The Forecast, Based On The Chaos Analysis Of Gas Emission

The gas is one of the most important factors that endanger the security of the mine production. It seriously threaten the lives of miners.At present, the mine has only realized the real-time monitoring of gas,but not the forcasting of gas.There are some limitation of the traditional forcasting method,such as modeling subjectivism and statistical forcasting,To aviod these,chaos theory is used in gas forcasting in this paper.The chaotic characteristies of gas time series is verified by calculating the power spectral,the largest Lyapunov index and correlation dimension,1500 gas data come from HeGang NanShan Mine.Noise is removed by the wavelets analysis, self similar structure of chaotic attractor is recovered in the gas time series. The phase space is reconstructed by the embedding dimension and the delay time,the embedding dimension of 8 is determined by correlation dimension method and Cao method,the delay time of 6 is selected by mutual information.In the phase space of reconstruction,the gas time series is predicted by the four models,which are the largest Lyapunov index and the weighted one-rank based on local model,the BP single step iteration method and BP multistep method based on global model, error functions are used to evaluate the forecast performance of four models,the results show that the methods based on global model is superior to the methods based on local model, especially the BP multistep method in step,accuracy and stability,it provide a more effective method for the dymanic prediction of gas time series.