多变量Hammerstein系统的极大似然参数估计方法  

Maximum Likelihood Identification Algorithm for Multivariable Hammerstein Nonlinear Systems

作　　者：张玮[1] 王冬青[1] 

机构地区：[1]青岛大学自动化工程学院,山东青岛266071

出　　处：《青岛大学学报（工程技术版）》2015年第2期19-25,共7页Journal of Qingdao University(Engineering & Technology Edition)

基　　金：国家自然科学基金资助项目(61104001)

摘　　要：针对有色噪声干扰的多变量Hammerstein系统存在维数高和模型难以参数化的问题,本研究提出了递推极大似然参数估计方法,并对极大化似然函数系统参数向量的估计公式进行推导,同时给出了极大似然参数的辨识原理,并对多变量Hammerstein模型的参数估计进行理论分析,并通过数值进行仿真实验。理论分析和仿真结果表明,本研究所提出的算法,能够有效实现有色噪声干扰下的多变量Hammerstein系统的参数估计。随着递推次数的增加,辨识精度总体上不断提高;而且噪声方差越小,参数估计精度越高。该算法原理简单、适用范围广,提高了参数的估计精度,且易于实现参数的在线辨识,因此该算法可以扩展应用于非均匀采样系统和多变量系统的辨识和参数估计中。In view of the problems of high dimensions and the model being difficult to be parameterized in multivariable Hammerstein systems with colored noises, a recursive maximum likelihood estimation algo- rithm is investigated in this paper. The estimated formulas of parameter vectors of the maximum likelihood system is derived, the identification principle of the maximum likelihood parameters is presented, the pa- rameter estimation of the multivariable Hammerstein model is analyzed, and the simulation is experimen- ted in this study. The analysis and simulation results indicate that the proposed algorithm could effectively realize the parameter estimation of multivariable Hammerstein systems with colored noises, capable of gen- erally improving the identification accuracy with the recursive number increasing. The smaller noise vari- ance, the higher accuracy of the parameter estimation is. The algorithm is simple in principle and suitable for many applications, can improve the parameter estimation accuracy, and is easy to be implemented on- line. Thus the algorithm can be applied to the identification and the parameter estimation of non-uniform sampling systems and multivariable systems.

关 键 词：多变量 HAMMERSTEIN系统 极大似然 参数估计 系统辨识 

分 类 号：O212.1[理学—概率论与数理统计]