NERC Interconnections Frequency Response Statistics, Typical Frequency Events Profiles, and Observations on NERC-FRI Report Statistics

Transcription

1 Consortium for Electric Reliability Technology Solutions NERC Interconnections Frequency Response Performance NERC Interconnections Frequency Response Statistics, Typical Frequency Events Profiles, and Observations on NERC-FRI Report Statistics For: NERC Resources Subcommittee, NERC Technical Staff By: Carlos Martinez, Rafael Campo CERTS/ASR Atlanta, Georgia, October 24-25,

2 Presentation Outline Review NERC Interconnections Frequency Events Datasets Available for Frequency Response Analysis NERC Interconnections Events Frequency Response Statistical Summaries and Trends NERC Interconnections Statistics and Trends for Events Loss MW and Frequency-C NERC Interconnections Typical Events Frequency Profiles Review CERTS Observations and Recommendations on NERC Frequency Response Initiative (FRI) Report statistics Version 09/16/2012 2

3 NERC Interconnections Candidate Frequency Events Datasets 3

4 NERC Interconnections Known Historical Events Datasets Eastern, ERCOT Interconnections Frequency Response Events and Corresponding Parameters 4 Interconnections NERC-RS Applications 1-Min., 1-Sec. Data 3 Interconnections Frequency Response J. Ingleson E. Allen ( Eastern) T. Bilke ( Eastern) S. Niemeyer ( ERCOT) NERC Archive & Min. 1-Sec. CERTS-LBNL ( FERC 2010 Report) RS-FWG Freq. Resp. Standard Form-1 4 Interconnections Frequency Response Events and Corresponding Parameters 4

5 NERC Interconnections Candidate Frequency Events Dataset - PMU 1-Sec. Data Year Interconnections Candidate Frequency Events Eastern Western ERCOT Hydro Quebec TOTAL Frequency Events Dataset including identification flags for: a. RS-FWG Selected Events, Form1 for: Eastern, Western, ERCOT, Hydro Quebec b. Eastern Events T. Bilke Dataset c. ERCOT Events S. Niemeyer Dataset CERTS RECOMMENDATION: Use the RS-FWG dataset as the Master Dataset for Interconnections Frequency Analysis, replication and validation of statistical analysis, and ALR events for each interconnection 5

6 NERC Interconnections 2009 to 2012 Events Frequency Response Statistics 6

7 Eastern Frequency Response Statistics 7

8 Western Frequency Response Statistics 8

9 ERCOT Frequency Response Statistics 9

10 NERC Interconnections Statistics for Events Loss MW and Events Frequency-C 10

11 Eastern, Western Frequency Events Loss MW Statistics Eastern Interconnection Western Interconnection 11

12 ERCOT, Hydro-Quebec Frequency Events Loss MW Statistics ERCOT Interconnection Hydro Q. Interconnection 12

13 Eastern, Western Frequency Events Frequency-C Statistics Eastern Interconnection Western Interconnection 13

14 ERCOT, Hydro-Quebec Frequency Events Frequency-C Statistics ERCOT Interconnection Hydro Q. Interconnection 14

15 NERC Interconnections Typical Events Frequency Profiles 15

16 16

17 CERTS Replication Analysis and Observations on NERC FRI Report Version 09/16/12 17

18 Rafael Campo Ph.D. Electrical Engineer; MSc and PhD in IEOR (Industrial Engineering and Operations Research). University of California, Berkeley; University Visiting Professor: University of Michigan (Ann Arbor), graduate course in Time Series Analysis and Forecast; South America and Caribbean Universities (Brazil, Colombia, Venezuela, Puerto Rico) - Courses in Operations Research; Consulting and Development Work for Electric Power Industry: Probability analysis of Frequency Response data, as collected by PMUs (CERTS); Grid Performance Monitoring Metrics using Phasors (CERTS); Hydro-thermal dispatch; Advanced applications in control centers (AEP and Systems Control Inc.); Time Series and Risk studies in electric power markets; Member of Market Surveillance Committees; Consulting work for: World Bank, IDB, IAEA, IEA, etc. 18

