基于改进自私羊群算法的光伏红外热图像热斑识别方法

Hot Spot Recognition Method of Photovoltaic Infrared Thermal Image Based on Improved Selfish Herd Algorithm

针对光伏发电中红外热图像识别准确率低,泛化能力差的问题,提出一种基于红外热图像与改进自私羊群算法的热斑识别方法。首先模仿深度学习分类训练的过程制作数据集,然后基于高斯分布提出一种热斑识别函数,接着将改进生存价值和训练过程后的自私羊群算法使用数据集对热斑识别函数中的位置参数进行寻优,随后导入各类测试图片,在经过双边滤波后使用热斑识别函数进行逐点计算,最后将计算的结果进行阈值分割,得到热斑检测结果图。实验结果表明,该模型训练得到的热斑识别函数由于高斯分布的集中性可以有效地对热斑进行诊断,同时抑制边缘干扰和突出细节特征。由于改进自私羊群算法优异的寻优能力,可以极大地提高模型的寻优效率,为基于红外热图像的光伏热斑识别提供了一种新的思路和方法。

Aiming at the problems of low accuracy and poor generalization ability of infrared thermal image recognition in photovoltaic power generation,a hot spot recognition method based on infrared thermal image and improved selfish sheep algorithm was proposed.By imitating the deep learning classification training process,datasets were made..Based on the gaussian distribution,a hot spot recognition function was presented.The survival value after the selfish herd algorithm was improved by using datasets of hot spot recognition function of the location parameters optimization.Then,all kinds of test images were imported,after bilateral filtering using hot spot recognition function point by point calculation.Finally,the calculated results were segmented by threshold,and the hot spot detection results were obtained.The experimental results show that the hot spot recognition function trained by this model could effectively diagnose hot spots due to the concentration of Gaussian distribution,and at the same time suppress edge interference and highlight details.Because of its excellent optimization ability,the selfish sheep algorithm could greatly improve the convergence rate of the model,and provide a new idea and method for photovoltaic hot spot recognition based on the infrared thermal image.