Research And Application Of Forecasting Model For Coal Mine Water Inrush Based On LSTM

Coal mine water inrush is one of the major disasters of safety production,it is affected by many factors,and these factors will change over time,so it’s a dynamic and complex system.Recurrent Neural Networks(RNNs),particularly those using Long Short-Term Memory(LSTM)hidden units,are powerful and increasingly popular models for learning from sequence data,because time series is introduced into the model by adding a self-join edge to the hidden layer of RNNs,a LSTM-based model is proposed to solve this problem.The main works include:(1)Summarized the factors that affect water inrush by study the water inrush mechanism and the traditional forecasting model.Through the method of normality test,feature selection based on Wrapper evaluation strategy and normalization,The data preprocessing is completed.(2)Research the RNN and LSTM model in both Model framework and training algorithm,built LSTM model to solve the problem of forecasting coal mine water inrush,and the optimal parameters are determined by a series of experiments.(3)Preprocessed data is send to LSTM model for further research,then use the Softmax classifier to output the prediction results.Finally,compared with improved BP and SVM by experiment,and evaluates the proposed model by prediction accuracy and speed.(4)After forecasting of water inrush,the spatial network model of coal mine roadway is established by graph and network theory,and the prediction model for range of water inrush is established according to calculation model of mine water inrush and tunnel water inrush path model,that is a reference for developing the water inrush prevention and the analysis of accident hazard degree.