Day-ahead electricity price forecasting using the relief algorithm and neural networks

In a competitive electricity market, forecast of energy prices is a key information for the market participants. However, price signal usually has a complex behavior due to its nonlinearity, nonstationarity, and time variancy. In spite of all performed researches on this area in the recent years, there is still an essential need for more accurate and robust price forecast methods. In this paper, a combination of a feature selection technique and neural network (NN) is proposed for this purpose. The feature selection method is a modified version of the relief algorithm, proposed for the feature selection of price forecasting. Then, by means of the most relevant, explanatory and irredundant features, a neural network (NN) predicts the next values of the price signal. The adjustable parameters of the whole method are fine-tuned by a cross-validation technique. The proposed method is examined on PJM electricity market, forecasting day-ahead locational marginal prices (LMPs), and compared with some of the most recent price forecast methods especially some other popular and validated feature selection techniques. These comparisons indicate the validity and robustness of the proposed forecasting method.