Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/574168985/awesome_PV_faults_diagnosis
literature review renewal of PV faults diagnosis, including papers, datasets, and codes
https://github.com/574168985/awesome_PV_faults_diagnosis
List: awesome_PV_faults_diagnosis
Last synced: 16 days ago
JSON representation
literature review renewal of PV faults diagnosis, including papers, datasets, and codes
- Host: GitHub
- URL: https://github.com/574168985/awesome_PV_faults_diagnosis
- Owner: 574168985
- License: mit
- Created: 2023-11-30T13:42:35.000Z (about 1 year ago)
- Default Branch: main
- Last Pushed: 2024-02-22T21:21:50.000Z (10 months ago)
- Last Synced: 2024-05-21T23:00:31.267Z (7 months ago)
- Size: 275 KB
- Stars: 13
- Watchers: 1
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.cn.md
- License: LICENSE
- Citation: citationsMémorables.md
Awesome Lists containing this project
- ultimate-awesome - awesome_PV_faults_diagnosis - Literature review renewal of PV faults diagnosis, including papers, datasets, and codes. (Other Lists / Monkey C Lists)
README
[](https://github.com/574168985/awesome_PV_faults_diagnosis/edit/main/README.md)
[](https://github.com/574168985/awesome_PV_faults_diagnosis/edit/main/README.fr.md)
[](https://github.com/574168985/awesome_PV_faults_diagnosis/edit/main/README.cn.md)
# 光伏组件故障监测
光伏组件故障监测文献综述，持续更新中，包括文献、数据集、代码
# 更新计划
在本文献综述中，我们关注:
- [ ] [基于图像的故障监测方法](#基于图像的故障监测方法)
- [ ] [基于多模态融合的方法](#基于多模态融合的方法)
- [ ] [基于图像方法的公开数据集](#基于图像方法的公开数据集)
# 基于图像的故障监测方法
图像类型, 图像的收集方法, 图像的处理方法, 故障类型及严重程度, 所应用的层次(cell, module, string, plant), 实验结果(仿真/实际实验), 准确率, 以及其他
加一个纵轴数据量 横轴准确率的方图
加一个ipynb demo
加上defect替换fault关键词搜索
加上可研究的点，想要实现的经济效果boge那一套
# 基于多模态融合的方法
图像类型, 图像的收集方法, 图像的处理方法, 如何融合, 故障类型及严重程度, 所应用的层次(cell, module, string, plant), 实验结果(仿真/实际实验), 准确率, 以及其他
其他的相关的文献综述文章
融合方法难找 从综述下手找已有归类或者从相对著名文章中筛选
# 基于图像方法的公开数据集
待续
## 【2023】基于图像的故障监测方法
- [|International Journal of Photoenergy，中科院4区] [|可见光图像] [|无人机] [|模糊、翻转、旋转、噪声、平移、缩放和扭曲] [|正常板块、蜗牛痕迹、层裂、玻璃破裂、变色和烧伤] [|光伏阵列] [|初始600，扩增到3150] [|CNN+Classifier(不使用AI分类器)] [|宣称在自建数据集上使用DenseNet-201+k-nearest达到100] [Enhancing Photovoltaic Module Fault Diagnosis with Unmanned Aerial Vehicles and Deep Learning-Based Image Analysis](https://www.hindawi.com/journals/ijp/2023/8665729/) **[报告|多分类任务中，非全连接分类器中k-nearest最优]**
- [|International Journal of Photoenergy，Q4] [|Image en lumière visible] [|drone] [|Flou, retournement, rotation, bruit, translation, zoom et distorsion] [|Plaques normales, marques d'escargots, fissures de couches, verre brisé, décoloration et brûlures] [|pannel] [|Initialement 600, étendu à 3 150] [|CNN+Classifier (sans utiliser le classificateur AI)] [|Prétend utiliser DenseNet-201+k-nearest sur un ensemble de données auto-construit pour atteindre 100] [Enhancing Photovoltaic Module Fault Diagnosis with Unmanned Aerial Vehicles and Deep Learning-Based Image Analysis](https://www.hindawi.com/journals/ijp/2023/8665729/) **[Rapport | Dans les tâches multi-classifications, k-le plus proche est le meilleur parmi les classificateurs non entièrement connectés]**
- [|IEEE Transactions on Industry Applications，中科院2区] [|红外成像+电致发光成像] [|?] [|?] [|11/2] [|光伏阵列] [|20000/2,624] [|MPViT（Multi-Path Vision Transformer for Dense Prediction）+自注意力层] [|宣称在dataset1和dataset2上的二元分类和多类分类实验中分别取得了94.1%和88.5%、90.7%和86.4%的准确度] [Photovoltaic Panel Fault Detection and Diagnosis Based on a Targeted Transformer-Style Model](https://ieeexplore.ieee.org/abstract/document/10274093) **[报告|局部细节特征可以提高模型对不同光伏电池板图像类别的区分能力]**
