摘要
为提升大坝变形预测能力,提出了一种基于粒子群算法(PSO)优化支持向量机(SVM)的混凝土重力坝变形预测模型。通过粒子群算法对支持向量机惩罚函数C与核函数σ进行寻优,避免了拟合过程中易陷入局部最优解的问题,提高了模型的拟合精度。以新疆北疆某碾压混凝土坝2014年~2019年变形监测数据为例,建立了逐步回归、SVM、PSO-SVM三种模型。结果表明,PSO-SVM模型预测时段复相关系数高达0.991,明显优于逐步回归与SVM模型,同时标准差也低于其他两种模型,说明PSO-SVM模型拟合及预测精度更高,有效验证了模型的可靠性及准确性。
A concrete gravity dam deformation prediction model based on support vector machine(SVM) optimized by particle swarm algorithm(PSO) is proposed in order to enhance the prediction capacity of dam deformation monitoring model.The particle swarm algorithm is applied to optimize the penalty function C and the nuclear function σ of support vector machine,thus avoiding the problems of falling into local optimal solutions in the fitting process,and then the fitting capacity of the model is enhanced.Taking the 2014 to 2019 deformation monitoring data of a RCC dam in northern Xinjiang as an example,three models of stepwise regression,SVM and PSO-SVM are developed to forecast dam deformation respectively.The results indicate that the multiple correlation coefficient of PSO-SVM is 0.991 in forecast period,which is significantly higher than that of stepwise regression and SVM models,and the standard deviation of PSO-SVM is also lower than that of other two models.The comparison shows that the fitting and prediction accuracy of PSO-SVM is higher,which effectively verifies the reliability and accuracy of the model.
作者
朱明远
吴艳
阮新民
周富强
美丽古丽
ZHU Mingyuan;WU Yan;RUAN Xinmin;ZHOU Fuqiang;Meiliguli(Xinjiang Institute of Water Resources and Hydropower Research,Urumqi 830049,Xinjiang,China;Xinjiang Ertix River Basin Development and Construction Management Bureau,Urumqi 830000,Xinjiang,China)
出处
《水力发电》
CAS
2022年第3期64-69,共6页
Water Power
基金
新疆维吾尔自治区水利科技专项(YF2020-05,XSKJ-2021-06)。
关键词
碾压混凝土坝
粒子群算法
支持向量机
变形监测
变形预测
RCC dam
particle swarm optimization algorithm
support vector machine
deformation monitoring
deformation prediction
