基于HP滤波的SARIMA中期电力负荷预测

Medium-term Power Load Forecasting Based on SARIMA Model with HP Filter

在用时间序列模型做电力负荷预测时,季节性是中期负荷预测重点分析的因素项。对全国2004年1月到2013年12月的电能消耗数据建立SARIMA模型,在模型定阶和参数显著性检验中发现,季节性因素的参数在调整过程中多数情况下不显著,这与季节性本身相矛盾,不利于SARIMA模型建立后的预测过程。鉴于这种情况,对原序列进行修整,将HP滤波法应用到ARIMA模型建立之前,提取不同频率的波谱序列。并运用OLS法分别建立模型进行分析,弱化趋势性、季节性等因素之间的相互作用,最后根据HP滤波原理对2014年1月到11月的电能消费量做出综合预测。从预测结果看,这种方法可以降低由序列趋势性、季节性等因素相互影响产生的相对误差,提高预测精度。

The seasonal component has been a key factor in time series models for medium-term power load forecasting. A Seasonal-ARIMA（ SARIMA） model is developed based on the electricity consumption data from January 2004 to December 2013. During the model order selection and the parameter significance testing,the parameters of seasonal components turn out to be quite non-significant in most cases. And that is not conducive to forecast the data. To address this issue,the hybrid time series model based on the HP filter is utilized to extract the spectrum sequences with different frequencies before the model is established. Then the sequences are separated into two parts,each for setting up a model to analyze interactions among various factors including trend component and seasonal component. Finally,based on the HP filter principle,an integrative forecast can be made for the electricity consumption from January to November in 2014. Experimental results demonstrate the new method reduces the relative error caused by the interaction between the trend component and the seasonal component and make more accurate forecast.