figure 4.8 presents a two dimensional view of the ESS optimization problem that ariseswhen deciding on energy storage allocation. The first step is to determine the railway vehicle’sphysical constraints. The gray area in Figure 4.8 shows the ESSs combinations that are notallowed given said physical constraints. The mass constraint in this study was assumed to be3500 kg, and the volume constraint to be 4000 L. The density of both the SC and the batteryis more than one, which is why the volume constraint line and the mass constraint line donot intersect making the mass constraint the only practical physical constraint as shown inFigure 4.8. The discrete ESS combinations selected for simulation are shown as dotted pointsin the same figure. The lightest combination considered for simulation is one where the SCbank mass was 500 kg, and the battery mass was 100 kg.As discussed earlier in Section 3.2, a railway vehicle’s mass is essential for proper traction.Therefore, we can alter the composition of the on-board ESS as long as we do not change thevehicle’s mass. This may necessitate the used of ballast. The ballast is assumed to be made ofsteel, and to have a density of 8.17 kg/L. The further the ESS mix is from the mass constraintline, the more ballast is needed. The use of ballast may not be necessary in BSC hybrids due to744.1. British Class 150: Trehafod to Treherberttheir density. Intuitively, ESS combinations that lie on the mass constraint line result in morerange of travel than the ones that do not since they use all available mass for energy storage.The use of ballast is expected to reduce the overall powertrain cost and therefore is includedin the study.0 500 1000 1500 2000 2500 3000 3500Battery Mass (kg)0500100015002000250030003500SC Mass (kg)Mass ConstraintVolume ConstraintFigure 4.8: A two dimensional illustration of the ESS sizing optimization problem that highlights thecases considered for simulation. The ESS in this study is a BSC hybrid, and the physical constraintsare for a British Class 150 DMU. The dotted circles represent the ESS cases selected for simulation.Figure 4.9 presents the maximum range of travel in kilometers on a single charge as afunction of the on-board battery mass. It was found that the maximum range of travel ona single charge was much more dependent on the on-board battery storage than on the on-board SC storage. This is mainly due to the higher energy density of lithium-ion batteries incomparison to SCs.754.1. British Class 150: Trehafod to Treherbert0 500 1000 1500 2000 2500 3000 3500Battery Mass (kg)0510152025Time (hours)500 kg1000 kg1500 kg2000 kg2500 kg3000 kgSmallest batterybank for 15 hoursof operationFigure 4.9: Maximum range of operation in kilometers travelled without recharging on 70% of thebattery bank’s charge as a function of the on-board battery bank mass for different SC masses (legend).Energy savings due to regeneration as a function of SC mass are presented in Figure 4.10.The smallest increment of battery bank mass considered was 100 kg, leading us to concludethat the battery mass threshold for maximum regeneration is less than 100 kg given that theregenerated energy is almost fixed for all battery sizes over 100 kg.500 1000 1500 2000 2500 3000SC Mass (kg)02468101214EnergyRegeneration(%)500 kg1000 kg1500 kg2000 kg2500 kg3000 kg100 kgFigure 4.10: Energy regenerated as a percentage of net energy consumed as a function of the on-boardESS mix for a BSC hybrid British Class 150 DMU.764.1. British Class 150: Trehafod to Treherbert4.1.4 Fuel Cell - Battery Series HybridIn this section, the results of the FCB power train simulation for the given case study arepresented and discussed. We begin by discussing single-trip plots followed by a discussion ofplots generated for longer periods of operation. The results of this section are highly dependenton how much power each source can contribute. Note that the results with each locomotivefitted with a 20 kW FC will not be the same as results with each fitted with a 100 kW FC. Asensitivity analysis of how the power of the FC stack impacts key trip parameters is presentedat the end of this section. For our trip analysis, we assumed that each DMU was powered bya 80 kW FC.Figure 4.11 presents the power demand profile during the 45 minute round trip from Tre-hafod to Treherbert on the British Class 150 DMU assuming a 80 kW FC system in each DMUand a 3 tonne battery bank. The first thing to be noticed is that the battery power is free tofluctuate while the FC power is constant throughout the trip. The power from the FC systemis controlled by a proportional controller that acts to sustain the charge of the battery bank.The fact that the FC is supplying full power through-out the trip means that the battery bankis below required charge levels throughout the trip.With the absence of a SC bank, the job of storing regenerated energy falls to the batterybank. Figure 4.11 shows the times during which a mechanical brake was required to prevent thevehicle’s speed from overshooting its desired value. We can see that the braking power neededis small when compared to the amount of regeneration power as presented in Figure 4.11.774.1. British Class 150: Trehafod to Treherbert0 5 10 15 20 25 30 35 40 45Time (minutes)-1000100200300Power(kW)Battery Power FC Power Net Power Frictional Brake PowerRegeneration PowerFigure 4.11: Trip power profile. The average FC power was limited to 80 kW in each DMU and thebattery bank mass in each DMU was 3000 kg.Figure 4.12 presents percentage change in the amount of hydrogen stored on-board the trainas a function of trip time, assuming that each DMU is carrying 3 hydrogen tanks. In the samefigure, battery SOC change is plotted. The small change in the battery bank’s SOC can beattributed to its size at 3 tonnes and to the fact that the FC system is constantly charging thebattery. We will demonstrate how battery size and SOC fluctuation are linked in later sections.Unlike the SOC plot, the hydrogen consumption plot cannot increase as hydrogen is not beingregenerated.784.1. British Class 150: Trehafod to Treherbert0 5 10 15 20 25 30 35 40 45Time (minutes)405060708090100Consumption(%)Bat tery SOC Hydrogen Consumpt ionFigure 4.12: The battery bank’s SOC, and the hydrogen consumption as a percentage change. Theaverage FC power was limited to 80 kW in each DMU, the battery bank mass in each DMU was 3000kg, and the hydrogen storage was limited to 3 tanks.The mean trip powertrain efficiency for this simulation was 84.35%. This is an expectedvalue since it represents the efficiency for the PMDC traction machine, the power electronics,and other powertrain inefficiencies. Figure 4.13 illustrates how the powertrain efficiency changeswith trip time. Similar to the previous study, the powertrain is more efficient when operatingin motoring mode than when operating in regenerative braking mode as Figure 4.13 illustrates.0 5 10 15 20 25 30 35 40 45Time (minutes)020406080100120(%)Instantaneous Average Moving Average EfficiencyEfficiencyEfficiencyEfficiencyFigure 4.13: Powertrain efficiency excluding the FC stack efficiency. The average powertrain efficiencywas 84.37%. The average FC power was limited to 80 kW in each DMU and the battery bank mass ineach DMU was 3000 kg.794.1. British Class 150: Trehafod to TreherbertThe above efficiency calculation does not account for the efficiency of the fuel cell system.This efficiency is an electromechanical efficiency. It is very important to include the efficiencyof the FC system in our calculations. The electrochemical efficiency of the FC system for thechosen trip averaged at 64% as Figure 4.14 illustrates. This is perhaps the maximum achievableFC system efficiency. The FC system operated at such a high efficiency because it was supplyingsteady power. If the FC system was allowed to fluctuate more aggressively, its efficiency woulddrop dramatically. The overall efficiency of the entire system