Research On Mining Pressure Prediction Method Of Working Face Based On LSTM Network

With the development of coal mine intelligence,the mine pressure monitoring data obtained every day has entered the era of big data.Aiming at the problem that the traditional mine pressure analysis and prediction method has been difficult to deal with the massive mine pressure monitoring data,through theoretical analysis,data preprocessing,model optimization and other methods,a computer automatic division method of the working cycle of the bracket is designed,and a long-and short-term The mining pressure prediction model of Long Short-Term Memory(hereinafter referred to as LSTM)neural network has built a mining pressure prediction cloud platform to realize the automatic analysis and prediction of the mining pressure of the working face.The main research results are as follows:(1)The computer automatic division method of the working cycle of the stent is designed.According to the pressure change of the support itself and the mutual influence between the supports,the change characteristics of the working resistance of the support are analyzed,the working cycle division method of the working face support is designed,and the Python language programming is used to realize the automatic division of the working cycle of the support by the computer,and the original mine pressure data is automatically processed.The working cycle of the bracket is corresponding to the coal cutting cycle one by one,and the initial support force and end-cycle resistance of the bracket are further obtained.(2)The mine pressure prediction model based on LSTM neural network is established.The mining pressure big data time series was constructed as the model training data set,and a control experiment was set up to study the influence of different historical data lengths and learning rates on the model prediction accuracy.The results show that the model prediction effect is relatively optimal when the length of historical data is 5000 and the learning rate is 0.001,and the relative error is 6.79%,which is within the allowable error range of engineering practice.(3)The reliability and generalization ability of the mine pressure prediction model based on LSTM neural network is verified.Three different supports a,b,and c of a coal mine face in Shendong mining area are selected for prediction,and the errors of the prediction results are 6.42%,6.80% and 6.79% respectively,and the model has strong generalization ability.The superiority of BP and RNN neural network in long sequence rock pressure prediction problem.The test results show that the mine pressure prediction method based on LSTM neural network is feasible,and it can assist coal miners to adjust the production plan in time in practical engineering applications.(4)A cloud platform for mine pressure prediction was built.The cloud platform visualizes the change of working resistance of the working face support and the thermal map of the mining pressure of the working face,predicts the value and change trend of the mining pressure in the future,and calculates the distribution of the initial support force of the working face,the initial support force compliance rate,and the cycle.Final resistance and to pressure criteria.The use of the cloud platform is verified by using the historical mine pressure data of the 42015 working face of the Buertai coal mine in the Shendong mining area.The results show that the analysis and prediction results of the mine pressure prediction cloud platform are accurate,which can effectively improve the work efficiency of coal miners.In this paper,there are 47 figures,10 tables,and 53 references.