双线性参数系统的极大似然递推参数估计  

Maximum Likelihood Recursive Parameter Estimation for Bilinear Parametric Systems

作　　者：刘海波[1] 江安宁 LIU Haibo;JIANG Anning(College of Automation and Electronic Engineering,Qingdao University of Science and Technology,Qingdao 266061,China)

机构地区：[1]青岛科技大学自动化与电子工程学院,山东青岛266061

出　　处：《青岛科技大学学报（自然科学版）》2024年第4期137-145,共9页Journal of Qingdao University of Science and Technology:Natural Science Edition

摘　　要：研究具有自回归滑动平均噪声的双线性参数系统,该系统结构复杂,噪声项的研究更具普遍意义。为实现系统参数的在线辨识,采用梯度搜索方法,推导了双线性参数系统的随机梯度算法。对于出现的未知项,基于递阶辨识原理使用其估计值进行替代。极大似然估计方法基于概率论,具有良好的一致性、渐近正态性和可用性,在引入极大似然估计方法后得到了相应的极大似然随机梯度算法,为进一步减小有色噪声对参数估计精度的影响,结合多新息辨识理论,将标量单新息扩展为多新息向量,研究了具有自回归滑动平均噪声的双线性参数系统的极大似然多新息随机梯度参数估计方法,并进行了仿真验证。In this paper we study bilinear parametric systems with autoregressive moving average noise.The structure of the system is complex,and the study of the noise term is of more general significance.In order to realize the on-line identification of system parameters,the stochastic gradient algorithm of bilinear parameter system is derived by using gradient search method.For unknown terms,the estimated value is used to replace them based on the principle of hierarchical identification.Maximum likelihood estimation method based on probability theory,has a good consistency,asymptotic normality and availability,after the introduction of maximum likelihood estimation method to get the corresponding maximum likelihood stochastic gradient algorithm,in order to further reduce the influence of colored noise on parameters estimation precision,combining many new interest identification theory,the extension of the scalar single new rates for new interest vector,A maximum likelihood multi-information stochastic gradient parameter estimation method for bilinear parametric systems with autoregressive moving average noise is studied and verified by simulation.

关 键 词：系统辨识 双线性参数系统 梯度搜索 极大似然 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]