including the efficiency of the FCsystem is therefore 54%, a value that is similar to what is to be expected from a diesel-electricrailway vehicle.0 5 10 15 20 25 30 35 40 45Time (minutes)020406080100(%)Instantaneous AverageEfficiencyEfficiency EfficiencyFigure 4.14: Fuel cell instantaneous and average efficiencies. The average FC power was limited to 80kW in each DMU giving an average efficiency of 64%, the battery bank mass in each DMU was 3000kg, and the hydrogen storage was limited to 3 tanks.Fifteen hour work day A single-trip analysis cannot be the only analysis upon whichfeasibility decisions are made. After all, trains usually operate for hours every day whichcalls for the examination of performance parameters given a wider window of operation. Theassumption that each DMU is powered by a 80 kW FC system is extended to this part of thestudy. This section assumes 15 hour work day.By extending the window of operation to 15 hours, a total of 20 consecutive trips can beexamined. For example, Figure 4.15 shows us that the 3 tank per DMU scenario presentedin the previous section would be more than 50% depleted by the end of the 15 hour work804.1. British Class 150: Trehafod to Treherbertday. It also shows that if each DMU carried just one tank of hydrogen, a tank refill wouldbe needed after approximately 8 hours, almost twice a day. However, the fuel economy isindependent of the number of tanks on-board the train. It depends on other factors, primarilythe control system, battery bank size and SOC, and FC power limit. A 100 kW FC consumesmore hydrogen than a 80 kW FC depending on the restrictions set by the control system. Forthis case, the fuel economy was found to be 0.112 tanks per hour or approximately 0.5 kg ofhydrogen per round trip (55.6 km/kg).0 5 10 15Time (hours)020406080100Consumption (%)1 Tank3 Tanks5 Tanks7 Tanks10 TanksFigure 4.15: Hydrogen consumption for a per DMU storage of 1 tank, 3 tanks, 5 tanks, 7 tanks and 10tanks after 15 hours of operation.Extending the window of operation allows us to make better informed decisions on howthe battery SOC changes. As Figure 4.16 demonstrates, the battery SOC change is linear andbounded. As long as the SOC does not reach any of its bounds at 0% or 100%, it linearlydecreases or increases depending on the FC power limit and the bank’s mass.From the figure we can deduce that for a FC power limit of 80 kW in each DMU, anybattery bank of mass less than 500 kg would need to be charged during the 15 hour operationalwindow. It can be argued that a lighter, easily depleted bank can also be easily swapped,eliminating the time needed for charging.814.1. British Class 150: Trehafod to Treherbert0 5 10 15Time (hours)01020304050SOC(%)100 kg500 kg1000 kg1500 kg2000 kg2500 kg3000 kg3500 kgFigure 4.16: Battery SOC for a per DMU battery storage of 100 kg, 500 kg, 1 tonne, 1.5 tonnes, 2tonnes, 2.5 tonnes, 3 tonnes, and 3.5 tonnes for 15 hours of operation.We do not aim to answer the question of which is the better use of available space on-boardthe DMU. We only aim to present a set of feasible scenarios. For example, if cost is to beminimized, a choice between a higher number of tanks combined with a smaller battery bankversus the opposite would have to be made. This choice depends on the answer to the questionof which is costlier, refueling a hydrogen tank or recharging a battery bank. The answer couldbe different if the aim is to minimize emissions. Life-cycle analysis of the manufacturing processof lithium-ion batteries, hydrogen tanks, and PEMFCs would determine which manufacturingprocess emits the most GHG. WTW analysis of hydrogen generation as compared to electricitygeneration for battery bank recharging would answer the question of which combination willdo the most harm to the environment on the long run.ESS Optimization study Similar to the BSC study discussed earlier, in this section wedemonstrate how the energy storage mix on-board the FCB DMU impacts certain key parame-ters such as number of continuous operational hours, regenerated energy, and net SOC change.Again, this section assumes a 80 kW FC on-board each DMU although it will be demonstratedlater how the results are fairly insensitive to the size and power limit of the FC stack. Fig-ure 4.17 presents a two dimensional view of the ESS optimization problem that arises whendeciding on energy storage allocation. The gray area in Figure 4.17 shows the ESSs combi-824.1. British Class 150: Trehafod to Treherbertnations that are not allowed given the physical constraints mentioned earlier. The discreteESS combinations selected for simulation are shown as dotted points in the same figure. Thelightest combination considered for simulation is one where the hydrogen storage is 1 tank, andthe battery mass is 100 kg.Figure 4.17 also shows the mass constraints for each source for a 15 hour operational day.Any combination enclosed between the volume constraint line, the mass constraint line, theminimum battery mass constraint, and the minimum hydrogen mass constraint is a combinationthat would allow continuous operation for 15 hours while obeying the physical constraints. Thatregion is our feasibility set. Simulation results will further determine how the energy storagemix on-board the DMU within that region impacts certain key parameters.The low density of hydrogen makes using it in fixed mass applications a challenge. The morethe hydrogen, the lighter the vehicle, and the more ballast is needed. This is not a problemwith batteries. The more batteries the heavier the vehicle.0 500 1000 1500 2000 2500 3000 3500Battery Mass (kg)02004006008001000Hydrogen Storage Mass (kg)Mass ConstraintMinimum Battery Mass ConstraintMinimum Hydrogen Mass ConstraintVolume ConstraintFigure 4.17: A two dimensional illustration of the ESS sizing optimization problem that highlights thecases considered for simulation. The ESS in this study is a FCB hybrid, and the physical constraintsare for a British Class 150 DMU. The dotted circles represent the ESS cases selected for simulation.Figure 4.18 presents the maximum hours of continuous operation without refueling as afunction of the on-board hydrogen storage. It was found that the maximum range of travelwithout refueling was much more dependent on the on-board hydrogen storage than on the834.1. British Class 150: Trehafod to Treherberton-board battery storage. This is mainly due to the higher energy density of hydrogen incomparison to lithium-ion batteries.1 2 3 4 5 6 7 8 9 10Hydrogen Storage (tanks)051015202530354045Time(hours)100 kg500 kg1000 kg1500 kg2000 kg2500 kg3000 kgFigure 4.18: Maximum number of hours of operation without refueling as a function of the on-boardhydrogen storage mass for a FCB hybrid British Class 150 motive unit for different battery bank masses(legend).Figure 4.19 presents the percentage change in the battery bank’s SOC per operational houras a function of battery mass for different hydrogen tank numbers assuming that each motiveunit had a 80 kW FC system as the prime mover. The figure shows that the battery bank’sSOC is independent of the on-board hydrogen storage. Since the energy supplied by the batterybank is fixed, increasing the battery bank’s mass by a factor decreases the hourly SOC changeby the same factor as presented in the figure.844.1. British Class 150: Trehafod to Treherbert0 500 1000 1500 2000 2500 3000 3500Battery Mass (kg)-20-18-16-14-12-10-8-6-4-20SOC(%/hour)1 tank3 tanks5 tanks7 tanks10 tanksFigure 4.19: The battery bank’s rate of loss of charge as a function of its mass for different hydrogentank numbers assuming that each motive unit had a 80 kW FC system as the prime mover.Energy savings due to regeneration as a function of battery mass for different hydrogenstorage combinations are presented in Figure 4.20. We can observe that for all the ESS com-binations considered, energy regeneration is fixed at approximately 12.2%. The smallest incre-ment of battery