改进的PSO-SVM模型在大坝变形预测中的应用  

The Application of Improved PSO-SVM Model in Dam Deformation Prediction

作　　者：徐朗 蔡德所[1] XU Lang, CAI De-suo(Three Gorges University, Water Resources and Enviromnent Institute ,Yichang 443002, Hubei Province, Chin)

机构地区：[1]三峡大学水利与环境学院,湖北宜昌443002

出　　处：《中国农村水利水电》2018年第3期120-123,128,共5页China Rural Water and Hydropower

基　　金：国家自然科学基金重点项目(51439003)

摘　　要：支持向量机(SVM)由于理论基础完善,在处理高维度非线性问题中,表现出了许多特有的优势。因此,支持向量机模型在处理大坝变形预测问题时具有明显的优越性,且常采用粒子群(PSO)算法对SVM模型的惩罚参数C和核函数σ进行寻优,但是标准的PSO算法存在早熟收敛以及易于陷入局部极小值等缺陷。引入柯西分布函数和密度函数,根据函数变化的性质对标准PSO算法中粒子位置和速度的更新公式进行改进,有效提升了算法的寻优性能。运用改进后的PSO-SVM模型对水布垭面板堆石坝的面板挠度变形进行预测分析,并与SVM模型、标准PSO-SVM模型的预测结果进行对比,结果显示,改进的PSO-SVM模型的拟合效果和预测精度都更加优秀,为进行大坝变形预测工作提供了一种性能优良,精度较高的预测模型。Vecto rmachine （SVM） is supported as the theoretical basis of improvement in the treatment of high dimension nonlinear problems, which shows many unique advantages. Therefore, it has the obvious superiority of the support vector machine model in the process of dam deformation prediction, and of ten particles warm optimization （PSO） algorithm for SVM model of the penalty parameter Candkernel function sigma optimization, but the standard PSO algorithm has premature convergence and is easy to fall into the local minimum defects. Cauchy distribution function and density function are introduced and modified according to the nature of function change in the updating formula of particle position and velocity in the standard PSO algorithm, effectively improve the performance of optimization algorithm. The improved pso SVM model is used to predict the panel deflection deformation of shuibuya panel rockfill dam and compare with the prediction results of SVM model standard pso SVM. The results show that the improved PSO-SVM model and the prediction accuracy is more excellent, which provides a good performance and high precision prediction model for dam deformation prediction.

关 键 词：大坝安全监测 支持向量机 改进的粒子群算法 面板挠度预测模型 

分 类 号：TV698.1[水利工程—水利水电工程]