基于轻量化YOLOv5s的光伏热斑检测定位方法  

PHOTOVOLTAIC HOT SPOT DETECTION AND POSITIONING METHOD BASED ON LIGHTWEIGHT YOLOv5s

作　　者：孙海蓉[1] 刘永朋 周黎辉 Sun Hairong;Liu Yongpeng;Zhou Lihui(Department of Automation,North China Electric Power University,Baoding 071003,China;Hebei Power Generation Process Simulation and Optimization Control Technology Innovation Center,North China Electric Power University,Baoding 071003,China;Zhangjiagang Scenauto Information Technology Co.,Ltd.,Zhangjiagang 215699,China)

机构地区：[1]华北电力大学自动化系,保定071003 [2]华北电力大学河北省发电过程仿真与优化控制技术创新中心,保定071003 [3]张家港迅见信息技术有限公司,张家港215699

出　　处：《太阳能学报》2024年第11期282-288,共7页Acta Energiae Solaris Sinica

基　　金：河北省省级科技计划(22567643H)。

摘　　要：针对目前目标检测技术在检测光伏热斑效应时模型检测速度低、计算复杂、模型结构复杂等问题,提出基于轻量化YOLOv5s的光伏热斑检测定位方法。首先,以YOLOv5s为基础模型,引入轻量网络ShuffleNetV2改进YOLOv5s的主干网络,利用其分组卷积和通道混洗的设计思想,减少模型参数和计算量,同时保持较高的准确率。其次,引入轻量级卷积GSConv优化YOLOv5s的Neck部分,利用其深度可分离卷积结合标准卷积的形式,降低计算复杂度,优化整体模型。最后利用数据集进行验证。结果表明,轻量化后的模型在保证较高精度的前提下,能够提高检测速度、减少参数量和计算量。A lightweight YOLOv5s based photovoltaic hot spot detection and localization method is proposed to address the issues of low model detection speed,complex computation,and complex model structure in current target detection technologies for detecting photovoltaic hot spot effects.Firstly,based on the YOLOv5s model,the lightweight network ShuffleNetV2 is introduced to improve the backbone network of YOLOv5s.By utilizing its design ideas of group convolution and channel shuffling,the model parameters and computational complexity are reduced while maintaining high accuracy.Secondly,introducing lightweight convolution GSConv to optimize the Neck part of YOLOv5s,utilizing its deep separable convolution combined with standard convolution to reduce computational complexity and optimize the overall model.Finally,use the dataset for validation.The results indicate that the lightweight model can improve detection speed,reduce parameter and computational complexity while ensuring high accuracy.

关 键 词：光伏组件 特征提取 红外热图像 图像识别 热斑检测 YOLOv5 

分 类 号：TM914.4[电气工程—电力电子与电力传动]