Abstract: The estimation of direct CO₂ emissions from three major industries and domestic energy consumption in Gansu province from 2000 to 2020 by emission factor method was carried out, to describe and analyze its evolution characteristics. Then we established a BP neural network model and forecasted the CO₂ emissions of Gansu province from 2021 to 2030. Finally, a STIRPAT expansion model of influencing factors of CO₂ emission in Gansu province was constructed, the influence degree and internal mechanism of each factor on CO₂ emissions were quantitatively explored by using multiple regression analysis, and the important influencing factors were further identified by combining random forests. The results showed that:the direct CO₂ emission from industrial and domestic energy consumption in Gansu province were generally fluctuating, and the secondary industry accounted for more than 70% of the total CO₂ emissions, which was the main source of carbon dioxide emissions. The prediction error of the BP neural network model was 2×10^-4, and the correlation coefficient was greater than 0.99, which had high accuracy for predicting CO₂ emissions in Gansu province, also it was concluded that the direct CO₂ emissions of energy consumption in Gansu province would reach the maximum value in 2026. The driving factors of CO₂ emissions in Gansu province were significantly different, the intensity of CO₂ emissions, economic development and urban and rural consumption had a greater positive effect on CO₂ emissions, and per capita consumption expenditure of urban residents was the main influencing factor. The growth of forest coverage inhibited CO₂ emissions, and the reduction of energy intensity contributed to the reduction of CO₂ emissions. The adjustment and transformation of the three major industrial structures could lead to a decline in CO₂ emissions. The research provided a theoretical basis and scientific basis for Gansu province to promote low-carbon emission reduction.