基于残差网络的光伏红外图像热斑识别方法

Thermal Spot Recognition of Photovoltaic Infrared Image Based on Residual Network

为了更有效地实现光伏阵列中热斑故障的识别,提出了一种残差结构和多尺度卷积的卷积神经网络模型算法。首先对光伏红外图像进行图像滤波,消除干扰信息,然后用等距分割方法提取光伏组件单元,最后利用改进的模型提取丰富的图像特征,完成红外图像热斑识别。实验结果表明,改进后的算法识别准确率优于原来的模型算法。

The harsh environment of photovoltaic array leads to frequent faults and is not easy to detect.In order to identify hot spot faults in photovoltaic array more effectively,a convolution neural network model algorithm with residual structure and multi-scale convolution was proposed.Firstly,the photovoltaic infrared image was filtered to eliminate the interference information.Then the equidistant segmentation method was used to extract the photovoltaic module unit,and the improved model was used to extract rich image features to complete the infrared image hot spot recognition.The experimental results showed that the recognition accuracy of the improved algorithm was better than that of the original model algorithm,which is of great significance to improve the automation level of photovoltaic fault detection.