Defect Detection System For Pv Module Based On Electroluminescence Imaging

With the continuous development of the new energy industry,the penetration rate of photovoltaic power generation is constantly improving,and the installed scale of photovoltaic power generation is rapidly expanding.The maintenance technology and maintenance cost of photovoltaic power station become one of the reasons that restricting its rapid development.Photovoltaic modules are susceptible to invisible defects that are not easily detected by the naked eye and are a common type of defect in operation and maintenance of photovoltaic power station.With electroluminescence(EL),high quality images of the interior of a module can be acquired.Therefore,EL defect detection is widely used for module interior defect detection.Current EL defect detection technology is mainly used in the process of PV module production.The collected module images have high quality and fixed angle.The computational performance of operating environment and the number of data sets could meet the requirements of the training neural network.However,at the photovoltaic power station,the collected images have low quality and random angle.The computing performance of the hardware device and the amount of data do not support the training and testing of neural network.More importantly,realtime detection results should be provided at the photovoltaic power station.In order to solve these problems,a solution is proposed which detects in real time at the photovoltaic power station and detects remotely in the cloud,respectively.First of all,this paper analyzes the demand of EL defect detection system at photovoltaic power station: real-time,practical and easy to operate.Then the solution that detects in real time at the photovoltaic power station and detects remotely in the cloud is proposed to detect EL defect.Next,the image acquisition scheme of electroluminescent(EL)imaging technology is determined,and the selection of power supply and camera for the image acquisition system is introduced.Real-time detection of EL defects was carried out at photovoltaic power station based on modules.According to the requirements of EL defect in module production process and the influence of EL defects on the output power of modules,the types of EL defects at the photovoltaic power station are determined as dark cells,light and shade cells,deep cracks isolating cell parts and dark spots.By using traditional digital image processing technology,module image segmentation and correction,image enhancement,connectivity detection are carried out to solve the problems of low quality and random angle of module images,and achieve real-time detection of EL defects.The algorithm can detect the EL defects of the module images collected from the roof photovoltaic power station in less than 110 ms.Tested in the images of other modules provided by the partner,the accuracy of EL defect detection is 95.45%,and the average time to detect an image is 110.17 ms.The remote detection of EL defects based on solar cells in the cloud is aimed at microcrack defects,which are easy to expand into dark cells and splinters that affects the output power of modules.It is necessary to give early warnings to the microcrack modules and make them as the key inspection target of the next operation and maintenance.Microcrack defects have the problems of inconspicuous features and difficult extraction,and there are few samples of onsite microcracked modules.Therefore,a solution is proposed to cut the module images to solar cells and use a deep neural network based on migration learning to train the model with the dataset collected in module production process.Then the model is fine-tuned and transferred to classify the class of dataset collected at the power station.The solar cells with or without microcracks are reclassified,and the cracked solar cells are labeled in the modules.The algorithm requires high computational performance and needs to run in a server in the cloud.The classification accuracy of the algorithm is 77%.There are certain false detections leading to false alarms,but the leakage rate is low,which can effectively warn microcracked modules and not miss defective modules.Finally,the real-time EL defect detection software is implemented based on Python and Py Qt5.The main interface of the software,image acquisition and display function,parameter setting function,function of calling EL defect detection algorithm are developed.The joint debugging with hardware equipment is carried out,which is demonstrated in the roof photovoltaic power station in the school.The remote microcrack detection algorithm based on battery cells was tested in a cloud server.The EL defect detection system designed in this paper solves the problems of low quality of images and the shortage of datasets in the power station,meets the demand of real-time detection,improves the efficiency of microcrack detection,and improves the level of automation and intellectualization of the operation and maintenance in the photovoltaic power station.