Accidents Analysis And Prediction Of Coal And Gas Outburst

Coal and gas outburst is one of the most serious and destructive accidents in coal mining and production.Coal and gas outburst is a complex,nonlinear and high dimensional problem which is affected by many factors.In reality,due to technical or human reasons,prominent accident data are prone to be missing.In the past studies,direct deletion was mostly adopted,but it is easy to cause the sample data to be in an unbalanced state.Therefore,improving the prediction accuracy of prominent samples under unbalanced data is a problem to be solved.Firstly,this paper carries out a multidimensional statistical analysis of the accident data in recent ten years,extracts the key words of the accident based on the data-driven idea,and summed up the coal safety problem new situation and new characteristics.Secondly based on the analysis of coal and gas outburst mechanism and influencing factors,the construction of coal and gas outburst index system was completed.Finally,in order to further accurately predict the outburst and reduce the incidence of coal and gas outburst accidents,the basic whale algorithm was improved,and the hybrid nuclear ultimate learning machine was constructed.The improved whale algorithm was used to optimize its parameters,and the model was applied to the prediction of coal and gas outburst risk.Through processing the sample data of Huainan coal mine,the data set before and after filling is constructed,and the selected influencing factors are input into IWRWOA-HKELM for 100 tests,and compared with other models such as HKELM,WOA-HKELM,etc.The results show that the improved whale algorithm proposed in this paper optimizes the parameter selection of hybrid kernel extreme learning machine,effectively improves the prediction accuracy of kernel extreme learning machine.At the same time through the comparative analysis of the prediction results before and after filling data,we can see that filling in data by miss Forest algorithm improves the accuracy of highlighting accidents.This shows that the method used in this paper is feasible and can provide a new method with theoretical and engineering significance for coal and gas outburst prediction.