STUDY ON CONTINIOUS-TIME SYSTEM IDENTIFICATION IN VIEW OF KNOWLEDGE OF PHYSICAL AND CHEMICAL LAWS

近年来，由于鲁棒控制理论的发展，存在模型不确定性的系统辨识问题引起了人们的广泛关注。由于以微机为中心的技术的进步，使得在更短的采样周期内进行处理成为可能，估计精度有望有相当大的提高。我们主要从以下几个方面进行了研究：(1)预滤波器如抗混叠滤波器对系统辨识的影响；(2)确定模型阶数方法的发展；(3)鲁棒控制系统辨识方法的评价；(4)多变量系统的推广；(5)非线性系统辨识方法的发展。我们的目的是开发适合控制器设计的识别方案，从模型质量的角度来看，模型的质量取决于模型在适当频率范围内充分跟踪特征的能力，并且预滤波器不仅是数据处理，而且是将植物的先验知识纳入模型。本研究充分表明，通过将辨识方法扩展到多变量系统，并考虑到植物的物理信息，发展非线性系统辨识，可以将辨识方法应用于大规模、复杂的实际系统。

Recently, much attention has been attracted for system identification, stimulated by develepment of robust control theory in the presence of model uncertainty. Considerable improvement of estimation accuracy is expected because processing in smaller sampling period has become possible due to the progress of the technology around microcomputer. We have studied mainly from the viewpoint of(1) effects of prefileter suce as antialias filter to system identification(2) development of the method to determine the order of a model(3) evaluation of system identification methods for robust control(4) extension to multivariate systems, and(5) development of methods for nonlinear systems.Our intention has been development of identification scheme suited well for controller design, from standpoint that the quality of the model depends on that the model has capability to trace adequately the characteristics in proper frequency range, and that prefilter is not only data processing but also ones taking a priori knowledge of the plant into the model. We showed totally in this research, that the identification schemes can be applied to large scale, complex real systems by extending them to multivariate systems and developing nonlinear systems identification, taking account of physical information of the plant.