A NOVEL ARIMA APPROACH ON ELECTRICITY PRICE FORECASTING WITH THE IMPROVEMENT OF PREDICTED ERROR

In power markets, accurate electricity price forecasting is a crucial issue concerned by all market participants. Several approaches have been proposed to attempt to improve the accuracy of price forecasting. However due to the complicated factors affecting electricity prices, experience shows that setting up single forecasting model is very difficult for improving the accuracy of forecasting. This paper proposes a new ARIMA approach on forecasting electricity price with the improvement of predicted errors for the first time. Except setting up a conventional price forecasting model, we also present forecasting error models by iterative method and use the predicted errors to update the forecasted prices so as to gradually improve the forecasting accuracy. An integrated ARIMA based forecasting model for daily average price is established for validating the effectiveness of the proposed methodology by the historical data of California Power Market. The results show the presented approach improves the accuracy of forecasting significantly and with the features of easy modeling and suitable for extending to forecasting market clearing price and electricity load, even other forecasting domains.