Journal of AppliedMath

Aiming at the difficulty of early warning of coal and gas outburst, the early warning index system of coal and gas outburst is analyzed and established, and an early warning method based on the Bi-directional Long Short-term Memory neural network algorithm (BiLSTM) is proposed. The Hunter-prey Optimization algorithm (HPO) is used to optimize the problem that BiLSTM is easy to fall into the local optimal solution, and the maximum or minimum value of the fitness function is obtained. Based on this, a coal and gas outburst early warning model based on the combination of Hunter-prey Optimization algorithm and Bi-directional Long Short-term Memory neural network is established. In the model, the dynamic optimization process of the position of the hunter and the prey in the HPO algorithm is used to realize the refined parameter adjustment of BiLSTM. With the help of coal mine measured data verification, the results show that the model can effectively warn of the danger of coal and gas outburst, and the early warning result is consistent with the danger of on-site coal and gas outburst. Compared with classical algorithms such as the Convolutional Neural Network model (CNN), Long Short-term Memory neural network model (LSTM) and Bi-directional Long Short-term Memory neural network model (BiLSTM), the Mean Absolute Error (MAE) of this model respectively reduced by 36.16%, 41.55%, and 56.86%, and the Mean Absolute Percentage Error (MAPE) were respectively reduced by 20.55%, 45.94%, and 55.18%, and the Root Mean Square Error (RMSE) was respectively reduced by 33.23%, 50.12% and 66.94%, and the average Relative Error (RE) is 1.57%. It is proved that the proposed Hybrid Deep Learning algorithm has a high early warning function and practical performance and can be used as an effective early warning model to guide coal mining work.