- [|Aswan University Journal of Sciences and Technology，中科院未收录] [|红外图像] [|实验室] [|？] [|？] [|光伏阵列] [|？] [|？] [|宣称在自建数据集上使用达到99.98%] [Solar Cell Anomaly Detection Based on Wavelet Scattering Transform and Artificial Intelligence](https://aujst.journals.ekb.eg/article_312683.html) **[报告|用小波散射变换（WST）更好地表示图像特征来处理红外图像中的多类异常检测]**
- [|Information Systems Research in Vietnam，中科院未收录] [|电致发光图像] [|？] [|？] [|？] [|光伏阵列] [|？] [|？] [|？] [Case Study: Utilising of Deep Learning Models for Fault Detection and Diagnosis of Photovoltaic Modules to Improve Solar Energy Constructions’ O&M Activities Quality](https://link.springer.com/chapter/10.1007/978-981-99-4792-8_5) **[报告|无监督学习用于检测和诊断模块中存在的视觉故障和缺陷]**
- [|Computational Intelligence and Neuroscience no longer accepts submissions，中科院未收录] [|电致发光图像] [|？] [|？] [|4] [|光伏电池] [|2,624] [|？] [|？] [Fault-Level Grading of Photovoltaic Cells Employing Lightweight Deep Learning Models](https://www.hindawi.com/journals/cin/2023/2663150/) **[报告|基于CPU的提出模型]**
- [|IET Renewable Power Generation，中科院4区] [|红外图像] [|？] [|？] [|？] [|光伏阵列] [|？] [|MobileNet-V3] [|？] [Automatic defect identification of PV panels with IR images through unmanned aircraft](https://ietresearch.onlinelibrary.wiley.com/doi/full/10.1049/rpg2.12831) **[报告|MobileNet V3>传统的卷积神经网络（CNN）]**
- [|Sustainability，中科院3区] [|红外图像] [|？] [|？] [|七种缺陷，包括积聚在PV模块上的沙子、覆盖的PV模块、破裂的PV模块、降解、脏的PV模块、短路的PV模块和过热的旁路二极管] [|光伏阵列] [|？] [|混合CNN–ML模型] [|？] [Embedded Hybrid Model (CNN–ML) for Fault Diagnosis of Photovoltaic Modules Using Thermographic Images](https://www.mdpi.com/2071-1050/15/10/7811) **[报告|Raspberry Pi 4部署，这种嵌入式解决方案在PV模块的诊断中是可行的]**
- [|MULTIMEDIA SYSTEMS，中科院4区] [|红外图像] [|？] [|？] [|热点] [|光伏阵列] [|？] [|改进的语义分割模型（Attention DeepLab）当作目标检测任务] [|宣称89.6%] [Hotspot defect detection for photovoltaic modules under complex backgrounds](https://link.springer.com/article/10.1007/s00530-023-01176-w) **[报告|在预测网络中添加一个用于微小热点缺陷的预测头。/通过合并多尺度特征图来修改特征融合网络中的路径聚合网络，以增强语义信息。/在特征融合网络中应用高效的通道注意模块，以消除混淆。]**
- [|Soft Computing，中科院3区] [|可见光图像] [|？] [|] [|] [|光伏阵列] [|？] [|YOLOv7-GX] [|宣称93.4%] [YOLOv7-GX-based defect detection for PV panels](https://www.researchsquare.com/article/rs-3118568/v1) **[报告|在训练过程中，为不同的图像和类别分配不同的权重，以解决样本分布不平衡的问题]**
- [|arxiv] [|可见光图像] [|无人机] [|] [|热点、蜗牛痕迹、裂缝] [|光伏阵列] [|模拟和真实图像数据集] [|无监督传感算法和3D增强现实可视化] [|] [Unveiling the Invisible: Enhanced Detection and Analysis of Deteriorated Areas in Solar PV Modules Using Unsupervised Sensing Algorithms and 3D Augmented Reality](https://arxiv.org/abs/2307.05136) **[报告|无监督传感算法与3D AR可视化的新组合]**
- [|JOURNAL OF CIRCUITS SYSTEMS AND COMPUTERS，中科院4区] [|可见光图像] [|无人机] [|] [|缺陷检测和颜色分类] [|] [|] [|U-Net] [|] [A Deep Learning-Based Surface Defects Detection and Color Classification Method for Solar Cells](https://www.worldscientific.com/doi/abs/10.1142/S0218126623501566) **[报告|分割小对象，可以实现对表面缺陷的检测]**
- [|Energies，中科院4区] [|] [|] [|使用K均值聚类算法来精化数据集] [|热点、裂缝和灰尘] [|] [|] [|轻量级的Inception残差卷积网络（LIRNet）] [|](https://www.mdpi.com/1996-1073/16/5/2112) **[报告|1、使用两阶段深度学习分类提高了训练稳定性并略微提高了识别准确性。2、通过减少学习类别数量以提高识别准确性]**
- [|Chinese Control Conference (CCC)] [|光伏图像] [|无人机] [|] [|] [|] [|] [|在YOLOv5的基础上用CPP-Neck替代了PANet颈网络。CPP-Neck包括CBPN和PA结构，用于增强主干网络特征信息的融合，同时提高在复杂背景中缺陷边界框的定位精度] [|宣称在mAP50方面取得了6.54%的改进] [YOLOv5-CPP: Improved YOLOv5-Based Defect Detection for Photovoltaic Panels](https://ieeexplore.ieee.org/abstract/document/10241184) **[报告|目标检测模型能够解决光伏红外图像中目标纹理特征信息较弱、图像背景复杂以及对比度较低等问题]**
- [|IEEE Access，中科院3区] [|红外图像] [|] [|] [|] [|] [|AI Hub提供的热成像数据集] [|为了解决在使用热图像时边缘信息丢失的问题，引入了新的对比增强模块，采用对比增强技术] [|宣称96.17%的准确率] [CECvT: Initial Diagnosis of Anomalies in Thermal Images](https://ieeexplore.ieee.org/abstract/document/10264866) **[报告|为了解决热图像中常见的边缘信息降级的问题，引入了一个对比增强模块]**
