Abstract: To improve the efficiency and accuracy of gas emission prediction in coal mine, in view of the multiple correlation and complexity of the influencing factors of gas emission, the principal component analysis method is proposed to reduce the dimension of the influencing factors. Aiming at the problem that BP neural network has slow convergence speed and is easy to fall into local optimal solution, annealing particle swarm optimization algorithm is introduced to optimize the weights and thresholds of BP neural network. The coupling algorithm of PCA-SAPSO-BP neural network is compiled and constructed by using Matlab software to predict the gas emission. The data of gas emission and its influencing factors in Qianjiaying Coal Mine in Kailuan Group are selected as samples, and the samples are predicted by BP neural network model, PSO-BP model and SAPSO-BP model. The results show that the average relative error of PCA-SAPSO-BP neural network model is 1.06%, PCA-PSO-BP model is 2.20%, PCA-BP model is 3.00%, SAPSO-BP model is 1.61%, PSO-BP model is 2.81%, BP model is 3.98%; the normalized mean square error of the prediction model is 0.002 5, the Hill inequality coefficient is 0.005 5, the average absolute error is 0.07 m3·min-1, and the judgment coefficient is 0.997 5. It is proved that the PCA-SAPSO-BP neural network model improves the prediction accuracy of gas emission of BP model.