Mine water inflow prediction based on DRN-BiLSTM model

Abstract

For the problem of low accuracy and applicability of the model prediction in the study of mine water inflow, a method of mine water inflow prediction based on bidirectional short and long memory network(BiLSTM) and deep residual network (DRN) is proposed. First, the data of mine water inflow is processed by wavelet decomposition and normalization to obtain trend item data and detail item data. Secondly, the trend item data was predicted by DRN network method, and the detail item data was predicted by BiLSTM network method. Finally, the two parts of the prediction results will be combined to get the mine water inflow prediction results. The results show that the DRN-BiLSTM model has higher prediction accuracy than a single model, indicating that the model has better generalization.

Keywords

• mine water inflow
• drn-bilstm model
• deep residual network (drn)
• bidirectional short and long memory network(bilstm)
• wavelet transform