bank mass considered was 100 kg, leading us to conclude that the battery massthreshold for maximum regeneration is less than 100 kg.100 500 1000 1500 2000 2500 3000Bat tery Mass (kg)02468101214EnergyRegeneration(%)1 t ank 3 tanks 5 tanks 7 tanks 10 tanksFigure 4.20: Energy regenerated as a percentage of net energy consumed as a function of the on-boardESS mix.854.1. British Class 150: Trehafod to TreherbertSensitivity to FC power per DMU So far in our analysis, we have assumed a fixed FCmaximum power of 80 kW. The FC system can adjust its power according to the battery bank’sSOC but cannot exceed 80 kW. We now investigate how sensitive are the key trip parametersdiscussed earlier to changes to the maximum allowable FC power.Figure 4.21 presents the results of the SOC sensitivity study. In this study, the batterybank’s SOC was recorded at different battery masses and FC system powers. The 100 kgbattery bank has a negative net SOC change for FC powers that are approximately less than22 kW. Not only does it have a positive net SOC change for FC powers higher than 22 kW, thenet change is relatively constant for most of the power range. This suggests that the batterybank is relatively insensitive to FC system power that are above a certain threshold. This is atestimony to the control system’s robustness.The wide gap between the 100 kg line and the 500 kg line as compared to the gaps between allthe other lines suggests that at a certain battery mass threshold, the SOC change is somewhatinsensitive to additional increases in the bank’s mass. It also shows that anything over 500 kgwill not have a significant net SOC change.20 40 60 80 100 120 140Fuel Cell Power (kW)-20-10010203040Δ SOC (%)100 kg 500 kg 1000 kg 1500 kg 2000 kg 2500 kg 3000 kgFigure 4.21: Battery bank SOC change per trip sensitivity to bank’s mass (legend) and FC system powerlimit.The efficiency of the FC system is a strong criteria when it comes to feasibility decisions.For this reason, Figure 4.22 is perhaps one of the most important figures. This is especiallythe case when the difference between the highest and the lowest efficiencies is almost 35%.864.1. British Class 150: Trehafod to TreherbertThe figure shows that at lower FC powers, the FC system efficiency is higher than at higherpowers. This is because higher power FC systems are more responsive to transients given thecontrol system employed. A 130 kW system needs only to supply full power for 15% of thetime needed for a 20 kW system to deliver the same energy. This means that the lower powerFC system would be delivering rated power at maximum efficiency for longer periods of time(see Figure 4.14), while higher power systems would turn on and off more often as they supplytoo much power, which reduces the system’s average efficiency.As mentioned earlier, FC systems are most efficient when operating at rated power. Giventhat all the simulations presented in this thesis relied on the 100 kW Honda FCX stack, withlower powers being examined through limiting the power output of the same stack, efficiencylevels should drop as lower FC powers are examined. By examining Figure 4.22 we can observethat this is true only when the FC power increases from 20 kW to 40 kW for battery masses ofover 500 kg. For battery masses of 500 kg and lower, the trend reverses, and the stack efficiencydrops from a higher value at 20 kW to a lower value at 40 kW. This is because the smaller thebattery bank is, the more aggressively its SOC will fluctuate given the same testing conditions(trip). Aggressive fluctuations in battery SOC will force to the FC system to react accordingly.A higher power FC system will react even more aggressively as it supplies a lot of power for avery short time.By examine the same figure, we can also observe that the stack efficiency drops dramaticallyafter 40 kW for all battery masses, suggesting that the best efficiency levels can be obtainedfor FC powers equal to or less than the average trip power which is somewhere between 40 kWand 60 kW.874.1. British Class 150: Trehafod to Treherbert20 40 60 80 100 120 140Fuel Cell Power (kW)203040506070Efficiency (%)100 kg 500 kg 1000 kg 1500 kg 2000 kg 2500 kg 3000 kgFigure 4.22: The sensitivity of the mean trip FC efficiency to changes in FC power and battery bankmass (legend).Figure 4.23 shows how sensitive fuel economy in tanks per hour is to FC system power andbattery bank mass. If there were no system to control the power output of the FC stack, andeach stack was allowed to operate at its maximum power for the same time period, then thehigher FC power would correspond to more energy delivered and more hydrogen consumed.This, however, is not the case. The system employed to control the power output of the FCsystem has one goal: to maintain the battery bank’s SOC. This means that the FC system hasto provide the average trip power.For FC powers less than the average trip power, fuel economy increases with stack power.This can be observed as the FC power is increased from 20 kW to 40 kW. For FC powers higherthan the average trip power, the power provided by the FC system will fluctuate according tothe battery bank’s SOC which should theoretically lead to higher fuel economies due to thereduced stack efficiency. However, if we observe Figure 4.23, we will see that for FC powershigher than 40 kW, the fuel economy fluctuates around an average of approximately 0.16. Areasonable generalized fuel economy is 0.16 tanks per hour, or 1 tank every 6 hours and 15minutes as deduced from Figure 4.23. A more robust control system would not result in suchfluctuations, some as extreme as 0.19 tanks per hour at 80 kW FC power.884.1. British Class 150: Trehafod to Treherbert20 40 60 80 100 120 140Fuel Cell Power (kW)0.10.120.140.160.180.2Fuel Economy (tanks/hour)100 kg 500 kg 1000 kg 1500 kg 2000 kg 2500 kg 3000 kgFigure 4.23: The sensitivity of the hourly rate of hydrogen consumption to changes in FC power andbattery bank mass (legend).Finally we examine how sensitive energy regeneration is to changes in battery mass and FCsystem power. We had previously shown that energy regeneration is not sensitive to incrementsin battery mass above a certain threshold value. It should come as no surprise that energyregeneration is not in any way dependent on the FC system, since FCs do not store regeneratedenergy. It also seems that the minimum required battery bank mass for full regeneration isbelow 100 kg.20 40 60 80 130Fuel Cell Power (kW)05101520Energy Regenerated (%)100 kg 500 kg 1000 kg 1500 kg 2000 kg 2500 kg 3000 kgFigure 4.24: The sensitivity of the regenerated energy as a percentage of the net energy to changes inFC power and battery bank mass.894.2. Intercity 125: King’s Cross to Newcastle4.2 Intercity 125: King’s Cross to Newcastle4.2.1 Train DescriptionBritish Class 43 is a locomotive that runs on a 1600 kW diesel engine. Two of these unitshauling 6-7 Mark 3 carriages form the Intercity 125 train. Each locomotive has two poweredtwo-axle bogies as shown in Figure 4.25. The case considered in this study is one where atrainset is formed of two locomotives and can carry 600 passengers seated. Table 4.2 containsthe train specifications. The specifications for the British Class 43 locomotive can be found inthe appendix. The space available for hybrid ESS installation was assumed to be approximately20 tonnes at 36000 Liters. This assumption was based on the combined mass and volume ofthe diesel engine at 12 tonnes, and the radiator and fuel compartment at 4 tonnes each.Table 4.2: The specifications of the Intercity 125 train as a single rigid body.Physical SpecsMass 404 tonnesRotational Allowance 15 %Davis Equation (kN) 3.22 + 0.1127v(m/s) + 0.0078v2(m/s)Performance SpecsMaximum Speed 200 km/hFigure 4.25: Intercity 125 train consist: two Class 43 locomotives hauling 8 Mark 3 carriages4.2.2 Route DescriptionThe route chosen for this study is the 432 kilometer route from London’s King’s