基于边缘粒子滤波的高速列车性能参数估计方法  

Performance parameter estimation method of high-speed train based on Rao-Blackwellised particle filter

作　　者：丁建明[1] 林建辉[1] 王晗[2] 黄晨光[1] 赵洁[1] 

机构地区：[1]西南交通大学牵引动力国家重点实验室,四川成都610031 [2]南车四方车辆有限公司,山东青岛266111

出　　处：《交通运输工程学报》2014年第3期52-57,共6页Journal of Traffic and Transportation Engineering

基　　金：国家自然科学基金项目(61134002;51305358);国家863计划项目(2011AA110501)

摘　　要：针对高速列车参数估计中参数增广为状态变量时所出现的非线性问题,提出一种基于边缘粒子滤波的参数估计方法。在Rao-Blackwellised(RB)框架下,将高速列车性能参数估计的概率模型进行分块化处理。应用卡尔曼滤波对线性的状态块进行一步预测和测量更新,应用粒子滤波对非线性的参数块进行一步预测与测量更新,实现参数的动态估计,并通过理论分析和高速列车参数估计实例验证了方法的有效性。分析结果表明:与经典的扩展卡尔曼滤波相比,提出的方法具有对初值免疫和算法稳定的特点;参数估计误差快速收敛到5%以内,且提出的参数估计方法是无偏估计,具有较好的工程适用性。In order to solve the nonlinear problems caused in augmenting the state vector of the performance parameters of high-speed train,a method of parameter estimation based on RaoBlackwellised particle filter（RBPF）was developed.Under the framework of Rao-Blackwellised（RB）principle,the probabilistic model of parameter estimation was divided.The Kalman filter（KF）was applied for the prediction time step and measurement update of linear state block and the RBPF was applied for the prediction time step and measurement update of nonlinear parameter block to realize the dynamic estimation.Through theoretical analysis and parameter estimation example of high-speed train,the validity of RBPF method for parameter dynamic estimation was verified.Analysis result shows that compared with the classical extended KF（EKF）method,RBPF method has the characteristics of initial immunity and algorithm stability.RBPF method shows its good engineering applicability for the minor parameter estimation error which is less than 5%,and inexistence of estimation deviation.7figs,15refs.

关 键 词：高速列车 性能参数 边缘粒子滤波 参数估计 

分 类 号：U260.11[机械工程—车辆工程]