Electricity Price Forecasting Model Based on Single Design Parameter Neural Network for Indian Energy Exchange

Abstract

The accuracy of electricity price forecasting is important to control and optimize the vertically unbundled electricity market power generation. In this research, based on the theories of the generalized regression neural network (GRNN), a new approach is proposed to predict the electricity prices with a single design parameter. In this paper, partial autocorrelation is applied on times series data to get correct input values. The data for historical prices are obtained from the energy exchange of India. The presented model is compared with other recent electricity price forecasting methods including artificial neural network (ANN), ANN-ANN-particle swarm optimization (PSO), Wavelet-based ANN, and Wavelet-based ANN-ANN-PSO. The results show that GRNN behaves best in determining accurate forecasts than other forecasting approaches, for instance, ANN, ANN-ANN-PSO, Wavelet-based ANN, Wavelet-based ANN-ANN-PSO, and Wavelet-based ANN-PSO (random initialization). The procedure is direct, and forecasting using this approach improves forecast precision. Subsequently, the proposed GRNN model can be used for an accurate forecast of market clearing price in any energy exchange.