Prediction of coal mine gas emission based on Markov chain improving IGA-BP model

Xiaoheng Yan^{1, 2}, Hua Fu¹, Weihua Chen¹ 

¹Department of Electrical and Control Engineering, Liaoning Technical University, Huludao Liaoning 125105, China

²Department of Safety Science and Engineering, Liaoning Technical University, Fuxin Liaoning 123000, China

There are a lot of factors that affect the gas emission, and among those there is a complicated and nonlinear relationship, so a BP neural network model based on immune genetic algorithm (IGA) was constructed to solve the problem of the traditional BP neural network such as, slow training speed, easy to be trapped into local optimums, and the premature convergence. In order to further improve accuracy of the prediction, the Markov chain was used to modify the residual series for the sample of bigger error. The correction result is more close to the measured value. The results showed that both the prediction accuracy and convergence speed of the IGA-BP model are better than the BP neural network model. The prediction after modified by Markov chain was further improved, the absolute average relative error of the prediction of the IGA-BP model is 2.40%.