基于数据挖掘模型的配电网故障定位诊断  

A New Approach for Fault Section Diagnosis of Distribution System Based on Data Mining Model

作　　者：廖志伟[1] 孙雅明[1] 杜红卫[1] 

机构地区：[1]天津大学电力自动化与能源工程学院,天津300072

出　　处：《天津大学学报（自然科学与工程技术版）》2002年第3期322-326,共5页Journal of Tianjin University：Science and Technology

基　　金：国家自然科学基金资助项目 (598770 1 6)

摘　　要：由于配电网故障定位所依据的故障信息来自于户外的 FTU,其运行环境较恶劣 ,元器件受损或信息丢失的可能性高 ,易形成变异故障模式 ,导致故障定位的错判 ,提出基于粗糙集 (RS)理论和遗传算法 (GA)相结合的数据挖掘 (DM)模型来处理实时输入信息的畸变和实现配电网的故障定位 .首先通过 RS对变异故障信息域的数据集进行划分 ,再用 GA挖掘出输入信息与故障定位诊断结果间冗余关系及内在关联性规则 .经仿真测试证明 ,基于 DM模型的故障定位与基于常规前馈神经网络 (FNN)故障定位原理相比 。Considering the working condition of feeder terminal units in practical application of fault section diagnosis for distribution networks,the damage of FTU elements and the distortion of information are unavoidable,and incorrect diagnosis may be caused by distorted fault patterns.This paper presents a data mining method,which is based on the combination of rough set(RS) theory and genetic algornhm(GA),to deal with distorted information and carry out the fault section diagnosis of distribution networks.In this approach,RS is used to analyze the generalized fault knowledge region and GA is used to mine the redundant relation and internal relevant rules among input information and fault section diagnosis results The high fault to1erance performance of the proposed approach is proved through comparison with that of feedforward neural networks model based fault section diagnosis.

关 键 词：数据挖掘模型 配电网 故障定位诊断 神经网络 容错性能 粗糙集 

分 类 号：TM726[电气工程—电力系统及自动化]