基于小波分析和GM-ARIMA模型的月度售电量预测  

Monthly Electricity Sales Forecast Based on Wavelet Analysis and GM-ARIMA Model

作　　者：樊娇[1] 冯昊[2] 牛东晓[1] 王筱雨[2] 刘福炎[2] 

机构地区：[1]华北电力大学经济与管理学院,北京102206 [2]国网浙江省电力公司经济技术研究院,浙江杭州310008

出　　处：《华北电力大学学报（自然科学版）》2015年第4期101-105,共5页Journal of North China Electric Power University：Natural Science Edition

基　　金：国家自然科学基金项目资助项目(71471059)

摘　　要：月度售电量直接反映电力企业的经营效益,准确的电量预测对于电力企业合理安排购售电方案、确定融资缺口具有重要意义。鉴于各地区月度售电量时间序列不但有随时间逐渐增长的趋势,还受节假日、气温等因素的影响存在随机项,为了提高电力企业月度售电量的预测精度,采用小波分析理论将月度售电量时间序列分解为近似序列和细节序列,并通过对分解后子序列的特征进行分析,分别采用相匹配的GM(1,1)模型和ARIMA模型对子序列进行预测,然后通过序列重构得到月度售电量的预测值。经实际算例验证,该组合预测方法的平均误差率为3.7%,与神经网络等常用单一预测方法相比能明显提高预测精度,具有较强的适应能力。Monthly electricity sales can reflect the business profit of power plants directly. Thus the accurate forecast of electricity sales can help those plants make reasonable arrangement of the electricity purchasing scheme and deter- mine the financing gap. The time series of monthly electricity sales not only tends to gradually increase with time, but also comes under the influence of the factors like temperature and holidays. In order to improve the prediction accuracy of the monthly sale of power plants, this paper employed the theory of wavelet analysis to decompose the time series of monthly electricity sales into approximation sequence and detail sequences. After analyzing the characteristics of de- composition sequences respectively, matching methods GM （1, 1 ） and ARIMA model, were employed to predict the decomposed subsequences. Then through the reconstruction of subsequences, the final prediction of electricity monthly sales with the average error rate of 3. 7% were obtained. Finally, it was verified by examples that the combi- nation forecasting method could obviously improve the prediction accuracy, with a strong ability to adapt.

关 键 词：小波分析 分解序列 灰色模型 ARIMA 

分 类 号：TM614[电气工程—电力系统及自动化]