- [|10th International Maritime Science Conference (IMSC 2023)] [|可见光图像] [|] [|] [|光伏（PV）电池板的短路电流（Isc）和最大功率点功率（Pmpp）] [|] [|] [|高斯过程回归（GPR）、支持向量机（SVM）、随机森林（RF）和神经网络（NN）] [|] [Using Machine Learning Techniques for Predicting Electrical Data of PV Panels from RGB Images](https://www.researchgate.net/profile/Ilija-Knezevic-2/publication/371672199_Using_Machine_Learning_Techniques_for_Predicting_Electrical_Data_of_PV_Panels_from_RGB_Images/links/648eb2fac41fb852dd0dada8/Using-Machine-Learning-Techniques-for-Predicting-Electrical-Data-of-PV-Panels-from-RGB-Images.pdf) **[报告|利用机器学习（ML）技术从RGB图像中预测PV电池的电气数据]**
- [|Journal of Optics-India，中科院未收录] [|可见光图像] [|] [|] [|机械缺陷如裂纹和针孔] [|] [|] [|利用图像处理和模糊逻辑技术进行缺陷检测。图像处理技术涉及阈值处理、数学形态学和边缘检测算子] [|] [Defect detection on solar cells using mathematical morphology and fuzzy logic techniques](https://link.springer.com/article/10.1007/s12596-023-01162-5) **[报告|使用Mamdani模糊模型，提高对个体和群体缺陷的识别准确率]**
- [|Journal of Optics-India，中科院未收录] [|] [|] [|] [|] [|] [|] [|使用多次形态学和带有适应参数的Canny边缘检测来提取和突显太阳能电池上的对象。然后，使用其特征和分类算法对检测到的对象进行分类，并检测不同的缺陷类型和组件] [|] [Automated visual inspection of solar cell images using adapted morphological and edge detection algorithms](https://link.springer.com/article/10.1007/s12596-023-01284-w) **[报告|视觉检测方法是无接触方法，优于有接触方法]**
## 【2022】基于图像的故障监测方法
- [|Chinese Control and Decision Conference (CCDC)] [|红外图像] [|无人机] [|定位和提取，旋转水平化，去除透视畸变，聚类分割。显著性] [|热点] [|光伏阵列] [|拍摄图片135，从中提取了1020个光伏模块，有96个正样本，表示这些PV模块样本中存在热点。另外，有924个负样本，表示这些PV模块是正常的] [|网格分割->聚类分割] [|宣称在自建数据集达到99.71%] [Hotspot detection of photovoltaic modules in infrared thermal image based on saliency analysis](https://ieeexplore.ieee.org/abstract/document/10033497) **[报告|二分类任务中，聚类分割>网格分割]**
- [|ICESEP-2022] [|红外图像] [|无人机] [|改变热点的位置和大小，裁剪出带热点的部分后对齐大小。报告旋转和翻转无用] [|热点] [|光伏阵列] [|初始39张带有热点的图像和235张没有热点的图像，扩增到292张带有热点的图像和292张没有热点的图像] [|CNN+Classifier(全连接分类器)] [|宣称在自建数据集达到96.58 ] [Hot Spot Detection of Photovoltaic Module Infrared Near-field Image based on Convolutional Neural Network](https://iopscience.iop.org/article/10.1088/1742-6596/2310/1/012076/meta) **[报告|二分类任务中，全连接分类器>朴素贝叶斯分类器]**
- [|IEEE Journal of Photovoltaics，中科院3区] [图像类型|电致发光] [|激光线扫描] [|图像堆栈->分类为背景、电致发光或激光像素->重建无接触传统电致发光,强调缺陷电致发光->校准] [|?] [|光伏阵列] [|?] [|?] [|?] [Reconstruction and Calibration of Contactless Electroluminescence Images From Laser Line Scanning of Photovoltaic Modules](https://ieeexplore.ieee.org/abstract/document/9736329) **[报告|提出了一种基于激光线扫描的无触点电致发光（EL）图像处理方法，生成两种类型的图像：对应于电压偏置EL（EBEL）的CCEL和突出缺陷的CHEL]**
- [|Sustainable Energy Grids & Networks，中科院3区] [|电致发光图像] [|暗室] [|？] [|有缺陷的(功率损失超过初始功率输出的3%被视为有缺陷的电池)，开裂，腐蚀] [|光伏阵列] [|制备了两个数据集，2624/1028] [|CNN+Classifier(SVM)] [|宣称在自建数据集上达到99.49] [A combined convolutional neural network model and support vector machine technique for fault detection and classification based on electroluminescence images of photovoltaic modules](https://www.sciencedirect.com/science/article/pii/S2352467722001916) **[报告|1、提出两个数据库D1（缺陷和正常）和D2（正常，裂纹和腐蚀）进行训练和评估，这两个数据库包含光伏电池的电致发光图像。2、位于电池角落的缺陷，这被认为是PV电池的一部分而不是缺陷]**
- [|IEEE Transactions on Industrial Informatics，中科院1区] [|电致发光] [|生产制造] [|？] [|10] [|光伏电池] [|36,543] [|] [|宣称在自建数据集上达到99.49] [PVEL-AD: A Large-Scale Open-World Dataset for Photovoltaic Cell Anomaly Detection](https://ieeexplore.ieee.org/abstract/document/9744494/algorithms?tabFilter=dataset#algorithms) **[报告|提供公开数据集。第一个为光伏太阳能电池异常检测提供框架真值的公共数据集（目标检测任务）]** 重点
