基于Blending多模型融合的短期负荷预测

Short-term Load Forecasting Based on Blending Multi-Model Fusion

针对电力负荷预测精度不高且泛化能力较弱等问题,提出一种基于多个模型融合Blending集成学习方式的短期负荷预测方法。在对数据预处理后,首先,分别设计实验对各单一模型LSTM、LightGBM、XGBoost、GBDT、KNN、SVM进行单独预测,同时用Pearson相关系数分析各模型误差相关性,优选预测性能好、相关性小的模型作为基学习器和元学习器,构建多个模型嵌入Blending集成学习方式的短期负荷预测模型,最后,通过中国南方某地区的真实负荷数据进行验证,算例表明,Blending预测模型能够充分发挥不同学习器的优势,提高泛化能力,且所提模型RMSE、MAPE值分别为102.97MW和0.67%,相较于单一预测模型在预测精度上有较大的提升。

To address the problems of low accuracy and weak generalization ability of electric load forecasting,a short-term load forecasting method based on multiple models fusion Blending ensemble learning approach is proposed.After preprocessing the data,firstly,experiments are designed to predict each single model LSTM,LightG⁃BM,XGBoost,GBDT,KNN,and SVM individually,while the error correlation of each model is analyzed by Pearson correlation coefficient,and the model with good prediction performance and small correlation is preferred as the base learner and meta-learner to construct multiple models embedded.Finally,the real load data of a region in southern China are used for validation,and the algorithm shows that the Blending prediction model can give full play to the ad⁃vantages of different learners and improve the generalization ability,and the RMSE and MAPE values of the proposed model are 102.97MW and 0.67%,respectively,compared with the single prediction model.The prediction accuracy is greatly improved compared with that of a single prediction model.