Time Sequence Prediction Method for Gas Concentration in High Outburst Coal Seams Based on GRU

In order to solve the difficult problem of accurate prediction of gas concentration in fully mechanized mining faces of high outburst coal seams and improve the timeliness of gas prevention and control in mines, this article combines mining working conditions, geological conditions, and spatial correlation characteristics to establish a gas concentration influence indicator system, the Recursive Feature Elimination Method (RFE) is adopted to screen the optimal feature, and the core prediction features are determined as the mixed amount of high extraction pipeline working conditions, high extraction pipeline pressure, mixed amount of return air corner working conditions, return air corner pipeline pressure, and gas emission quantity. A gas concentration prediction model is established based on Gated Recurrent Unit (GRU). The research results indicate that the fitting degrees of the predicted values and the on-site actual values of GRU time sequence model are both greater than 0.86, their Mean Absolute Error (MAE) are both less than 0.015, and the Root Mean Square Error (RMSE) is controlled within 0.02. Compared with the traditional model prediction results, the MAE is reduced by 40%, effectively improving the prediction accuracy and stability, and providing reliable technical support for gas early warning and prevention and control of mines.