- [|mdpi data，中科院未收录] [|可见光图像] [|] [|？] [|动态重配置和最大功率点跟踪] [|光伏电池] [|5211] [|采用基于图像的方法来估算光伏板的电特性] [|] [Dataset for Detecting the Electrical Behavior of Photovoltaic Panels from RGB Images](https://www.mdpi.com/2306-5729/7/6/82) **[报告|介绍了一个数据集，将RGB图像与不同辐照度和阴影条件下的光伏板电数据进行关联；此外，数据集还提供了补充的天气数据和其他图像特征，以支持估算模型的训练。具体而言，该数据集被设计用于支持基于图像的电数据估算模型的设计，以替代大量的传感器]**
- [|Chinese Automation Congress (CAC)] [|电致发光] [|] [|？] [|裂纹和碎片] [|光伏电池] [|402] [|改进YOLOv5s，通过整合空间和通道信息使网络专注于图像中的关键区域；通过增强通道注意力来减少混叠效应] [|宣称在构建的缺陷图像数据集上实现了0.923的平均精度（mAP），与原始YOLOv5s算法相比精度提升了3.3%] [Defect detection for PV Modules based on the improved YOLOv5s](https://ieeexplore.ieee.org/abstract/document/10055183) **[报告|通过引入视觉注意机制改进了YOLOv5s网络中的交叉阶段部分网络操作单元和空间金字塔池化操作单元]**
- [|IEEE International Conference on Power and Power and Renewable Energy (ICPRE)] [|红外] [|] [|对已有的太阳能电池故障公共数据集进行灰度处理和数据增强] [|10] [|] [|36,543] [|引入SENet注意力机制到轻量级网络ShuffleNetV2中] [|宣称在公共数据集上达到84.06%] [Fault Diagnosis Method for Photovoltaic Panels Based on Improved ShuffleNet V2 and Infrared Images](https://ieeexplore.ieee.org/abstract/document/9960396) **[报告|由于光伏模块故障种类繁多且特征差异较小，导致个别故障类型诊断效果较差]**
- [|Elsevier Energy，中科院1区] [|红外] [|] [|转换和校正、去除银网、非线性插值、等效分割和聚类] [|面板表面不均匀灰尘积累] [|光伏电池] [|36,543] [|提出了一个新的误差评估方法，称为“误差循环”，用于分析测量结果与实验结果之间的一致性。] [|宣称R2和平均绝对误差（MAE）分别为78.7%和3.67] [A deep residual neural network identification method for uneven dust accumulation on photovoltaic (PV) panels](https://www.sciencedirect.com/science/article/pii/S0360544221025500) **[报告|提出了一种新的光伏面板表面不均匀灰尘积累的识别方法，用于定量分析灰尘状态（浓度和分布）；仍需考虑：灰尘浓度和类型以及不同照明条件的影响]**
- [|onlinelibrary.wiley] [|红外] [|] [|边缘检测和霍夫变换] [|10] [|光伏电池] [|36,543] [|算法分为两个阶段。在第一阶段，进行太阳能电池板局部化，并进行特征提取和分析。在第二阶段，分析光伏电池板表面积退化对光伏系统功率输出的影响。] [|] [Photovoltaic Module Fault. Part 1: Detection with Image Processing Approaches-Chapter 3](https://onlinelibrary.wiley.com/doi/abs/10.1002/9781119873785.ch3) **[报告|分析光伏电池板表面积退化对光伏系统功率输出的影响]**
- [|Journal of Quality Technology，中科院3区] [|红外] [|无人机] [|使用Transform Invariant Low-rank Textures（TILT）方法来处理光伏图像的非平滑背景，然后应用边缘检测来剪裁和对齐图像；图像去噪和分割的后处理程序] [|] [|] [|20个正常（无异常）的训练样本和100个测试样本] [|使用Robust Principal Component Analysis (RPCA)方法，该方法可以从低秩背景中分离出稀疏损坏的异常成分，从而实现同时检测和隔离异常] [|宣称在自建数据集上召回率为0.80，并检测到重要异常的最大召回率为1。在F1分数方面，该方法比两种基准方法分别提高了44.5%和114.3%] [Online automatic anomaly detection for photovoltaic systems using thermography imaging and low rank matrix decomposition](https://www.tandfonline.com/doi/full/10.1080/00224065.2021.1948372) **[报告|在图像底部边缘通常存在正常的热点，它们不应被检测为异常]**
- [|International Journal of Photoenergy，中科院4区] [|普通图像和红外图像] [|生产制造] [|？] [|10] [|] [|] [|] [|] [Infrared Thermal Images of Solar PV Panels for Fault Identification Using Image Processing Technique](https://www.hindawi.com/journals/ijp/2022/6427076/) **[报告|热成像可以利用Hough变换技术快速定位热点]**
- [|International Journal of Photoenergy，中科院未收录] [|红外] [|生产制造] [|？] [|10] [|光伏电池] [|36,543] [|CNN+Classifier(SVM)] [|宣称在自建数据集上达到99.49] [Machine learning framework for photovoltaic module defect detection with infrared images](https://link.springer.com/article/10.1007/s13198-021-01544-7) **[报告|]**
- [|International Journal of Photoenergy，中科院3区] [|红外] [|] [|？] [|] [|] [|] [|使用基于模糊逻辑的边缘检测技术来检测PV模块上异常的方向。此外，灰度共生矩阵用于提取图像的纹理特征。这些提取的特征被标记并与支持向量机分类器进行训练，以对PV模块中的故障类型进行分类] [|宣称94.4%] [Deep learning-based model for fault classification in solar modules using infrared images](https://www.sciencedirect.com/science/article/pii/S221313882200162X) **[报告|利用纹理特征分析和监督式机器学习方法进行光伏（PV）模块缺陷检测]**
