Different-source prediction of gas emission based on SAPSO-ELM and its application

In order to improve the accuracy of gas emission prediction, in view of the multiple correlations and complexity of the influencing factors of gas emission, principal component analysis and separated source prediction theory were combined, the gas emission data of the mining layer, adjacent layer, and goaf were respectively subjected to principal component analysis to reduce dimensionality, and the predictor was obtained. Aiming at the problem that the input weight matrix and hidden layer threshold of the extreme learning machine were generated randomly, the simulated annealing particle swarm optimization algorithm was used to optimize the parameters of the extreme learning machine, and the gas in a coal mining face in Xinjiang was optimized. The output and influencing factors were used as the input of the SAPSO-ELM model for training, and then the trained SAPSO-ELM model was used to verify and predict the gas emission of a coal mining face in Shaanxi, and the prediction results of the original ELM model was compared. The results show that the average relative error of the SAPSO-ELM model is 3.45%, and the average relative error of the ELM model is 8.81%. Compared with the ELM model, the prediction accuracy and efficiency are better than the original ELM model. The combination of source prediction theory and principal component analysis effectively solves the multiple correlations among multiple factors and reduces the complexity of the prediction model. Meanwhile, the SAPSO-ELM prediction model realizes the rapid and accurate prediction of gas emission, which plays a guiding role in preventing gas accidents and ensuring safe and efficient mining of coal mines.