基于动态组合残差修正的预测方法

本文建立了一种基于残差修正的组合预测方法,并基于该方法证明了针对多个单一的预测方法根据其在某个时间段的相对预测误差的大小选择组合选项可以进一步提高预测精度.提出了针对不同时间段可根据各种单项预测模型的相对预测误差的大小动态选取相对预测误差最小的两种模型构成组合残差来修正基本方法的预测误差,以提高预测精度.最后通过实际空调负荷预测对其进行了验证,结果表明这种动态组合残差修正的预测方法相对于基于多个固定单一预测方法的组合预测方法,可以进一步改善预测效果.

Forecasting model with dynamical combined residual error correction

In order to further improve the accuracy, a forecasting method with combined residual error correction is used in this paper. Based on the forecasting method, the residual error correction model and its combination ways are analyzed among some single forecasting methods, and the very high accuracy of the combination method with two minimum relative prediction errors is proven within a certain period. And then one forecasting method with dynamical combined residual error correction is proposed, based on choosing two minimum relative prediction errors to form the correction model during different periods. A study case indicates that the combination ways proposed in this paper can further improve the accuracy of combination forecast.