- [|International Journal of Photoenergy，中科院未收录] [|红外] [|三脚架上安装的热成像仪] [|？] [|] [|] [|] [|] [|] [An efficient optical inspection of photovoltaic modules deploying edge detectors and ancillary techniques](https://ijece.iaescore.com/index.php/IJECE/article/view/28435) **[报告|实验系统使用安装在三脚架上的红外热成像相机，同时结合实验室内实时测试测得的电气参数]**
- [|Privacy and Data Analytics: Select Proceedings of ISPDA] [|] [|] [|？] [|] [|] [|36,543] [|DenseNet] [|] [Defect Analysis of Faulty Regions in Photovoltaic Panels Using Deep Learning Method](https://link.springer.com/chapter/10.1007/978-981-16-9089-1_5) **[报告|采用DenseNet架构证明其在效率上优于CNN和ResNet-50]**
- [|Energy Reports，中科院2区] [|红外] [|] [|？] [|热点] [|光伏面板] [|36,543] [|改进的YOLOv5] [|宣称平均精度（mAP）达到了87.8％，平均召回率为89.0％，F1分数达到了88.9％] [A novel detection method for hot spots of photovoltaic (PV) panels using improved anchors and prediction heads of YOLOv5 network](https://www.sciencedirect.com/science/article/pii/S235248472201575X) **[报告|通过K均值聚类算法对数据标注框的长度宽度比进行聚类，添加一组较小数值的锚点以优化聚类数，从而提高YOLOv5网络在不同尺度上对PV面板热点的检测精度]**
- [|Advances in Science and Engineering Technology International Conferences (ASET)] [|] [|] [|？] [|] [|] [|几百张] [|迁移学习，MobileNetV2作为特征提取器] [|在elpv-dataset上报告了85%的故障分类准确率，在磁砖和Kolektor Surface-Defect数据集上都报告了92%的故障分类准确率] [Fault Detection and Identification Based on Image Processing and Deep Learning](https://ieeexplore.ieee.org/abstract/document/9734799) **[报告|推广在电力换向器、太阳能模块和磁砖等不同故障检测问题，通向通用故障检测和定位模型]**
- [|Sustainable Energy, Grids and Networks，中科院3区] [|] [|在MATLAB/Simulink环境中设计了一个PV系统，使用从国家技术研究所阿加尔塔拉分部的光伏系统获取的实时辐照度和温度数据。构建的数据集用于提取特征，包括一个新的指标，以在Python 3.7中训练这些算法。] [|？] [|] [|光伏电池] [|36,543] [|分类提升（CatBoost）、轻量级梯度提升方法（LGBM）和极端梯度提升（XGBoost）] [|宣称平均检测和分类准确率分别为99.996%和99.745%] [Performance assessment of selective machine learning techniques for improved PV array fault diagnosis](https://www.sciencedirect.com/science/article/pii/S2352467721001454) **[报告|]**
- [|International Journal of Applied Earth Observation and Geoinformation，中科院2区] [|红外] [|生产制造] [|？] [|10] [|] [|] [|CNN+Classifier(SVM)] [|宣称在自建数据集上达到99.49] [Developing a deep learning-based layer-3 solution for thermal infrared large-scale photovoltaic module inspection from orthorectified big UAV imagery data](https://www.sciencedirect.com/science/article/pii/S0303243421003597) **[报告|LGBM>CatBoost、XGBoost和RF]**
- [|MDPI Electronics，中科院3区] [|] [|] [|？] [|] [|光伏阵列] [|] [|基于邻近像素的虚拟成像（NPBVI）技术] [|] [Neighboring-Pixel-Based Maximum Power Point Tracking Algorithm for Partially Shaded Photovoltaic (PV) Systems](https://www.mdpi.com/2079-9292/11/3/359) **[报告|提出一种最大功率点跟踪（MPPT）方法。在高分辨率图像测试时，不确定性水平增加。这可能是由于从高分辨率镜头捕获图像时考虑到的小图像细节]**
- [|International Conference on Image Processing and Capsule Networks] [|EL图像] [|] [|图像去噪] [|光学或机械缺陷，如微裂纹、裂纹的大小，以及在图像采集过程中引起的电气或电机械干扰的副作用] [|] [|] [|通过语义分割，进行缺陷的检测] [|] [Defect Analysis of Electroluminescence Images of PV CELL](https://link.springer.com/chapter/10.1007/978-3-030-84760-9_57) **[报告|尽管微裂纹的数量与PV电池的效率没有直接关系，但由于裂纹的大小、宽度、位置以及太阳能电池组件的运行年限，微裂纹对测得的输出功率产生了重大影响]**
- [|Remote Sensing，中科院2区] [|红外] [|航拍图像] [|？] [|] [|] [|] [|] [|] [RID—Roof Information Dataset for Computer Vision-Based Photovoltaic Potential Assessment](https://www.mdpi.com/2072-4292/14/10/2299) **[报告|引入了两个新的多类别数据集，分别用于屋顶分割和屋顶结构的语义分割，重点是从航拍图像中提取屋顶信息，尤其是在光伏潜力分析方面的应用。虽然基于2D航拍图像的深度学习光伏潜力分析包含了结构信息，但缺乏坡度和阴影信息。因此，有必要研究从航拍图像学习这些信息或将2D和3D数据合并以进一步提高潜力估计准确性。除了光伏潜力分析，丰富的3D模型还可以推动其他研究和应用，如城市噪声扩散建模、屋顶隔热评估或建筑能源需求估算等方面的发展。]**
- [|Energies，中科院4区] [|红外] [|航空红外热成像（aIRT）] [|？] [|] [|] [|] [|] [|] [Automatic Inspection of Photovoltaic Power Plants Using Aerial Infrared Thermography: A Review](https://www.mdpi.com/1996-1073/15/6/2055) **[报告|综述航空红外热成像（aIRT）框架在光伏电站中自动化不同任务：故障检测和分类，飞行路径规划的优化，简化在现场定位故障的PV电站正射投影，检测模块污染并将其与PV模块上的实际故障区分开]**
