非理想数据条件下非线性状态空间系统辨识方法研究

辨识数据作为系统辨识的核心要素之一，它能直接影响辨识算法的性能。因数据获取方式不同、数据采集设备故障等因素，再加上外部干扰随机不确定，导致过程变量间存在未知时滞、过程数据中包含异常值、输出数据中包含非对称干扰噪声等问题。为此，本项目拟针对上述辨识问题，开展辨识数据不理想条件下，非线性状态空间系统的辨识方法研究：设计有效的辨识算法估计非线性状态空间系统模型参数，并建立稳定可靠的辨识框架。具体的研究内容包括：（1）未知时变时滞存在条件下双率系统辨识算法研究；（2）状态过程和量测过程均受到异常值干扰情况下的鲁棒辨识算法研究；（3）输出数据中包含非对称干扰噪声条件下的辨识算法研究；（4）结合仿真和实验验证上述辨识算法的有效性。项目所提出的辨识算法不仅能丰富和完善现有的辨识算法，也可促进非线性状态空间模型在控制器设计、观测器设计和输出预测等领域的应用。

The performance of the identification algorithms is usually influenced by the identification data, which is often treated as one of the core elements of system identification. For practical system identification, the quality of the identification data may be deteriorated by many unavoidable factors. For example the common problems such as the unknown time delay, the outliers and the asymmetric output noise may be imposed by the external system disturbances, different data acquisition methods and hardware malfunctions and so on. In this project, the identification of nonlinear state space systems based on non-ideal identification data is deeply studied. The main content lies in the following four parts: (1) The identification of nonlinear state space systems under dual-rate sampled data and time-delays; (2) The identification of nonlinear state space systems under outlier-contaminated state noise and measurement noise; (3) The identification of nonlinear state space systems under asymmetric measurement noise; (4) Verify the proposed algorithms through the simulation and experimental tests. Some practical and common issues are considered in this project and the corresponding effective algorithms are developed as well. The proposed identification algorithms can extend the existing algorithms and they can also facilitate the applications of nonlinear state space model in system controller design, system observer design, output prediction and so on.