深度学习在单相接地故障识别的应用

Deep Learning Applied to Earth Fault Detection in Ineffectively Neutral Point Grounding System

配电网的单相接地故障识别一直是个充满挑战的话题,一方面在于配电网通常采用的小电流接地方式导致单相接地故障不够明显和稳定;另一方面,即使是采用大电流接地系统的接地故障也往往由于存在较大的过渡电阻而弱化了本有的故障特征。本文重点介绍了深度学习与继电保护和馈线自动化经验相结合,开发了基于人工智能的接地故障方向判别和选线的新算法,并成功部署在常规继电保护装置的嵌入式系统中。

The identification of single-phase ground faults in distribution networks has always been a challenging topic.On the one hand,the ineffective grounding scheme usually used in distribution networks makes single-phase ground faults not obvious and stable.On the other hand,even ground faults in effective grounding systems tend to weaken the inherent fault characteristics due to the presence of large fault resistance.This paper focuses on the combination of deep learning and relay protection and feeder automation experience and development of a new algorithm for ground fault direction discrimination and line selection based on artificial intelligence,which is successfully deployed in the embedded system of conventional relay protection devices.