- [|Procedia Computer Science，中科院未收录] [|红外] [|] [|？] [|] [|光伏电池板] [|] [|分类+目标检测] [|宣称ResNet-50迁移学习模型获得了最高的85.37%的F1，Faster R-CNN获得67%的MAP] [Solar panel hotspot localization and fault classification using deep learning approach](https://www.sciencedirect.com/science/article/pii/S1877050922008225) **[报告|分别采用resnet-50和faster-rcnn实践分类和目标检测]**
- [|International Journal of Photoenergy，中科院4区] [|可见光图像+红外图像] [|] [|？] [|热点] [|] [|] [|] [|] [Infrared Thermal Images of Solar PV Panels for Fault Identification Using Image Processing Technique](https://www.hindawi.com/journals/ijp/2022/6427076/) **[报告|太阳能电池板中的大多数故障都因内部电阻增加而被记录为热点]**
- [|Remote Sensing，中科院2区] [|红外] [|无人机] [|？] [|] [|光伏电池板] [|] [|分割模型，分别是DeepLabV3+、Feature Pyramid Network（FPN）和U-Net] [|宣称采用DeepLabV3+、FPN和U-Net这三种模型，分别获得了79%、85%和86%的交并比（IoU），以及87%、92%和94%的Dice系数] [Photovoltaics Plant Fault Detection Using Deep Learning Techniques](https://www.mdpi.com/2072-4292/14/15/3728) **[报告|U-Net>DeepLabV3+，FPN]**
- [|Progress in Energy，中科院未收录] [|红外] [|] [|？] [|] [|] [|] [|] [|] [Outdoor luminescence imaging of field-deployed PV modules]([https://iopscience.iop.org/journal/2516-1083](https://iopscience.iop.org/article/10.1088/2516-1083/ac9a33/meta) **[报告|室外电致发光]**
## 【2021】基于图像的故障监测方法
- [|Energy Conversion and Management，中科院1区] [|红外] [|包括通过无人机和地面操作员获取的图像] [|各种预处理策略，包括像素的标准化和均匀化、灰度化、阈值化、离散小波变换以及Sobel Feldman和盒状模糊滤波] [|10] [|] [|分类任务：1000，分割任务：200] [|] [|宣称在自建数据集上达到100] [Deep learning strategies for automatic fault diagnosis in photovoltaic systems by thermographic images](https://www.sciencedirect.com/science/article/pii/S019689042100491X) **[报告|引入数据增强技术和Dropout层以减少过拟合，但根据结果，这些技术并没有提供实质性的改善，甚至可能导致准确性下降几个百分点]**
- [|Fault Detection in aerial images of photovoltaic modules based on Deep learning] [|] [|无人机] [|？] [|伤痕、层裂、变色、玻璃破裂、良好面板和蜗牛痕迹] [|光伏模块] [|] [|卷积神经网络（CNN）] [|] [Fault Detection in aerial images of photovoltaic modules based on Deep learning](https://iopscience.iop.org/article/10.1088/1757-899X/1012/1/012030/meta) **[报告|应用cnn]**
- [|International Conference on Environment and Electrical Engineering (EEEIC)] [|红外] [|无人机] [|？] [|] [|] [|] [|CNN] [|宣称在自建数据集上达到100] [Thermal anomalies detection in a photovoltaic plant using artificial intelligence: Italy case studies](https://ieeexplore.ieee.org/abstract/document/9584494) **[报告|图像通常由合格的操作员逐个检查。人工智能技术的应用使得即使是非专业用户也能够检测过热，并减少系统检查所需的时间]**
- [|Renewable Energy，中科院1区] [|红外] [|] [|为了处理不平衡的数据集，研究采用了欠采样和过采样的方法。通过使用数据增强的过采样，模型的泛化性能相比欠采样提高了约6％。] [|11，一些PV故障（如开裂、污染、植被和阴影）在类别内和类别间的变异性很高，这使得在小样本数的不平衡数据集中正确分类这些故障变得困难] [|] [|] [|] [|] [Automatic fault classification in photovoltaic modules using Convolutional Neural Networks](https://www.sciencedirect.com/science/article/pii/S0960148121010752) **[报告|评估红外热像和机器学习方法的组合应用于PV模块不同缺陷类别分类的挑战和效率；对低变异性故障类别： 将故障类别减少到八个具有较低内部和类别间变异性的类别对自动分类器的性能产生了很大影响，其准确度达到了78.85％；对特定故障类别的易分类性： 研究发现一些故障类别，如二极管组合和二极管，可以导致PV系统最多降低66％的功率，具有非常具体的模式，因此可以通过所提出的模型轻松分类。这种类型的故障可以以超过90％的准确度正确分类。]**
- [|Processes，中科院3区] [|] [|] [|？] [|] [|] [|] [|考虑到光伏发电的非连续性特点，我们构建了相应的非连续回归模型，基于XGBoost算法，并通过k均值聚类的结果进行优化] [|] [Intelligent and Data-Driven Fault Detection of Photovoltaic Plants](https://www.mdpi.com/2227-9717/9/10/1711) **[报告|一种无监督方法，不依赖于标记的故障数据]**
- [|IEEE Journal of Photovoltaics，中科院3区] [|红外] [|] [|？] [|过热区域] [|] [|] [|首先，从具有过热区域的热成像图像中转换出多个大规模图像，以精确检测过热区域目标。然后，从这些图像中提取感兴趣的区域，以界定可能存在过热区域的区域。最后，使用深度联合学习模型从这些区域中识别过热区域的类型和位置] [|] [Automated Overheated Region Object Detection of Photovoltaic Module With Thermography Image](https://ieeexplore.ieee.org/abstract/document/9314927) **[报告|识别单个光伏模块的热成像图像中不同潜在过热区域目标的特定类型和准确位置]**
