Loading...
HomeMy WebLinkAboutPillar Mountain High Penetration Wind Flywheel and Battery Energy Storage System Evaluation - Oct 2011 - REF Grant 7050803WWW.EPSINC.COM PHONE (425) 883-2833 4020 148th AVE NE, SUITE C, REDMOND, WASHINGTON 98052 FAX (425) 883-0464 PHONE (907) 522-1953 3305 ARCTIC BLVD., SUITE 201, ANCHORAGE, ALASKA 99503 FAX (907) 522-1182 st Flywheel and Battery Energy Storage System Evaluation for Kodiak Electric Association, Inc. October 13, 2011 John D.L. Hieb James W. Cote, Jr., Ph.D., P.E. Daniel C. Rogers, P.E. Kodiak Flywheel and Battery Energy Storage System Evaluation WWW.EPSINC.COM PHONE (425) 883-2833 4020 148th AVE. NE, SUITE C, REDMOND, WASHINGTON 98052 FAX (425) 883-0464 PHONE (907) 522-1953 3305 ARCTIC BLVD., SUITE 201, ANCHORAGE, ALASKA 99503 FAX (907) 522-1182 Summary of Changes Revision Revision Date Revision Description 0 October 13, 2011 Initial Release Table of Contents 1 Executive Summary .............................................................................................................. 1 2 Introduction ........................................................................................................................... 1 3 Study Preparations ................................................................................................................ 2 3.1 Initial Assumptions .......................................................................................................... 2 3.2 Wind Data Analysis ........................................................................................................ 3 4 Flywheel Evaluation .............................................................................................................. 6 4.1 Increasing Terror Lake Integral Gain .............................................................................. 7 4.2 Automatic Generation Control ........................................................................................ 8 4.3 Flywheel VAR Capability ................................................................................................ 8 4.4 Flywheel Parameter Tuning ............................................................................................ 9 4.5 Minimum Number of Flywheel Modules ....................................................................... 10 5 Battery Evaluation ............................................................................................................... 13 5.1 BESS controls .............................................................................................................. 13 5.2 BESS Extreme Ramp Results ...................................................................................... 14 5.3 Battery Expected Life Calculations ............................................................................... 16 6 Conclusions ......................................................................................................................... 20 Appendix A – Wind Data Analysis .............................................................................................. 21 Appendix B – Severe 2-Minute Simulations with Flywheels ....................................................... 22 Appendix C – Severe 2-minute Ramp Simulations with BESS ................................................... 23 Appendix D – Xtreme Power BESS Loss of Life Calculation ...................................................... 24 1 Executive Summary EPS recommends that KEA pursue the battery technology over the flywheel technology based on cost and performance. Comparable performance can be achieved using approximately 12 flywheel modules versus two BESS systems. One battery system initially costs twice that of a flywheel module. The battery bank replacement cost is a comparably low $300,000. The expected life cycle of the battery packs is more than 16 years. The preferred controls are based on an aggressive droop type response. The response should have a deadband around 60 Hz to prevent constant storage charge/discharge. Based on the simulation results, the droop setting should be set to 0.85% with a frequency deadband of 59.90 – 60.02 Hz. These parameters should be evaluated during commissioning and revisited based on system operational experience. 2 Introduction Electric Power Systems Inc. (EPS) was contracted to perform a study to determine the system requirements for an energy storage system designed to counteract the negative effects wind has upon the Kodiak Electric Association (KEA) frequency regulation. EPS evaluated the PowerStore™ flywheel modules from PowerCorp as well as the Xtreme Power battery energy storage system DPR™ 15-100C. The base PowerStore™ module has ratings of 1 MW and 18 MW-seconds of energy. The base DPR™ module has ratings of 1.5 MW with 1 MW-hour of energy. EPS was tasked with determining the best device to assist with the frequency regulation of the KEA system. Due to the technological differences between the flywheel and the battery systems, the sizing and control systems were investigated separately. This allowed the control schemes to be tuned to best utilize the strengths and