DAM Delays- Regression Analysis January 25th, 2021 WMWG ERCOT Staff

DAM Delays- Regression Analysis January 25th, 2021 WMWG ERCOT Staff

Overview Objective: To determine inputs that have significant influence on the total time taken to complete DAM Optimization Performed simple linear regression analysis on 22 variables against DAM run time and ranked them based on R-squared (R-squared is a statistical measure of how close the data are to the fitted regression line) Performed multiple linear regression using 6 most significant variables against DAM Run Time and provide Model Interpretations PUBLIC 2

Methodology Performed simple linear regression on 22 variables against DAM run time and ranked them based on model’s goodness of fit (R-squared) PTP-related variables PTPTOTAL INT PTP 10 MWs PTPTOTAL DLY PTP multiHourBLK PTP 5.1-10 MWs PTP 0-5 MWs Energy-Only Offer-related variables EOO INT EOO NOFixedQ NOmultiHourBLK EOO FixedQTY MultiHourBLK EOO MultiHourBLK EOO DLY EOO FixedQTY NoMultiHourBLK EB FixedQTY MultiHourBLK EB INT EB NOFixedQ NOmultiHourBLK EB FixQ NoMultiHourBLK EB MultiHourBLK EB DLY Three Part Offers TPO INT TPO DLY Other variables Outages Count Number of DAM Constraints (binding) R-squared – – – PUBLIC Energy Bid-related variables the amount of variability in the dependent variable (i.e. DAM Run Time) explained by the independent variable (i.e. EOO INT) On a scale of 0 to 1, higher value implies a better model fit If R squared is 1, it means that the two variables are perfectly correlated. 3

Simple Linear Regression Analysis (1) Independent Variable Intercept Coefficient PTPTOTAL INT 0.4026 -9.41769 0.000921 **Num DAM Constraint 0.3824 -11.3779 0.148031 PTP 10 MWs 0.3502 9.054788 0.003267 PTPTOTAL DLY 0.3125 13.05524 0.006401 PTP multiHourBLK 0.2718 77.63386 0.002753 PTP 5.1-10 MWs 0.2549 56.43301 0.002732 PTP 0-5 MWs 0.2396 30.66995 0.00106 EOO INT 0.1995 48.91647 0.006436 EOO NOFixedQ NOmultiHourBLK 0.1889 54.68494 0.006528 EOO FixedQTY MultiHourBLK 0.1175 86.16995 0.116792 EOO MultiHourBLK 0.112 83.37832 0.094864 EOO DLY 0.0453 82.97662 0.026324 *Sorted in descending order by R-squared PUBLIC ** Num DAM Constraint is not an input to DAM 4

Simple Linear Regression Analysis (2) Independent Variable Intercept Coefficient EB FixedQTY MultiHourBLK 0.0291 99.70958 0.033693 EB INT 0.0211 98.29238 0.001793 EB NOFixedQ NOmultiHourBLK 0.02 100.8517 0.001844 EB FixQ NoMultiHourBLK 0.012 105.1406 0.027555 Outages 0.0092 104.3915 0.003682 TPO DLY 0.0051 96.87542 0.107963 EB MultiHourBLK 0.004 124.7751 -0.01009 EB DLY 0.003 112.0172 0.004541 EOO FixedQTY NoMultiHourBLK 0.0027 125.6841 -0.01979 TPO INT 0.0017 105.4514 0.00285 *Sorted in descending order by R-squared PUBLIC 5

Multicollinearity- When an independent variable is highly correlated with one or more independent variable (s), it can increase the variance of the coefficient estimates, causing coefficient estimates to be unstable (i.e. switched sign( /-)) and sensitive to minor change to the model. Independent Variable Intercept Coefficient PTPTOTAL INT 0.4026 -9.41769 0.000921 Num DAM Constraint 0.3824 -11.3779 0.148031 PTP 10 MWs 0.3502 9.054788 0.003267 PTPTOTAL DLY 0.3125 13.05524 0.006401 PTP multiHourBLKStrongly correlated with PTPTOTAL INT 0.2718 77.63386 0.002753 PTP 5.1-10 MWs 0.2549 56.43301 0.002732 PTP 0-5 MWs 0.2396 30.66995 0.00106 EOO INT 0.1995 48.91647 0.006436 EOO NOFixedQ NOmultiHourBLK 0.1889 54.68494 0.006528 EOO FixedQTY MultiHourBLK Strongly0.1175 correlated with 86.16995 EOO INT PUBLIC 0.116792 EOO MultiHourBLK 0.112 83.37832 0.094864 EOO DLY 0.0453 82.97662 0.026324