- [|Proceedings of 2021 International Conference on Autonomous Unmanned Systems (ICAUS 2021)] [|] [|] [|通过Mosaic数据增强方法扩展数据集] [|] [|] [|] [|选择ResNet50作为特征提取网络以提高效率并减少计算量。通过分析卷积特征，引入Attention算法以扩大模型的感受野，并采用FPT结构增强不同层次相同位置和同一层不同位置的特征信息融合。设计了一种基于损失函数最小化的二分图匹配算法，以加强地面实况框分布的重要性，并改进损失函数以缓解对非极值预测边界框的抑制问题] [|] [An Optimized One-Stage Detector for Autonomous PV Defect Detection](https://link.springer.com/chapter/10.1007/978-981-16-9492-9_168) **[报告|优化网络的特征融合部分和损失函数，旨在改善传统一阶段目标检测模型在红外图像中对光伏缺陷的性能不佳]**
- [|SENSORS，中科院3区] [|红外] [|] [|？] [|对PV面板缺陷进行五类分类，包括鸟粪、单一缺陷、拼接、水平排列的串和块] [|光伏面板] [|] [|isolated convolution neural model (ICNM)] [|宣称达到97.62%] [Photovoltaic Panels Classification Using Isolated and Transfer Learned Deep Neural Models Using Infrared Thermographic Images](https://www.mdpi.com/1424-8220/21/16/5668) **[报告|应用了孤立的深度卷积神经网络（ICNM）]**
- [|Energy，中科院1区] [|电致发光图像] [|] [|数据增强、适当的模型大小、L2权重正则化、批量归一化和dropout的组合，旋转、翻转、裁剪、对比度和模糊处理，这些数据增强操作有助于将模型准确度提高约6.5%左右] [|] [|光伏电池] [|] [|使用轻量级卷积神经网络（CNN）] [|宣称93.02%] [CNN based automatic detection of photovoltaic cell defects in electroluminescence images](https://www-sciencedirect-com.ezproxy.universite-paris-saclay.fr/science/article/pii/S0360544219320146) **[报告|使用轻量级卷积神经网络（CNN）在公开数据集【elpv-dataset】上SOTA]**
- [|Energies，中科院4区] [|] [|] [|] [|对地短路和串中导线断路，二极管短路、内部断路和内部寄生串联电阻的降解] [|] [|] [|由诊断算法、重配置算法和分布式开关矩阵三个主要部分组成的自动故障管理系统] [|宣称在计划实验中恢复了超过90%的功率损失，并且在诊断准确性和灵敏性方面达到了100%] [Automated Fault Management System in a Photovoltaic Array: A Reconfiguration-Based Approach](https://www.mdpi.com/1996-1073/14/9/2397) **[报告|基于行为规则的诊断方法仅在样本延迟中处理数据。不需要数据训练，这适用于使用微控制器或物联网设备进行实施]**
- [|Chinese Automation Congress (CAC)] [|红外] [|无人机] [|] [|] [|光伏板] [|] [|先分割，再目标检测] [|] [Deep Learning-based Method for PV Panels Segmentation and Defects Detection with Infrared Images](https://ieeexplore.ieee.org/abstract/document/9728350) **[报告|建立一个基于无人机红外图像的太阳能光伏阵列分割和模块级缺陷检测的流程。在该框架中，先分割，再识别（目标检测而非分类），使用了U2-Net对光伏板进行精确分割，而YOLOv4则被训练用于识别光伏模块的故障信息]**
- [|Energies，中科院4区] [|] [|] [|] [|] [|] [|] [|] [|] [A Review on Artificial Intelligence Applications for Grid-Connected Solar Photovoltaic Systems](https://www.mdpi.com/1996-1073/14/15/4690) **[报告|综述了对太阳能光伏价值链中系统层面使用不同人工智能技术进行各种目的的研究的观察结果：在优化的情况下，进化算法，如粒子群优化（PSO）和遗传算法（GA），被广泛使用，而在时间序列数据的情况下，比如辐照度预测或功率预测，人工神经网络被成功地应用。此外，在控制和维护方面广泛使用的推理模型、数据驱动算法和学习方法被认为更具体适用于某种条件、问题或数据集，并且在相同方面缺乏普遍适用性。然而，即使大多数论文报告取得了巨大成功，也必须注意到这种成功是基于调整特定模型的特定参数，同时保持其他模型的默认参数。此外，基于输入数据提出的一些模型可能并非具有普适性。目前，行业中似乎存在数据不一致的问题，但随着物联网解决方案的出现，大量传感器的部署，无人机在维护中提供的视频流以及自然语言处理技术的出现，缺乏数据的问题可能会消失。]**
- [|Applied Sciences-Basel，中科院4区] [|可见光] [|] [|] [|] [|光伏电池] [|] [|] [|] [Evaluation of Artificial Intelligence-Based Models for Classifying Defective Photovoltaic Cells](https://www.mdpi.com/2076-3417/11/9/4226) **[报告|使用五种不同分类器模型（KNN、SVM、RF、MLP和CNN）的太阳能光伏电池缺陷分类器]**
- [|Applied Sciences-Basel，中科院4区] [|红外和可见光] [|无人机] [|灰度转换、滤波、三维温度表示、概率密度函数和累积分布函数分析] [|] [|光伏电池] [|] [|提出了一个四步的图像分析过程。在第一步中，原始红外图像经过滤波并转换为灰度图像。然后，在第二步中，图像被转换为3D颜色，以放大温差。第三步中建立了热温度分布的直方图，以找出高温区域的百分比。最后一步建立了温度分布的累积图表，以确定可能存在异常或缺陷的电池数量] [|] [Using UAV to Detect Solar Module Fault Conditions of a Solar Power Farm with IR and Visual Image Analysis](https://www.mdpi.com/2076-3417/11/4/1835) **[报告|图像识别可以提高红外图像的清晰度]**
