基于鲁棒模糊C-均值回归的TS建模方法及其在热工过程中的应用  

TS Modeling Based on Robust Fuzzy C-Regressions and Its Application for Thermal Process

作　　者：赵阳[1] 王培红[1] 苏志刚 李益国[1] 朱晓瑾 ZHAO Yang;WANG Peihong;SU Zhigang;LI Yiguo;ZHU Xiaojin(Key Laboratory of Energy Thermal Conversion and Control of Ministry of Education （Southeast University）, Nanjing 210096, Jiangsu Province, China;Zhejiang Zheneng Taizhou Second Electric Power Generation Co. Ltd., Taizhou 318000, Zhejiang Province, China)

机构地区：[1]能源热转换及其过程测控教育部重点实验室(东南大学),江苏省南京市210096 [2]浙江浙能台州第二发电有限责任公司,浙江省台州市318000

出　　处：《中国电机工程学报》2018年第7期2063-2069,共7页Proceedings of the CSEE

基　　金：国家自然科学基金资助项目(51476028);科技部支撑计划(2015BAA03B02)~~

摘　　要：热工过程往往具有强耦合、大惯性和非线性等特点,且在运行过程中易受到诸种不确定性因素的干扰,导致常规建模方法难以获得令人满意的效果。针对非线性热工对象特性建模问题,文中改进了模糊C均值回归算法（FCR）的误差函数,提出一种鲁棒模糊C均值回归算法（RFCR）,并将其用于热工对象的TS建模前件辨识过程中。一方面,相比于FCM聚类算法,RFCR作为一种超平面型聚类算法,更加符合TS建模局部线性化的特点,能够更加精确地辨识TS模型的前件参数;另一方面,RFCR算法能够克服离群点和噪音点的影响,增强了TS模糊辨识的鲁棒性,从而更加适宜于热工过程建模。仿真实例表明,应用文中新算法所建立的煤气炉及机炉协调系统的TS模型具有精确性高、鲁棒性强的特点,证明了算法的有效性和实用性。The thermal process is often of strong coupling, great delay and nonlinearities, and it can be disturbed by uncertain factors easily, which make the conventional modeling method difficult to obtain satisfactory results. For this situation, This paper improved the error function of the fuzzy C-regressions（FCR） and proposed robust fuzzy C-regressions （RFCR）, Furthermore, TS modeling method based on RFCR was presented. On the one hand, as a hyper-plane-shaped clustering algorithm, RFCR is more suitable in the fuzzy space partition for the identification of TS modeling, and the parameters of the TS model can be identified more accurately; On the other hand, RFCR can overcome the influence of outliers or noises, and thus enhance the robustness of the TS fuzzy identification. The performance of the proposed modeling method was evaluated by some examples, and the results demonstrate that our algorithm has the advantages of higher accuracy and robustness on the gas furnace and boiler-turbine System, which shows the effectiveness and practicality of our proposed algorithm.

关 键 词：热工过程 TS模糊建模 前件辨识 模糊C-均值回归 鲁棒性 

分 类 号：TK172[动力工程及工程热物理—热能工程]