X11 Seasonal Adjustment Combined With SVM For Short-term Electric Load Forecasting

Electric load can be influenced by various factors such as the weather, seasons, holi-days, social and cultural factors. It exhibits strong nonlinear characteristics with seasonal effects included, which makes accurate forecasting of electric load very challenging. Sup-port vector machines (SVM) are commonly used to solve nonlinear regression and time series problems because of its outstanding capabilities of prediction through nonlinear mapping. However, in order to analyze the seasonal time series accurately and to reduce the effect of seasonal and irregular factors, X11 seasonal adjustment method is useful in addressing the seasonal effects in time series. This paper proposes a new hybrid model for short-term electric load forecasting by combining a SVM model and X11 seasonal adjustment technique. X11 seasonal adjustment method is expanded to deal with the seasonal time series whose period is one-day and sampling frequency is one sample point per half of an hour. The experimental results show that the proposed model is a promising alternative for short-term electric load forecasting.