Classifying day-ahead electricity markets using pattern recognition for demand response

Abstract

In this paper, we model the Demand Response eXchange (DRX) as an entity that facilitates the trading of demand response (DR) in the existing bulk power market through DR aggregators (DRA). DR as a service is used by the independent system operator (ISO) only when the market is settled inefficiently. A simple threshold on locational marginal price (LMP) cannot be set as the market clearing price depends on various factors, such as the transmission network availability, weather, and available resources. To detect inefficiency in the day-ahead market, we use statistical pattern recognition and compare among the various available techniques to integrate the DRX into a fully deregulated day-ahead market clearing method. We use support vector machines (SVM) to detect market inefficiencies during market clearing using real-data from the PJM ISO, and validate on the IEEE 24-bus system. We show that the DRX can be integrated into the existing bulk power market, with the ISO using pattern recognition techniques to detect market inefficiencies and trigger the DRX during such hours.