Day-Ahead electricity price forecasting using WT, ANN and Chaotic gravitational search model

With the deregulation of electric power systems, market participants are facing an important task of bidding energy to an Independent System Operator (ISO). Modeling and electricity prices forecasting in competitive market relative with its characteristics such as inability to store, non-stationary and time variant behavior and seasonality violation should be considered. A model with more accuracy and less error will be more efficient which a price forecast with a less prediction errors, yields maximum profits for market players. To achieve more accurate and robust price forecast, in this paper, a new hybrid forecast technique based on Wavelet Transforms (WT), feature selection technique, Artificial Neural Network (ANN) and Gravitational Search Algorithm (GSA) combined with Chaotic Local Search (CLS) model is proposed for day-ahead electricity price forecasting. The feature selection method is an improved version of the Mmutual Information (MI) technique. The superiority of this proposed method is examined by using the data acquired from the Iran and market clearing price (MCP) of Spanish market. Empirical results show that the proposed method performs better than some of the other price forecast methods. Also, the exploitation and exploration of the proposed intelligent algorithm for tuning weight and bias parameters of ANN is improved.