基于YOLOv4的绝缘子爆裂故障识别研究

Insulator Burst Fault Identification Based on YOLOv4

提出了一种基于YOLOv4目标检测方法的绝缘子爆裂故障智能识别模型,以某供电局一年内正常绝缘子及爆裂绝缘子图像为样本,对故障识别模型进行训练,得出识别模型的权重。采用该模型对绝缘子及其爆裂故障进行识别测试,结果表明,该模型识别绝缘子的平均精度(AP)为92.6%,识别绝缘子爆裂故障的平均精度(AP)为91.78%,模型每秒处理帧数(FPS)为46 frame/s,与Faster R-CNN、SSD模型比较可得,提出的绝缘子爆裂故障识别模型能够准确、快速地对绝缘及其爆裂故障进行识别。

Based on the YOLOv4 target detection method,in this study,we proposed an intelligent insulator burst fault recognition model.Considering images of normal and burst insulators in a power supply bureau within one year as samples,the proposed model was trained to obtain its weight.The proposed model was further used to identify insulators and their bursting faults.Experimental results showed that the proposed model had an average precision of insulator positioning of 92.6%,an average precision of insulator burst fault location of 91.78%,and the model’s resolution was 46 frame/s.Compared with Faster R-CNN and SSD models,the constructed insulator burst fault identification model can accurately and quickly identify insulators and their burst faults.