基于改进PSO-SVM-MC模型的大坝变形监测模型

Dam Deformation Monitoring Model Based on Improved PSO-SVM-MC Model

分析与处理大坝变形监测资料在大坝安全监测中意义重大。支持向量机(SVM)在大坝安全监测建模中应用广泛,但采用标准粒子群(PSO)算法对SVM参数寻优过程中,易陷入局部最优,且残差也会影响模型的预测精度。为提高大坝监测模型的泛化能力和预测精度,采用改进后的自适应位置PSO(APPSO)算法,对SVM模型的参数进行寻优,并利用马尔科夫链(MC)模型修正PSO-SVM模型的残差。工程实例分析表明,PSO-SVM-MC模型可提高模型预测的泛化能力和精度。

It is very important to analyze and deal with dam deformation monitoring data for dam safety monitoring.The SVM model is commonly applied to dam safety monitoring.But parameters of SVM model are optimized by PSO,which is easy to trap local extremum.In order to reduce the influence of residual error on forecasting precision and improve generalization ability of dam monitoring model,we set up PSO-SVM-Markov chain model.Parameters optimization is based on improved PSO.And Markov chain model is used to correct residual errors of PSO-SVM model.Instance results indicate that the improved PSO-SVMMarkov chain model can enhance the generalization ability and prediction accuracy of the model.