WAVELET ANALYSIS BASED ARIMA HOURLY ELECTRICITY PRICES FORECASTING APPROACH

In the electricity markets, the electricity price has special periodical characteristics, which fluctuates in the period of day, week and year, and the longer period involves the shorter one. Because wavelet transform can decompose the signals with different frequency into different blocks with different frequency bands, a wavelet analysis based Auto Regressive Integrated Moving Average (ARIMA) model is applied to short-term electricity price forecasting. Firstly, as a stochastic sequence the electricity price is decomposed by wavelet transform and the general picture sequence in low frequency band and the detail sequence in high frequency band are obtained, on this basis the sub-sequences of electricity price are forecasted by ARIMA respectively. At stationary time interval of electricity price the forecasting result of general picture sequence is directly taken as the forecasted electricity price; at non-stationary time interval of electricity price the forecasted results of all sub-sequences are reconstructed and taken as the final forecasted result. To compare the forecasted results, two kinds of electricity price forecasting are made, one of them directly uses ARIMA forecasting model and another uses wavelet analysis combined with ARIMA (WARIMA) forecasting model. The California historical hourly electricity prices are used to test the proposed approach, the results show that WARIMA approach improves the accuracy of forecasting obviously.