China’s power industry is a major energy consumer, with the carbon dioxide (CO2) generated by coal consumption making the power industry one of the key emission sectors. Therefore, it is crucial to explore energy conservation and emissions reduction strategies suitable for China’s current situation. Taking a typical cogeneration enterprise in North China as an example, this paper aims to establish a generalized regression prediction model for carbon emissions of coal-fired power plants, which will provide a reference for China to seek strategies for carbon peaking and carbon neutralization in the future. Firstly, in terms of the selection of influencing factors, this paper uses objective index screening methods, simulation means, and the eXtreme Gradient Boosting algorithm (XG-Boost) to analyze the feature importance of various influencing factors. It is concluded that the relevant influencing factors of the boiler feed water system have a strong correlation and characteristic importance with the carbon emissions results of coal-fired power plants. Therefore, this paper proposes to introduce these factors into the regression prediction model as auxiliary variables to more scientifically reflect the carbon emissions results of coal-fired power plants. Secondly, in the aspect of regression prediction model establishment, inspired by the sparrow’s foraging behavior and anti-predation behavior, this paper selects the sparrow search algorithm (SSA) with strong optimization ability and fast convergence speed to optimize the super parameters of the long short-term memory network algorithm (LSTM). It is proposed to use the SSA-LSTM algorithm to establish the carbon emissions regression prediction model of coal-fired power plants. The advantage of the SSA-LSTM algorithm is that it can effectively simplify the super parameter selection process of the LSTM algorithm, effectively solve the global optimization problem, prevent the model from falling into overfitting and local optimization, and make the carbon emissions regression prediction model of coal-fired power plants achieve a better fitting effect. By comparing the performance indicators of the model before and after the improvement, it is found that the regression prediction effect of the SSA-LSTM coal-fired power plant carbon emissions regression prediction model, which introduces boiler feed water influencing factors, has been effectively improved. Therefore, the model proposed in this paper can be used to conduct a comprehensive impact factor analysis and regression prediction analysis on the carbon emissions intensity of China’s coal-fired power plants, formulate targeted carbon emissions reduction countermeasures, and provide a theoretical basis for energy conservation and emissions reduction of China’s coal-fired power plants.