ANN Daily Peak Forecast for Peak Demand Charges Management

Abstract

Demand charges (DC) is one of the major utility charges that represents a considerable portion of the electricity bill, especially in the case of large electricity consumers. Predicting the monthly peak of the facility helps manage peak demand charges (PDC). Forecasting of the monthly peak demand value of the facility is required to manage the PDC component of the energy bill. The monthly peak of a given class-A facility is a very specific and unique problem that requires an individual forecasting module for each facility because each facility is unique in its pattern of operation, energy consumption, and load profile. This paper proposes a methodology based on artificial neural networks (ANN) to forecast the daily peak demand of a given class-A facility to help manage its PDC. Based on the market regulations of Ontario (Canada), and the use of battery storage systems (BSSs) and real data for a large class-A Canadian electricity consumer in Ontario, the simulation results demonstrate the effectiveness of the proposed forecasting module in minimizing the DC cost of the class-A electricity customer. Using real class-A electricity consumer demand data, we show that our algorithm module is more consistent from day to day and provides a solution to peak demand problems.