Research on Dam Deformation Prediction Model Based on Wavelet Neural Network

Abstract

The BP traditional neural network dam deformation prediction model has the disadvantages of slow convergence speed, easy to fall into local extreme points and poor generalization ability, a dam deformation prediction model based on wavelet neural network is constructed by combining the time-frequency local analysis ability of wavelet and the self-learning and self-organizing characteristics of neural network; the measured value of settlement of Lixi barrage in Guangzhou is selected as the research object, From January 2016 to June 2019, a total of 557 sets of data were used as model training set to establish the model; from July 2019 to October 2019, 95 sets of data were used as model verification set to evaluate the model performance. The test set prediction results of wavelet neural network and BP traditional neural network show that the mean square error of deformation prediction model based on BP traditional neural network is 0.138 and the certainty coefficient is 0.748; based on wavelet analysis, the mean square error of the coupled neural network deformation prediction model is 0.025 and the certainty coefficient is 0.927. Wavelet neural network has fast convergence speed, high prediction accuracy and strong generalization ability in dam deformation prediction And other characteristics.