Day-ahead electricity price forecasting based on rolling time series and least square-support vector machine model

Abstract

Considering the electricity price's volatility and various elements which affect the price in the electricity market, the paper presents hybrid model for the day-ahead electricity market clearing price forecasting. The paper adopts autoregressive moving average (ARMAX) model to reveal the linear relationship between power load and electricity price; the generalized autoregressive conditional heteroskedasticity (GARCH) model to reveal the heteroskedasticity properties of residual. Simultaneously the paper presents the inexactness and irrationality that modeling by the historical data long ago to forecast the price with the change of the time, then presents the rolling forecast that constantly using the latest data to modeling the ARMAX-AR-GARCH model. To reveal the nonlinear relationship between power load and electricity price, the paper adopts least squares support vector machine (LS-SVM). Using the proposed method, the day-ahead electricity prices of California electricity market are forecasted, prediction results show the efficiency of the proposed method.