Abstract: For the hot spot of PV modules caused by local shadow shading,a hot spot diagnosis method based on Complete Neighborhood Preserving Embedding(CNPE)was proposed.The method used flow learning to extract nonlinear features from PV module operation data and achieved highly sensitive detection of module status without the help of infrared images.Compared with traditional methods,the proposed method could reflect the module operation status more accurately and then realize fault detection and classification.Relying on the experimental platform of the State Grid Electric Power Research Institute,the performance of the proposed method with 15 groups of hot spot samples of different degrees was validated and compared,and the result showed that the proposed method could effectively achieve fault detection and evaluation with good results.

Key words: data-driven, photovoltaic module, hot spot, evaluation function

摘要： 针对局部阴影遮挡引起的光伏组件热斑,提出一种基于完全邻域保持嵌入的热斑故障诊断方法。该方法采用流形学习从光伏组件运行数据中提取非线性特征,在不借助红外图像的前提下实现组件状态的高灵敏检测。所提方法较传统方法能够更精准地反映组件运行状态,进而实现故障检测与故障分级。依托国家电网电力科学研究院实验平台,通过对15组不同程度的热斑故障样本进行性能验证与对比,所提方法能够有效实现故障检测与评估。

关键词: 数据驱动, 光伏组件, 热斑故障, 评估函数