CrossStation to Newcastle in the UK. It was the chosen route for our intercity rail study. As thetrain leaves the King’s Cross station, it gains 40 meters in elevation only to end up 15 metersbelow London’s altitude after 100 km. The net elevation change in 432 kilometers is less than5 meters (see Figure 4.26), meaning that both stations are at approximately the same elevationlevel. Peak to peak elevation change is slightly over 80 meters with a maximum gradient of 10904.2. Intercity 125: King’s Cross to Newcastlemeters per kilometer or 1%.0 100 200 300 400 500 600 700 800Distance (km)-40-2002040Altitude(m)Figure 4.26: Intercity 125 Train Velocity Profile: London’s King’s Cross to Newcastle Round-trip.As illustrated in Figure 4.27, the train makes 2 stops between King’s Cross Station andNewcastle resulting in a one way trip time of approximately 160 minutes. Unlike the previouscase study, in this case study the train is capable of reaching its top speed of 200 km/h assuminga fast as possible trajectory as computed using methods discussed in Section 3.3.050100150200250Velocity(km/h)Actual Speed Desired Speed1701750 100 200 300 400 500 600 700 800Distance (km)Station StopsFigure 4.27: Intercity 125 Train Velocity Profile: King’s Cross to Newcastle. The average FC power waslimited to 400 kW in each Class 43 locomotive and the battery bank mass to 5 tonnes. The inset plothighlights the perturbation that occurs when the system changes the mode of operation from motoringmode to regenerative braking mode and back to motoring again.914.2. Intercity 125: King’s Cross to Newcastle4.2.3 Battery - Supercapacitor Parallel HybridIn this section, the results of the BSC power train simulation for the given case study arepresented and discussed. Unlike the previous case study, this case study presents a trip that islong enough for single-trip analysis to be sufficient. Another reason for not examining longerdurations is that the best range achieved relying on the on-board ESS was less than 100% ofthe full round trip. Similar to the previous BSC study, the results of this section are highlydependent on the mass of each source. A sensitivity analysis of how the ESS mix impacts keytrip parameters is presented at the end of this section.The results of the conducted simulations demonstrate that battery technology can be suc-cessfully employed in heavy vehicles just like they are in light vehicles. Figure 4.28 presents thepower demand profile during the 318 minute round trip from London’s King’s Cross Stationto Newcastle on the Intercity 125 assuming a 16.5 tonne battery bank in each locomotive andtwo and a half tonne SC bank. The trip shown in the figure was powered by two Class 43locomotives, each containing a battery bank of 609,525 UPF454261 lithium-ion cells arrangedin 1,505 parallel branches of 405 series-connected cells each, and a supercapacitor bank of 5,000BCAP4500 SC units arranged in 9 parallel branches of 526 series-connected capacitors.0 50 100 150 200 250 300Time (minutes)-5000500100015002000Power(kW)Bat tery Power SC Power Catenary PowerFigure 4.28: Intercity 125 Train Velocity Profile: London‘s King’s Cross to Newcastle Round-trip.Similar to the previous BSC case study, the battery bank is capable of providing all thetransient power demand, making the SC bank unnecessary in motoring mode. On the other924.2. Intercity 125: King’s Cross to Newcastlehand, the SC bank overtakes the battery bank when regeneration is needed. It is also noticeablethat the train spends the majority of the trip in the motoring mode, making regeneration a lessimportant feature. The figure also shows the point in time when the battery bank is depletedand the power from the catenary is required.Examining an electric railway vehicle’s fuel economy is essential to any feasibility study. Atapproximately $ 4.3 - 5 million USD per kilometer [64], on-board ESS must significantly reducethe required infrastructure for discontinuous electrification to be feasible. Figure 4.12 presentspercentage change of the ESS stored on-board the train as a function of trip time. The firstthing to be noticed is the almost linear fashion by which the battery bank loses its charge.Since batteries should not be overly charged or overly depleted as mentioned in Section 2.4,the battery SOC is only allowed to swing between 80% to 10% using only 70 % of its charge.Although the SC bank loses slightly over 20% of its initial charge, it is important to rememberthat for this example trip we used a very small SC bank.0 50 100 150 200 250 300Time (minutes)020406080100SOC(%)Bat tery SOC SC SOCFigure 4.29: Intercity 125 Train Velocity Profile: London‘s King’s Cross to Newcastle Round-trip.The mean trip powertrain efficiency for this simulation was 86%. This is very close to the84% efficiency obtained in the previous case study. This is expected since both models are madeup of the same powertrain components. Figure 4.30 illustrates how the powertrain efficiencychanges with trip time.934.2. Intercity 125: King’s Cross to Newcastle0 50 100 150 200 250 300Time (minutes)020406080100(%)Average EfficiencyInstantaneous EfficiencyEfficiencyFigure 4.30: Powertrain efficiency which averaged at 86%.ESS Optimization study In this section we demonstrate how the energy storage mix on-board the locomotive impacts certain key parameters such as the maximum catenaryless rangeof travel in kilometers, regenerated energy, and net SOC change.Figure 4.31 presents a two dimensional view of the ESS optimization problem that ariseswhen deciding on energy storage allocation. Following the same steps in the previous study, thefirst step is to determine the railway vehicle’s physical constraints. The gray area in Figure 4.31shows the ESSs combinations that are not allowed given given the volume and mass constraintsof the vehicle. Similar to the Trehafod to Treherbert BSC study presented earlier, the volumeconstraint line and the mass constraint line do not intersect. The density of both the SCand the battery is more than one, which is why the mass constraint becomes the only physicalconstraint. A fact that is highlighted in Figure 4.31. The discrete ESS combinations selected forsimulation are shown as dotted points in the same figure. The lightest combination consideredfor simulation is one where the SC bank mass is 2 tonnes, and the battery mass is 500 kg.944.2. Intercity 125: King’s Cross to Newcastle0 5 10 15 20 25 30Battery Mass (tonnes)051015202530SC Mass (tonnes)Mass ConstraintVolume ConstraintFigure 4.31: A two dimensional illustration of the ESS sizing optimization problem that highlights thecases considered for simulation. The ESS in this study is a BSC hybrid, and the physical constraints arefor a British Class 43 locomotive. The dotted circles represent the ESS cases selected for simulation.Figure 4.32 presents the maximum range of travel in kilometers without refueling as afunction of the on-board battery mass. It was found that the maximum range of travel withoutrefueling was much more dependent on the on-board battery storage than on the on-board SCstorage. This is mainly due to the higher energy density of lithium-ion batteries in comparisonto SCs.954.2. Intercity 125: King’s Cross to Newcastle0 2 4 6 8 10 12 14 16 18Batter Mass (tonnes)0100200300400500600700Range(km)19500 kg17000 kg14500 kg12000 kg9500 kg7000 kg4500 kg2000 kgFigure 4.32: Maximum range of operation in kilometers traveled without recharging on 70% of thebattery bank’s charge as a function of the on-board battery bank mass for different SC bank sizes.Energy savings due to regeneration as a function of battery mass are presented in Fig-ure 4.33. The smallest increment of battery bank mass considered was 500 kg, leading us toconclude that the battery mass threshold for maximum regeneration is less than 500 kg.0.5 3 5.5 8 10.5 13 15.5 18Bat tery Mass (tonnes)00.10.20.30.40.5EnergyRegeneration(%)19.5 tonnes17 tonnes14.5 tonnes12 tonnes9.5 