STUDY ON HYBRID MODEL FOR SYSTEM MARGINAL PRICE FORECASTING IN ELECTRICITY MARKET

Price forecasting has been a useful tool for market participants to provide important economic information. But the complicated influence factors of the price make the forecasting more difficult. So a novel hybrid model for forecasting system marginal price (SMP) in spot market is presented, which integrates independent component analysis with support vector machine, called ICA-SVM. First, this paper designs the ICA self-adapting iteration method by constructing hybrid optimal transform function, and a new de-redundancy function based on higher-order statistic information. Then the feature extraction of SMP influencing factors is realized, which produces SMP effective influencing factor sample set. After the training of regress SVM with the obtained sample set, SMP forecast model is built whose accuracy is enhanced with the generalization ability of support vector machine and the feature extraction ability of independent component analysis. Finally, real-word data of spot market in California is employed to demonstrate the validity of the proposed approach.