基于可见光图像特征的光伏板碎裂状态分析与研究  

Analysis and Research on the Crack State of Photovoltaic Panels Based on Visible Light Image Features

作　　者：钟泳松 徐凌桦[1] 周克[1,2] 李凯飞 ZHONG Yongsong;XU Linghua;ZHOU Ke;LI Kaifei(The Electrical Engineering College,Guizhou University,Guiyang 550025,China;The Department of Brewing Engineering Automation,Moutai Institute,Zunyi 564507,China)

机构地区：[1]贵州大学电气工程学院,贵阳550025 [2]茅台学院酿酒工程自动化系,遵义564507

出　　处：《微处理机》2022年第2期38-41,共4页Microprocessors

基　　金：国家自然科学基金(61861007);贵州省工业攻关项目(黔科合支撑【2019】2152)2020年;贵州大学混合式课程建设项目“计算机控制技术”(2020030)。

摘　　要：光伏板表面裂纹会导致光伏系统发电效率降低,并引发腐蚀、灼烧等并发故障,进行光伏板碎裂状态智能识别与分析对光伏电站发电效率和高效运维有重要意义。利用碎裂光伏板在可见光图像上具有显著纹理特征这一特点,提出一种光伏板碎裂状态的识别和分析方法。以嵌入了注意力机制的残差网络来辨识不同碎裂状态程度的光伏板图像,建立不同碎裂程度光伏板和发电损失率的对应关系。实验表明:改进算法具有优良的识别精度,可以批量快速并定量地分析光伏板碎裂对光伏发电效率的影响,为光伏电站智能巡检和运维技术提供参考与新思路。Cracks on the surface of photovoltaic panels will reduce the power generation efficiency of photovoltaic systems, and lead to corrosion, burning and other concurrent failures. Intelligent identification and analysis of the crack state of photovoltaic panels is of great significance to the power generation efficiency and efficient operation and maintenance of photovoltaic power plants. A method for identifying and analyzing the crack state of photovoltaic panels is proposed, which takes advantage of the remarkable texture characteristics of cracked photovoltaic panels in visible light images. The residual network embedded with attention mechanism is used to identify the images of photovoltaic panels with different degree of crack, and the corresponding relationship between photovoltaic panels with different degree of crack and power generation loss rate is established. Experiments show that the improved algorithm has excellent recognition accuracy, and can quickly and quantitatively analyze the impact of photovoltaic panel crack on photovoltaic power generation efficiency in batches, which provides reference and new ideas for intel-ligent inspection and operation and maintenance technology of photovoltaic power plants.

关 键 词：光伏板 碎裂状态识别 残差网络 注意力机制 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TM615.2[自动化与计算机技术—计算机科学与技术]