一种负荷预测模型预测能力的评价标准

An evaluation criterion for the forecast accuracy of load forecasting models

常用的负荷预测模型预测能力的评价标准(如MSE和MAE)在使用过程中存在一定的局限性。引入Diebold-Mariano(DM)检验,给出了判断不同预测模型预测能力是否存在显著差别的定量分析方案。算例分析表明,基于DM检验的现代评价标准可以缓解样本中随机干扰造成的影响,有效地甄别模型的预测能力,为更全面地比较负荷预测模型提供了参考。

Common load forecast models evaluation criteria,such as MSE and MAE,have limitation in application to some degrees.In this paper,the Diebold-Mariano（DM） test is introduced.Furthermore,the scheme of quantitative analysis,which can discriminate the significant difference of forecasting accuracy between different models,is proposed.Case study indicates that this morden evaluation criterion based on DM test can relieve the influence of random sample disturbance,and can also evaluate the performance of load forecasting models.Moreover,the proposed method can provide reference to the comprehensive evaluation for load forecasting models.