Intelligent Detection Of Hotspots Of Photovoltaic Modules Based On Infrared Images

In the last two decades,the photovoltaic(PV)industry has developed rapidly.The installation and production of photovoltaic plants increased widely.As the PV system operation time grew,more and more defect occurred on the PV modules,and different types of faults have gradually appeared.These faults not only affect the efficiency of power generation,as well as bring hidden safety hazards which may cause safety accidents such as fire disaster.Therefore,the operation and maintenance and inspection of photovoltaic power plants are very important.At present,due to the convenience,quickness,and low price of unmanned aerial vehicles(UAV),the method of inspecting photovoltaic plants through infrared cameras carried by UAVs has attracted increasing attention from researchers.In this paper,two methods are proposed for the recognition and detection of photovoltaic modules in UAV infrared images towards different situations.One is the traditional image processing algorithm,which extracts photovoltaic modules through template matching,image pyramid structure,Hough line detection,and interference line filtering.Then use edge detection and other algorithms for fault detection and filter out environmental interference through near-circle rate,component gray variance,etc.,finally try to adapt parameters according to shooting environment.The second is an object detection algorithm based on deep learning.Based on Faster-RCNN and SSD algorithm,it is optimized using ResNet18 backbone network,feature fusion method,Focal Loss function and other methods.The data set was collected and labeled.The method of pre-training and fine-tuning was used to solve the problem of insufficient data set.Models with different sizes and different hardware requirements are given for different platforms.Field tests on photovoltaic power plants in different environments verify the effectiveness of the algorithm.