Abstract: “Carbon Peak” and “carbon neutrality” are a social revolution affecting all mankind. How to accurately predict and reduce carbon emissions is the key to the success of this revolution, and also gives the research on this topic a more difficult mission. In order to accurately predict the carbon emissions in the energy sector, based on the energy security considerations such as power balance, power amount balance and peak balancing, a hybrid calculation model is adopted to optimize the installed structure of electric power and energy consumption structure in the process of carbon emission prediction, so as to obtain the optimal minimum carbon emission value based on the installed cost of electric power and the cost of energy consumption structure according to the optimized energy consumption structure. The whole forecasting process is mainly divided into three parts. Firstly, the influencing factors of carbon emissions in the energy sector are analyzed according to the LMDI decomposition method. GDP has a positive driving effect on carbon emissions, while energy consumption intensity and industrial structure have a negative driving effect on carbon emissions. Secondly, the structure of the carbon system is preliminarily optimized and the initial carbon emission value is predicted based on the preliminary optimization results. Finally, based on the rationality and accuracy of the process, a model evaluation method is proposed. The key index verification method is adopted to perform feedback verification on the binding indicators such as the proportion of electric energy in terminal energy consumption, the proportion of non-fossil energy consumption and the cumulative decline rate of carbon intensity in the prediction process. After the key index verification and multiple feedback adjustment, the carbon emission value meeting the constraints of the multi-objective function and each constraint index is obtained. The feasibility and reliability of the optimization and prediction process are verified by using the actual data of a province in central China. The results show that only in the peak reaching scenario, the province can meet the target of carbon peak and pass the verification of three key indicators.