Research Of Surface And Domain Combined Coal And Gas Outburst Prediction Based On PCA-AKH-BP Neural Network Model And Application

Coal and gas outburst accidents are one of the major accidents in coal mines.In the statistical analysis of many years' data of the world,China is the the most serious country in coal and gas outburst,accounting for about one-third of the world.Coal and gas outburst not only seriously affect the safety of coal production in China,but also is a serious threat to the safety of miners,which caused enormous economic losses.Therefore,it is significant that coal and gas outburst's accurate forecast is.As the current understanding of coal and gas outburst mechanism is not perfect,people can not fully use a single physical index and use the analytical methods to accurately predict coal and gas outburst.Therefore,in order to improve my country's coal and gas outburst accident-prone situation and the accuracy of coal and gas outburst prediction,based on the basic mechanism and the basic factors that influence the outburst,this paper combines the regional(domain)forecast method and the face(surface)forecast method's index system stipulated in the “Provisions on Prevention and Control of Coal and Gas Outbursts”,and put forward a new prediction index system and a new prediction method of coal and gas outburst based on PCA-AKH-BP neural network which is combined with surface and domain.The main conclusions are as follows:1.This paper summarizes the current coal and gas outburst mechanism and the research status of coal and gas outburst prediction methods,and analyzes the shortcomings of some existing hypothesis and forecasting methods.What's more,the necessity and practical significance of the prediction method of coal and gas outburst combined surface and domain prediction are expounded.2.A new prediction index system,combined surface and domain,of coal and gas outburst is put forward on the basis of analysis of traditional forecasting indexes and the influencing factors of coal and gas outburst.It is proposed to use the change rate instead of a single numerical index to eliminate the effect of the dimension on the accuracy of the prediction result.It is better than the index value to reflect the impact on the outburst.On the other hand,according to the practical conditions,adding some index to predict the outburst risk,which is measured easily and isn't used in old prediction method.What's more,the principal component analysis method is used to pretreat the index system,which eliminates the correlation among each index.3.It is pointed out that BP neural network is easy to fall into the local minimum,with slow convergence and instability in coal and gas outburst prediction.In this paper,some improvement programs are summarized,and the advantages and disadvantages of each method are analyzed.The BP neural network is optimized by the modified krill algorithm and the prediction model based on PCA-AKH-BP neural network of coal and gas outburst of the mining face area is established.4.According to measured value of outburst indicators in the one mining area of Xipo coal mine(4+5)coal seam,based on the PCA-AKH-BP neural network,the outburst risk of the mining face is predicted,and the prediction result is consistent with the actual one.Compared with the traditional BP neural network,it is concluded that PCA-AKH-BP is faster and more convenient than traditional BP neural network,and the prediction result is more accurate.