摘要
在电力市场中,准确的电价预测是各市场参与方共同关心的重要问题。已经提出多种理论和方法尝试提高电价预测精度,然而由于影响电价的因素十分复杂,实践证明靠建立单一的电价预测模型来提高预测精度是非常困难的。该文在分析电价波动特性和现有预测方法的基础上,首次提出结合误差预测校正电价预测来提高预测精度的新思路。在建立常规电价预测模型的基础上,对预测后的残差形成的随机序列也迭代地建立预测模型,并用预测的误差修正电价预测结果。该文采用ARIMA方法建立电价预测和误差预测模型,并用加州电力市场的历史数据建立基于ARIMA的日平均电价预测模型,预测结果表明所提方法能明显改善预测精度,而且方法简捷明了,能够推广到小时电价预测、负荷预测和其它预测领域。
In power markets, accurate electricity price forecasting is a crucial issue concerned by all market participants. Several approaches have been proposed to attempt to improve the accuracy of price forecasting. However due to the complicated factors affecting electricity prices, experience shows that setting up single forecasting model is very difficult for improving the accuracy of forecasting. This paper proposes a new ARIMA approach on forecasting electricity price with the improvement of predicted errors for the first time. Except setting up a conventional price forecasting model, we also present forecasting error models by iterative method and use the predicted errors to update the forecasted prices so as to gradually improve the forecasting accuracy. An integrated ARIMA based forecasting model for daily average price is established for validating the effectiveness of the proposed methodology by the historical data of California Power Market. The results show the presented approach improves the accuracy of forecasting significantly and with the features of easy modeling and suitable for extending to forecasting market clearing price and electricity load, even other forecasting domains.
出处
《中国电机工程学报》
EI
CSCD
北大核心
2004年第12期63-68,共6页
Proceedings of the CSEE
基金
高等学校骨干教师资助计划(GG-470-10079-1001)香港特别行政区政府资助基金(RGC)。