通过对已有的各种电价预测方法的深入研究,提出一种带置信区间的混合式电价预测方法。比较了自相关函数和偏自相关函数判断平稳性及残差方差图定阶法,认为游程检验判断电价序列平稳性方法和AIC准则模型定阶法可避免预测过程中的主观因素,提高预测精度。通过引入电价误差序列异方差判断,结合ARIMA和GARCH,提出含置信区间的电价预测方法,极大地克服了单点预测缺点,增加算法的灵活性,使得市场竞价者可以根据自己对电价预测精度的期望选择电价波动的范围。采用PJM市场数据,验证了算法的有效性。
Base on research on existing methods for electricity price forecasting, a new combined method with confidence intervals is proposed. The run test was used to judge the price series stationarity and the AIC criterion method was applied to determine the order, so the subjectivity can be avoided and the forecasting accuracy can be improved. Moreover, the ARIMA and GARCH were effectively combined though heteroscedasticity determination for electricity price error series. Also, the price forecasting method with confidence intervals is proposed, which overcomes the shortcomings of single-point forecasting and increases the algorithm flexibility to enable the participators to choose the price fluctuation range according to their expectations on price forecasting accuracy. The model is proved effective by data on PJM market.