minimize the weaknesses of each control system. Upon determining the sizing and control system needs, the flywheel and battery systems were compared using their expected life-cycle costs. This report is the last in a series of progress reports and is meant close issues brought up in the previous reports as well as to provide final conclusions. The previous progress reports are listed below: July 13, 2011 – Progress report #1 for Flywheel and Battery Energy Storage System Evaluation August 15, 2011 – Progress report for Flywheel and Battery Energy Storage System (BESS) Evaluation Additionally this report will reference results from several previous studies conducted by EPS for KEA leading up to this project. The previous studies are listed below with a quick abstract for each: July 8, 2010 – Third Terror Turbine / Additional 4.5 MW Wind Generation Study o This was the first study looking at the addition of 4.5 MW of additional wind to the KEA system. This study evaluated whether the addition of a third turbine at the Terror Lake Hydro Generation station would be able to provide enough additional frequency regulation to compensate for the addition of another 4.5 MW of wind. This study performed transient stability analysis such as unit trips and line faults. This study also benchmarked the simulated KEA frequency regulation as compared to actual system recordings for a 15-minute system snapshot. The conclusion was the third Terror Lake turbine would improve system frequency regulation, but would not fully compensate for the additional wind effects on frequency regulation. August 20, 2010 – Supplementary Wind Benchmark Cases o This report serves as an addendum to the July 8, 2010 report. The purpose of this study was to determine how the system frequency regulation of today compares to the proposed cases with an additional Terror Lake turbine and 1, 2, or 3 additional wind turbines. The conclusion of this report was that by adding the third Terror Lake turbine and one more wind turbine, the frequency regulation would be better than today. However, adding two wind turbines would have a negative impact on the system frequency regulation. March 4, 2011 – Inertia and Needle Valve Sensitivity for the Third Terror Lake Turbine o Based to the results of the August 20, 2010 study, this study focused on improving the expected frequency regulation by improving the response of the Terror Lake generation station. The study investigated adding inertia to the third Terror Lake turbine and decreasing the needle opening times for the Terror Lake turbine governor system. This report also analyzed the impact of developing and using a Terror Lake unit in condense mode. This report determined that the additional inertia would have limited impact on the minute-by-minute frequency regulation, and that changing the needle opening times would have no impact on the frequency regulation due to the existing governor tuning. This result was troubling since it highlighted the issue that in order to prevent oscillation between the electric system and the Terror Lake tunnel, the governor may need to be “de- tuned” when the third Terror Lake turbine is added. April 25, 2011 – Initial Flywheel Evaluation o This study was intended to provide a proof of concept for an energy storage solution for the KEA frequency regulation issue. This study used the PowerStore™ flywheel system to assist the Terror Lake generation station in regulating the KEA frequency. It showed that there is significant benefit to having an energy storage system for the frequency regulation. 3 Study Preparations 3.1 Initial Assumptions EPS has made the following assumptions as part of the performance criteria for the various phases of this and previous studies: 1. With the 3 additional wind turbines and an additional Terror Lake Turbine, o The frequency regulation shall be no worse than the current frequency regulation assuming the use of a proposed energy storage system. o Diesels will not be used to assist the frequency regulation unless required for maximum power or energy needs. o The Terror Lake governors will need to be “detuned”. This was modeled in our studies by multiplying the integrator gain constant in the governor speed droop control by 0.67. o The actual wind power output from the wind farms will be twice the values seen today, based on the following assumptions: Geographical diversity of the wind turbines was not accounted for in this study as the number of turbines is very small as is the geographic footprint of the wind farm. This assumption also provides some margin in the study results as any wind diversity will likely have a positive impact on the system frequency regulation. 