Electricity Price Prediction with Support Vector Machine and Bacterial Foraging Optimization Algorithm for Day-Ahead model

Abstract

Predicting the price of electricity is an important aspect in the operation and planning of power systems. However, predicting the price of electricity is a relatively challenging task as it faces very uncertain conditions. Hence, this study proposes a hybrid Least Square Support Vector Machine (LSSVM) and Bacterial Foraging optimization Algorithm (BFOA) for day-ahead electricity price forecast. The main contribution of this work is the multistage optimization approach of LSSVM-BFOA that can improve the forecasting accuracy and efficiency. This is achieved by optimizing the input features and parameters of LSSVM at the same time. The input features have been reduced by six optimization levels in order to avoid losing any significant input. At the same time, the average MAPE is observed and the second stage of optimization is carried out. These processes are performed until there is no improvement in MAPE is observed. This model is examined in the Ontario power market. The LSSVM-BFOA model developed showed higher prediction accuracy with less complex model structure than most existing models. The day ahead price forecast is beneficial for both power generators and consumers in bidding for electricity prices.