19 CERTS FR Performance Replication - FRI Report CERTS REPLICATION DESCRIPTIVE STATISTICS AND TRENDS - DIFFERENCES WITH FRI REPORT Eastern Frequency Response Performance From 2009 to

20 Use of Statistics for Reliability Performance Analysis Additional variables introduced by electricity markets, integration of renewable generation and Smart Grids make necessary the use of statistical and risk analysis for reliability performance analysis and for defining adequate performance metrics for reliability standards; The statistical and risk analysis process described in the next slide is recommended for reliability analysis, including cross-validation and replication; In our view, the statistical processes used in the NERC FRI Report are sound; We have reservations on the FRI conclusion of Eastern FR upward trend during the period (Slide 24). Replication (Slide 19) does not confirm this trend; We make recommendations to complement the FRI analysis and improve robustness of the performance results; 20

21 Recommended Statistical and Risk Processes for Interconnections Reliability Performance Analysis Define and agree on a common interconnections frequency and frequency events dataset based on SCADA and PMU data (CERTS recommends using the NERC RS-FWG dataset) Prepare, expand, and clean interconnections frequency and frequency events data from recommended dataset Identify and fit appropriate statistical models such as Robust Linear Regressions, Survival, etc. Evaluate the statistical model fit using accepted statistical tests and confidence margins Cross-validate statistical models preferable by external organizations Identify and assess related risks, and investigate and recommend mitigations if required NERC Committees/Subcommittees to review and approve draft reports Not Approval Approval Produce final Reliability Performance and Trends reports, and/or update websites for Stakeholders 21

22 CERTS Observations on FRI Report Summary CERTS replication boxplots and regression show that Frequency Responses in 2010 and in 2011 were lower than in Frequency response in 2011 appears slightly lower than in The scatter plot from the FRI report and CERTS replication reveal large variability and consequently large uncertainty in the data, suggesting difficulties in fitting appropriate linear regression models. The R- square obtained in the FRI analysis (0.0184) is too low. The FRI report specifically points this characteristic:..changes in time explain only 1.8% of variability of Frequency Response.. The FRI report postulates that the distribution of the FR is (truncated) normal. However, the histogram-density of the FR distribution from CERTS replication, does not support this assertion. In addition, CERTS recent reports on FR performance provide rather strong evidence of the presence of fat tails in the distributions of MW and frequency events for the Eastern Interconnection. The FRI correlations report of the linear regression explanatory variables (date, season, and pre-disturbance frequency) with the FR, as well as their coefficients of determination, are rather low to draw definitive conclusions. 22

23 Detailed Observations on FRI Report Version 09/16/12 Data used in FRI Report has very large variance; The more complete available events data set was not used (Eastern = 391 events available vs. 158 used observations); No accounting was made for measurement errors and measurement uncertainties; Note: CERTS has made analysis of PMU measurement uncertainties; Additional explanatory variables might be more relevant than the ones used; Results using RS-FWG data set fail to confirm FRI Report Eastern FR uptrend 23

24 FRI Report Scatter Plot of Eastern FR (Using 2009, 2010 and 2011 data) 24

25 FRI Report - FR vs. Time Model Mathematical regression model fit to the data, with time as the only explanatory variable: FR(t) = k + a*t + ε (t) ( k is the intercept and a the slope); ε (t) iid Normal random variables; Standard deviation of ε (t) = 602; therefore, with 95% probability actual values are (Linear regression) fitted values ± 1,204 MW/.1Hz (Mean value about 2,400 MW/.1Hz) 25

26 Observations on FRI Model: FR vs. Time Statistical tests confirm that residuals have normal distribution, with zero mean. The regression hypothesis are then satisfied; P-value used for testing flat versus positive slope (0.083) is larger than commonly used significance level (0.05) At 5% significance level, therefore, the slope is flat; Model coefficient of determination (R 2 ) too low (1.8%) Very low dependence of FR on time, as the paper indicates; F tests at 5% significance level No variation in time of FR from 2009 to 2010 to 2011; We advice caution in concluding FR dependence on time; 26

