Better Forecasting by ERCOT has Reduced the ORDC LOLP Even As Intermittent Generation Capacity has Increased

Abstract

The suggestion of multiplying the standard deviation () used in the Electric Reliability Council of Texas’ (ERCOT) Operating Reserve Demand Curve (ORDC) loss of load probability (LOLP) by a factor greater than 1.0 has been posited to solve the “missing money argument.” Historical data from 2015 through 2017 supports this theory, although the calculated factor is only 1.02. This working paper compares the ERCOT 2015 LOLP distributions with the 2018 distributions. 21 of 24 distributions have seen their standard deviation decrease, despite the historical data that shows that the standard deviation of distribution of one hour change in wind generation has increased. Additionally, the change in the mean of the summer time distributions result in a shifting away from the minimum contingency level. These two outcomes both produce a lower LOLP for a given reserve level in 2018 compared to 2015. This is caused by more accurate forecasting by ERCOT. The greater forecasting accuracy overcomes the increase in change of wind generation over an hour resulting in a reduced LOLP. Further evidence suggests that in the windy, low load months, when wind generation makes up a larger percentage of total energy produced, the change in the mean of the distributions does reflect the increase in LOLP caused by intermittent generation. An analysis of the proposal to “shift the mean error by one standard deviation” is performed showing that is identical to increasing MCL by one standard deviation. Variations on this proposal are analyzed. ERCOT 2015 data (linked in references and recorded with Mendeley) is evaluated with the proposal to shift the mean error of the distribution by a standard deviation and the variations, and the increase electricity cost from the proposals ranges from $2.4 billion to $5.75 billion per year. Finally, the foundational problem of using a forecasted error distribution to model a scarcity pricing mechanism instead of a market driven change in price model is briefly discussed.