Abstract: Shanxi, as a major province of energy use and carbon emission, has an important demonstration significance for the whole country to promote the "dual carbon" strategy. The carbon emissions of Shanxi Province from 2000 to 2020 were calculated based on IPCC emission coefficient method. Tapio decoupling model was used to analyze the decoupling relationship between carbon emissions and economic development, LMDI method was used to decompose the factors affecting carbon emission changes, and PSO-BP neural network model was used to simulate and forecast the carbon emissions of Shanxi Province. The results showed that the carbon emission in Shanxi Province increased during 2000-2020, while the carbon emission intensity decreased, and the decoupling coefficient was 0.585, indicating a weak decoupling state as a whole. Economic growth was the determining factor of carbon emission growth, and the optimization and adjustment of industrial structure and energy intensity was the leading factor to restrain carbon emission. The introduction of particle swarm optimization (PSO) improved the prediction accuracy of BP neural network effectively. The predicted results showed that carbon emissions in Shanxi Province would peak in 2032, 2029 and 2027 under three scenarios: baseline scenario, low carbon scenario and intensive low carbon scenario, respectively. In view of the forecast results, relevant policy suggestions were put forward.