浅埋深综采工作面低氧致因分析及预测模型研究OA

The problem of low oxygen in the fully-mechanized mining face of coal mines is serious and threatens production safety.Its causes involve the coupling effect of multiple factors and the dynamic evolution law of gas.Taking the No.34205 fully-mechanized working face in No.3 Underground Mine of China Coal Pingshuo Group as the research object,through the dynamic analysis of environ-mental parameters such as annual gas volume fraction monitoring(O2,CH4,CO,CO2)and temperature and humidity pressure,combined with four machine learning prediction models of support vector regression(SVR),random forest(RF),decision tree(DT)and gradient boosting(GB),the causes of low oxygen in the return corner were revealed and the prediction model was constructed.Research has shown that sudden drops in ground pressure and slow propulsion speed exacerbate low oxygen phenomena through the"breathing effect"of goaf and the oxidation of residual coal,respectively.The lowest O2 volume fraction during the morning shift is 17.58％;comparison of prediction models shows that the GB model has the best prediction performance(MSE=0.011 6,R2=0.936 4),with an error reduction of 86.57％ compared to SVR,and can accurately capture low oxygen dynamic fluctuations.The study provides a theoretical basis for low oxygen warning and management in coal mines through the analysis of gas dynamic evolution laws,low oxygen causation,and multi mod-el comprehensive prediction optimization.