tonnes7 tonnes4.5 tonnes2 tonnesFigure 4.33: Regenerated energy as a percentage of net energy consumed as a function of the on-boardESS mix.964.2. Intercity 125: King’s Cross to Newcastle4.2.4 Fuel Cell - Battery Series HybridIn this section, the results of the FCB power train simulation for the given case study arepresented and discussed. Similar to the BSC study, this case study presents a trip that is longenough for a single-trip analysis to be sufficient. Yet, analysis for longer durations are includedfor completeness purposes. All the results in this section assume that each British Class 43locomotive is powered by eight 100 kW Honda FCX FC stacks and carries a 5 tonne batterybank to handle transient power demands. A sensitivity analysis of how the power of the FCstack impacts key trip parameters was not conducted. Instead, we extend the conclusions ofthe sensitivity analysis conducted in the Trehafod to Treherbert case study to this case study.Figure 4.34 presents the power demand profile during the 318 minute round trip fromLondon’s King’s Cross Station to Newcastle on the Intercity 125. Similar to the results presentedin Subsection 4.1.4, the battery power fluctuates aggressively. The power from the FC systemis controlled by a proportional controller that acts to sustain the charge of the battery bank.The fact that the FC power is fluctuating, although slightly, means that there is potential forFC stack downsizing.0 50 100 150 200 250 300Time (minutes)-1500-1000-5000500100015002000Power(kW)Bat tery Power FC PowerFigure 4.34: The power supplied by the FC system, and the battery bank. The average FC power waslimited to 800 kW in each Class 43 locomotive and the battery bank mass to 5 tonnes.For an in depth analysis, it is perhaps best to examine the power demand profile of eachsource at a much smaller time scale. Figure 4.35 presents the power demand profile of bothsources starting at minute 144 to minute 154 in the 318 minute trip. In this time span, the974.2. Intercity 125: King’s Cross to Newcastlepower from the FC system fluctuates most aggressively as compared to the rest of the trip. Acloser look at minute 146 reveals that a sudden power drop was encountered. By comparingboth power demand plots at minute 146, we can observe how much faster the battery bankreacts. The fact that the FC system is turned off for the duration of slightly less than 2minutes including a minute during which the battery power profile was negative shows us thatthe system was in regeneration mode from minute 146 to 147. At minute 147, the batterypower profile is positive indicating that the system is in motoring mode, but the FC systemwas slow to react as intended.Another region of interest is the region between minute 151 and 152. We can see that theFC system is providing constant full power while the battery bank’s power makes a massiveswing going up to 800 kW and down to -800 kW. This proves that there was a significant dropin power demand by the traction machines and all the FC power had to be dumped into thebattery bank.145 146 147 148 149 150 151 152 153Time (minutes)-50005001000Power(kW)Bat tery Power FC PowerFigure 4.35: The power supplied by the FC system, and the battery bank for a 10 minute period startingat minute 144 to minute 154.. The average FC power was limited to 800 kW in each Class 43 locomotiveand the battery bank mass to 5 tonnes.Figure 4.36 shows how the powertrain efficiency was unaffected throughout all the conductedsimulations presented in this thesis. The average powertrain efficiency of 86.5% is very similarto the values obtained for the BSC (Figure 4.30) study of the same case study, as well as theBSC and FCB (Figure 4.13) studies of the Trehafod to Treherbert case study.984.2. Intercity 125: King’s Cross to Newcastle0 50 100 150 200 250 300Time (minutes)020406080100120(%)Instantaneous Average Moving AverageEfficiency Efficiency EfficiencyEfficiencyFigure 4.36: Powertrain efficiency excluding the FC stack efficiency which averaged at 86.521%. Theaverage FC power was limited to 800 kW in each Class 43 locomotive and the battery bank mass to 5tonnes.FC hybrids have worse efficiency levels when compared to battery or SC hybrids. Theoverall efficiency of the system is the product of the powertrain efficiency and the FC systemefficiency. As discussed earlier, the powertrain efficiency is almost fixed for all simulations ataround 85%-86%. For this case study, the average FC system efficiency was observed to be54.15%. If the efficiency of converting hydrogen and oxygen to electrical energy is 54.15%,and the efficiency of converting that electrical energy to mechanical energy at the wheel/railinterface is 86.5%, then the overall efficiency of the system is 46.8% which is almost half of theBSC system efficiency for the same case study.Perhaps the most important aspect of any vehicle is its fuel economy. Figure 4.12 presentspercentage change in the amount of hydrogen stored on-board the train as a function of triptime, assuming that each locomotive is carrying 86 hydrogen tanks. In the same figure, batterySOC change is plotted. The small change in the battery bank’s SOC can be attributed to itssize at 5 tonnes and to the fact that the FC system is constantly charging the battery. We willdemonstrate how battery size and SOC fluctuation are linked in later sections. Unlike the SOCplot, the hybrogen consumption plot cannot increase as hydrogen is not being regenerated.994.2. Intercity 125: King’s Cross to Newcastle0 50 100 150 200 250 300Time (minutes)020406080100Efficiency(%)Instantaneous Efficiency Average EfficiencyFigure 4.37: Fuel cell instantaneous and average efficiencies. The average FC efficiency for the trip was54.15%. The average FC power was limited to 800 kW in each Class 43 locomotive and the batterybank mass to 5 tonnes.0 50 100 150 200 250 300Time (minutes)020406080100Consumption(%)Bat tery SOC Hydrogen Consumpt ionFigure 4.38: The battery bank’s SOC and the hydrogen fuel economy as a percentage change. Theaverage FC power was limited to 800 kW in each Class 43 locomotive, the battery bank mass to 5tonnes, and the hydrogen tanks to 86.1004.2. Intercity 125: King’s Cross to NewcastleFifteen hour work day As mentioned earlier, single-trip analysis presents a restricted viewof system performance. By extending the window of operation to slightly less than 16 hours,a total of 3 consecutive trips can be examined. For example, Figure 4.39 shows us that the 86tank per locomotive scenario presented in the previous section would be approximately 35%depleted by the end of the 16 hour work day. The fuel economy is independent of the numberof tanks on-board the train. It depends on other factors, primarily the control system, batterybank size and SOC, and FC power limit. For this case, the fuel economy was found to be 1.9tanks per hour or approximately 56 kg of hydrogen per round trip per locomotive.Extending the window of operation allows us to make better informed decisions on how thebattery SOC behaves. As Figure 4.39 demonstrates, the net battery SOC change is linear andbounded. As long as the SOC does not reach any of its bounds at 0% or 100%, it linearlydecreases or increases depending on the FC power limit and the bank’s mass. From the figurewe can deduce that for a FC power limit of 800 kW in each locomotive, a 5 tonne battery bankwould lose 10% every 3 round trips.0 5 10 15Time (minutes)405060708090100Consumption(%)Bat tery SOC Hydrogen Consumpt ionFigure 4.39: The battery bank’s SOC and the hydrogen fuel consumption as a percentage change for 3consecutive trips. The average FC power was limited to 800 kW in each British Class 43 locomotive,the battery bank mass to 5 tonnes, and the hydrogen tanks to 86.ESS Optimization study In this section we demonstrate how the energy storage mix on-board the locomotive impacts certain key parameters such as the number of continuous opera-tional hours without