Several-hours-ahead electricity price and load forecasting using neural networks

Abstract

In daily power markets, forecasting electricity prices and loads are the most essential task and basis for any decision making. An approach to predict the market behaviors is to use the historical prices, loads and other required information to forecast the future prices and loads. This paper introduces an approach for several-hours-ahead electricity price and load forecasting using artificial intelligence method, such as neural network model, which uses publicly available data from NEMMCO web site to forecast electricity prices and loads for the Victorian electricity market. An approach of selection of similar days is proposed according to which the load and price curves are forecasted by using the information of the days being similar to that of the forecast day. A Euclidean norm with weighted factors is used for the selection of similar days. Two different ANN models, one for several-hours-ahead load forecasting and another for several-hours-ahead price forecasting have been proposed. The forecasted price and load from the neural network is obtained by adding a correction to the selected similar days, and the correction is obtained from the neural network. MAPE results show that several-hours-ahead electricity price and load in the deregulated Victorian market can be forecasted with reasonable accuracy.