Short-term Load Forecasting Based on Blending Multi-Model Fusion
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.
NETL
NSTL
万方数据