27 FRI Report - Multiple Linear Regression Time as an explanatory variable has very low coefficient of determination and does not seem to pass the t-test; Season and pre-disturbance frequency have larger explanatory value and pass the t-test; Overall coefficient of determination (15%) low; standard deviation of residuals (565 MW/.1Hz) high; As indicated above, need additional data and/or further explanatory variables, for more robust results; 27

28 FRI Report - Notes on Normality Assumption for FR CERTS fit and tested Power Law distributions to extreme (MW and delta frequency) events; The results prove the presence of fat tails in the statistical distribution of these events; Extreme events are the most important from a reliability point of view; They would not be captured with the usual regression analysis; 28

29 CERTS Suggestions and Recommendations 29

30 Suggestions and Recommendations - Summary Use the available RS FWG Frequency Events dataset for uniformity, for increasing event sample size and for improving statistical confidences; Use Statistical Analysis for Interconnections Frequency Analysis, but be certain the analysis can be replicated; To add robustness to the performance results we suggest: Accounting for measurement errors and measurement uncertainty; Trying additional explanatory variables, example: Time-of-day (see next slide); MISO tried Top-of-the-Hour; Sundays/No Sundays and Average Temperatures, for a Δfreq analysis using 2011 and 2012 data (first 6 months); they obtained R 2 = 0.58; Research practicality and utilization of the Survival Probability models as an alternative for Linear Regression models 30

31 Eastern FR vs. Hour of Day 31

32 Alternative: Survival Probability Plots for Comparisons Method widely used in reliability analysis; Empirical survival functions are compared for (non negative) data sets involved (2 at a time): X and Y; Y X (stoch.); iff Pr(Y a) Pr(Y a) for all a; In this case, mean of Y is above mean of X; Compare plots of (1 F) for Y and X; Different from regression, method does not require equal number of observations for X and Y and is less influenced by outliers; Additional comparisons can be made visually; Statistical tests can be performed; Next slide compares Δ freq (C A) of 2011 and

33 Probability Use of Survival Plots for Comparing Eastern Events Frequency-C for 2011 and Empirical Survival Function Prob(Event 2011 > x mhz) Prob(Event 2012 > x mhz) Frequency-C below 2011 Frequency-C Except for Large Values; K-S Test Passed at 5% Significance Level 33 x

34 11/4/2012 PAS Perspectives on ALR1-12 Metric and Variable Generation Metrics Matthew Varghese NERC Staff RS Meeting October 24, 2012 AGENDA Results presented in 2012 SOR &FRI Report SAS Time Trends for EAST & WEST interconnections Data Quality Issues Deletion & Selection of Events Data Submission and Update Enhancement Variable Generation Metric Proposals 2 RELIABILITY ACCOUNTABILITY 1

35 11/4/2012 Results presented in SOR & FRI reports 3 RELIABILITY ACCOUNTABILITY Results presented in SOR & FRI FRI Report PDF Page 34 4 RELIABILITY ACCOUNTABILITY 2

36 11/4/2012 Results presented in SOR & FRI FRI Report PDF Page 35 5 RELIABILITY ACCOUNTABILITY Results presented in SOR & FRI FRI Report PDF Page 35 6 RELIABILITY ACCOUNTABILITY 3

37 11/4/2012 Results presented in SOR & FRI FRI Report PDF Page 36 7 RELIABILITY ACCOUNTABILITY Results presented in FRI FRI Report PDF Page 37 8 RELIABILITY ACCOUNTABILITY 4

38 11/4/2012 Tim e Trends of ALR1-12 on NERC Site 9 RELIABILITY ACCOUNTABILITY Eastern Interconnection Frequency Response on NERC idashboards site RELIABILITY ACCOUNTABILITY 5

39 11/4/2012 Western Interconnection Frequency Response on NERC idashboards site RELIABILITY ACCOUNTABILITY Data Quality Issues - Deletion and Selection of Frequency Response Events 12 RELIABILITY ACCOUNTABILITY 6

