Abstract:

In order to improve the accuracy of the risk prediction of coal spontaneous combustion in the goaf, a PSO algorithm was used to improve the connection weight and threshold values of BPNN. A coal spontaneous combustion risk prediction model (PSO-BPNN model) coupling the PSO algorithm with BPNN was constructed. The PSO-BPNN model was compared and analyzed with the prediction results of the standard BPNN model and SVR model respectively. The results show that the optimized mean relative error, mean absolute error and root mean square error are reduced by 9.35%, 0.170 7 and 0.205 6, respectively, and the coefficient of determination is increased by 0.116 9 compared with the BPNN model. Compared with the SVR model, the optimized mean relative error, mean absolute error and root mean square error are reduced by 5.41%, 0.115 2 and 0.171 5, respectively, and the coefficient of determination is increased by 0.0891. It was demonstrated that the PSO-BPNN model has higher prediction accuracy than the standard BPNN model and SVR model.

Key words: coal spontaneous combustion, particle swarm optimization(PSO), back propagation neural network(BPNN), support vector regression(SVR), prediction model, goaf