The analysis of coal safety production monitoring data based on deep learning

Abstract

With the integration of informatization and industrialization in the coal industry, the production automation system, safety production monitoring and control system and production scheduling system have been gradually improved. In view of the coal mine production environment, the analysis and auxiliary prediction of real-time data were the actual needs of coal mine on-site staff. Therefore, this paper analyzes the convolution neural network in deep learning, and combines the LSTM to train the monitoring data of coal safety production. The monitoring data of coal safety production were collected within 1 month, and were totaling 7 dimensions and 17568 data. The network initial value was 1e-3, droupout was in the range of 0.5∼1.0, batch_size was 20, the training time step was 60, the number of neurons in the hidden layer was 300, LSTM's lays number was 3, at this condition of LSTM's parameters, the results was best, the average deviation was 8.478%. Therefore, it was feasible to use this model to predict gas in this system. So as to predict the safety production data of the mine, set the range of the normal value, and if it exceeds the normal range, use the sound and light alarm way to provide the prediction and early warning assistance for the coal staff.