【作者】 张继承；

【导师】 石红瑞；

【作者基本信息】 东华大学 ， 模式识别与智能系统， 2006， 硕士

【摘要】 多速率系统的模型辨识,包括提升模型的辨识,及对提升模型的辨识结果进行抽取得到快速单速率模型,是多速率采样系统研究中的一个重要方向。本文针对多速率系统辨识的若干理论问题展开了研究,以提升技术和传统的模型辨识理论作为工具,进行模型辨识的研究,并得到最终的快速单速率模型辨识结果,将辨识结果应用于预测控制的研究中,并能得到满意的结果。具体内容如下: 1.讨论系统的可控性和可观性对多速率系统辨识的影响,对可能存在的系统时延进行估计,并分析估计结果的准确性。考察多速率过程因果约束条件,并利用最小二乘方法辨识提升模型。 2.针对模型状态未知的情况,同时基于获得的状态估计和提升系统的输入和输出数据,递归估计出模型参数,然后基于获得的参数估计,进一步计算系统的状态。基于前面的提升模型辨识结果,抽取出快速的单速率模型。在这里,首先分析了快速单速率模型估计结果的唯一性,并分别利用频域和时域方法计算单速率模型。 3.利用预报误差辨识方法辨识提升模型,并将辨识结果应用于双速率预测控制器设计中。 4.根据串级系统外层和内层的不同采样周期关系,提出了一种不同的基于双速率系统的单控制器设计方法。更多还原

【Abstract】 The theoretic research direction of multi-rate systems model identification focuses on lifted model identification and extracting fast-rate model. All of these are one of key points of this field. In this thesis, some theoretic problems in multi-rate sampled systems are studied. The study of model identification is analyzed based on the lifting technology and the traditional system identification algorithm. The final fast-rate single-rate model identification results are derived. According to the identification results, the multi-rate predictive control algorithm is investigated. This thesis is composed of follows:1. Controllability and observability of the lifted system are analyzed, which are essential to the identifiability issues. The time delay of the process from the liftd model is estimated, the accuracy of this estimation process is analyzed. This thesis discusses the causality issues involved in the lifed models and proposes a subspace-based least squares algorithm for identifying the state-space models of the lifted systems.2. For cases where states are unmeasured, based on the hierarchical identification principle and the least squares algorithm, the unknown parameters and states of the lifted model are estimated. Based on the results studied in the proceeding sections, the fast-rate single-rate model is extracted. In this section, time domain and frequency domain approaches to compute a fast-rate model are presented.3. By using the prediction error approach based on dual-rate systems, the lifted model is identified. The lifted model identification result is used for the design of dualrate predictive controller.4. In cascade control system, the sample rate of outer loop and inner loop are always different. So a single model predictive controller that could handle the multirate cascade control strategy in a straight forward manner is proposed.更多还原