3.2 Wind Data Analysis KEA provided one year’s worth of wind output and frequency data from the year 2010. This data combined with the original 15-minute worst case wind fluctuation event (Case 1 provided by KEA) was used to determine the limiting conditions on the KEA system. The frequency data was analyzed to determine the relative severity of the original Case 1 event. The data was analyzed by computing the standard deviation of the frequency in each 15 minute interval for the whole year, as discussed in previous reports. The results were highlighted based on severity to visualize how often severe frequency changes occurred. Figure 1 shown below is an example of this review. The spreadsheet used to create this figure is attached in Appendix A. The yellow highlighted cells are frequency standard deviations close to, but not as severe as the originally Case 1. The red cells represent time periods when the standard deviation of frequency is worse than observed in Case 1. The results in the figure show these frequency results for January 5th to January 12th 2010. Figure 1: January Frequency Regulation Visualization Unfortunately, there were many 15-minute time intervals in 2010 with standard deviations of frequency that were more severe than Case 1. EPS used Case 1 as a severe condition test case and based the control strategy around maintaining the frequency regulation for this case. Since there were time intervals that were more severe than Case 1, any proposed solution had to be able to withstand the most severe wind ramp events. Using the standard deviation of frequency calculations, the standard deviation values were sorted by severity, and counted. Table 1 shows the number of 15 minute time intervals that have frequency deviations for each of the ranges (types) shown in the left most column. As the standard deviation value increases, the number of time intervals decreases. Each count in Table 1 represents one 15-minute time frame from 2010. STD. DEV Range JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC Type 10 0.05 1791 1807 1861 2062 1736 813 623 1047 1818 1790 1445 726 Type 20.05 0.075 575 493 596 467 676 1086 916 682 599 690 649 815 Type 30.075 0.10 256 216 239 169 247 485 720 615 291 278 332 417 Type 40.10 0.125 168 116 132 81 152 234 437 460 116 104 267 472 Type 50.125 0.15 106 36 75 70 82 109 203 129 36 43 139 418 Type 60.15 0.175 49 15 30 27 54 73 54 32 15 33 41 104 Type 70.175 0.20 18 2 19 4 21 47 18 8 4 23 7 17 Type 80.20 0.225 4 0 15 0 2 28 4 2 1 12 3 3 Type 90.225 0.25 6 0 1 0 4 4 1 1 0 3 1 3 Type 10>0.25 3 3 4 0 2 1 0 0 0 0 0 1 Table 1: Wind Types Based on Standard Deviation of Frequency In order to compare the life-cycle costs of the competing technologies, 30 different time frames were selected based on the standard deviation of frequency and are shown in Table 2. These standard deviations were spread throughout the year to capture seasonal differences in wind regimes. Three snapshots were taken for each “type” listed in Table 1. These three snapshots represent high, medium and low standard deviations within the specified range. For example, the Type 4 row in Table 2 is represented by April 7, 22:00, August 24, 12:15, and December 19, 22:00. These fifteen minute simulations have standard deviations of frequency of 0.1009, 0.1124, and 0.1247 Hz respectively. Type Std. Dev Range Low Std. Dev Mid Std. Dev High Std. Dev 1 0 –0.05 Jan 2, 2:45 May 18, 16:30 Sep 25, 13:15 2 0.05 –0.075 Feb 16, 1:00 Jun 26, 12:45 Oct 20, 8:00 3 0.075 –0.10 Mar23, 16:15 Jul 6, 9:15 Nov 26, 18:45 4 0.10 –0.125 Apr 7, 22:00 Aug24, 12:15 Dec19, 22:00 5 0.125 –0.150 Jan 12, 8:00 May 30, 1:00 Sep 24,20:15 6 0.150 –0.175 Feb 14, 21:45 Jun 2, 14:15 Oct 8, 18:15 7 0.175 –0.20 Mar6, 19:15 Jul 5, 20:15 Nov 10, 14:45 8 0.20 –0.225 Mar9, 14:30 Aug23, 10:30 Dec8, 7:00 9 0.225 –0.250 Jan 11, 16:45 May 31, 22:00 Oct 1, 21:31 10 0.250 –Higher Jan 11, 17:15 Jun 2, 19:00 Mar4, 0:30 Table 2: Selected Snapshots for Life-Cycle Calculation The 30 selected cases in Table 2 were then simulated with each proposed energy storage solution in place. The battery vendor, Xtreme Power then evaluated the impact each case would have on battery life cycle. EPS also analyzed the wind data looking for system trip events where multiple wind units tripped in rapid succession. This was done by selecting each 5-second time frame in the annual data and determining if the wind power output decreased by 2 MW or more. These instances only occurred 4 times: January 5 at 11:58, February 12 at 4:05, Feb 28 at 11:13, and Aug 8 at 7:46. Each of these time frames were faults or system outages and were not the result of wind ramp. As such these time frames will not be used as design criteria for the potential storage solution. In order to find the worst-case wind ramp conditions for the flywheel storage device, the data was analyzed looking for the largest change in wind power over a 2-minute period. EPS only selected the cases where the wind dropped more than 2 MW in 2 minutes. This would be equivalent to a 4 MW drop if the number of wind turbines were to double. These time frames would provide the most severe condition for the flywheels which are limited on their energy storage. Table 3 contains the dates and times of these severe ramps. Date 5 January 10 January 11 January 11 January 8 February 12 February 22 February 28 February Time 11:58 15:09 17:00 23:10 21:20 4:00 14:20 11:10 Date 3 March 4 March 4 March 6 March 8 March 9 March 11 March 13 March Time 18:30 0:20 22:55 14:10 10:15 5:35 16:45 6:00 Date 25 April 1 June 11 July 23 August 23 August 24 September 24 