一种光伏阵列串联电弧故障智能检测方法  

An Intelligent Detection Method for Series Arc Fault of Photovoltaic Array

作　　者：金辉 高伟[1] 杨耿杰[1] JIN Hui;GAO Wei;YANG Geng-jie(College of Electrical Engineering and Automation,Fuzhou University,Fuzhou 350108,China)

机构地区：[1]福州大学电气工程与自动化学院,福建福州350108

出　　处：《电工电气》2023年第1期43-47,66,共6页Electrotechnics Electric

摘　　要：由于串联电弧故障特征表现不足以及样本不平衡的问题,导致传统的诊断算法检测效果不佳。提出了一种基于图像识别的光伏阵列串联电弧故障诊断方法 :利用格拉姆角和场(GASF)将发生串联电弧故障时的暂态电流数据编码为二维图像,从而放大电弧故障的本质特征;深度卷积生成对抗网络(DCGAN)被用来增扩电弧故障GASF特征图像,以均衡正常与故障样本数量;训练一个LeNet-5诊断模型完成电弧故障的识别。经过实验验证,所提方法有效提升了光伏阵列串联电弧故障的辨识度,且具备优秀的抗干扰能力,对实测数据的整体识别准确率高达99.5%。The traditional diagnostic algorithms have poor performance because of inadequate characteristic expression of series arc fault(SAF) and sample imbalance. A detection method for series arc fault of photovoltaic array based on image recognition is put forward. First,according to Gramian angular summation field(GASF), the paper encodes the transient current data of SAF into two-dimensional image which amplifies the essential characteristics of SAF. Second, the deep convolution generative adversarial network(DCGAN) is adopted to enlarge GASF fault characteristic expression image of SAF to achieve balance between normal and fault sample numbers. Finally, a LeNet-5diagnostic model is trained to recognize SAF. The experimental results show that this method efficiently improves the SAF of photovoltaic arrays identification accuracy to 99.5% and has great anti-interference ability.

关 键 词：光伏阵列 串联电弧故障 格拉姆角和场 深度卷积生成对抗网络 

分 类 号：TM615[电气工程—电力系统及自动化]