SSA-LSTM Based Deformation Prediction Modeling of Concrete Dams

Dam deformation prediction is essentially a time series prediction problem with non-stationarity and multiple external influences that are difficult to quantify. In order to improve the prediction accuracy and solve the problems of large randomness and difficult parameter selection in LSTM prediction, this paper proposes a concrete dam deformation prediction model (SSA-LSTM) using sparrow search algorithm (SSA) to optimize LSTM parameters. Firstly, taking the concrete dam deformation monitoring data as an example, the historical monitoring data are preprocessed by cleaning, normalization and denoising. Subsequently, the hyperparameters of the Long Short-Term Memory (LSTM) neural network model are fine-tuned utilizing the Salp Swarm Algorithm (SSA) to enhance the congruence between the input data and the network architecture. Finally, the deformation of concrete dam body is predicted. The results show that the SSA-LSTM model proposed in this paper significantly enhances the prediction performance, provides a new solution for dam deformation prediction, and is effective in long time series deformation prediction, providing a useful reference for engineering practice and decision making.