September 1 October Time 23:00 13:45 12:40 10:07 12:10 17:20 18:55 13:00 Date 1 October 1 October 3 October 12 October 12 October 12 October 13 October 11 November Time 21:30 22:00 9:00 0:30 0:50 3:10 3:20 11:50 Date 16 November 22 November 25 November 25 November 26 November 28 November 29 November 29 November Time 13:30 21:00 0:20 2:30 8:40 18:05 3:30 10:40 Date 29 November 30 November 30 November 30 November 30 November 2 December 3 December 3 December Time 23:45 0:20 2:30 5:45 8:10 6:40 8:30 17:05 Date 8 December 9 December 13 December 13 December 13 December 13 December 14 December 14 December Time 6:55 21:00 2:05 4:10 9:05 15:10 3:44 7:05 Date 14 December 14 December 14 December 18 December 19 December 22 December 22 December 23 December Time 8:00 11:00 19:25 12:10 5:55 11:10 12:20 21:00 Table 3: Severe 2-Minute Wind Ramp Events 4 Flywheel Evaluation After the initial evaluation of PowerCorp’s PowerStore™ flywheel, it was fairly clear that the technological limitation of the device would be its energy storage capacity. With a base rating of 18 MW-sec, the flywheel energy can be quickly absorbed by a system the size of Kodiak’s, especially considering the fluctuations of a 9 MW wind plant. EPS spent a significant amount of time tuning the flywheel control system parameters to minimize the number of flywheels that are necessary to provide the same frequency regulation in the future as is experienced today. The flywheel variables and system improvements that were investigated include: 1. Increasing the integral gain of the Terror Lake governors 2. Using an Automatic Generation Control (AGC) system and tuning 3. Using the Flywheel VAR capabilities 4. Tuning the flywheel parameters for droop response 5. Changing the number of flywheel modules 4.1 Increasing Terror Lake Integral Gain Due to the concerns over the lack of energy in the flywheels, EPS questioned whether the Terror Lake generation station could be tuned to be more responsive. If the Terror Lake unit could be more responsive, then the flywheel energy requirements would be reduced. This issue was brought up in the March 4, 2011 report. The main concern of over-improving the Terror Lake governor response is the potential of power system oscillations occurring within the Terror Lake tunnel system. EPS ran a series of simulations with varying values of the Terror Lake governor integral gain. The system event was a wind turbine trip. These simulations were intended to show the governor manufacturer what the expected power oscillations would be with different integral gain values. The simulation results sent to the governor manufacturer are shown below in Figure 2. Figure 2: System Response to Varying Governor Gain The simulations shown above imply that there is some room for improvement, but KEA informed EPS that the governor manufacturer would not feel comfortable recommending a significant change to the governor gains at this time and would need to retune the governor if any gain changes are made. As such, EPS concluded our investigation of this potential system improvement. EPS recommends that KEA and the governor manufacturer work to maximize the Terror Lake governor response when commissioning the proposed third Terror Lake turbine. 4.2 Automatic Generation Control EPS further investigated the possibility of using signals from the flywheel to trigger additional movement of the Terror Lake units via an automatic generation control (AGC) type of system. The AGC would use the power and state of charge signals of the flywheel system as inputs. A classical AGC would have an Area Control Error or ACE. This signal would be a combination of the system frequency error and tie line flow error. Since KEA is an island the tie line flow error would always be zero. EPS set up an AGC model for the KEA system using frequency, flywheel power output, and flywheel state of charge as the inputs to create an ACE signal. The system was setup such that the ACE would be calculated using the following equation: ACE = (Frequency Error) – 1*(Flywheel MW) - 2*(Max Charge - Flywheel SOC) AGC would have a deadband for ACE where no action is taken. For a negative ACE, the generators would be given commands to raise their generation. The AGC parameters that could be adjusted are ACE deadband, 1, and 2. Additionally, the EPS AGC model could use a “panic” mode in which the unit command pulses could be doubled. The AGC controls need to be coordinated with the flywheel controls in order to provide the optimum system-wide response. Initially, the AGC simulations were providing better frequency regulation compared to the simulations with one Terror Lake unit operating in isochronous mode. However, this was due to the operation of the deflector control in the turbine / governor. Since the AGC needs to have some dead-band to prevent control oscillations around 60 Hz, there are steady state conditions that are marginally above 60 Hz. The deflector controls are setup such that they will quickly enter the water stream when the frequency goes above 60 Hz to protect the unit from over- speed. These simulations were overly optimistic because the Terror Lake units at times ended up in steady-state water wasting mode. As the wind power dropped, the deflectors would pull away from the stream resulting in near immediate response. This made the simulations seem much improved, but steady state water wasting should be avoided as it defeats the purpose of integrating additional wind. When EPS made the necessary corrections to the Terror Lake governor gains to prevent steady state water wasting, the AGC system control no longer provided significant frequency regulation benefit. 