Research And Application On Early Warning Technology Of Coal Spontaneous Combustion In Goaf Based On GA-SVM

China is a major consumer and producer of coal,but most coal mines have spontaneous combustion problems especially the coal spontaneous combustion disasters in the goaf,which seriously restricts the safety of coal mines.So the prevention and management of spontaneous combustion disasters are great significance.In this paper,taking the 24321 working face of a mine as the research object,through the programmed heating experiments of coal samples,the change law of each gas with coal temperature was summarized and analyzed.The critical temperature of coal samples was determined in a range 50?60?,and the dry cracking temperature was 100?110? at 24321 working face.Based on the programmed temperature rise experiment and the mathematical model,it was found that the shortest spontaneous combustion period of coal in this working face was 43 days.The seepage field of the goaf was simulated by using FLUENT software combined with the actual parameters of 24321 working face,the spontaneous combustion dangerous area of the goaf of the 24321 working face was divided by using the method of on-site testing.Combining the two results,the oxidation zone on the intake side was 55?106m and the width was 51m;on the return side was 41?84m,and the width is 43m,which provided a basis for the establishment of a coal spontaneous combustion warning model in the goaf.In view of the characteristics that coal spontaneous combustion early warning gas indicators were various and they influence mutually,the main component analysis method was used to select the coal spontaneous combustion dynamic early warning indicators in the goaf.The newly generated main components retained the main information of the original indicator data,greatly simplifying the calculation process and improved the analysis efficiency;in case of the multiple and non-linear characteristics of coal spontaneous combustion early warning indicators,a genetic algorithm(GA)optimized supporting vector machine(SVM)for the coal spontaneous combustion dynamic early warning model of the goaf was constructed.And the model with field data was verified,the accuracy rate of early warning results reached 92%.Based on the connection among the advancement speed of the working face,the shortest coal spontaneous combustion time and the width of the oxidation heating zone,an early warning model of coal spontaneous combustion trend in the goaf was built to predict the special time point and coal spontaneous combustion status during the mining of the working face.Based on the coal spontaneous combustion dynamics and trend early warning model of the goaf,the coal spontaneous combustion early warning software of the goaf were researched and developed.Combined with the actual conditions of the working face,the fire monitoring layout plan was determined,and the software was applied to the 24321 working face,which realized the aim of collecting and monitoring the characteristics of coal spontaneous combustion information,face management and warning and other functions,and the software had good data processing capabilities to meet the real-time warning requirements for coal spontaneous combustion in goaf area.