Research On Multiple Parameter Fusion Prediction And Real Time Warning Of Coal Mine Gas

With coal mining going deeper,coal mine gas disasters occur frequently,which seriously threatens the sustainable development of coal industry and the safety of coal miners.Gas concentration is an important factor causing gas disasters.When gas concentration continues to rise to the alarm limit,open fire will cause gas explosion or gas outburst,resulting in casualties and economic losses.Therefore,in view of the need of coal mine safety production,accurate prediction of gas concentration in coal mining face and real-time warning of mine safety state are very important research topics..The main influencing factors of mine gas concentration include CO concentration,wind speed and temperature.Because of the complex non-linear relationship among the influencing factors,if the single sensor is used to predict the gas concentration,the prediction accuracy is low,and it can not effectively reflect the real mine environment.Therefore,multi-sensor fusion technology is used to predict gas concentration,improve the accuracy of gas concentration prediction,and build a multi-parameter gas concentration prediction model.Firstly,the model collects the original data of the influencing factors through various sensors in the mine.Then,the improved wavelet threshold denoising method is used to denoise the collected time series of gas concentration.The dimension of influencing factors of gas concentration is reduced by equidistant mapping algorithm,and the low-dimensional features are extracted.Then the extracted low-dimensional features are fused by LS-SVM.A parallel dual adaptive AIS-PSO optimization algorithm is proposed by combining the adaptive PSO algorithm with the adaptive AIS algorithm with cloning and mutation operators.The Gauss kernel parameter ? and regularization parameter ? of LS-SVM are optimized.Finally,the gas concentration in the upper corner is used as the output of the prediction model to predict the gas concentration,and the simulation experiments are carried out with PSO-LSSVM and LS-SVM methods.The results show that the proposed gas concentration prediction model has higher accuracy than the other two methods..Because the prediction model of gas concentration only evaluates the future trend of gas concentration,and does not judge the safety condition of mine.In order to evaluate the mine safety status in real time,a mine gas early warning model based on CS algorithm optimization SVM was established.According to the early warning range of gas concentration,CO concentration and wind speed,the coal mine safety situation is divided into four warning levels: safety,safety,alarm and danger.The measured values of each sensor in mine are used as input of SVM classifier and four warning levels are used as output.Compared with PSO-SVM,SVM and BP methods,the results show that CS-SVM has higher classification accuracy than the other three methods.