大坝变形监测的粒子群优化高斯过程预测
|
摘要
针对大坝变形数据的非平稳非线性特点,传统预测模型受到了一定限制。鉴于高斯过程(Gaussian Process,GP)对非平稳数据具有高自适应性,考虑到其自身在协方差函数选取以及超参数优化方面存在不足,为提高高斯过程模型的预测精度,文中通过粒子群算法(Particle Swarm Optimization,PSO)优化其超参数并选择最优协方差函数。通过实例验证分析,比较多元回归分析、GP、PSO-GP三种模型在大坝变形监测数据处理中的预测精度,表明大坝非线性预测模型粒子群优化高斯过程算法具有较高的预测精度,是一种有效的大坝变形分析预测方法。
For the non-stationary and nonlinear characteristics of dam deformation data series,the traditional prediction models have some limitations.Gaussian process(GP)is a new machine learning technology which has been developed recently,with high adaptability.In order to improve the prediction accuracy of Gaussian process algorithm,the optimal covariance function suitable for engineering cases should be selected,and in this study its super-parameters are optimized by particle swarm optimization.Experimental results of three models with multiple regression、GP and PSOGP showed that the PSO-GP dam deformation prediction model is a valuable new method with higher accuracy for future dam deformation prediction.
For the non-stationary and nonlinear characteristics of dam deformation data series,the traditional prediction models have some limitations.Gaussian process(GP)is a new machine learning technology which has been developed recently,with high adaptability.In order to improve the prediction accuracy of Gaussian process algorithm,the optimal covariance function suitable for engineering cases should be selected,and in this study its super-parameters are optimized by particle swarm optimization.Experimental results of three models with multiple regression、GP and PSOGP showed that the PSO-GP dam deformation prediction model is a valuable new method with higher accuracy for future dam deformation prediction.
引文
[1]周世健,罗亦泳,鲁铁定.局部均值分解(LMD)方法在多尺度变形分析中的应用分析探讨[J].江西科学,2014(4):461-466.
[2]王杰,徐锡杰,解斐斐.基于PSO算法的BP神经网络对植被叶片氮素含量的预测[J].北京测绘,2018,32(11):1289-1292.
[3]周昀琦,王奉伟,周世健,等.顾及邻近点变形因素的高斯过程建模及预测[J].测绘科学,2018,43(4):114-121.
[4]赵洪波.支持向量机在隧道围岩变形预测中的应用[J].岩石力学与工程学报,2005,24(4):649-653.
[5]刘嘉.经验模态分解和最小二乘支持向量机在大坝变形预测中的应用[J].北京测绘,2019,33(1):101-105.
[6]邹宗民,王冠,徐磊,等.非等时距灰色-马尔科夫链模型在深基坑变形预测中的应用[J].北京测绘,2019,33(1):84-89.
[7]王奉伟.改进的局部均值分解和高斯过程在变形监测数据处理中的应用研究[D].江西南昌:东华理工大学,2016.
[8]范千,王新洲,许承权.大坝变形预测的支持向量机模型[J].测绘工程,2007,16(6):1-4.
[9]罗亦泳,姚宜斌,张立亭,等.基于高斯过程的GPS高程转换模型[J].测绘通报,2015(11):11-14,59.
[10]RASMUSSEN C,WILLIAMS CKI.Gaussian Processes for Machine Learning[M].MIT Press,2005.
[11]ASSAREH E,BEHRANG M,ASSARI M,et al.Application of PSO(Particle Swarm Optimization)and GA Techniques on Demand Estimation of Oil in Iran[J].Energy,2010,35(12):5223-5229.
[12]WANG Chunshan,HAN Cunliang,QIE Zhihong,et al.WITHDRAWN:The TAM and PSO-SVM Methods for Assessing the Crack Damage of Concrete Slab[J].OptikInternational Journal for Light and Electron Optics,2014:S0030402614001405.
[13]GARCIA Nieto,GARCIA Gonzalo,ALONSO Fernández,et al.Hybrid PSO-SVM-based Method for Long-term Forecasting of Turbidity in the Nalón River Basin:A Case Study in Northern Spain[J].Ecological Engineering,2014,73:192-200.
[14]SEEGER M.Gaussian Process for Machine Learning[J].International Journal of Neural Systems,2008,14(2):69-106.
[15]徐冲,刘保国,刘开云,等.基于粒子群-高斯过程回归耦合算法的滑坡位移时序分析预测智能模型[J].岩土力学,2011,32(6):1669-1675.
[16]SAIN Stephan.The Nature of Statistical Learning Theory[J].Technimetrics,1996,38(4):409-409.
[17]任超,梁月吉,庞光锋,等.基于经验模态分解和支持向量机的大坝变形预测[J].水电能源科学,2014(12):67-70.
[2]王杰,徐锡杰,解斐斐.基于PSO算法的BP神经网络对植被叶片氮素含量的预测[J].北京测绘,2018,32(11):1289-1292.
[3]周昀琦,王奉伟,周世健,等.顾及邻近点变形因素的高斯过程建模及预测[J].测绘科学,2018,43(4):114-121.
[4]赵洪波.支持向量机在隧道围岩变形预测中的应用[J].岩石力学与工程学报,2005,24(4):649-653.
[5]刘嘉.经验模态分解和最小二乘支持向量机在大坝变形预测中的应用[J].北京测绘,2019,33(1):101-105.
[6]邹宗民,王冠,徐磊,等.非等时距灰色-马尔科夫链模型在深基坑变形预测中的应用[J].北京测绘,2019,33(1):84-89.
[7]王奉伟.改进的局部均值分解和高斯过程在变形监测数据处理中的应用研究[D].江西南昌:东华理工大学,2016.
[8]范千,王新洲,许承权.大坝变形预测的支持向量机模型[J].测绘工程,2007,16(6):1-4.
[9]罗亦泳,姚宜斌,张立亭,等.基于高斯过程的GPS高程转换模型[J].测绘通报,2015(11):11-14,59.
[10]RASMUSSEN C,WILLIAMS CKI.Gaussian Processes for Machine Learning[M].MIT Press,2005.
[11]ASSAREH E,BEHRANG M,ASSARI M,et al.Application of PSO(Particle Swarm Optimization)and GA Techniques on Demand Estimation of Oil in Iran[J].Energy,2010,35(12):5223-5229.
[12]WANG Chunshan,HAN Cunliang,QIE Zhihong,et al.WITHDRAWN:The TAM and PSO-SVM Methods for Assessing the Crack Damage of Concrete Slab[J].OptikInternational Journal for Light and Electron Optics,2014:S0030402614001405.
[13]GARCIA Nieto,GARCIA Gonzalo,ALONSO Fernández,et al.Hybrid PSO-SVM-based Method for Long-term Forecasting of Turbidity in the Nalón River Basin:A Case Study in Northern Spain[J].Ecological Engineering,2014,73:192-200.
[14]SEEGER M.Gaussian Process for Machine Learning[J].International Journal of Neural Systems,2008,14(2):69-106.
[15]徐冲,刘保国,刘开云,等.基于粒子群-高斯过程回归耦合算法的滑坡位移时序分析预测智能模型[J].岩土力学,2011,32(6):1669-1675.
[16]SAIN Stephan.The Nature of Statistical Learning Theory[J].Technimetrics,1996,38(4):409-409.
[17]任超,梁月吉,庞光锋,等.基于经验模态分解和支持向量机的大坝变形预测[J].水电能源科学,2014(12):67-70.