Carbon Emission Prediction on the Power Generation Side Based on the CNN-LSTM Model

As computer technology advances at a breakneck pace, its applications are becoming more and more prevalent in diverse fields. Particularly in addressing global environmental issues, computer modeling and data analysis techniques provide robust support for accurate predictions and decision-making. To address a series of issues related to global warming caused by greenhouse gases, China has proposed the “dual carbon” goals to achieve carbon emission reductions across various industries. As the single industry with the highest proportion of carbon emissions in China, the carbon reduction process of the power industry will directly affect the further realization of the “dual carbon” goals. This paper analyzes and selects the main factors affecting carbon emissions based on the carbon emission data of the power industry in a certain area from 1995 to 2022. Subsequently, it constructs a Convolutional Neural NetworksLong Short-Term Memory (CNN-LSTM) model and a STIRPAT model. By comparing the fitting degree R2 and the Mean Absolute Percentage Error (MAPE), the model with higher precision is selected for scenario prediction, resulting in forecasts of future carbon emissions from the power generation side under different scenarios.