Research On Hybrid Grey-Neural Network For Roof Pressure Forecasting In Coal Mine

In recent years, coal mine accidents occur frequently in most areas of China.According to statistics, many mines have appeared phenomenon of abnormal roofpressure, resulting in many serious accidents, and leading to great economic losses tothe coal mining enterprises. Therefore, it is implicitly important for the realization ofefficient production of coal mine safety to further systematically study the mechanismof abnormal coal mine roof pressures display to find out the conditions of itsoccurrence and forecasting methods to study the corresponding control measures.According to the mining area to conduct its roof pressure prediction plays a major rolein effectively preventing the occurrence of roof accidents, hence ensuring mine safety.Roof pressure in coal mine is affected by various natural and man-made factors,with dynamic and fuzzy non-linear relationship exists among them. At present most ofcoal mine roof pressure prediction methods also depend on the expert experience andengineering analogy that brought more or less limitations and blindness in guidance ofproduction. Besides, the coal mine safety monitoring system in China focus more onroof pressure monitoring, lacking of systematic prediction measures, this alsocontributes to the difficulty of coal mine safety guarantee.At present, with the development of computer technology and data processing,the grey theory and neural network forecasting model has been successfully applied indata mining, pattern recognition and machine learning fields. Neural network systemhas the characteristics such as parallel processing, self organizing adaptive, intelligentalgorithm and the strong nonlinear mapping ability, which brought it unparalleledadvantage in the mine roof pressure prediction compared with traditionalmathematical methods can not match the applicability and superiority. However, theneural network model is also easy to fall into local minimum and the shortcomings ofslow convergence speed. In order to overcome the shortcoming of neural network,this paper presents an improved grey-neural network model, combining the greytheory and neural network algorithm organically. The improved algorithm not onlyovercomes the drawbacks of neural network model that is easy to fall into localminimum with slower convergence, but also overcomes the of deficiency of greynetwork model which lacks of self-feedback regulation Finally, this paper selects Xuzhou8133coal mine roof pressure test data as theexperimental data to implement simulation. According to simulation results, obviously,the improved grey-neural network model increases prediction precision, shortens thetraining time, and enhances the network efficiency. Good performance of theproposed prediction model demonstrates important significance for coal mine safetyin future.