Research on the Dynamic Fitting and Prediction Model of Underground CBM Extraction Capacity Based on Improved Wavelet Neural Network Method

Abstract

According to the difficult question that how to calculate the productivity of coal mine underground extraction-CBM, introduced the method of wavelet neural network, for different pre-draining time, the productivity of CBM was calculated with dynamic fitting. Because wavelet neural network exists convergence speed slow and easy to fall into local minimum, therefore a parameter correction of the improved algorithm was put forward. The first neuron function of output layer used hyperbolic tangent function instead of traditional Sigmod function, Secondly used the way that added the momentum term in adjustment type of weights to select the learning step, in order to improve the efficiency of network learning. Using the proposed improvement method, the productivity prediction model of coal mine underground extraction-CBM was established. The results showed that: the improved wavelet neural network model can be good for accurate prediction of the productivity model of coal mine underground extraction-CBM, the prediction accuracy and generalization ability was better than the BP neural network and the wavelet neural network model, it had important guiding significance to CBM extraction engineering of deployment and mine gas management.