基于特征挖掘的电网故障诊断方法  

Method of Power Grid Fault Diagnosis Based on Feature Mining

作　　者：李再华[1] 白晓民[1] 周子冠[1] 许婧[1] 李晓珺[1] 张霖[1] 孟珺遐[1] 朱宁辉[1] 

机构地区：[1]中国电力科学研究院,北京市海淀区100192

出　　处：《中国电机工程学报》2010年第10期16-22,共7页Proceedings of the CSEE

基　　金：国家重点基础研究发展计划资助项目(973项目)(2004CB217904)~~

摘　　要：专家系统在应用方面的主要瓶颈是:规则库的维护;推理的速度和准确度的协调。分析了故障信息序列中必有或特有的信息,提出了基于特征挖掘的关联规则挖掘方法。结合电网故障信息的特征,改进了频繁模式(frequent pattern,FP)–算法:考虑了故障信息的特征,如时序和因果关联关系、故障性质、严重故障、稀有故障等因素;增加了规则的"或"逻辑;改进了FP-树的修剪技术。算例表明该算法能够大量减少无效挖掘,推理速度和准确度显著提高,适用于在线诊断。The two main bottlenecks in the application of expert system are： the maintenance of rule base; the coordination of speed and accuracy of reasoning. Features and key events of fault events were analyzed, then a novel method of association rule mining based on feature mining was presented, the method was originated from frequent pattern （FP）-algorithm and was improved. The improvements include： features of fault information are utilized, such as the time sequence and causality of events, fault type and serious fault or unusual fault; OR logical function of rules is added; prune technique of FP-tree is improved. Use case shows the improved algorithm could reduce invalid mining largely, and the speed and accuracy of the reasoning is heightened prominently. The algorithm is fit for being used online.

关 键 词：数据挖掘 关联规则 特征挖掘 频繁模式一算法 故障诊断 专家系统 

分 类 号：TM72[电气工程—电力系统及自动化]