4.3 Flywheel VAR Capability EPS was tasked with determining the ability of the flywheel to provide reactive power support to the KEA system. The model provided by PowerCorp has the capability of this type of control. During the PowerCorp visit to the EPS office in Redmond, WA, the reactive power settings were discussed with Dr. Andrew Tuckey. He advised against using the flywheel system for reactive support in steady state conditions. The reasoning provided by Dr. Tuckey was due to the behavior of the limiter in the flywheel module. The limiter is based on the thermal ratings of the power electronics. These limits, in turn, are based on the current flowing through the power electronics circuitry. He explained that by having steady state reactive current flowing through the device, the ability of the system to respond to a system frequency disturbance will be reduced. As the response to frequency disturbances is the primary purpose of this device, he recommended against this practice. He advocated the use of switched capacitor banks which could be automatically controlled using the PowerCorp system and would ultimately be more cost-effective. As such, EPS recommends against using the flywheel system for steady state reactive compensation. 4.4 Flywheel Parameter Tuning The model provided to EPS by PowerCorp has four main parameters that describe the control of the flywheel power output. The four parameters are: 1. Static Frequency Gain (Droop) 2. Dynamic Frequency Gain (Rate of change of frequency component) 3. Frequency Deadband Hi 4. Frequency Deadband Lo When tuning these parameters EPS started with a simple, aggressive droop with a tight deadband around 60 Hz. This approach provided poor results. The flywheel quickly ran out of energy before the Terror Lake units had a chance to respond. Since the Terror Lake units are operating on a droop line or in isochronous mode, the primary input is frequency. If the flywheel regulates the frequency close to 60, the Terror Lake units will not move, and the flywheel will eventually run out of energy. When the flywheel runs out of energy it quickly ramps down to 0 MW. When the flywheel ramps down, the event appears more severe than the original wind ramp. Therefore it is important to find a control strategy that will avoid exhausting the flywheel energy. The next approach was to have a larger deadband around 60 Hz. Initially 59.85-60.15Hz was selected. This approach provided significantly better results since only the Terror Lake units responded to frequencies down to 59.85 Hz and would continue ramping up as long as the frequency is below 60. However, this scheme did have a problem in that the flywheel never recharged. When the frequency goes above 60 Hz, the deflectors enter the stream to prevent overspeed. Hence, the frequency was rarely above the deadband value of 60.15 and without an opportunity to recharge, the flywheels eventually ran out of energy. The next approach was to have the upper frequency deadband reduced to 60 Hz, such that any time the frequency moved above 60 Hz, the flywheel would recharge. This control change had three positive effects. 1) The flywheel had a chance to recharge in preparation for the next downturn in wind generation. 2) The flywheel would spend much more time fully charged and be better prepared for a downturn in wind generation. 3) By absorbing power above 60 Hz, the flywheel can prevent some water wasting at the Terror Lake generation station since it reduces the net upward slope of wind generation. For these reasons, the upper frequency deadband setting should be set to slightly above 60 Hz. The dynamic frequency gain was not used since its application would be better suited to unit trip events. This control scheme would allow for additional flywheel power based on the slope of the system frequency decay. EPS recommends discussion with the manufacturer on the proper use of this parameter if flywheels are the selected technology. However, considering the low energy concerns, any additional power output would use that energy more quickly. This left the lower frequency deadband and the static frequency gain to be determined. Several simulations with a mix of these settings at both high and low values were simulated. The control strategy with a deadband of 59.85 Hz and a static frequency gain of 1.15 appeared to be the best to use. Again, these settings made sense with the behavior of the Terror Lake governors which are highly sensitive to frequency. The flywheel controls allow Terror Lake to continue providing the majority of the frequency regulation, but provide support for large wind