Electricity price forecasting with effective feature preprocessing

Forecasting electricity price in a deregulated market is important for the market participants. Feature preprocessing technique in a forecasting model is essential and can significantly influence the forecasting accuracy. Although several feature preprocessing techniques have been applied in electricity price forecasting problem, no systematic research is conducted to study and compare their performances. Therefore, how to select effective feature preprocessing techniques in price forecasting models remains a problem. In this paper, the authors conduct a comprehensive study of existing feature preprocessing techniques and their empirical performances in price forecasting problem. It is demonstrated in our experiment that, effective feature preprocessing can greatly enhance the forecasting accuracy. Moreover, it can identify the potential related factors of electricity price. The authors also propose a discussion about selecting suitable feature preprocessing techniques in price forecasting problem, which can be useful for the researchers who want to integrate effective feature preprocessing techniques in their price forecasting models