40 11/4/2012 Data Quality - Event deletion Example: Out of 79 WI events, 13 were deleted Adelta DT(Local) Day TZ from 60 Point A T2-16 Point B - MW Point A Loss Response Problems MW/.1Hz 4/19/2009 4:08 Sun PDT ,505 MW Loss + frequency 6/10/ :01 Wed PDT ,498 RAS Event 7/12/ :49 Sun PDT ,102 MW Loss + Runback 9/20/ :44 Sun PDT ,840 Frequency 9/25/2009 9:26 Fri PDT ,549 Frequency 12/3/2009 3:31 Thu PST ,934 MW Loss 12/19/2009 1:29 Sat PST ,779 MW Loss 1/18/ :43 Mon PST ,603 RAS Event 2/20/ :35 Sat PST ,051 MW Loss 6/18/2010 8:07 Fri PDT ,764 MW Loss 11/4/2010 9:46 Thu PDT ,315 MW Loss 1/13/ :06 Thu PST ,899 Frequency 7/14/ :46 Thu PDT ,678 MW Loss 13 RELIABILITY ACCOUNTABILITY Data Quality -Event selection Example: Out of 9 WI 2012 events, 7 selected by WECC PWG UTC Date/Time Pacific Date/Time PAC Day Zone A Value B Value FMA B A C Value FMA Point C MW Loss Frequency Response Actual MW Loss 2/23/2012 0:57 2/22/ :57 Tue PST , /1/2012 4:53 2/29/ :53 Tue PST , /30/ :04 3/30/ :04 Fri PDT N/A N/A N/A 975 Runback 4/4/ :25 4/4/2012 9:25 Wed PDT , /6/ :38 4/6/2012 9:38 Fri PDT ,400 1, /9/ :52 5/9/2012 8:52 Wed PDT , /10/ :10 5/10/ :10 Thu PDT , /10/ :33 5/10/ :33 Thu PDT , Load Rejection Test 5/15/2012 2:01 5/14/ :01 Sun PDT N/A N/A N/A 633 Insufficient Data 14 RELIABILITY ACCOUNTABILITY 7

41 11/4/2012 Data Submission and Update Enhancement Reporting Procedure Enhancements. New NERC Online Reporting Tool available in 2Q RELIABILITY ACCOUNTABILITY Proposed Variable Generation Metrics 16 RELIABILITY ACCOUNTABILITY 8

42 11/4/2012 Increase in wind & solar from 2010 to RELIABILITY ACCOUNTABILITY Separation of Two Types of Reserves Load Following and Regulation Load Following in MW = difference between Actual Load and Hour Ahead Scheduled Regulation in MW = difference between Actual Load and 5 minute Ahead Scheduled REFERENCE: California ISO Renewable Integration Report RenewableResourcesReport.pdf 18 RELIABILITY ACCOUNTABILITY 9

43 11/4/2012 Proposed VG Metric VG metric = Change in Load Following / Change in wind & solar Where Load following = Max value of load following time trend 19 RELIABILITY ACCOUNTABILITY Metric Reporting Templates REFERENCE: NERC PAS Reliability Metric Reporting Template 20 RELIABILITY ACCOUNTABILITY 10

44 Phasor Measurement Unit s Frequency Calculations Eddy Lim FERC Office of Electric Reliability NERC Resources Subcommittee Meeting October 23-24,

45 Motivation Observed Frequency Spike in SONGS PMU Data after Path 44 Separation Is this real or a bad data point? Was it possible the units experienced this? 2

46 What is a Phasor? In an AC system, voltage and current are sinusoidal Can be represented as: A sine wave changing with time A vector rotating in a circle Both are equivalent, the second is mathematically easier 3

47 What is a Phasor? Need three values to describe a phasor Frequency (speed of rotation) Magnitude (length of arrow) Phase angle (angle between two arrows) This example shows 3 phasors with: Different magnitudes Same frequency Constant phase angles 4

48 Phasor Calculation PMU samples voltage and current waveforms at a fixed rate PMU computes a phasor each time a new sample is taken 24 points per cycle (1440 points/second) Phasors are averaged and positive sequence phasors are reported from PMU and streamed to servers Typical reporting rate of 30 or 60 times a second 5

49 Phasors has to be Measured Relative to Something For phasors reported by PMUs, angle is relative to a 60 Hz waveform that peaks at the top of the second Per IEEE Standard C Using the GPS as the timing clock, we know exactly where the top of the second is 6

