基于组合模型的江苏省居民用电量预测分析

Forecast and Analysis of Residential Electricity Consumption in Jiangsu Province Based on Combined Model

研究利用时间序列基本分析方法ARIMA模型分析法、指数平滑ETS模型和神经网络自回归模型对江苏省居民每月用电量进行数据分析、处理、拟合、检验及预测,以2004年1月至2017年12月用电计量数据作为分析样本,使用R软件对该时间序列进行建模。对给出的数据建立ARIMA模型、ETS模型和NNAR神经网络自回归模型,接着利用MAE、RMSE、MAPE三个评价指标来衡量模型的优良度。尝试通过组合模型对2018年江苏省居民12个月的用电量进行预测,与实际值进行对比验证,发现权重模型的误差最小,选择作为最终预测模型。最后得出结论,组合模型的预测效果要优于非组合模型。

The study uses the basic analysis methods of time series ARIMA model analysis method,exponential smoothing ETS model and neural network autoregressive model to analyze,process,fit,test and predict the monthly electricity consumption of residents in Jiangsu Province.The electricity consumption data from January to December 2017 is used as the analysis sample,and the R software is used to model the time series.Establish the ARIMA model,ETS model and NNAR neural network autoregressive model for the given data,and then use the three evaluation indicators of MAE,RMSE,and MAPE to measure the goodness of the model.An attempt was made to predict the 12-month electricity consumption of residents in Jiangsu Province in 2018 through a combined model,and compared with the actual value for verification,it was found that the error of the weight model was the smallest,and it was selected as the final prediction model.Finally,it is concluded that the predictive effect of the combined model is better than that of the non-combined model.