Error Analysis of Customer Baseline Load (CBL) Calculation Methods for Residential Customers

Federal Energy Regulatory Commission (FERC) 745 order has created an environment that allows demand response owners to sell their load reduction in the wholesale market. One of the main challenges that independent system operators and utilities face is developing customer baseline load (CBL) calculation methods that work satisfactorily in this new environment. Consequently, it is critical that these methods need to be evaluated from the error performance's perspective. In this paper, error analysis of CBL calculation methods for residential customers is carried out theoretically and empirically. To perform theoretical analysis, the utility function of customers is analyzed to determine the existence of the economic incentives for gaming and inefficient consumption as well as studying the impact of inaccuracy on the social welfare loss. Furthermore, to perform the empirical analysis, well-established CBL calculation methods, HighXofY (New York ISO, well known as NYISO), LowXofY, MidXofY, exponential moving average (New England ISO, well known as ISONE), and regression are first introduced and, then, utilized to calculate the CBL. A dataset consisting of 262 residential customers is used for this analysis. In addition, the error analysis is performed using accuracy and bias metrics. To reach a valid conclusion about the overall performance of CBL methods, an economic analysis of a hypothetical peak time rebate (PTR) program is carried out. According to the results of the case study, the utility pays at least half of its revenue as a rebate solely due to inaccuracy of CBL methods. In addition, it is demonstrated that PTR creates inefficiencies in the residential sector because of the failure of CBL calculation methods to accurately predict the customers' load profile on the event day.