refueling, regenerated energy as a percentage of the net consumed energy,1014.2. Intercity 125: King’s Cross to Newcastleand net SOC change.Figure 4.40 presents a two dimensional view of the ESS optimization problem that ariseswhen deciding on energy storage allocation. he discrete ESS combinations selected for simu-lation are shown as dotted points in the same figure. The lightest combination considered forsimulation is one where the hydrogen storage mass was 2.5 tonnes, and the battery mass was1 tonne.The use of ballast is necessary in FCB hybrids due to the low density of hydrogen. OnlyESS combinations that lie on the mass constraint line do not require the use of a ballast. Thefurther from the mass constraint line a combination is, the more ballast is needed.0 5 10 15 20Battery Mass (tonnes)012345678910Hydrogen Mass (tonnes)Mass ConstraintVolume ConstraintFigure 4.40: A two dimensional illustration of the ESS sizing optimization problem that highlights thecases considered for simulation. The ESS in this study is a FCB hybrid, and the physical constraintsare for a British Class 43 locomotive. The dotted circles represent the ESS cases selected for simulation.Similar to the conclusions made in the Trehafod to Treherbet FCB study presented earlier,Figure 4.41 highlights how the number of hours of continuous operation (range of travel) is afunction of the on-board hydrogen mass. This is mainly due to the higher energy density ofhydrogen gas in comparison to lithium-ion batteries.1024.2. Intercity 125: King’s Cross to Newcastle0 10 20 30 40 50 60 70 80 90 100Hydrogen Storage (tanks)05101520253035404550Time(hours)2 tonnes3 tonnes5 tonnes6.5 tonnes8 tonnes10 tonnes15 tonnesFigure 4.41: Number of hours of continuous operation without refueling as a function of the on-boardhydrogen storage mass.Figure 4.42 presents the percentage change in the battery bank’s SOC per operational houras a function of battery mass assuming that each motive unit had a 800 kW FC system as theprime mover. The figure shows that the battery bank’s SOC is independent of the on-boardhydrogen storage. Since the energy supplied by the battery bank is fixed, increasing the batterybank’s mass by a factor decreases the hourly SOC change by the same factor as presented inthe figure.0 2 4 6 8 10 12 14 16 18 20Battery Mass (tonnes)-2-1.8-1.6-1.4-1.2-1-0.8-0.6-0.4-0.20SOC(%/hour)24 tanks34 tanks38 tanks48 tanks57 tanks67 tanks76 tanks86 tanksFigure 4.42: The battery bank’s rate of loss of charge as a function of its mass.Energy savings due to regeneration as a function of battery mass are presented in Fig-ure 4.43. The smallest increment of battery bank mass considered was 100 kg, leading us toconclude that the battery mass threshold for maximum regeneration is less than 100 kg.1034.2. Intercity 125: King’s Cross to Newcastle0 2 4 6 8 10 12 14 16Bat tery Mass (kg)00.10.20.30.40.50.6EnergyRegeneration(%)24 tanks34 tanks38 tanks48 tanks57 tanks67 tanks76 tanks86 tanksFigure 4.43: Regenerated energy as a percentage of net energy consumed as a function of the on-boardESS mix.104Chapter 5Conclusions and RecommendationsStated broadly, the goal of the research work presented in this thesis was to assess thetechnical feasibility of various clean power sources stored on-board moving railway vehicles.In particular, battery/supercapacitor hybrids versus fuel cell/battery hybrids were examined.Two case studies that present the extremes of trip length were chosen. Simulink models forthe different powertrains were built and simulated. The discretization of component volumesand corresponding masses enabled the production of a set of feasible solutions. Results ofmodel simulation at the chosen discrete points were plotted, which enabled the examination ofthe impact of component sizing on key trip parameters. It is important to remember that allthe simulated case studies assumed a fast-as-possible velocity profile. This assumption has asignificant impact on energy consumption, most notably on the ability to regenerate energy.Lithium ion batteries proved to be very capable in handling all the required transientpower demand. In regeneration, supercapacitors outperformed lithium-ion batteries and almosteliminated the need for frictional brakes. A study to measure the potential financial savingsfrom reduced mechanical brake wear and tear versus the cost of installing supercapacitor banksis therefore recommended. The studies conducted lead us to conclude that maximizing batterystorage on-board a railway vehicle maximizes the range of catenaryless operation. In fact, ithas been demonstrated that complete elimination of the overhead electrification infrastructureis possible if a sufficient number of batteries are used, moreover, space on-board was not thelimiting factor.For a fixed mass BSC ESS, volume restriction was not a problem. The most volume re-stricting ESS combination occupied less than 50% of the available volume. The more batteries,the less supercapacitors, and the less volume occupied. Energy regeneration was unaffected byESS sizing mix over a specified minimum. Overall, energy regeneration was insignificant for105Chapter 5. Conclusions and Recommendationsthe longer trip, but that is case specific.Although very inefficient when compared to lithium-ion batteries and supercapacitors, hy-drogen fuel cells still managed to outperform other sources when it came to range extension.However, they proved unable to handle transient power demand and had to be hybridized witha more dynamic power source. The robustness of the control system employed in the simulatedFCB models kept the fuel economy almost constant as fuel cell stack power was increased. Thiswould lead us to conclude that a more robust control system could enable the use of fuel cellstacks with maximum power levels lower than the average trip power demand. This may reducepowertrain cost without affecting the fuel economy.Since a railway vehicle’s mass must remain unaffected by ESS sizing changes for tractionpurposes, ballast was required in the case of hydrogen storage. The low density of hydrogen,even when pressurized, meant that the volume restricted compartment on-board the railwayvehicle limited the amount of hydrogen stored. Yet, enough hydrogen could be stored to outlastBSC hybrids.The main conclusion is that clean ESS technology on-board moving railway vehicles arehighly feasible from a technical point of view as a gateway technology to electrify the NorthAmerican railway fleet. It is therefore recommended that an independent cost benefit analysisfor the given case studies be conducted. For a more meaningful feasibility assessment, WTWemissions analysis of the proposed solutions is recommended.Future work:• It is recommended to conduct a study on the potential financial savings from reducedmechanical brake wear and tear versus the cost of installing supercapacitor banks.• An independent cost-benefit study for the case studies presented in this thesis is recom-mended.• WTW emissions analysis of the proposed solutions is recommended.• A study on the application of the proposed technologies to freight rail systems is recom-mended.106Bibliography[1] E. Boozarjomehri, E. Morrison, I. Roth, and G. Lovegrove, “Moving AwayFrom Diesel and Towards All-Electric Locomotives in North America: Planningand Logistics of Ultra-Capacitor/Battery Technology,” in 2012 Joint Rail Conference,2012, p. 777. [Online]. Available: http://www.scopus.com/inward/record.url?eid=2-s2.0-84892649247{&}partnerID=tZOtx3y1 → pages 2[2] M. Ogasa, “Application of energy storage technologies for electric railway vehicles-exampleswith hybrid electric railway vehicles,” IEEJ Transactions on Electrical and ElectronicEngineering, vol. 5, no. 3, pp. 304–311, 2010. → pages 2, 9, 12, 13[3] L. Mir, I. Etxeberria-Otadui, D. Arenaza, I. Perez, I. Sarasola, and T. Nieva, “A Superca-pacitor Based Light Rail Vehicle : System design and operations modes,” in IEEE EnergyConversion Congress and Exposition, vol. 146. San Jose, CA: IEEE, 2009, pp. 1632–1639.