Research on Hot Spot Detection Method of Photovoltaic Panel Based on Thermal Infrared Image
Photovoltaic(PV)panels,which have been in outdoor environment for a long time,are highly susceptible to hot spot effect due to stain shading,which in turn affects the safe and efficient operation of PV power plants.Aiming at this problem,research is carried out on hot spot detection based on traditional image processing and machine learning-based target detection algorithm.Based on traditional image processing,the region segmentation algorithm and edge detection algorithm are utilized to test and study the effect of hot spot detection.Based on machine learning,an improved you only look once version 4(YOLOv4)hot spot detection method is proposed.In this case,the dataset is obtained by field photographing the hot spots of PV panels paired with simulated hot spots.The experimental results show that the improved YOLOv4 model achieves 92.31%intersection over union(IoU)of hot spot detection indexes and 93.42%average precision(AP)in the dataset,which are both better than the effect of YOLOv4 model.This research has certain value for engineering application.
Photovoltaic(PV)panel hot spotTraditional image processingMachine learningTarget detectionYou only look once version 4(YOLOv4)Fault detection
NETL
NSTL
万方数据
