基于符号有向图节点定量化模型的故障诊断方法  

Research on Fault Diagnosis Approach Using Quantitative Knowledge of the Node of Signed Directed Graph

作　　者：曹文亮[1] 王兵树[1] 马进[1] 马良玉[1] 辛宇峰 

机构地区：[1]华北电力大学自动化系,河北省保定市071003 [2]阳城国际发电有限责任公司,山西省晋城市048102

出　　处：《中国电机工程学报》2006年第7期67-71,共5页Proceedings of the CSEE

摘　　要：基于系统衍生节点的状态不可能比要对它施加影响节点状态差这一假定,该文研究了一种将SDG定性定量模型、主元统计法(PCA)和遗传算法相结合的电站热力系统智能故障诊断方法。通过用SDG构造系统模型,并结合节点的定量信息对节点之间的因果联系进行描述,形成SDG节点定量化模型;用PCA监控不同的变量,并考虑变量间的相关关系,有效地解决了确定节点故障发生的可能性大小的问题;用遗传算法对故障传播路径进行搜索,可以有效处理SDG定性模型中不可测节点干扰、计算量大和规则组合爆炸等问题。案例研究表明,该方法具有较强的故障诊断能力。Based on the postulation of the condition of a node should not be worst than those of its descendent nodes in signed directed graph （SDG） model of power plant thermal process, this paper studies a hybrid intelligent fault diagnosis method that integrates SDG, principal component analysis （PCA） and genetic algorithm together, to search the possible fault propagation paths. First, the cause-effect relation between nodes is described by combining the quantitative knowledge, to form the qualitative and quantitative model （QSDG）, and the PCA that can monitor the correlation among different variables is used to handle uncertainty in the system, and then the genetic Mgorithm is used to deal with the disturbance of unmeasured nodes and shorten the calculating time. The case studies show the QSDG has better resolution in fault diagnosis.

关 键 词：符号有向图 定性定量模型 主元统计 遗传算法 故障诊断 

分 类 号：TK267[动力工程及工程热物理—动力机械及工程]