50 An Example (Blue is GPS, Green is PMU) θ = 45 θ = 45 θ = T 0 =0 T=T 0 +45º Time (Cycles) 7

51 Frequency Calculation Positive sequence phasor looks something like this: A = Magnitude e = Euler number (approx ) Phase j = operator, shift 90 B = Constant t = Time f = Frequency Easy to calculate frequency from phase! 8

52 Off-nominal Frequency θ = 36 θ = 36 θ = 36 θ = 9 θ = 45 θ = 81 θ = 117 T 0 = Time (Cycles) 9 Angle between the waveforms changes as time passes Angle grows at a constant rate (36 per cycle)

53 Frequency Calculation 60 Hz 6 Hz = 54 Hz Green waveform is at 54 Hz 10

54 What Happens During a Transient? Typically apply a low-pass filter with cutoff frequency around 400 Hz. Filter will remove the noise of switching surges, but not step changes in voltage or angle. 11

55 A Simple Transient θ = 0 θ = 45 θ = 0 θ = 0 θ = 45 θ = 45 θ = 0 θ = 0 θ = 0 T 0 = T=T 0 +45º Time (Cycles) 12 If f is measured from dθ/dt, it will be zero for all data points except one. At that one point ( θ = 45 ), f will be very large

56 Transients in the Bus Angle Appears when a Line Trips Assume system runs slightly under 60 Hz Watch for step change in angle when the line trips θ Hz f Degrees θ Time 13

57 Actual Bus Angle Data from South of SONGS Separation (SONGS is reference bus) 14 Angle Sudden (4-6 cycle) change in angle at separation :38:20 15:38:22 15:38:24 15:38:26 Devers Palo Verde 20 15:27 15:29 15:31 15:33 15:35 15:37 Time (PDT)

58 Actual MW and Frequency Data from South of SONGS Separation Power (MW) 1,750 1,500 1,250 1, SONGS 2 Power SONGS 3 Power Frequency Frequency (Hz) 0 15:38:20 15:38:22 15:38:24 15:38:26 Time (PDT)

59 PMU Calculated SONGS Frequency This is an invalid point that results from how the PMU calculates frequency. SONGS units did not actually accelerate over 61 Hz Actual acceleration approx. 1 Hz / second 16

60 Simulations of a Separation Using a Simple 4- Bus System to Compare Calculation Methods 100 MW Bus Angles 13 kv 230 kv -100 MW 17

61 Calculation of frequency using the derivative of the bus angles 100 MW Frequency 61.1 Hz 13 kv 230 kv 59.3 Hz -100 MW 18

62 Comparison of Frequency Calculations Methods, Using Results from Transient Stability Run (Derivative of Bus Angles and the Output of Frequency Channel) 100 MW 61.1 Hz Frequency 13 kv 230 kv -100 MW 19

63 Comparison of Frequency Calculations Methods, SONGS Separation During Southwest Blackout (Derivative of Bus Angles and the Output of Frequency Channel) Frequency Frequency calculation using derivative of bus angles (PMU) Frequency calculation from transient stability run PMU Reported Frequency Time (s)

64 Summary PMU frequency calculations are generally accurate except for one to three readings during transient events. Particularly with large sudden changes in phase angles 21

65 Series Capacitor Insertion 60.1 Frequency Data from PMUs Frequency (Hz) Time (s), event around t=10

66 Series Capacitor Insertion 60 Bus Angle Data from PMUs Delta Bus Angle (degrees) Time (s), event around t=10

67 Questions?

68 More Detailed Math Taken from Synchronized Phasor Measurements and Their Applications by Phadke and Thorp (2008)

69 11/4/2012 Consortium for Electric Reliability Technology Solutions NERC Applications Status NERC Applications Status for Resources Subcommittee Gil Tam Atlanta, Georgia October 24-25, 2012 Agenda NERC Application Status Summary Inadvertent Application Update Frequency Response Event Values Definition Update Per RS Decision at the July 2012 RS Meeting Page 1 1

