摘要
大坝监控过程中,大坝变形的实测值是一个非线性且非平稳的时间序列,支持向量机(SVM)适用于解决小样本、非线性问题,在SVM算法的基础上建立了改进的大坝变形监控模型,利用差分自回归移动平均模型(ARIMA)解决非平稳时间序列问题的优势,对SVM模型的残差进行处理,并采用粒子群算法(PSO)优化支持向量机(SVM)中的核函数。实例分析表明,优化后的组合模型预测结果可靠,且精度较SVM模型有所提高。
The monitoring data of dam deformation is a nonlinear and non-stationary time series. The Support Vector Machine( SVM) is suitable for solving the problems of small sample and nonlinear effectively,so dam deformation model is established on the basis of SVM and came up with improved methods. With the aim of solving the non-stationary issue effectively,the paper used ARIMA model of SVM model fitting to predict the residual item correction,then used Particle Swarm Optimization( PSO) to improve the accuracy of nuclear parameter optimization of SVM. Using the measured distortion monitoring data of a domestic dam as an example,the model is tested to be of high precision and credible results and it is worthy of being applied to the dam deformation forecast or other hydropower projects.
引文
[1]顾冲时,吴中如.大坝与坝基安全监控理论和方法及其应用[M].南京:河海大学出版社,2006:56-73.
[2]SU Huaizhi,WEN Zhiping,WU Zhongru.Study on an Intelligent Inference Engine in Early-Warning System of Dam Health[J].Water Resources Management,2011,25(6):1545-1563.
[3]李经纬,包腾飞.ARIMA-GRNN模型在大坝安全监测中的应用[J].水电能源科学,2013,31(7):48-50.
[4]王利,张双成,李亚红.动态灰色预测模型在大坝变形监测及预报中的应用研究[J].西安科技大学学报,2005,25(3):328-332.
[5]黄丽.BP神经网络算法改进及应用研究[D].重庆:重庆师范大学,2008:30-35.
[6]丁世飞,齐丙娟,谭红艳.支持向量机理论与算法研究综述[J].电子科技大学学报,2011,40(1):2-10.
[7]邓乃杨,田英杰.数据挖掘中的新方法:支持向量机[M].北京:科学出版社,2005:47-52.
[8]常甜甜.支持向量机学习算法若干问题的研究[D].西安:西安电子科技大学,2010:28-32.
[9]杜传阳,郑东健.基于SVM理论的大坝变形监测模型改进方法研究[J].三峡大学学报(自然科学版),2015,37(2):10-14.