多变量受控自回归滑动平均系统的极大似然辨识方法  

M aximum likelihood identification method for a multivariable controlled autoregressive moving average system

作　　者：高艳普 王向东 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 [2]青岛海大海洋仪器开发有限公司,山东青岛266003

出　　处：《山东大学学报（工学版）》2015年第2期49-55,74,共8页Journal of Shandong University（Engineering Science）

基　　金：国家自然科学基金资助课题(61104001)

摘　　要：提出了一种针对多变量受控自回归滑动平均(controlled autoregressive moving average system-like,CARMAlike)系统的极大似然参数估计算法。将CARM A-like系统分解成为m个辨识模型(m是输出量的个数),使每一个辨识模型仅包含一个需要估计的参数向量,通过极大似然方法估计每个辨识模型的参数向量,从而得到整个系统的参数估计值。仿真结果验证了该算法的有效性。An algorithm of maximum likelihood method for parameters estimate was presented aimed at multivariable controlled autoregressive moving average （CARMA-like）. The algorithm transform the CARMA-like system into m identification models （m is the output numbers）, each of which only had a parameter vector which needed to be esti- mated, and then through maximum likelihood method for estimating parameter vectors of each identification model, and all parameters estimate of the system were obtained. Simulation results verified the effectiveness of the proposed algorithm.

关 键 词：参数辨识 多变量系统 极大似然法 最小二乘法 CARMA模型 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]