70 11/4/2012 NERC Applications Status Summary Application Resource Adequacy (ACE Frequency) Inadvertent Area Interchange Error (AIE) NERC Applications Status and Authorized Users (Many companies have several authorized users) Release 7.0 Current Production version 170 authorized users. Release Current Production version 272 authorized users New website design to resolve existing application interface with Window 7 is near completion. EPG and NERC are working to install and test website and target for completion by November. (details on following slide) Release 1.0 Current production version 122 authorized users Intelligent Alarms Frequency Monitoring and Analysis (FMA) Automated Reliability Reports (ARR) Release 1.0 Current production version 138 authorized users Release 2.5 Current production version 118 authorized users Release 1.0 Current production version 63 authorized users Monthly reports through September 2012 and Seasonal reports through Summer 2012 have been posted in ARR website. (ARR website access details on following slide) Page 2 Overview of Activities Support Users of CERTS NERC Applications and Update Applications as needed Support RS and NERC staff on Frequency Event Excursion Analysis Inadvertent Data Entry Website Development, Implementation and Training Frequency Response Event Detection/Analysis and Generate Monthly Reports with Frequency Plots Generate Monthly BAAL Reports and Submit Reports to the Balancing Authority Reliability based b Control lstandard ddrafting Team (BARCSDT) in Support of Proof of Concept Field Trial Standards Development Analytic Support Page 3 2

71 11/4/2012 ARR Website Access As a subscriber and recipient of the Daily Automatic Reliability Reports, you can also access the ARR Monthly, Quarterly and Yearly reports as well as the archived Daily reports at the following website: To access the above website, use your EPG Username (xxxxxxxxx) and password to login Note: If you do not already have a password (or forgot your username or password) and need to activate one for the first time, please go to the EPG Login Page at and click on the Forgot your password or username? link on This Page and follow instructions If you have any questions, please contact EPG at or via support@electricpowergroup.com You can also use this link to contact EPG: and use the Contact Us form Page 4 New Inadvertent Data Entry Website Implementation EPG has completed software design and sent software to NERC for implementation NERC IT staff is installing software Testing is in progress Target is to complete and deploy in early November 2012 EPG will send notices and user instructions to all authorized users NERC Will Provide Hosting Services Web Address User Security Data Backup User Training EPG will schedule and provide a one hour training session via WebEx Page 5 3

72 11/4/2012 Frequency Response Event Values Definition Update Per RS decision at the July 25 26, 2012 RS meeting, the definitions of the following values for the Frequency Response events have been revised and implemented effective with the month of July 2012 report: The definition of value A has been revised as the average frequency value between T( 16) and T( 1) The Criteria for Point C for all Interconnections is the maximum or minimum frequency value not to exceed 12 seconds after T(0) The value A and Point C frequency value have been updated retroactive to January 2009, and all the frequency yplots have also been updated retroactive to July 2011 The event detection delta frequency threshold for ERCOT has been revised from 90 mhz to 150 mhz effective August 2012 Page 6 Frequency Event Detection Methodology Delta Frequency Detection Methodology A frequency event is detected and captured if during a 15 second rolling time window the change frequency exceeds the frequency threshold shown in the table below for each interconnection. The thresholds in the table are being monitored and tuned to ensure the appropriate significant events are captured Freq. Delta Rolling 15 sec window Revised: September 2012 Page 7 4

73 11/4/2012 1

74 11/4/2012 2

75 11/4/2012 3

76 11/4/2012 4

77 11/4/2012 5

78 11/4/2012 ERCOT Frequency Control & Time Error Correction Report Resources Subcommittee Meeting Atlanta, GA October 25, 2012 ERCOT Time Error Corrections 2012 Time Corrections Total Hours on Avg Hours Avg Corrections % Time on Year Month Days Fast Slow Count Control Per Correction Per Day Correction 2012 Jan % Feb % Mar % Apr % May % Jun % Jul % Aug % Sep % 3% Oct 31 Nov 30 Dec 31 1

79 11/4/2012 Monthly Hours on Time Correction Total Hours on Correction Hours on Control Monthly Time Correction Summary Number of Corrections Slow Fast 2

80 11/4/2012 Frequency Profile, CPS1 & 2, Daily RMS1 & ERCOT Total Energy and Wind Generation. ERCOT FREQUENCY CONTROL PERFORMANCE Comparing July 2010 vs July 2012 profile of frequency in 5 mhz bins Jul-10 Jul

