Short-term Load Forecasting Based on Asymmetric ARCH Models

The analysis on the characteristics of volatility can help to give a more precise description in load time series and may contribute to load forecasting performance. In this study, first, asymmetric effect in load time series is investigated, and the asymmetric ARCH type models, including EGARCH, TARCH, PARCH, and TCGARCH, are introduced as feasible methods for short-term load forecasting. Second, the estimation of all these asymmetric models on two empirical distributions (normal, GED) in load time series is presented. Third, with the help of asymmetric parameter, reverse leverage effect is proposed and the mechanism of asymmetric effects between different shocks is discussed. Finally, a comprehensive comparison of the forecast performance among the asymmetric ARCH type models is presented. Empirical results indicate that PARCH-GED model outperforms others in terms of some standard statistics indices, and it may be a promising statistical method for short-term electric power load forecasting.