Particle Swarm Optimization Based GM(1,2) Method on Day-Ahead Electricity Price Forecasting with Predicted Error Improvement

Under deregulated environment, accurate electricity price forecasting is a crucial issue concerned by all participants. Experience shows that single forecasting model is very difficult to improve the forecasting accuracy due to the complicated factors affecting electricity prices. A particle swarm optimization (PSO) based GM(1,2) method on day-ahead electricity price forecasting with predicted error improvement is proposed, in which the moving average method is used to process the raw series, the PSO based GM(1,2) model to the processed series and the time series analysis to further improve the predicted errors. The numerical example based on the historical data of the PJM market shows that the method can reflect the characteristics of electricity price better and the forecasting accuracy can be improved virtually compared with the conventional GM(1,2) model. The forecasted prices accurate enough to be used by market participants to prepare their bidding strategies.