基于神经网络和元件关联分析的电网故障诊断

An Approach Integrating Neural Networks with Topology Analysis of Electric Power Networks for Fault Diagnosis

利用人工神经网络（ＡＮＮ）和元件关联分析进行电力系统故障诊断，提出了一种电 力系统多重复杂故障的诊断方法。该方法采用面向元件的ＡＮＮ模型，将电力系统的元件分 为３类，即线路、变压器和母线，对每类元件都有一个特定的ＡＮＮ处理其报警信息，确定 故障位置。对于同一跳闸区域中面向各个元件的ＡＮＮ的诊断输出，通过定义一个故障指标 函数，根据各元件的故障指标函数值的大小来识别同一跳闸区域内的多重故障。该方法所使 用的ＡＮＮ模型规模小，通用于网络的所有元件，且故障识别的方法简单，适用于大规模电 力系统的故障诊断。

This paper proposes a new approach for multiple fault diagnosis of power systems by integrating artificial neural networks (ANN)with topology analysis of electric power networks In the proposed approach, three element-oriented feed-forward ANNmodules are employed to locate three types of fault elements-transmission lines, transformers and bus-bars Each ANN produces its own di- agnosis results using the relative alarm messages when a fault occurs in the power system All diagnosis results produced by all the ANNs are synthesized through defining a fault index function based on top- ology analysis of electric power networks to diagnose multiple faults in a blackout area The values of the index function are used as possibility indicators of different faults The proposed approach is promis- ing for on-line fault diagnosis in large-scale power systems