Displacement prediction model for earth-rock dams based on Bayesian optimization algorithm and BiGRU

The prediction of earth-rock dam displacement is one of the core contents of dam safety monitoring. Traditional methods such as statistical models and machine learning models have certain effects, but there are some problems such as low efficiency of parameter optimization and insufficient extraction of time series features. In recent years, Bayesian optimization algorithm (BO) has been widely used in the field of engineering prediction due to its strong global search capability and adaptability to high-dimensional parameter spaces. Bidirectional Gated Recurrent Unit (BiGRU) can capture the forward and backward dependency relationships of time series simultaneously, making it more suitable for complex time series modeling than traditional GRU. Aiming at the complicated nonlinear and nonstationary characteristics of the time series of earth-rock dam displacement monitoring, this paper proposes a displacement prediction model for earth-rock dam (BO-BiGRU) which combines Bayesian optimization and BiGRU, automatically adjusts the superparameters of BiGRU model through Bayesian optimization, and realizes high-precision prediction by combining the measured displacement data. The results show that compared with the BO-GRU, GRU, BO-SVM, SVM and SRA models, the RMSE and MAE of the BO-BiGRU model decreased by 18.9%、23.2%、33.9%、39.4%、48.7%and 19.6%、23.9%、31.1%、41.1%、49.9% respectively, and the mean R² reached 0.98. The results show that, compared with other prediction methods, BO-BiGRU model has higher prediction accuracy and better stability, and can better reflect the long-term deformation trend of earth-rock dam and its fluctuation characteristics with water level, which provides a new method for displacement prediction of earth-rock dam.