At present, Industrial energy in China comes mainly from coal, with the power industry, which is the main consumer of coal, accounting for 40% of the country's total CO2 emissions each year and this share may increase further in the future. At present, the calculation of carbon emissions in the power industry mainly relies on the 2006IPCC National Greenhouse Gas Guidelines, which take into account the influence of relevant element in the production process of power plants on CO 2 emissions, while ignoring the influence of other factors. In this paper, considering the influence of power generation, power supply, coal consumption, carbon content of base element received, carbon content per unit calorific value, and carbon emission intensity of power supply on carbon emissions, the RBF neural network prediction model is established with them as input variables and carbon emissions of coal-fired power plants as output variables. The analysis of the RBF network forecast results leads to the conclusion that the optimal mean square error of the RBF network at the 15th iteration is 0.0070 and that it has a high forecast accuracy suitable for predicting CO2 emissions from coal-fired power plants.