ramp events. With four flywheels using these settings, the future Case 1 standard deviation of frequency is the same as the current Case 1 configuration. A plot of the simulated future case with 3 Terror Lake generators and 6 wind turbines is shown below in Figure 3 Figure 3: Case 1 - 3 Terror, 6 Wind, 4 Flywheels, Same Standard Deviation as Present Case 1 The top left traces show system frequency from the original recording (purple) and the simulation (red). The top right black trace is the flywheel state of charge. The bottom left shows the wind (orange), Terror Lake (blue/green), and flywheel (black) power outputs. The bottom right traces show the power fraction in red and black. The power fraction is equal to one minus the portion of water that is deflected. The standard deviation of frequency for the case shown is 0.1547 Hz. The present case with 2 Terror Lake units and 3 wind turbines had a simulated standard deviation of 0.1566 Hz. This control strategy accomplishes the frequency goal, and will be referred to as control strategy G. 4.5 Minimum Number of Flywheel Modules With the control strategy problem solved, EPS next determined the minimum number of flywheels necessary to respond to all wind ramp events through the year. Due to the minimal energy storage of the flywheel systems, the most severe events are the prolonged wind ramp events. However, a trip event which results in a low frequency for a short period of time is easily handled by the flywheel system. The time frames used to determine the minimum number of flywheel systems is shown in Table 3. Control strategy G was used for these simulations, and the number of flywheels was increased until the simulations did not run out of energy. An example showing the results for 12 flywheels is shown below in Figure 4. Figure 4: Extreme Ramp - 12 Flywheels with Control Strategy G The bottom left set of traces show the power outputs of the different units. The orange trace is the wind plant, the black is flywheel output, and the blue and green represent the Terror Lake units. In this simulation, the wind output moves from 8 MW down to 0 in approximately 9 minutes. The top right black trace shows the flywheel state of charge. It can be seen that the flywheel runs out of energy near the end of the simulation, but the wind is at zero and the flywheel does not need energy to cover any additional loss of wind. The top left traces show system frequency from 58 to 60.5 Hz. The purple trace shows the recorded frequency from 2010, and the red trace shows the simulation frequency with double the wind. The results of the other severe 2-minute ramp events are shown in Appendix B. Several of the events in the severe ramp set of data are likely unit trip events in rapid succession. This is a situation where doubling the wind output is unlikely to be a realistic situation when KEA doubles the number of wind turbines. While the recordings suggest that two of the turbines can trip in rapid succession, it is even less likely that the equivalent of four turbines would trip in the same time frame when the wind is doubled. However, the extreme ramp shown in Figure 4 can be expected, and must be the design criteria for the flywheels since these events appear to happen approximately 60 times per year. Without the system improvements discussed earlier such as increasing governor gain or implementing supplementary automatic generator control, 12 flywheel systems are recommended. The flywheels come in modules with one energy storage size with several different power output levels. The power requirements for the proposed control strategy are approximately 3 MW. As such, EPS recommends using the flywheel configuration with the smallest power rating to minimize the project cost. Currently the smallest power rating PowerCorp produces is a 0.5 MW unit. With 12 of these, the power rating would be 6 MW which is more power than is necessary. The cost of 12 flywheel modules seems prohibitive. 5 Battery Evaluation After the initial evaluation of the BESS, it was clear that the energy issues seen with the flywheels were not going to be an issue since one BESS module contains 1 MWhr of energy which is 200 times the energy of the flywheel module. However, the battery system does have loss of life issues. As the batteries charge and discharge, there is a predictable system degradation. It is this degradation that will be the major factor in defining the life-cycle cost of the BESS technology. 