Prediction Based On Bp Neural Network Of Gas Emission

The gas is one of the most important factors which harm the mine-shaft safety production. It has the great significance that carrying on the prediction of the amount of mine gas gushing to the region which has not mined, it would be effectively prevent the gas out of gauge and the gas explosion, and it could be ensure the safety of the mine production.The relation between the amount of mine gas gushing and its impact factors is dynamic, dark and nonlinear. Its change behaves as a complex nonlinear dynamic. In the researching we find that the traditional linear prediction method is very difficult to establish a prediction model of the amount of mine gas gushing. Because Neural Network have the characteristic of self- organization, self-adaptation, parallel process and have the ability of Input/Output nonlinear mapping, it have the adaptability and superiority that other traditional prediction method can not compare with in prediction of the amount of mine gas gushing, establish the prediction model by approximation the function between the amount of mine gas gushing and its influencing factors using BP neural network.To the model of the amount of mine gas gushing, improve the traditional BP neural network. Using the Momentum method and the Strategy of Study rate self-adaptation to overcome the weakness of the slow of the BP network training velocity; Using the Heredity Algorithm to optimize the BP neural network model's weight and valve to overcome the weakness of running into the local minimum value easily.To the different mining area, the main control factor of the coal bed gas gushing out quantity is different. Finding the research area's main control factor of the gas gushing out quantity is propitious to understand the gas storage rule in the research area clearly and improve operation efficiency. There are two methods that are used to confirm the main control factor of gas gushing out quantity: Gray Correlation Analysis and Weight Technical Progress Factor Analysis. Making the main control factor of the main mine gas gushing out quantity as the main input node of the BP neural network, and forecast the coal bed gas gushing out quantity in the unknown region.We implement the system with the C++ language, but not the method that use the Matlab toolbox in the implementation process. This system can configure factors easily, they are the input node number, the output node number, the hidden layer number, the training precision, the maximal training number, the training length of stride, the dynamic factor and so on. We can carry on the single sample prediction and the continuous sample real-time prediction. We can observe the training result of the system through the single sample prediction. The real-time prediction need to be connect to database, we can carry on the real-time prediction through set the parameters, they are database server address, database password, database name and so on. By the contrast, we find the improving BP neural network arithmetic can do the high accuracy prediction to the amount of mine gas gushing.