基于轻量AlexNet的电容型电压互感器故障诊断

Lightweight AlexNet-based Fault Diagnosis for Capacitor Voltage Transformers

摘要:

电容型电压互感器（CVT）是重要的一次侧电压监测元件。针对环境温度、湿度以及元件老化等因素造成的电容型电压互感器一次侧电容上下臂击穿或互感器二次侧短路等故障，提出了一种基于轻量AlexNet的电容型电压互感器故障诊断方法。该方法利用Matlab建立了CVT电路模型，分别对高压臂电容击穿、低压臂电容击穿以及互感器二次侧短路3种典型的故障进行仿真。采集CVT二次侧电压数据，利用马尔可夫变迁场将其转化为特征矩阵，最后使用轻量化的AlexNet神经网络对电压特征矩阵进行故障分类。仿真实验证明，所提方法在不拆除CVT的情况下，能准确检测出CVT的故障类型。

Abstract:

Capacitor voltage transformers (CVT) are important primary voltage monitoring components, but due to the influence of ambient temperature, humidity, aging of the components and other factors caused by capacitor upper and lower arm breakdown in primary side of capacitor voltage transformer and short circuit in secondary side of the transformer and other faults, a light-weight AlexNet-based fault diagnosis method for capacitor voltage transformer is proposed. This method uses Matlab to build a CVT circuit model and simulates three typical faults, namely, capacitance breakdown of high-voltage arm, capacitance breakdown of low-voltage arm and short circuit in secondary side of the transformer. The voltage data in secondary side of CVT are collected and transformed into feature matrices using Markov transition fields.Finally the voltage feature matrices are classified into faults using a light-weight AlexNet neural network. The simulation experiments prove that the proposed method can accurately detect the fault type of CVT without removing the CVT.