Author(s)

Shuoqin Lin, Zhengsheng Zhan, Shukai Cheng, Canjian Huang, Dangyue Lai

Corresponding Author

Shuoqin Lin

Abstract

Against the background of global energy transition and “double carbon” target, power system load forecasting is crucial in ensuring energy security, optimizing resource allocation and promoting sustainable economic and social development. In order to cope with the complex changes of power system load, improve the prediction accuracy, and provide accurate support for grid operation and scheduling, the establishment of scientific and effective load forecasting models has become a key research topic. In this paper, based on the NAR model and LSTM model improved by ARIMA and CEEMD algorithms, the short-term and long-term power load forecasting methods are proposed respectively, and the PSO-XGBoost algorithm is used to construct the electric load fitting model to study the abnormal sudden changes in the power load data. Using the constructed model, the maximum and minimum values of daily loads of each industry in the region for the next three months are predicted and the prediction accuracy is analyzed.

Keywords

Electric Load; ARIMA-NAR; LSTM; PSO-XGBoost; Double Carbon