中文摘要：

分析了子空间法的基本原理和辨识步骤，提出用子空间法在线识别主导特征值并指导控制器协调设计。在各机组AVR参考点施加伪随机激励，通过广域量测采集多点输出量，用于提取多机系统的降阶MIMO模型。基于辨识模型可进行小干扰稳定性估计，估计结果非常接近线性化方法的理论值；辨识模型还可用来设计全局协调的控制器，通过求解Levine-Atlans方程组得到各机协调的最优分散反馈增益。对8机4区域测试系统的研究表明，所提方法在模型提取，小干扰稳定分析及协调励磁控制的实现3个环节均可行、有效。

英文摘要：

In this paper a subspace algorithm is applied to identify the MIMO model of power systems from measured input-output data and then to design expected controllers. The identified model is proved to be valid for extracting electromechanical eigenvalues on line through comparing estimated results with their theoretical values, which are acquired from linear model of simulated system. Each generator is excited by pseudorandom binary signal at the AVR＇s reference, and its response such as speed, voltage and active power are sampled by phasor measurement unit. The identified low-order MIMO model is applied to design decentralized and coordinated excitation laws, which are decided by optimal feedback coefficients and derived from the solution of a Levine-Athans equation. At last, an eight-machine test system is studied, satisfied model cross validation illustrates the identification is successful. similar to their actual values. Also, the excitation laws based on identified model Estimated eigenvalue is very have ensured good dynamic characteristics.