基于深度卷积神经网络的光伏组件热斑检测

热斑效应是造成光伏组件损坏的主要原因之一,提早发现光伏组件热斑效应并及时解决,可有效减少损失.该文针对热斑效应问题提出改进的Faster R-CNN红外热斑图像检测方法,该方法是基于SpotFPN多尺度特征学习模块,将SpotFPN应用在二阶段目标检测网络中提高了模型的检测精度,改善热斑的识别准确率.同时为解决数据集不...

SOLAR PHOTOVOLTAIC MODULES HOT SPOT DETECTION BASED ON DEEP CONVOLUTIONAL NEURAL NETWORKS

Hot spot problem is one of the main causes of damage to solar photovoltaic panels.It is helpful to find hot spot in time to solve this problem.In this study,an improved Faster R-CNN detection method is proposed for infrared hot spot image detection,which can speed up the training and identification of model that quickly and accurately locate hot spots.First,we are using synthetic infrared hot spot dataset to solve the problem of insufficient infrared hot spot dataset;Second,we propose SpotFPN multiscale feature learning module to improve the model accuracy.The application of SpotFPN in the two-stage object detection network can improve the detection accuracy of the model and the recognition accuracy of hot spots.At the same time,in order to solve the over-fitting problem caused by insufficient dataset,the data enhancement technology is used to effectively expand the infrared hot spot dataset.The model uses pretraining weights to learn the hot spot dataset,and its average detection accuracy higher about 3%than before.