基于小波分析的短期电价ARIMA预测方法

电力市场中的电价具有特殊的周期性,以天、周、年为周期波动,且大周期中嵌套小周期.作者提出一种基于小波分析的累积式自回归滑动平均方法(WARIMA)用于短期电价预测,首先利用小波变换能将交织不同频率成份的混合信号分解成不同频带上的块信号的特性,将电价这一随机序列进行小波分解,得到低频上的概貌序列和高频上的细节序列,并在此基础上对各个子电价序列分别利用累积式自回归滑动平均模型(ARIMA)进行预测,然后在电价平稳时段用概貌序列预测结果直接作为电价预测结果,而在电价非平稳时段将各子序列预测结果重构作为最终的预测结果.为了对比分析,将直接使用ARIMA模型的预测结果和采用WARIMA方法的预测结果进行了比较,表明引入小波分析对提高预测精度是有益的.

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.