基于机器学习算法的光伏组串故障诊断技术研究

Research on Photovoltaic String Fault Diagnosis Technology Based on Machine Learning Algorithms

常规缺陷检测方法主要依据光伏电站面板的异常状态数据来检测面板缺陷,但检测结果存在一定的随机性,导致缺陷检测结果不够清晰。因此,本文采用了无人机影像技术来设计光伏电站面板缺陷检测方法。首先,本文从图像中提取出缺陷特征,然后结合无人机影像技术,通过灰度共生矩阵将缺陷图像与完整图像分开,以识别可见光图像中的缺陷位置。接着,我们将缺陷图像放入光伏面板缺陷检测模型中进行进一步检测,使图像纹理特征和形状特征高度融合,从而实现光伏电站面板缺陷的精准检测。最后,本文采用对比实验的方式验证了该检测方法具有检测置信度和检测精准度更高的优点,使其能够应用于实际生活中。

Traditional defect detection methods for photovoltaic(PV)panels rely on abnormal state data of PV station panels,but the results often suffer from randomness,leading to unclear defect detection outcomes.Addressing this issue,this paper designs a PV panel defect detection method using unmanned aerial vehicle(UAV)imaging technology.Initially,defect features are extracted from images.Combining UAV imaging technology,the gray-level co-occurrence matrix is employed to distinguish defect images from complete ones,identifying defect locations in visible light images.Subsequently,these defect images are further analyzed using a PV panel defect detection model,which integrates image texture and shape features for precise defect identification in PV station panels.Finally,comparative experiments demonstrate that this detection method offers higher confidence and accuracy in detection,making it applicable in real-world scenarios.