Time Series Analysis about Gas Emission of Working Face Based on the Coupled Model of Chaos and Neural Network
The actual gas emission system contains many nonlinear factors, but at the present time, most of analysis and forecast methods of gas emission mostly suppose that the relationship of each variable is linear. This kind of limitation makes it hard to accurately analyze and forecast the emission quantity of working faces in practical application. Based on this, the coupling model of chaos and neural network was put forward. It was found that the relative error of forecast data was less than 3％ after the real data of gas density was forecasted by the method. It is showed that using the reconstruction phase space theory to the time series forecast modeling of gas emission in working face can well reflects the intrinsic movement mechanism of this series, reveal the complicated movement regulation and nonlinear characteristic of dynamic system of gas emission and taking the saturated embedded dimension of reconstruction phase space as the number of input layer of network can overcome the randomicity of the selection of node of network input layer. This result provides a kind of new basis for the predictability of the gas gush of the excavation and the choice of the predictable model, and it has important meaning for the prevention and control for gas explosion accidents of coal mine and effective utilize for gas.
chaos neural network working faces gas emission time series phase space reconstruction
HE Liwen SONG Yi SHI Shiliang LIU Zhengcai
Energy Engineering College, Xiangtan University, Xiangtan 411100, Hunan, China Human Provincial Key Human Provincial Key Laboratory of Safe Mining Techniques of Coal Mines, Hunan University of Science School of Energy and Safety Engineering, Hunan University of Science and Technology, Xiangtan 411201 Energy Engineering College, Xiangtan University, Xiangtan 411100, Hunan, China
国际会议
The 2010 International Symposium on Safety Science and Technology(2010 安全科学与技术国际会议)
杭州
英文
1599-1604
2010-10-26（万方平台首次上网日期，不代表论文的发表时间）