摘要：Dam break will cause serious social and environmental impacts, such as the life loss, vegetation damage and water pollution. To ensure safe operation of dam, Long Short-Term Memory (LSTM) network is introduced to forecast the dam deformation with historical monitoring data. Moreover, to extend the length period of forecasting, an LSTM network with dynamic update strategy (DLSTM) is investigated. First, multi-step forecasting is implemented by training the LSTM network. Then, the network structure is dynamically adjusted with a certain step threshold and multi-step forecasting is implemented again. Finally, a case is given to verify the effectiveness and superiority of DLSTM network.
