Day ahead demand response model with algorithm-based consumption classification and tariff planning

Abstract

Demand Response (DR) programs effectively support the supply-demand balance in power systems. This study proposes a new Time-of-Use (ToU) tariff model, a type of price-based demand response (DR). The proposed model aims to shape the daily consumption profile of the distribution region by encouraging consumers through tariffs. Generally, in ToU demand response, various presumptions are being adopted to define consumption periods and consumer classes. It is unlikely that these presumptions are suitable for every day of the year and every consumer in the system. Achieving these definitions and classifications based on algorithms, free from human error, will increase the success and comprehensiveness of the DR program. To maximize success in this context, a deep learning model has been leveraged for day-ahead consumption forecasting. The forecasted daily consumption profile is partitioned into ToUs by using the Moving Boundary method. In addition, consumers within the region have been segmented based on the similarities in consumption behavior. Tariffs have been priced for each cluster through an optimization model that considers their participation rates in total consumption and contributions to consumption in ToU regions. The proposed model has been tested over a dataset belonging to a distribution system where different types of consumers coexist. Two sample days have been chosen as one is a weekday and the other is a weekend. The results prove that shifting consumption as encouraged by the proposed TOU method reduces consumers' costs. While, the profitability of the system operator is also preserved.