Fundamental Problems with ERCOT's Operating Reserve Demand Curve and a Proposed Solution

Abstract

The methodology employed by the Electric Reliability Council of Texas (ERCOT) in its Operating Reserve Demand Curve (ORDC) as its scarcity pricing mechanism assumes that ERCOT’s distribution of the hour-ahead forecasted reserve level error can be used to accurately model a loss of load probability (LOLP). Because the hour-ahead forecasted reserve level error varies greatly throughout the day and year, ERCOT attempts to compensate by producing six curves a day, four seasons a year for a total of 24 different loss of load probability distributions. By introducing six different curves each day, the ORDC introduces a pricing change that acts identically like a reliability unit commitment (RUC). The ORDC, the scarcity pricing mechanism used by ERCOT, introduces out-of-market actions (pricing RUCs) six times a day. The out-of-market actions are equivalent to 100’s of MWs of RUCs, and with the recent retirement of thousands of MWs of generating capacity, the RUC effects will only be greater. The methodology employed by ERCOT to calculate the LOLP incorrectly applies the hour-ahead forecasted reserve level error distribution to the real-time reserve level. There is no basis in probability theory to apply a forecasted reserve level error distribution to the real-time reserve level. Consequently, the LOLP calculated by the ORDC is incorrect and without a mathematical foundation. Basically, it is a math mistake (explained in the body with empirical proof provided in the appendix). The proposed solution is a simple curve fit of the form f(x)=e^(-(x/σ)^2 ) as a better model of loss of load probability and satisfies the boundary conditions of: as real-time reserve level approaches zero, loss of load probability approaches 1.0 and as real-time reserve level approaches infinity (or very large) loss of load probability approaches zero. Finally, a third boundary condition equating the rate of change of system lambda with the rate of change of LOLP at a given reserve level is introduced. The rate of change of system lambda can be determined empirically by using the historical values of system lambda as reserve level changes. By correlating the rate of change of system lambda at a given reserve level with the slope of the curve fit at that same reserve level (f^' (x)= -2x/σ^2 e^(-(x/σ)^2 )) the rate of change of LOLP will be equal to the rate of change of system lambda. Keywords: Electric Reliability Council of Texas, ERCOT, Operating Reserve Demand Curve, ORDC, loss of load probability, LOLP, forecasted reserve level error, real-time reserve level, out-of-market, reliability unit commitment, RUC, scarcity conditions, finite limits, normal distribution, first principles, boundary conditions.