Research On Dam Deformation Prediction Based On Improved Sparrow Search Algorithm With Optimized GRU

The number of dams in China is the highest in the world and is increasing.Dams play an important role in promoting national economic development,while accidents in dams can bring great disasters,so it is important to monitor the safety of dams and establish an accurate deformation prediction model.In this paper,we address the shortcomings of the traditional deformation prediction model,which requires high monitoring data or difficult to describe quantitative relationships,and build an optimized gated-cycle unit(GRU)model based on the improved sparrow search algorithm(ISSA)and apply it to dam deformation prediction.First,four improvement strategies are proposed for the traditional sparrow search algorithm(SSA).The first is to use Sobol sequence initialization to obtain a more uniformly distributed and widely traversed initial population;the second is to use adaptive weighting factors to avoid falling into local optimum;the third is to use Corsi variational perturbation to enhance the global search capability of the algorithm;the fourth is to introduce a dynamic spiral search strategy to enhance the search capability while speeding up the late convergence of the algorithm to improve the efficiency of the algorithm.According to the above four strategies,the traditional SSA algorithm is improved to construct ISSA algorithm,and the performance of ISSA algorithm is tested by benchmark function.Next,a dam deformation prediction model based on the ISSA algorithm for GRU parameter search optimization is constructed.The three parameters(learning rate,number of input layer neurons,and number of hidden layer neurons)of GRU are optimized by the ISSA algorithm,and the root mean square error of the prediction results of the dam deformation prediction model is used to construct the objective function.Finally,the ISSA-GRU dam deformation prediction model constructed in this paper is applied in practice and compared with other models.The measured deformation monitoring point data of the vertical displacement of Long Yangxia Dam in Qinghai is used as the research data.The experimental results show that among the four accuracy evaluation index values of root mean square error,mean absolute error,mean absolute error percentage,and coefficient of determination,the GRU model is 0.8051 mm,0.6607 mm,10.06%,and 0.9060,respectively,and the SSA-GRU model is 0.5015 mm,0.4017 mm,5.96%,and 0.9635,respectively.0.4777 mm,0.3824 mm,5.80%,0.9669 for the MSSA-GRU model,0.5100 mm,0.4172 mm,6.25%,0.9623 for the ABC-GRU model,and 0.8653 mm,0.7521 mm,12.01%,0.8914 for the BP neural network model,respectively.While the four indexes of ISSA-GRU model are0.3854 mm,0.3090 mm,4.71%,0.9785,all indexes are better than other comparison models.the first three indexes of RMSE,MAE and MAPE are reduced by 23.2%,23.1% and 21.0%respectively when comparing ISSA-GRU model with SSA-GRU model.Therefore,the ISSAGRU dam deformation model constructed in this paper has the advantage of high prediction accuracy,thus providing a new method for dam deformation prediction.