Shortterm load forecasting based on the Kfold crossvalidation and stacking ensemble

Authors

Wenguang ZHU, Economic and Technological Research Institute, State Grid Jiangxi Electric Power Company, Nanchang 330000 , China
Yingxue LI, Economic and Technological Research Institute, State Grid Jiangxi Electric Power Company, Nanchang 330000 , China
Weiqun YANG, Economic and Technological Research Institute, State Grid Jiangxi Electric Power Company, Nanchang 330000 , China
Xiaochun LIU, Economic and Technological Research Institute, State Grid Jiangxi Electric Power Company, Nanchang 330000 , China
Ning XIONG, Economic and Technological Research Institute, State Grid Jiangxi Electric Power Company, Nanchang 330000 , China
Cheng ZHOU, Economic and Technological Research Institute, State Grid Jiangxi Electric Power Company, Nanchang 330000 , China
Li WANG, Economic and Technological Research Institute, State Grid Jiangxi Electric Power Company, Nanchang 330000 , China

Keywords

shortterm load forecasting, pearson correlation coefficient, Kfold cross validation, stacking ensemble

Abstract

Shortterm load forecasting is of great significance for the economic dispatching and operation of power systems. In order to improve the accuracy of shortterm load forecasting, a shortterm load forecasting method based on kfold cross validation and Stacking ensemble is proposed. Firstly, the Pearson coefficient method is utilized to screen multiple features affecting shortterm load, and redundant features are eliminated. Secondly, the kfold validation crossvalidation method is applied to train the submodels of the first level, and the prediction results of each submodel are taken as new features to train the second level model. Thirdly, the results of the submodel are stacked, and the shortterm load forecasting results obtained by the second layer model. Finally, the validity of the proposed method is verified by the actual data set from New England. The simulation results show that the proposed kfold crossvalidation method can effectively improve the generalization ability of the model. Stacking fusion can not only improve the average accuracy of prediction, but also reduce the maximum prediction error, which is more advantageous than single model prediction.

DOI

10.19781/j.issn.16739140.2021.01.010

First Page

87

Last Page

95

Recommended Citation

ZHU, Wenguang; LI, Yingxue; YANG, Weiqun; LIU, Xiaochun; XIONG, Ning; ZHOU, Cheng; and WANG, Li (2021) "Shortterm load forecasting based on the Kfold crossvalidation and stacking ensemble," Journal of Electric Power Science and Technology: Vol. 36: Iss. 1, Article 10.
DOI: 10.19781/j.issn.16739140.2021.01.010
Available at: https://jepst.researchcommons.org/journal/vol36/iss1/10