Renmin Zhujiang (Jan 2022)
Dam Deformation Prediction Model Based on Artificial Electric Field Algorithm-Extreme Learning Machine Model
Abstract
This paper studies a prediction method combining the artificial electric field algorithm (AEFA) and extreme learning machine (ELM) to improve the accuracy of dam deformation prediction.With the 72nd dam settlement data of Guandi Hydropower Station as an example,three ELM prediction models with a delay time of 1,and embedding dimensions of 2D,3D,and 5D are constructed.AEFA is applied to optimize the ELM input layer weights and hidden layer bias to construct three different AEFA-ELM dam deformation prediction models with different embedded dimensions and build the corresponding AEFA-support vector machine (SVM) and AEFA-BP as prediction comparison models.AEFA-ELM,AEFA-SVM,and AEFA-BP models with nine embedding dimensions are used for the training and prediction of the deformation data of the example dam.The results show that the AEFA-ELM model with embedded dimensions of 2D,3D,and 5D has an average relative error of 3.94%,4.08%,and 3.67% of the dam deformation prediction for the next 10 periods of the example,respectively.The prediction errors are all smaller than those of AEFA-SVM and AEFA-BP models.The proposed model possesses high prediction accuracy and has a certain reference value for dam deformation prediction research.