→ pages[4] Z. Gao, J. Fang, Y. Zhang, D. Sun, L. Jiang, and X. Yang, “Control Research of Super-capcitor Energy Storage System for Urban Rail Transit Network,” in Information Science,Electronics and Electrical Engineering (ISEEE), 2014 International Conference on, 2014,pp. 181–185. → pages[5] I. Szenasy, “Capacitive and hybrid energy storages for metro railcars,” IEEE InternationalElectric Vehicle Conference, pp. 8–12, 2013. → pages 2[6] A. R. Miller, K. S. Hess, D. L. Barnes, and T. L. Erickson, “System design of a largefuel cell hybrid locomotive,” Journal of Power Sources, vol. 173, no. 2 SPEC. ISS., pp.935–942, 2007. → pages 3, 14, 24107Bibliography[7] C. R. Akli, B. Sareni, X. Roboam, and A. Jeunesse, “Integrated optimal design of a hy-brid locomotive with multiobjective genetic algorithms,” International Journal of AppliedElectromagnetics and Mechanics, vol. 30, no. 3-4, pp. 151–162, 2009. → pages 3, 26[8] A. Jaafar, C. R. Akli, B. Sareni, X. Roboam, and A. Jeunesse, “Sizing and energy man-agement of a hybrid locomotive based on flywheel and accumulators,” IEEE Transactionson Vehicular Technology, vol. 58, no. 8, pp. 3947–3958, 2009. → pages 3, 24[9] M. Meinert, P. Prenleloup, S. Schmid, and R. Palacin, “Energy storage technologies andhybrid architectures for specific diesel-driven rail duty cycles: Design and system integra-tion aspects,” Applied Energy, vol. 157, pp. 619–629, 2015. → pages 4, 11[10] L. Yu, J. H. He, H. T. Yip, F. Du, Z. Q. Bo, and J. Hu, “Simulation of regenerative brak-ing in DC railway system based on MATLAB/Simulink,” Universities Power EngineeringConference UPEC 2010 45th International, no. iv, pp. 1–5, 2010. → pages 8[11] S. M. Mousavi Gazafrudi, A. Tabakhpour Langerudy, E. F. Fuchs, and K. Al-Haddad,“Power Quality Issues in Railway Electrification: A Comprehensive Perspective,” IEEETransactions on Industrial Electronics, vol. 62, no. 5, pp. 3081–3090, 2015. → pages 8[12] A. Ogunsola, A. Mariscotti, and L. Sandrolini, “Estimation of stray current from a DC-electrified railway and impressed potential on a buried pipe,” IEEE Transactions on PowerDelivery, vol. 27, no. 4, pp. 2238–2246, 2012. → pages 9[13] L. Bertolini, M. Carsana, and P. Pedeferri, “Corrosion behaviour of steel in concrete inthe presence of stray current,” Corrosion Science, vol. 49, no. 3, pp. 1056–1068, 2007. →pages 9[14] M. Steiner, M. Klohr, and S. Pagiela, “Energy storage system with Ultracaps on board ofrailway vehicles,” European Conference on Power Electronics and Applications, 2007. →pages 9, 13[15] F. Kiessling, Contact Lines for Electric Railways. Planning, Design, Implementation,Maintenance. Erlangen : Publicis. → pages 9108Bibliography[16] R. Hou, Y. Yang, and A. Emadi, “Hybrid electric locomotive powertrains,” IEEE Trans-portation Electrification Conference and Expo, ITEC Asia-Pacific, pp. 1–6, 2014. → pages10[17] “Electrical Transportation Annual Report of Committee on Transportation,” Detroit,Michigan, Tech. Rep., 1927. → pages 10[18] E. I. Altman, “Predicting railroad bankruptcies in America,” The Bell Journal of Eco-nomics and Management Science, vol. 4, no. 1, pp. 184–211, 1973. → pages 10[19] T. Murray, The Milwaukee Road. St. Paul, Minn. : MBI Pub. Co. → pages 10[20] A. Hoffrichter, J. Silmon, F. Schmid, S. Hillmansen, and C. Roberts, “Feasibility of discon-tinuous electrification on the Great Western Main Line determined by train simulation,”Proceedings of the Institution of Mechanical Engineers, Part F: Journal of Rail and RapidTransit, vol. 227, no. 3, pp. 296–306, 2012. → pages 10[21] R. Cousineau, “Development of a Hybrid Switcher Locomotive,” IEEE Instrumentation &Measurement Magazine, no. February, pp. 25–29, 2006. → pages 11[22] S. Subhedar, P. Majumdar, R. Kilaparti, and D. Schroeder, “Duty cycle and transportphenomenon analysis of battery storage system for hybrid locomotive application,” inASME 2012 Rail Transportation Division Fall Technical Conference, Omaha, Nebraska,2012, pp. 169–179. → pages 24[23] O. Mitsuyuki, T. Shinji, F. Takehito, T. Eiichi, K. Takashi, and S. Motomi, “Developmentof a New Energy Train hybrid type ,” in The World Congress on Railway Research 2003,2003, pp. 48–54. → pages 11[24] H. Hirose, K. Yoshida, and K. Shibanuma, “Development of catenary and storage batteryhybrid train system,” in Electrical Systems for Aircraft, Railway and Ship Propulsion.IEEE, oct 2012, pp. 1–4. → pages vii, 12[25] G. D. Marin, G. F. Naterer, and K. Gabriel, “Rail transportation by hydrogen vs. electrifi-cation - Case study for Ontario, Canada, II: Energy supply and distribution,” InternationalJournal of Hydrogen Energy, vol. 35, no. 12, pp. 6097–6107, 2010. → pages 14109Bibliography[26] ——, “Rail transportation by hydrogen vs. electrification - Case study for Ontario, Canada,I: Propulsion and storage,” International Journal of Hydrogen Energy, vol. 35, no. 12, pp.6097–6107, 2010. → pages 14[27] A. R. Miller, “Tunneling and mining applications of fuel cell vehicles,” Fuel Cells Bulletin,vol. 3, no. 22, pp. 5–9, 2000. → pages 14[28] C. R. Akli, X. Roboam, B. Sareni, and A. Jeunesse, “Energy management and sizing of ahybrid locomotive,” European Conference on Power Electronics and Applications, vol. c,pp. 1–10, 2007. → pages 14, 24[29] A. Hoffrichter, P. Fisher, J. Tutcher, S. Hillmansen, and C. Roberts, “Performance evalua-tion of the hydrogen-powered prototype locomotive ’Hydrogen Pioneer’,” Journal of PowerSources, vol. 250, pp. 120–127, 2014. → pages 15[30] W. Chen, F. Peng, Z. Liu, Q. Li, and C. Dai, “System integration of China’s first PEMFClocomotive,” Journal of Modern Transportation, vol. 21, no. 3, pp. 163–168, 2013. → pages15[31] O. Tremblay, “Experimental Validation of a Battery Dynamic Model for EV ApplicationsExperimental Validation of a Battery Dynamic Model for EV Applications,” vol. 3, no.October, pp. 289–298, 2015. → pages vii, 16[32] S. N. Motapon, O. Tremblay, and L.-A. Dessaint, “Development of a generic fuel cell model:application to a fuel cell vehicle simulation,” International Journal of Power Electronics,vol. 4, no. 6, pp. 505–522, 2012, pMID: 52427. → pages vii, 17, 20[33] J.-J. Hwang, Y.-J. Chen, and J.-K. Kuo, “The study on the power management system ina fuel cell hybrid vehicle,” International Journal of Hydrogen Energy, vol. 37, no. 5, pp.4476–4489, 2012. → pages vi, vii, 17, 20, 117[34] J. Cho, H. S. Kim, and K. Min, “Transient response of a unit proton-exchange membranefuel cell under various operating conditions,” Journal of Power Sources, vol. 185, no. 1,pp. 118–128, 2008. → pages vii, 18, 19110Bibliography[35] C. Liu and L. Liu, “Optimal power source sizing of fuel cell hybrid vehicles based on Pon-tryagin’s minimum principle,” International Journal of Hydrogen Energy, vol. 40, no. 26,pp. 8454–8464, 2015. → pages 18[36] C. H. Zheng, N. W. Kim, and S. W. Cha, “Optimal control in the power managementof fuel cell hybrid vehicles,” International Journal of Hydrogen Energy, vol. 37, no. 1, pp.655–663, 2012. → pages 18[37] R. L. Borup, J. R. Davey, F. H. Garzon, D. L. Wood, and M. A. Inbody, “PEM fuel cellelectrocatalyst durability measurements,” Journal of Power Sources, vol. 163, no. 1 SPEC.ISS., pp. 76–81, 2006. → pages 18[38] W. S. Lin and C. H. Zheng, “Energy management of a fuel cell/ultracapacitor hybridpower system using an adaptive optimal-control method,” Journal of Power Sources, vol.196, no. 6, pp. 3280–3289, 2011. → pages 19[39] M. C. Kisacikoglu, M. Uzunoglu, and M. S. Alam, “Load sharing using fuzzy logic controlin a fuel cell/ultracapacitor hybrid vehicle,” International Journal of Hydrogen Energy,vol. 34, no. 3, pp. 1497–1507, 2009. → pages 19[40] P. Thounthong, S. Rae¨l, and B. Davat, “Energy management of fuelcell/battery/supercapacitor hybrid power source for vehicle applications,” Journalof Power Sources, vol. 193, no. 1, pp. 376–385, 2009. → pages 19, 25, 