Coal Mine Fire Prediction Dased on Long Short Term Memory Network

The fire caused is one of the most difficult fires in coal mines by coal seam spontaneous combustion, it seriously endangers the coal mine production and workers' life safety. In view of the concealment and uncertainty characteristics of high temperature points of coal seam residual coal, the LSTM neural network timing memory, self-learning and adaptive ability are used to accurately analyze the spontaneous combustion gas in coal mine. The multiple gas factors and historical data factors of coal mine spontaneous combustion are comprehensively considered, and the potential relationship between the data is excavated, so as to identify the nonlinear influence of each gas factor on coal mine fire in a long time range, and solve the problem of single traditional prediction gas factor and low prediction accuracy. In this paper, the LSTM neural network model is built and the time step length and depth of the model are analyzed. The super parameters of the model are optimized, and the accuracy and reliability of coal mine fire prediction and early warning are improved.