考虑误差修正的MC-BP短期电力负荷预测方法

MC-BP Short-term Power Load Forecasting Method Considering Error Correction

准确的短期负荷预测是电网日常调度的重要依据。针对目前短期电力负荷预测精度问题,提出了一种考虑误差滚动修正的MC-BP短期电力负荷预测方法。首先,建立了基于逐步试错法的BP负荷预测模型,分析了预测误差的概率密度分布,构建了基于蒙特卡洛(Monte Carlo,MC)的日负荷误差滚动修正策略;其次,选用了某地区2015-2019年的负荷数据进行预测,比较了CNN-BiLSTM、LSTM和BP模型的预测结果,3种预测模型的测试集NRMSE分别为5.97%、6.49%和5.5%;最后,对比了BP和LSTM预测方法修正误差、线性回归方法修正误差和误差滚动修正方法的误差修正策略的修正效果,对后一天的误差修正NRMSE的相对变化率分别为-26.68%、-28.81%、-43.90%、-88.64%。预测结果表明:所提出的考虑误差滚动修正的MC-BP短期电力负荷预测方法具有良好的预测效果。

Accurate short-term load forecasting is an important basis for daily grid scheduling.In response to the current problem of low precision in short-term power load forecasting,this paper proposes a correction of error rolling with MC-BP method for short-term load forecasting.Firstly,a BP load forecasting model based on the stepwise trial and error method is established,and the probability density distribution of the forecasting error is analyzed.Then,a Monte Carlo-based daily load error rolling correction strategy is constructed.Load data from a certain region during 2015-2019 was used to compare the prediction results of the CNN-BiLSTM,LSTM,and BP models.The NRMSE of the test sets for the three models were 5.97%,6.49%,and 5.5%,respectively.The error correction strategies of BP and LSTM prediction methods,linear regression method and error rolling correction method were compared.The relative change rate of NRMSE for error correction on the following day was-26.68%,-28.81%,-43.90%and-88.64%respectively.The results show that the proposed MC-BP method considering error rolling correction has good prediction performance for short-term power load forecasting.