Research on pressure predictive control algorithm of hydraulic support in working face based on numerical simulation and GA-BP neural network

In view of the difficulty in perceiving the pressure of hydraulic support in working face, this paper takes 10401 fully mechanized mining face as the engineering background, uses FLAC 3D to establish a three-dimensional numerical model of working face-hydraulic support, and obtains continuous pressure samples of 100 hydraulic supports under 17 push progress. Taking the support position, pushing progress and continuous historical pressure as input, a three-step advanced pressure prediction model of BP neural network optimized by genetic algorithm is constructed. Combined with the predicted pressure utilization rate, pressure growth rate and prediction uncertainty, a four-level mine pressure risk determination and hierarchical control method is established. The results showed that the MAE, RMSE and MAPE of the GA-BP model test set were 1.78 MPa, 1.94 MPa and 7.089%, respectively, which were 36.29%, 37.93% and 48.66% lower than those of the ordinary BP model, and the determination coefficient was increased from 0.795 to 0.921. Based on the time hysteresis and spatial linkage correction, the four-level risk identification accuracy of 300 test samples reached 87.33%, and the macro average F1 value was 85.68%. When level III and IV are defined as high-risk states, the high-risk recall rate reaches 100%, and no high-risk missed alarm samples appear in the current test data. The constructed algorithm can provide reference for the adjustment of initial support force, advance speed control and adjacent support of hydraulic support in fully mechanized mining face.