非均匀采样多率ARMAX系统RELS算法性能分析  

Performance Analysis of RELS Identification for Non-uniformly Sampled Mutlirate ARMAX Systems

作　　者：刘艳君[1] 谢莉[1] 

机构地区：[1]江南大学控制科学与工程研究中心,江苏无锡214122

出　　处：《控制工程》2009年第4期409-411,415,共4页Control Engineering of China

基　　金：国家自然科学基金资助项目(60574051)

摘　　要：针对一类有色噪声干扰的非均匀采样多率ARMAX系统的辨识问题,基于增广参数维数理论,将系统模型参数化,将信息向量中含有的不可测噪声项用其估计残差代替,推导了非均匀采样ARMAX系统的递推增广最小二乘(RELS)算法;利用鞅收敛定理对该算法的收敛性进行了理论分析,结果表明该算法在噪声方差有界和广义持续激励的条件下能够收敛到真参数。仿真例子验证了该算法具有良好的收敛速度与估计精度。To the identification of non-uniformly sampled multirate ARMAX systems, a recursive extended least squares algorithm is presented based on the idea of replacing the unmeasurable noise terms with the estimated residuals. The convergence properties of the proposed algorithm are studied by using the martingale convergence theorem. It is shown that the parameter estimation error consistently converges to zero under the generalized persistent excitation condition and bounded noise variance. The simulation results show that the algorithm proposed is effective.

关 键 词：递推辨识 参数估计 多率系统 状态空间模型 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]