Prediction Technolgy Of Coal Seam Gas Emission Quantity Based On GA-SVM Neural Network

The thesis is aimed at introducing a prediction model of gas emission quantity of the working face based on GA-SVM neural network. GA can obtain the global optimal solution and avoid falling into the local minimum value. Therefore, using GA to find the optimal punishment parameter c and kernel function parameter g of SVM. The prediction precision of SVM is better than BP and other neural network, and it also has the fast training speed and the characteristics of good performance of generalization. To sum up, we establish the prediction model for gas emission quantity of the working face based on GA-SVM. This new model has received advantages of both SVM and GA. The main influencing factor for gas emission quantity of the working face include:the depth of coal layer, the thickness of coal layer, the dip angle of coal layer, the primitive gas content of mining, the spacing of mining layer, the height of coal mining, the gas content of near coal layer, the thickness of near coal layer, rock properties between the coal seam, the length of working face, the advancing speed, the rate of production, the daily output of gas emission quantity. The influencing factor of them for gas emission quantity is complex and nonlinear. Therefore, we use them as the impact parameters for prediction. The gas emission quantity as the target parameter. We use the impact parameters and target parameter as the input value and output value of the GA-SVM to train, after training, we use the reciprocal of error absolute value of prediction output and expected output as the fitness function value of the GA to optimize parameters. The research results show that:As the prediction model of SVM combined with GA, the maximum relative error is 5.91%, the minimum relative error is 0.92%, and the average relative error is 2.2%. Compared with uncoupled and other prediction models, we have established the new model has the better generalization ability and higher prediction precision, and the practical application at the group level of coal mine, Tiefa coal industry group, the maximum relative error is 6.80%, the prediction of the minimum relative error is 0.47%, the average relative error was 2.89%.The result show that the model has practical application value.