基于输入扩张的闭环系统子空间辨识及其一致性分析.PDFVIP

基于输入扩张的闭环系统子空间辨识及其一致性分析* 杨 华 李少远 上海交通大学自动化研究所，上海，200240 摘 要 子空间辨识作为一种新的辨识算法，由于其对先验知识要求较少，对于多变量系统辨识广泛适用， 以及在数值计算上的优势，受到了控制和辨识领域的广泛关注。在开环情况下，由于不可测噪声与输入 之间无关性条件成立，使得辨识结果满足一致无偏性要求。但是，由于工艺或者稳定性需要，许多工业 过程必须在闭环条件下进行辨识，当系统中存在反馈，上述无关性不再成立，成为子空间方法应用于闭 环辨识的主要障碍。本文结合线性代数和几何学的基本概念，将输入输出误差序列包含至输入子空间中， 基于输入扩张的状态空间构造方法，在子空间辨识框架内提出一种新的闭环辨识算法；解决开环算法应 用于闭环系统辨识时产生有偏估计，甚至不能正确辨识的问题；实现闭环条件下对系统状态空间矩阵的 一致估计，并理论证明该辨识算法的渐进一致性；最后通过仿真实例来验证本算法的有效性。 关键词 闭环辨识，子空间方法，输入扩张，渐进一致性 Closed-Loop Subspace Identification Based on Augmented Input with Consistency Analysis YANG Hua, LI Shaoyuan (Institute of Automation, Shanghai Jiao Tong University, Shanghai 200240) Abstract: As a kind of new identification methods, subspace methods have attracted much attention and stimulated a lot of research in the field of control and identification. They have many merits such as the minimize request of prior knowledge, applicability for MIMO process, and the robustness of numerical computation. Under open-loop condition, the independence between immeasurable noise and inputs makes the results are consistence and unbiased. However, closed-loop identification has to be taken in the need of practical industry processes. When the feedback is included in the system, the independence mentioned above will never satisfied, which becomes the crucial problem that subspace method has to be faced with. According to the basic problem of closed-loop identification a new closed-loop identification algorithm is proposed in the framework of subspace method combined with linear algebra and geometry. In order to implement a new reconstruction of the state sequence, the output and in