5.1 BESS controls Many of the assumptions made with the flywheel systems are still valid with the battery system. Mainly, the control characteristics are quite similar. This is due to the fact that both systems employ power electronics to convert the stored energy (battery charge, or flywheel rotating mass) into power injections into the grid. The user models for both energy storage systems came with a droop characteristic with some sort of frequency deadband. The main difference between the models, of course, is the energy content. Since many of the flywheel control parameters were defined by the operation characteristics of the Terror Lake generation station, a very similar control strategy should be used for the BESS. EPS ran a simulation using the same droop and deadband characteristics used for the flywheel. The results are shown below in Figure 5. The frequency is shown in the top left. The unit power outputs are shown in the bottom left where the green trace is the combination of the wind and BESS output. This signal shows the actual change in wind output as seen from the Terror Lake generation station. The top right plot shows unit reference signals and BESS state of charge. The state of charge signal is zoomed in to 45-55%. The bottom right shows the Terror Lake power fraction (per unit amount of water not being deflected, or hitting the turbine). This simulation was run with a 3 MW capacity and 1 MWhr of energy storage. The simulation is faulty in that the XP modules come in sizes of 1.5 MW with 1 MWhr of energy. So this simulation shows 2 modules with half the energy storage. The only difference between our simulations and those with the larger amount of storage is that the state of charge curve would move half as much. Figure 5: BESS Results with Control Strategy G As before, the standard deviation of frequency seen in this future case simulation is equivalent to the simulation of today’s case. The BESS will not run out of energy in a 15 minute simulation as it has enough energy storage to provide 40 minutes at full power output. The maximum power output from the BESS is approximately 1.8 MW for this disturbance so the 3 MW for the two BESS modules should be sufficient. 5.2 BESS Extreme Ramp Results In order to confirm that the control strategy works properly for the worst case conditions seen in 2010, the extreme 2-minute ramps were simulated with the BESS in place. These are the same scenarios used to determine the necessary flywheel energy. Because the BESS should have no trouble with the energy storage, this evaluation was used to determine if the power rating of the BESS was sufficient. Figure 6 shown below has the results of the November 30th ramp event that was shown above in Figure 4 but with the battery system instead of the flywheel system. Figure 6: Extreme Ramp - BESS with Control Strategy G It can be seen from the simulation that the BESS system has enough energy (top right black trace) and that the power output does not reach its maximum value for this extreme case. Despite the loss of 5 MW of wind in the first 90 seconds, the frequency only dips to 59.5 Hz. With the proposed BESS control strategy, there were some conditions that required the BESS to reach its maximum power rating. The BESS does still provide the 3 MW rated power until the frequency recovers. During these events, only the January 11 at 17:00 event causes the frequency to fall below 59.0 Hz. This event is shown below in Figure 7. The rest of the results are shown in Appendix C. Figure 7: Extreme Ramp BESS Maxed Power Output Since the BESS has a comparably large energy storage rating, the BESS sizing requirement will be based on the required power output. The wind ramp event shown in Figure 7 causes trouble due to the near instantaneous loss of 6 MW of wind power. The BESS contributes 3 MW resulting in a net loss of 3 MW of wind power. The resulting frequency is approximately 58.6 Hz. Seeing as this is the only wind ramp event that causes the BESS to reach its power limit and still have a frequency dip below 59 Hz, it should not require an additional BESS module to increase the maximum power output to 4.5 MW for this one case. This is especially true when compared to current operation such that when a wind turbine trips near full output, the frequency decays below 59.0 Hz and occurs much more frequently than once per year. With these considerations, EPS recommends that two BESS modules be used to assist KEA with frequency regulation. 5.3 Battery Expected Life Calculations In order to determine the life-cycle cost, EPS needed to know how often the batteries would need to be replaced. Figure 8, taken from Xtreme Power’s website, describes their warranty characteristic. For small state of charge movements, these batteries can provide hundreds of thousands of cycles, but far fewer cycles for larger movements. EPS next determined how large, and how many charge / discharge cycles should be expected in a given year. Figure 8: Xtreme Power Battery Warranty Curve Using the selected time frames from Table 2, simulations were run with double the wind output. Using these different types of wind regimes, the simulations provided an idea of how many cycles the batteries would go through. The thirty selected time frames represent three samples of each type of wind event. Type 1 events are the least severe with type 10 events being the most severe. The yearly distribution of these events shown in Table 1 can then be used to determine the expected number of cycles for the year. For example, if a 15-minute type 1 event causes one 1% SOC cycle, and there are 17,519 type 1 events in a year, then type 1 events would cause 17,519 1% SOC cycles. The task is to determine the number of cycles for each event type and combine them into a loss-of-life calculation. EPS provided