Application of extension neural network algorithm and chaos synchronization detection method to partial discharge diagnosis of power capacitor

Abstract

The power capacitor is an important equipment of a power system, which must run in a high-temperature and high-voltage environment for a long time, in order to maintain the stability of system. Hence, the failure rate of the power capacitor increases over time. Conventional capacitor testing methods mostly require multiple sensors and signal parameters, which increase system cost and complexity. This study attempted to propose a novel method for capacitor fault recognition. It developed a fault diagnosis system for power capacitor by employing the extension neural network (ENN) algorithm and the chaos synchronization detection method. In terms of signal acquisition, partial discharge was measured by hardware circuits, such as high-frequency current transformer (HFCT), high-pass filter, noninverting amplifier circuit, and high frequency oscillography. The ENN and the chaos method were integrated with hardware circuits to develop a human-machine interface fault diagnosis system designed with LabVIEW. The proposed method was also compared with extension method and artificial neural network algorithm. According to the results, the ENN has the best recognition result, and the huge data could be reduced greatly by the data pre-processing of the chaos synchronization detection method. Any subtle changes in the power capacitor discharge signal could be detected effectively, thus achieving an accurate operating state of the power capacitor.