| Along with the amounts of our country’s coal production and consumption increasing, the depth of the mine exploitation and production also increase. The series of safety problems caused by gas need to be solved quickly. The gas emission is an important reference index which affects every safety design process of mine. The deeper mine extend, the more complex the influencing factors of gas emission are. The original linear prediction methods can’t predict gas emission real and accurately because of their applicable conditions and the existing defects. Therefore, seeking a science and technology method with low cost and high reliability to predict gas emission has important guiding significance to improve efficiency, security of mine.To this end, the paper puts forward a difference GM-RBF neural network combination model to predict gas emission of coal seam, and the following several research works have been done. According to the merits and demerits of the grey system and neural network theory, the measure of combining GM (1, N) model with RBF neural network model by difference has been put forward. Then taking the No.13-1coal seam No.1472(3) working face of No.13-1coal seam at Pansan coal mine in Huainan Mining Group for example, GM(1,N) model, RBF neural network model and difference GM-RBF combined model have been constructed to predict gas emission. Using the MATLAB software, the three kinds of model have been analogue simulated and compared. The results shows that relative to the GM (1,7) model and RBF neural network model, the MAE value of difference GM-RBF combination model has been reduced by42.82%and42.82%respectively. MAPE has been reduced by43.48%and16.89%respectively. RMSE has been reduced by40.4%and19.78%and RRMSE has been reduced by39.41%and39.41%respectively. So the difference GM-RBF combination model has higher prediction accuracy and better stability. It can achieve real-time, dynamic and accurate prediction of gas emission in working face. |
PDF Full Text Request