simulation results of the thirty 15-minute simulations to Xtreme Power engineers for this loss-of-life analysis. The data provided to Xtreme Power was the power output and state of charge in one second increments. Xtreme Power engineers analyzed the data and provided EPS with a spreadsheet with the loss-of-life for each of the 15 minute simulations based on the percentage of life lost. The spreadsheet they provided is attached in Appendix D. Table 4 shows the relevant results of their calculations as well as the calculation performed by EPS to determine the cumulative loss-of-life. The Case columns list the selected 15-minute simulations corresponding to the times listed in Table 2. The Total Impact column lists the impact each individual 15-minute simulation has on the battery life. The Type Mean column lists the average Total Impact for each frequency standard deviation type. The Type Max column lists the maximum Total Impact for each type. There are 34,040 15-minute samples throughout the year, and the number of each wind event type is listed in the Samples column. The yearly usage columns list the Type Mean or Type Max multiplied by the Number of Samples to provide the impact on BESS life for each frequency type. The bottom of Table 4 lists the yearly usage and the life expectancy of the batteries. The results provided by Xtreme Power confirm the selection criteria used by EPS. Specifically, the worst case conditions (types 7-10) had much more battery usage per 15-minute simulation when compared to the lower variability conditions (types 1-3). These results confirm the EPS selection criteria and add confidence that the standard deviation of frequency correlates with the BESS loss of life calculations. With this confirmation, EPS could continue with the loss-of-life analysis. Samples Total Impact Type Mean Type Max #of Type Type Mean Type Max hi 0.0000000% mid 0.0002303% lo 0.0000000% hi 0.0002449% mid 0.0002449% lo 0.0002449% hi 0.0002449% mid 0.0002604% lo 0.0002449% hi 0.0002768% mid 0.0002604% lo 0.0005217% hi 0.0004906% mid 0.0004751% lo 0.0002604% hi 0.0002449% mid 0.0002604% lo 0.0000000% hi 0.0003130% mid 0.0002768% lo 0.0004000% hi 0.0003762% mid 0.0009804% lo 0.0003130% hi 0.0010288% mid 0.0003130% lo 0.0005433% hi 0.0009512% mid 0.0003538% lo 0.0009823% 6.205241% 9.625221% 16.12 10.39 Case 1 15 Minute Yearly Usage 0.0000768% 0.0002303% 17519 1.3448126% 4.0344377% 0.0002500% 0.0002604% 4265 1.0663518% 1.1104017% 0.0002449% 0.0002449% 8244 2.0186238% 2.0186238% 0.0620271% 0.0752012% 0.0003530% 0.0005217% 2739 0.9667403% 1.4288936% 0.0004087% 0.0004906% 1446 0.5910007% 0.7094674% 0.0001684% 0.0002604% 527 0.0887489% 0.1372056% 0.0106741% 0.0137517% Yearly Usage (% of Battery Life) 0.0005565% 0.0009804% 74 0.0411817% 0.0725467% 0.0006283% 0.0010288% 24 0.0150801% 0.0246912% 2 3 4 5 6 7 8 9 10 Battery Life (Years) 0.0007624% 0.0009823% 14 0.0003299% 0.0004000% 188 Table 4: BESS Loss of Life Calculation The two methodologies were selected to represent an expected life-expectancy (Mean column) and a pessimistic life-expectancy (Max column). It can be seen that depending on the method used to calculate the life-expectancy, the batteries should last between 10 and 16 years. Additionally, the provided case had ½ the energy capacity that the actual system would have so the life expectancy would be even larger as the depth of discharge in the cycles would be half as large. The life-expectancy should be used in combination with the expected BESS power losses, and expected maintenance costs to determine the total life-cycle costs. However, considering that 12 flywheel modules would be required to provide a similar system frequency response, the BESS system makes more sense. The BESS makes even more sense when looking at the system end-of-life when only the spent batteries need to be replaced. The battery replacement cost estimate of $300,000 is only a small portion of the initial capital cost of the battery system. 6 Conclusions EPS makes the following recommendations based on the results of this study: EPS recommends that KEA pursue the battery technology over the flywheel technology based on cost. o Approximately 12 flywheel modules would be necessary to provide the required energy if flywheel systems are used. o One battery system initially costs twice that of a flywheel. o The battery bank replacement cost is a comparably low amount of $300,000. o The expected life of the battery packs is 16 years. The recommended control strategy is based on droop with a deadband around 60 Hz. o The recommended droop setting for the BESS is 0.85%. o The recommended deadband setting for the BESS is from 59.9 to 60.02 Hz. o This would result in full power output at 59.39 Hz assuming a power rating of 3 MW. The recommended settings for the BESS control should be evaluated with system experience, but based on these settings the battery technology is clearly the better technology for KEA’s system. Appendix A – Wind Data Analysis Appendix B – Severe 2-Minute Simulations with Flywheels Appendix C – Severe 2-minute Ramp Simulations with BESS Appendix D – Xtreme Power BESS Loss of Life Calculation