基于Jackknife模型平均的社会用电量预测研究

Social Electricity Consumption Forecasting Based on Jackknife Model Averaging

针对社会用电量波动的复杂性,文章将Jackknife模型平均理论引入社会用电量分析与预测研究中,通过加权平均不同模型的预测值,最大限度减少有用信息的遗失,以提高社会用电量预测准确度.通过选取中国和美国不同时期社会用电量数据集,并使用各类预测误差指标以及Diebold-Mariano检验法,来验证所提出的Jackknife模型平均方法的有效性.研究结果表明:Jackknife模型平均方法可以有效降低单个社会用电量预测模型的预测误差,为用电量预测提供了一种新的建模框架.

The accurate prediction of electricity consumption points out the fluctuation range of electricity consumption in a given time window in the future,which not only provides important information for power supply enterprises,but also an important basis for power departments to formulate relevant policies.In view of the complexity of electricity consumption fluctuations,the Jackknife model average(JMA) theory is employed for electricity consumption forecasting.This technique maximizes the utilization of various information by weighting the predicted values of different models,and finally improves the accuracy of electricity consumption prediction.Furthermore,the forecasting performance of the JMA method is evaluated and compared with seven benchmark models on the basis of accuracy measures and Diebold-Mariano test by selecting the monthly electricity consumption data sets of China and the United States in different periods.The experimental results show that the Jackknife model average method can effectively reduce the prediction error of a single electricity consumption prediction model and is an effective electricity consumption prediction model.