1. ERCOT Planning WMS 10/20/2010 Target Reserve Margin and Effective Load Carrying Capability of Installed Wind Capacity for the ERCOT System – Methodology and Results

2. 2 Overview The scope of this study is to evaluate the impact of system volatility on the relationship between generation reserve levels and system reliability. The following components of system volatility are being considered: –The forced outage and derating of generating facilities. –Load forecast uncertainty related to weather. –The intermittent nature of wind power. 10/20/2010WMS

3. 3 Reliability Indices The following are the three main reliability indices used in this study: –Loss-of-load events (LOLEV): The number of times in a year that available generation was incapable of meeting demand. LOLEV provides information about the frequency of events. –Loss-of-load hours (LOLH): The number of hours in a year that available generation was incapable of meeting demand. LOLH provides information about the duration of events. –Expected unserved energy (EUE): The total amount of MWh in a year of demand that could not be met by available generation. 10/20/2010WMS

4. 4 Input Data – Generation All existing, as well as future resources with a signed interconnection agreement, that are expected to be in service in year 2012 are being considered, including units under reliability must-run (RMR) review. –Includes generation and load from private use networks –Does not include DC ties or hydroelectric capacity. Monthly capacity multipliers are applied in order to model the seasonal capacity ratings of thermal units. The seasonal values from the RARF are used for that purpose. Forced and scheduled outages are being modeled in accordance with available NERC GADS data. Unit Specific data has been used whenever provided. The transmission network is not being modeled. 10/20/2010WMS

5. 5 Input Data – Load Hourly chronological load profiles for year 2012 prepared by ERCOT. –Five load scenarios were developed in order to capture weather related uncertainty: extreme summer with a 10% probability of occurrence, warmer than average with a 23% probability, average with a 34% probability, cooler than average with a 23% probability, much cooler than average with a 10% probability. –The associated probability of occurrence is to be used later in the calculation of reliability indices. –The economic growth assumption behind all scenarios is based on Moody’s base economic forecast. 10/20/2010WMS

6. 6 Input data - Wind Representative hourly wind energy availability data for each wind plant provided by AWS Truewind. –Based on the wind generation assessment report prepared for ERCOT by AWS Truewind provided for the CREZ analysis. –Since the transmission network is not being considered, the wind-farm-specific hourly profiles are aggregated. To capture the randomness of wind generation, daily wind profiles are generated by randomizing the available daily profiles using a MATLAB model. –Forced outages of individual wind turbines are not being modeled. 10/20/2010WMS

7. 7 Generator outage modeling Outages are modeled sequentially, using random draws from two exponential distributions. The time on outage for each unit is randomly drawn from an exponential distribution with mean equal to the mean time to repair (MTTR): –MTTR = (FOH + EFDH) / # of FO occurrences The time in service for each unit is randomly drawn from an exponential distribution with mean equal to the mean time to failure (MTTF): –MTTF = SOAF x MTTR x [(1/ EFORd) – 1 )] –SOAF is the scheduled outage adjustment factor equal to [(8784 – SOH) / 8784] which is applied while calculating MTTF to account for any loss of outage time due to overlap of forced outages with scheduled outages. The outage modeling described above results in unit unavailability due to forced outage equal to EFORd. 10/20/2010WMS

8. 8 For every unit, build sequences of generator availability and unavailability periods using MTTF and MTTR respectively. For every day, create random daily wind profiles. To randomly choose a day’s wind profile, a span of +/- 7 days is used. Generate hourly resource profiles by summing up the hourly capacity available (wind and non-wind). Once hourly resource profiles are available, the margin is calculated. A negative margin indicates a loss of load hour and it’s value the amount of load that could not be served. A sequence of loss of load hours is treated as one loss of load event. Flowchart The reliability metrics (LOLEV, EUE and LOLH) are updated based on the hourly margin obtained. Based on the probability associated with each of the load scenarios, calculate the study-wide reliability indices. Once a (maximum) number of iterations has been run or the stopping criteria have been met, proceed to the next load scenario. See next slide for the stopping criteria. Start Read generation, wind and load data from an EXCEL file Generate random outages and randomize available wind capacity for a pre-specified number of years Proceed with the next load scenario Update reliability metrics (LOLEV EUE, LOLH) For every load scenario Calculate probability based reliability indices Print results Stop Check whether the stopping criteria are met nono yes Evaluate the hourly margin between resource and load (Margin = Resources – Demand) 10/20/2010WMS

9. 9 Stopping criteria Each load scenario is simulated for a (maximum) number of iterations (set to 10,000) or until the stopping criteria are met. The stopping criteria are: –A minimum number of iterations (set to 1,000). –The LOLEV halfwidth for a 95% confidence interval is less than a percentage (set to 5%) of the LOLEV average. –The total number of loss of load events is greater than a minimum value (set to 1). 10/20/2010WMS

10. 10 ELCC calculation Methodology for the ELCC calculation: –ERCOT will be evaluating the ELCC of wind units by comparing them to the 2012 planned fleet. –First, the reliability metrics are evaluated for a base case scenario. Then, the wind units are removed and a multiplier is applied to the capacity of each non-wind generator in the fleet. The capacity multiplier is adjusted accordingly and the simulation is re-run until the level of the appropriate reliability metric is equivalent to that in the base case. –The ELCC value is equal to the ratio of additional non-wind capacity (capacity multiplier-1 times the installed non-wind capacity) divided by the total installed capacity of wind, in per cent. 10/20/2010WMS

11. 11 Comparison to the 2007 study by Global Energy Decisions The current study differs from the 2007 Global Energy Decisions study in the following ways: –ERCOT used a family of load profiles representing different weather conditions. –All hours of the year are being modeled instead of a representative week from each month. –Generator outages are modeled sequentially. –Wind volatility is modeled in a more dynamic way. –The ELCC of existing wind is compared to the reliability of the existing fleet rather than hypothetical new generation. –A much higher number of iterations was performed in order to ensure the statistical significance of the results. 10/20/2010WMS

12. 12 Study Results The following random variables are estimated for year 2012 and for different reserve margin levels: –The annual loss-of-load events (LOLEV). –The annual loss-of-load hours (LOLH). –The annual expected unserved energy (EUE). The calculated ELCC of wind generation is reported based on LOLEV. The calculated target reserve margin is reported based on 0.1 LOLEV per year. –This is equivalent to 1 loss of load event every 10 years. The ELCC for wind from the 2010 study using the LOLEV metric of 0.1 is 12.2% (12.5% for the median scenario). The Target Reserve from the 2010 study using the LOLEV metric of 0.1 is 14.1% (13.3% for the median scenario). 10/20/2010WMS

13. 13 LOLEV across all load scenarios WMS10/20/2010

14. 14 LOLEV for the median load scenario WMS10/20/2010

15. 15 LOLH across all load scenarios WMS10/20/2010

16. 16 LOLH for the median load scenario WMS10/20/2010

17. 17 EUE across all load scenarios WMS10/20/2010

18. 18 EUE for the median load scenario WMS10/20/2010