A CNN-BiLSTM-attention model for time series prediction of microseismic events in coal mines

Microseismic monitoring is very important to realize the risk assessment and pre-warning of coal mine impact disasters. To enhance the precision of the early warning for coal mine bump disasters, a microseismic time series prediction approach that integrates the attention mechanism of the convolutional neural network with the bidirectional long short-term memory network model is put forward. Firstly, feature extraction is carried out on the data of the pickup sensor using CNN. Secondly, the extracted feature sequences are input into BiLSTM for time series modeling, and the attention mechanism is incorporated to allocate weights to the hidden layers of BiLSTM so as to enhance the precision of model prediction. To confirm the efficacy and stability of the proposed method, the model is validated and contrasted with the data from the pickup sensor on the 103-coal mine operation face in Xinjiang. According to the experimental outcomes, the prediction precision and generalization ability of this model outperform those of other models.