基于有限记忆变遗忘因子的子空间辨识算法

针对传统递推子空间辨识算法对时变参数跟踪速度慢的问题,基于自适应变遗忘因子机制提出一种新的子空间辨识算法.为此首先设计了变遗忘因子作用下输入输出Hankel矩阵的更新机制；然后运用系统矩阵特征值空间欧氏距离信息实现变遗忘因子的自适应更新；最后为隔断历史数据的作用,采用有限记忆法进一步改进算法.理论及仿真结果表明,新算法跟踪速度快、跟踪效果好.

Subspace identification algorithm based on finite-memory variable forgetting factor

A novel subspace identification algorithm is proposed based on self-adaptive variable forgetting factor to deal with the problem of low convergence rate in traditional algorithms. The update form of input-output data Hankel matrices is redesigned. The self-adaptive forgetting factor is realized with the help of Euclidean-distance of eigenvalues of the identified system matrix. In order to eliminate the effect of old data, a modified algorithm is designed based on the finite-memory method. Theoretical proof and simulation results show that the tracking response of the modified algorithm is faster and the performance is better than the traditional algorithms.