Multicollinearity- When an independent variable is highly correlated with one or more independent variable (s), it can increase the variance of the coefficient estimates, causing coefficient estimates to be unstable (i.e. switched sign( /-)) and sensitive to minor change to the model. Independent Variable Intercept Coefficient EB FixedQTY MultiHourBLK 0.0291 99.70958 0.033693 EB INT 0.0211 98.29238 0.001793 Mildly correlated 0.02 with EB NOFixedQ NOmultiHourBLK EB FixedQTY MultiHourBLK 100.8517 0.001844 EB FixQ NoMultiHourBLK 0.012 105.1406 0.027555 Outages 0.0092 104.3915 0.003682 TPO DLY 0.0051 96.87542 0.107963 EB MultiHourBLK 0.004 124.7751 -0.01009 EB DLY 0.003 112.0172 0.004541 EOO FixedQTY NoMultiHourBLK 0.0027 125.6841 -0.01979 TPO INT 0.0017 105.4514 0.00285 PUBLIC

PTPTOTAL INT PUBLIC Strong positive relationship between PTPTOTAL INT and DAM Run Time 8

Number of Constraints PUBLIC Strong positive relationship between Number of constraints and DAM Run Time 9

Energy Only Offer – EOO INT PUBLIC Mildly positive relationship between EOO INT and DAM Run Time is identified 10

Energy Bid- EB FixedQTY MultiHourBLK PUBLIC Small positive relationship between EB FixedQTY MultiHourBLK and DAM Run Time is identified 11

Outage PUBLIC No strong relationship between Outage and DAM run time is identified. 12

Three Part Offer- TPO DLY PUBLIC No strong relationship between TPO DLY and DAM Run Time is identified 13

PTPTOTAL INT vs. Number of Constraints PUBLIC Strongly positive relationship between PTPTOTAL INT and Number of Constraints is identified 14

Outages vs. Number of Constraints PUBLIC Mildly positive relationship between Outage Count and Number of Constraints is identified 15

Multiple Regression Analysis Dependent Variable : DAM Run Time Predictors Estimates 90 % Confidence Interval p value (Intercept) - 57.11 -85.78 -28.44 0.001* PTPTOTAL INT 0.0005 0.0004 0.0006 0.001* EOO INT 0.0016 0.0008 0.0025 0.002* Number of DAM Constraints 0.0873 0.0719 0.1026 0.001* TPO DLY 0.0966 -0.0143 0.2074 0.152 Outages -0.0019 -0.0048 0.0010 0.278 Observations 636 0.492 *The effect of PTPTOTAL INT, EOO INT and Number of DAM Constraints on DAM Run Time are found to be significant at 5% statistical significance level PUBLIC 16

Regression Model Interpretation For every 10,000 unit increase in PTPTOTAL INT (Max: 208,512), DAM Run Time is increased by 5 minutes (4 6 minutes) For every 1,000 unit increase in EOO INT (Max: 19,915), DAM Run Time is increased by 1.6 minutes (0.8 2.5 minutes) For every 100 unit increase in the Number of Constraints (Max: 1,242), DAM Run Time is increased by 8.3 minutes (6.76 9.81 minutes) ** Model Interpretations for Outage and TPO DLY are omitted since their effect on DAM Run Time are deemed to be insignificant at 5% significance level. PUBLIC 17

Average Volume of PTP Intervals by Year Average yearly PTP intervals have increased from 2011 (consistently since 2016) Year Avg. PTP intervals 2011 35,059 2012 46,719 2013 70,791 2014 58,892 2015 72,080 2016 61,645 2017 79,983 2018 123,674 2019 128,762 2020 147,232 PUBLIC 18

Summary Total PTP Intervals, Total number of Constraints and Total Energy Only Offer (EOO) intervals appears to be the three most significant variables affecting the Total DAM Run time – Of these three variables Total PTP Intervals is the most influential variable affecting Total DAM Run time, and – Growing trend of Total PTP Intervals will exasperate the DAM performance issue – Total number of Constraints is not an input to the DAM Optimization – Total Number of Constraints appears to be highly correlated with the volume of total PTP intervals Outages does not appear to be significant variable affecting the Total DAM Run time. There was some very small positive correlation between Outage Count and Number of Constraints ERCOT at the January 6th WMS meeting had presented removing block bids as a way to reduce complexity and improve DAM Run time. The data does not show limiting block bids will be most effective in solving the performance problem. PUBLIC 19