基于机器学习的光伏组件热斑图像检测方法研究

Research on Machine Learning Based Hot Spot Image Detection Method for Photovoltaic Modules

摘要:

热斑效应是造成光伏组件损坏的主要原因之一，为提早发现光伏组件热斑效应，利用机器学习算法优化设计光伏组件热斑图像检测方法。根据光伏组件的工作原理以及热斑的产生机理，设置光伏组件热斑图像检测标准。采用红外成像的方式采集光伏组件热斑图像，通过颜色空间转换、滤波去噪、背景干扰去除和图像增强等步骤，实现初始图像的预处理，利用机器学习中的卷积神经网络算法，从轮廓和颜色两个方面提取热斑图像特征，根据特征匹配结果得出光伏组件热斑图像的检测结果。通过性能测试实验得出结论:与传统检测方法相比，优化设计方法的漏检率和误检率明显更低，且光伏组件热斑面积检测误差较低。

Abstract:

The hot spot effect is one of the main causes of damage to photovoltaic modules. In order to detect the hot spot effect of photovoltaic modules in advance, machine learning algorithms are used to optimize the design of photovoltaic module hot spot image detection methods. According to the working principle of photovoltaic modules and the mechanism of hot spot generation, set the detection standards for hot spot images of photovoltaic modules. The infrared imaging method is used to collect the hot spot image of photovoltaic module, and the initial image pre-processing is achieved through color space conversion, filtering denoising, background interference removal, image enhancement and other steps. The convolutional neural network algorithm in machine learning is used to extract the hot spot image features from both contour and color, and the detection results of the hot spot image of photovoltaic module are obtained according to the feature matching results. Through performance testing experiments, it is concluded that compared with traditional detection methods, the optimized design method has significantly lower missed detection and false detection rates, and the detection error of photovoltaic module hot spot area is lower.