Multi-scale combined prediction model of concrete dam deformation based on VMD-LSTM-ARIMA

The deformation of concrete dam can be regarded as the result of the synergistic action of hydraulic component, temperature component and aging component. According to the different component characteristics of deformation and the correlation of different time scales, a multi-scale combined prediction model for concrete dam deformation based on VMD-LSTM- ARIMA is proposed. Firstly, using the adaptive analysis function of VMD, the trend term and cycle term of dam deformation are decomposed. Secondly, LSTM model is used to effectively predict the cycle term and trend term under different scales, and ARIMA model is used to identify the effective information of the remaining term. Finally, based on a practical project, the effectiveness and superiority of the proposed model are verified by comparing with the conventional combination algorithm. The calculation results show that the combined model fully considers the characteristics of the dam deformation, and can effectively fit and predict the dam deformation.