Prediction of gas emission time series based on W-RBF
Hongzhen Yu · Meitan xuebao · 2008
Gas forepassed data collected from coal mine monitoring system were analyzed for their character.Then,small-data-sets were used to determine gas time series was chaos time series.Furthermore,the input neuron number of RBF neural network was determined by the character of chaos time series.A novel method of gas time series prediction method based on W-RBF was proposed on those processes,this method combine the multi-differentiation character of wavelet with RBF neural network.The simulation results indicate that the prediction results of W-RBF is accurate.The maximal absolute error of W-RBF is 0.1%,and 92 sampling gain 0 error.