A novel text-based framework for forecasting coal power overcapacity in China from the industrial correlation perspective

The accurate forecasting of coal power overcapacity in China has significant impact on the promotion of the dual carbon strategy. However, due to the complexity of the causes of coal power overcapacity, existing forecasting methods are unable to comprehensively identify influencing factors and incorporate them into forecasting models. Given this, we propose a text-based framework for forecasting coal power overcapacity. Specifically, a topic model and a sentiment analysis method are used to identifies and quantifies the influencing factors from policies and news texts of coal power and its related industries. In addition, under the dual carbon target, the development trend of coal power overcapacity under different environmental constraint scenarios was examined. The empirical test shows that the performance of the forecasting models based on the text data of the related industries is better than that based on the text data of the single coal power industry and the numerical data of the single objective variable, and the changes in capacity of the upstream and alternative industries and the industrial policy support of the downstream industries are the main causes of overcapacity. The scenario forecasting results indicate that the coal power overcapacity scale shows an upward trend from 2021 to 2060.