57[41] A. Arce, A. J. del Real, and C. Bordons, “MPC for battery/fuel cell hybrid vehiclesincluding fuel cell dynamics and battery performance improvement,” Journal of ProcessControl, vol. 19, no. 8, pp. 1289–1304, 2009. → pages 20[42] Z. Yu, D. Zinger, and A. Bose, “An innovative optimal power allocation strategy for fuelcell, battery and supercapacitor hybrid electric vehicle,” Journal of Power Sources, vol.196, no. 4, pp. 2351–2359, 2011. → pages 20[43] C. Y. Li and G. P. Liu, “Optimal fuzzy power control and management of fuel cell/batteryhybrid vehicles,” Journal of Power Sources, vol. 192, no. 2, pp. 525–533, 2009. → pages20111Bibliography[44] P. Thounthong, P. Tricoli, and B. Davat, “Performance investigation of linear and nonlin-ear controls for a fuel cell/supercapacitor hybrid power plant,” International Journal ofElectrical Power and Energy Systems, vol. 54, pp. 454–464, 2014. → pages 20[45] T. Q. Hua, R. K. Ahluwalia, J. K. Peng, M. Kromer, S. Lasher, K. McKenney, K. Law,and J. Sinha, “Technical assessment of compressed hydrogen storage tank systems forautomotive applications,” International Journal of Hydrogen Energy, vol. 36, no. 4, pp.3037–3049, 2011. → pages 21[46] J. Baert, S. Jemei, D. Chamagne, D. Hissel, S. Hibon, and D. Hegy, “Sizing of a hybridlocomotive,” in Vehicle Power and Propulsion Conference. IEEE, 2011. → pages 24[47] D. Fares, R. Chedid, F. Panik, S. Karaki, and R. Jabr, “Dynamic programming techniquefor optimizing fuel cell hybrid vehicles,” International Journal of Hydrogen Energy, vol. 40,no. 24, pp. 7777–7790, 2015. → pages 26[48] D. Tarsitano, L. Mazzola, S. Arrigoni, F. L. Mapelli, and F. Cheli, “Energy managementalgorithms comparison for an electric bus with an hybrid energy storage system by meansof dynamic programming,” IEEE Vehicular Technology Conference, 2014. → pages[49] S. Lu, S. Hillmansen, and C. Roberts, “A power-management strategy for multiple-unitrailroad vehicles,” IEEE Transactions on Vehicular Technology, vol. 60, no. 2, pp. 406–420,2011. → pages 26[50] D. Fares, R. Chedid, S. Karaki, R. Jabr, F. Panik, H. Gabele, and Y. Huang, “Optimalpower allocation for a FCHV based on linear programming and PID controller,” Inter-national Journal of Hydrogen Energy, vol. 39, no. 36, pp. 21 724–21 738, 2014. → pages26[51] V. Herrera, A. Saez-de Ibarra, A. Milo, H. Gaztanaga, and H. Camblong, “Optimal En-ergy Management of a Hybrid Electric using Genetic Algorithm,” Electrical Systems forAircraft, Railway, Ship Propulsion and Road Vehicles (ESARS), 2015 International Con-ference on, no. 6, pp. 3–5, 2015. → pages 26112Bibliography[52] M. Jain, C. Desai, and S. S. Williamson, “Genetic algorithm based optimal powertraincomponent sizing and control strategy design for a fuel cell hybrid electric bus,” VehiclePower and Propulsion Conference, 2009. VPPC ’09. IEEE, pp. 980–985, 2009. → pages26[53] J. Ribau, R. Viegas, A. Angelino, A. Moutinho, and C. Silva, “A new offline optimizationapproach for designing a fuel cell hybrid bus,” Transportation Research Part C: EmergingTechnologies, vol. 42, pp. 14–27, 2014. → pages 26[54] C. Dextreit and I. Kolmanovsky, V, “Game Theory Controller for Hybrid Electric Vehi-cles,” IEEE Transactions on Control Systems Technology, vol. 22, no. 2, pp. 652–663, 2014.→ pages 26[55] J. Shen and A. Khaligh, “A Supervisory EnergyManagement Control Strategy in a Bat-tery/Ultracapacitor Hybrid Energy Storage System,” IEEE Transactions on Transporta-tion Electrification, vol. 1, no. 3, pp. 223–231, 2015. → pages 26[56] M. Jain, C. Desai, N. Kharma, and S. S. Williamson, “Optimal Powertrain ComponentSizing of a Fuel Cell Plug-in Hybrid Electric Vehicle Using Multi- Objective Genetic Al-gorithm,” IECON ’09. 35th Annual Conference of IEEE on Industrial Electronics, pp.3741–3746, 2009. → pages 26[57] Q. Wu, S. Luo, and C. Cole, “Longitudinal dynamics and energy analysis for heavy haultrains,” Journal of Modern Transportation, vol. 22, no. 3, pp. 127–136, 2014. → pages 33[58] S. Lu, S. Hillmansen, T. K. Ho, and C. Roberts, “Single-train trajectory optimization,”IEEE Transactions on Intelligent Transportation Systems, vol. 14, no. 2, pp. 743–750,2013. → pages 34[59] J. Bernard, M. Hofer, U. Hannesen, A. Toth, A. Tsukada, F. N. Bu¨chi, and P. Dietrich,“Fuel cell/battery passive hybrid power source for electric powertrains,” Journal of PowerSources, vol. 196, no. 14, pp. 5867–5872, 2011. → pages 64[60] P. Garcia, L. M. Fernandez, C. A. Garcia, and F. Jurado, “Energy management system of113Bibliographyfuel-cell-battery hybrid tramway,” IEEE Transactions on Industrial Electronics, vol. 57,no. 12, pp. 4013–4023, 2010. → pages 64[61] a. Ravey, N. Watrin, B. Blunier, and A. Miraoui, “Energy sources sizing for hybrid fuel cellvehicles based on statistical description of driving cycles,” Ieee Transactions on VehicularTechnology, vol. 60, no. 9, pp. 4164–4174, 2010. → pages 64[62] P. Thounthong, V. Chunkag, P. Sethakul, B. Davat, and M. Hinaje, “Comparative studyof fuel-cell vehicle hybridization with battery or supercapacitor storage device,” IEEETransactions on Vehicular Technology, vol. 58, no. 8, pp. 3892–3904, 2009. → pages 64[63] A. Fadel and B. Zhou, “An experimental and analytical comparison study of power man-agement methodologies of fuel cell-battery hybrid vehicles,” Journal of Power Sources, vol.196, no. 6, pp. 3271–3279, 2011. → pages 64[64] D. R. Peterman, J. Frittelli, and W. J. Mallett, “High Speed Rail (HSR) in the UnitedStates,” Tech. Rep., 2009. → pages 93114Appendix115Appendix ATablesTable A.1: The specifications of Panasonic’s UPF454261 Lithium-ion battery.UPF454261 specificationsRated capacity: 1450 mAhNominal voltage: 3.7 VWeight: 27.0 gEnergy density: Volumetric 462 Wh/lGravimetric 199 Wh/kgTable A.2: The specifications of Maxwell’s BCAP3400 supercapacitor.BCAP3400 specificationsRated capacitance: 3400 FRated capacity: 1400 mAhNominal voltage: 2.85 VWeight: 520 gEnergy density: Volumetric 7.9e-6 Wh/lGravimetric 7.7 Wh/kg116Appendix A. TablesTable A.3: The specifications of the 100 kW Honda FCX PEMFC model used in this study as obtainedfrom [33].PEMFC specificationsStack power: Nominal 85.5 kWMaximum 100 kWInternal resistance: 0.17572 ΩNernst voltage: 1.1729 V/cellNominal utilization: Hydrogen 95.24 %Oxidant 50.03 %Nominal consumption: Fuel 794.4 slpmAir 1891 slpmExchange current: 0.024152 AExchange coefficient: 1.1912Fuel composition: 99.95 %Oxidant composition: 21 %Fuel flow rate at nominalhydrogen utilization:Nominal 374.8 lpmMaximum 456.7 lpmAir flow rate at nominaloxidant utilization:Nominal 1698 lpmMaximum 2069 lpmSystem temperature: 368 KFuel supply pressure: 3 barAir supply pressure: 3 barTable A.4: The specifications of British Class 43 Locomotive.Physical SpecsLength 17.79 mWidth 2.74 mHeight 3.8 mWheel Diameter 1.016 mMass 70.25 tonnesFuel Capacity 4500 LPerformance SpecsMaximum Speed 200 km/hMaximum Engine Power 1600 kW117Appendix A. TablesTable A.5: The gradient profile of the Trehafod to Treherbert trip.Distance(km) Grade(m/km)0 9.900990.14079 14.705880.33387 13.698630.46259 21.27660.51086 3.690040.57522 2.272730.6074 2.50.83266 2.72480.96138 10.204081.04183 9.009011.21882 7.194241.54062 9.259261.58889 11.764711.63716 10.526321.68543 5. .. .. .. .. .. .. .11.17853 8.5470111.40379 9.7087411.66123 7.9365111.75777 5.7471312.16002 6.8965512.54618 7.8740212.75535 6.8493212.94843 8.2644613.54376 6.9930113.70466 10.2040813.84947 11.90476118Appendix A. TablesTable A.6: The speed limit profile of the Trehafod to Treherbert trip.Distance (km) Velocity (km/h)0 48.270.30168 48.270.36202 64.36. .. .1.63094 64.361.72967 40.225. .. .1.99661 40.2251.99935 32.182.03866 32.182.07797 64.36. .. .4.18339 64.364.58564 80.45. .. .9.01042 80.459.31211 56.3159.5882 56.3159.58911 56.3159.61379 96.54. .. .12.94879 96.5413.79717 80.4513.83739 80.4513.87396 80.4513.9023 0119

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.24.1-0320830/manifest

Comment

Related Items