Long-term electricity consumption forecasting in China-based on a combined model of KPCA and linear regression

Abstract

Total electricity consumption is a barometer of a country's economy. Long-term forecasting of total electricity consumption in the whole society can effectively track a country's economic development and monitor the implementation of energy conservation and emission reduction policies. How to effectively forecast the long-term total electricity consumption is an important topic in the academic and industrial fields. The combined model of kernel principal component analysis (KPCA) and linear regression (LR) proposed in this paper can accurately predict the changes in total electricity consumption over time, even if the sample size is small. Meanwhile, the model results have strong interpretability and practical value. Further, through the correlation analysis of principal components obtained from KPCA dimensionality reduction, this paper finds that the most important features affecting the total electricity consumption are the economy feature and production efficiency feature. Finally, this paper predicts that China's total social electricity consumption will reach 1.83 trillion KWH in 2035, which is more optimistic than the prediction of Oxford experts, which is consistent with the reality that China has achieved an overall victory in the fight against COVID-19.