基于部分信息融合的分布式故障诊断策略在分布式发电系统中的应用  

Distributed Fault Diagnosis Strategy Based on Partial Information Fusion for Distributed Generation System

作　　者：朱琳[1] 李双宏[1] 杨煜普[1] 

机构地区：[1]上海交通大学电子信息与电气工程学院

出　　处：《化工自动化及仪表》2017年第4期351-356,共6页Control and Instruments in Chemical Industry

基　　金：国家自然科学基金项目(61273161)

摘　　要：将模型的故障监测框架和基于数据的学习方法相结合,提出了一种新的故障监测方法,实现了串级工业过程中的故障监测和定位。首先,对串级工业系统进行分析,得出整个系统的划分方法,并提供了构建子系统的条件。然后,采用分布式主元分析(PCA)方法在实际测量数据集中提取子系统的特征信息,用于TS模糊推理的建模。再提出基于二级贝叶斯的模糊模型实现故障的非线性识别。通过比较模型输出和每个子系统的实际测量值构造残差监测故障,实现定位。最后,通过固体氧化物燃料电池(SOFC)系统仿真实验,验证了所提出的故障监测模型的有效性和可行性。Through having model＇ s fault detection framework combined with data-based learning method, a new fault monitoring method was proposed to achieve fault detection and location in serial connection process. Firstly, having serially connected system analyzed to obtain partition method of the whole system and to provide conditions to construct subsystems; and then, having the distributed principal component analysis （PCA） a- dopted to extract subsystems＇ feature information from real measurement data for the modeling of TS fuzzy in- ference ; and finally, having two-stage Bayes theory-based fuzzy model proposed to realize non-linear identifica- tion of the faults. Through comparing output signals of the TS models and the real measurements of each sub- system, evaluating the residuals and examining the fault occurrence with location information were implemen- ted. Simulating solid oxide fuel cells （SOFC） proves both feasibility and efficiency of the method proposed.

关 键 词：故障诊断 分布式PCA TS模糊模型 串级系统 固体氧化物燃料电池 

分 类 号：TH865[机械工程—仪器科学与技术]