81 11/4/2012 Comparing August 2010 vs August 2012 profile of frequency in 5 mhz bins Aug-10 Aug Comparing September 2010 vs September 2012 profile of frequency in 5 mhz bins Sep-10 Sep-12 4

82 11/4/ ERCOT Frequency Profile Comparison January through December of each Year One Minute Occurances ERCOT CPS1 12 Month Rolling Average CPS1 = e CPS1 Average Nov-05 Sep-05 Jul-05 May-05 Mar-05 Jan-05 Nov-04 Sep-04 Jul-04 May-04 Mar-04 Jan-04 Nov-03 Sep-03 Jul-03 May-03 Jul-09 May-09 Mar-09 Jan-09 Nov-08 Sep-08 Jul-08 May-08 Mar-08 Jan-08 Nov-07 Sep-07 Jul-07 May-07 Mar-07 Jan-07 Nov-06 Sep-06 Jul-06 May-06 Mar-06 Jan-06 Sep-12 Jul-12 May-12 Mar-12 Jan-12 Nov-11 Sep-11 Jul-11 May-11 Mar-11 Jan-11 Nov-10 Sep-10 Jul-10 May-10 Mar-10 Jan-10 Nov-09 Sep Monthly Average 12 Month Rolling Average 5

83 11/4/ ERCOT CPS2 Score* *ERCOT as a single control area is exempt from CPS2. These scores are For Information Only CPS Oct-06 Jul-06 Apr-06 Jan-06 Oct-05 Jul-05 Jan-09 Oct-08 Jul-08 Apr-08 Jan-08 Oct-07 Jul-07 Apr-07 Jan-07 Jul-12 Apr-12 Jan-12 Oct-11 Jul-11 Apr-11 Jan-11 Oct-10 Jul-10 Apr-10 Jan-10 Oct-09 Jul-09 Apr-09 CPS2 Month Trend (Monthly CPS2 Score) Daily RMS1 of ERCOT Frequency /1/2000 7/1/2000 1/1/2001 7/1/2001 1/1/2002 7/1/2002 1/1/2003 7/1/2003 1/1/2004 7/1/2004 1/1/2005 7/1/2005 1/1/2006 7/1/2006 1/1/2007 7/1/2007 1/1/2008 7/1/2008 1/1/2009 7/1/2009 1/1/2010 7/1/2010 1/1/2011 7/1/2011 1/1/2012 7/1/2012 6

84 11/4/2012 Daily RMS1 of ERCOT Frequency /1/2004 4/1/2004 7/1/ /1/2004 1/1/2005 4/1/2005 7/1/ /1/2005 1/1/2006 4/1/2006 7/1/ /1/2006 1/1/2007 4/1/2007 7/1/ /1/2007 1/1/2008 4/1/2008 7/1/ /1/2008 1/1/2009 4/1/2009 7/1/ /1/2009 1/1/2010 4/1/2010 7/1/ /1/2010 Daily RMS1 of ERCOT Frequency 1/1/2011 4/1/2011 7/1/ /1/2011 1/1/2012 4/1/2012 7/1/ /1/2007 4/1/2007 7/1/ /1/2007 1/1/2008 4/1/2008 7/1/ /1/2008 1/1/2009 4/1/2009 7/1/ /1/2009 1/1/2010 4/1/2010 7/1/ /1/2010 1/1/2011 4/1/2011 7/1/ /1/2011 1/1/2012 4/1/2012 7/1/2012 7

85 11/4/2012 ERCOT Total Energy 45,000,000 40,000,000 35,000,000 30,000,000 MWH 25,000,000 20,000,000 15,000,000 10,000,000 5,000,000 0 January February March April May June July August September October November December ERCOT Total Energy from Wind Generation 3,500,000 3,000,000 2,500, MWH 2,000,000 1,500,000 1,000, ,000 0 January February March April May June July August September October November December

86 11/4/2012 ERCOT % Energy from Wind Generation 14.00% 12.00% 10.00% 00% 8.00% 6.00% 4.00% 2.00% 0.00% January February March April May June July August September October November December