Abstract: In order to improve the identification accuracy of the T-S fuzzy model, and to solve the problem of determining the clustering centers in the traditional clustering algorithms, a Nearest Neighbor Fuzzy Clustering-based (NNFC) T-S fuzzy identification method is proposed. Firstly, the proposed nearest neighbor clustering approach decreases the subjectivity of the artificial presetting for the initial parameters of clustering. And the computation efficiency of clustering is also increased. Furthermore, the result of nearest clustering is afforded for the initial parameters of the Fuzzy c-Means (FCM) algorithm. Thus the premise parameters in the fuzzy rules are identified accurately. Finally, the Stable Kalman Filter (SKF) method is combined with the presented NNFC to estimate the consequent parameters quickly. The effectiveness of proposed method is verified by the classic chemical pH neural process. 

Key words: Nearest neighbour, clustering initialization, fuzzy clustering, T-S fuzzy identification

摘要： 为提高T-S模糊模型的辨识精度和计算效率，并针对传统聚类算法存在的聚类中心选取问题，提出一种基于最近邻模糊聚类的T-S模糊辨识方法。首先，利用所提的最近邻聚类法降低了传统计算中人为预设聚类初始参数的主观性，并提高了聚类效率，其结果作为模糊c均值算法的初始参数，来实现对模糊规则中前提参数的准确辨识，最后结合稳态卡尔曼滤波算法快速估计规则的后件参数。所提方法的有效性通过典型化工过程pH中和过程的建模得以验证。

关键词: 最近邻, 聚类初始化, 模糊聚类, T-S模糊辨识