Day-ahead electricity price forecasting using artificial intelligence

Accurate day-ahead electricity price forecasting (DEPF) has significant meanings in deregulated electrical power market due to its profitable function for all the participants to make reasonable decisions during the market business activities. However, the DEPF with satisfactory precision is difficult to be gained because of the violent volatility of electricity price caused by many factors. In this study, a multilayer perceptron artificial neural networks model is constructed for the DEPF in spot market of Nord Pool which is one of the most successful electrical power markets in the world. The major influencing factors are chosen by statistical methods called auto correlation function (ACF) and cross correlation function (CCF), and the standard error back-propagation algorithm is improved by using self-adaptive learning rate and self-adaptive momentum coefficient algorithm to make the training process more efficient both in global optimization and time saving. The most suitable structure of the network is determined by a trial-and-error experiment minimizing MAPE and MSRE of the network and several commonly used error indicators are employed to evaluate the goodness of fit performance of the model. The case study indicates that the DEPF of the proposed model is more reasonable and accurate, comparing with that of traditional ARIMA model.