- [|IOP Conference Series: Earth and Environmental Science] [|红外] [|] [|利用掩码处理在原始图像中隐藏无关的背景区域，并使用labelme软件标记图像数据] [|热点] [|光伏电池] [|] [|SegNet] [|] [Efficient Region Segmentation of PV Module in Infrared Imagery using Segnet](https://iopscience.iop.org/article/10.1088/1755-1315/793/1/012018/meta) **[报告|为了提高热点检测的精度，先对光伏模块红外图像进行有效区域分割是非常重要的]**
- [|Journal of Mechanical Engineering Research and Developments，中科院未收录] [|] [|] [|] [|] [|] [|] [|] [|] [Defect detection by automatic control in the photovoltaic panel manufacturing process](https://jmerd.net/Paper/Vol.44,No.5(2021)/258-262.pdf) **[报告|基于Matlab]**
- [| Campus Tecnológico Local San Carlos] [|红外和可见光] [|无人机] [|？] [|部分阴影、污染和电气故障] [|] [|] [|] [|] [Comparison of the effectiveness among different photovoltaic fault detection techniques](https://repositoriotec.tec.ac.cr/handle/2238/14050) **[报告|不同故障检测方法的有效性：定量比较光伏系统中三种故障检测技术的效果：a) 视觉图像（VI），b) 红外热像（IRT），和c) 电气分析（EA）。结果显示，三种技术在检测到的总故障数量上没有显著差异；然而，VI在检测污染方面表现最佳，而在检测电气故障方面表现最差。此外，所有技术中最容易检测到的是部分阴影。因此，为了尽可能多地检测故障，VI应与IRT或EA相结合。可以为fusion-based方法张本]**
- [|Annual Conference of Industrial Electronics Society] [|] [|] [|？] [|局部阴影和低辐照度] [|] [|] [|] [|通过MATLAB/Simulink模拟验证了该方法在各种大气和局部阴影条件下检测故障的有效性] [Power based Fault Detection Method for PV Arrays](https://ieeexplore.ieee.org/abstract/document/9589163) **[报告|针对太阳能光伏系统在均匀和局部阴影条件下的故障]**
- [|Fifteenth International Conference on Quality Control by Artificial Vision] [|红外] [|] [|] [|] [|] [|] [|基于不同特征的方法，通过使用减法网络的预测结果应用投票策略来确定输入图像中的异常电池板] [|构建了两个数据集来评估他们的方法：一个由手动提取的电池板图像构成的清晰数据集，以及一个由自动提取方法提取的包含噪声的数据集。该方法在这两个数据集上都实现了超过90%的分类准确度] [Identify solar panel defects by using differences between solar panels](https://www.spiedigitallibrary.org/conference-proceedings-of-spie/11794/1179415/Identify-solar-panel-defects-by-using-differences-between-solar-panels/10.1117/12.2586911.short#_=_) **[报告|由于正常电池板和有缺陷的电池板在温度特征上可能非常相似，从单个电池板图像中识别异常变得困难]**
- [|arxiv] [|] [|] [|] [|] [|] [|3200个训练样本和800个测试样本（每种类型200个）] [|] [|] [Graph neural network-based fault diagnosis: a review](https://arxiv.org/abs/2111.08185) **[报告|综述了了基于图神经网络（Graph Neural Network，GNN）的故障诊断（Fault Diagnosis，FD），从多个应用领域获取的数据可以通过图的方式进行有效表示，而与传统的故障诊断方法相比，这种表示形式能够实现更优越的性能]**
## 【2023】基于多模态融合的方法
- [|Energies，中科院4区] [|使用SDP处理信号，生成具有特定雪花状特征的图像] [|] [|] [|焊接不良、裂纹和旁路二极管失效] [|光伏电池] [|] [|结合了对称化点图案（SDP）和AlexNet] [|宣称99.8%] [Fault Diagnosis for PV Modules Based on AlexNet and Symmetrized Dot Pattern](https://www.mdpi.com/1996-1073/16/22/7563) **[报告|比较了五种算法的检测性能，分别是三种机器学习算法：SDP+HOG+SVM、SDP+HOG+ENN和SDP+HOG+BPNN，以及两种深度学习算法：SDP+CNN和SDP+AlexNet。这篇论文将信号的微小变化转化为更明显变化的雪花图表。然后，这些图表被输入到训练和测试速度都很快的AlexNet中进行故障诊断]**
- [|Applied Sciences-Basel，中科院4区] [|红外和可见光图像] [|] [|？] [|] [|] [|] [|] [|] [Infrared and Visible Image Fusion: Methods, Datasets, Applications, and Prospects](https://www.mdpi.com/2076-3417/13/19/10891) **[报告|红外和可见光图像融合综述，包含pv方向03年的一篇经典融合文章[Fault diagnosis for photovoltaic array with the technique of infrared/visible image fusion](https://www.spiedigitallibrary.org/conference-proceedings-of-spie/5286/0000/Fault-diagnosis-for-photovoltaic-array-with-the-technique-of-infrared/10.1117/12.539825.short)]**
## 【2022】基于多模态融合的方法
- [|MODERN ELECTRIC POWER，中科院未收录] [|] [|] [|] [|直流串联电弧故障] [|光伏系统] [|] [|] [|] [A Multi-feature Fusion-Based Method to Recognize Series DC Arc Fault in Photovoltaic System](http://xddl.ncepujournal.com/en/article/Y2022/I5/529) **[报告|通过在时间域、频率域和能量熵中提取的特征构建了故障电弧的特征空间]**
- [|Energies，中科院4区] [|使用SDP处理信号，生成具有特定雪花状特征的图像] [|] [|？] [|焊接不良、断裂和旁路二极管失效] [|光伏模块] [|3200] [|将对称化点图案（SDP）与卷积神经网络（CNN）相结合] [|宣称在自建数据集上达到99.88％] [A Fault Detection Method Based on CNN and Symmetrized Dot Pattern for PV Modules](https://www.mdpi.com/1996-1073/15/17/6449) **[报告|]**
## 【2021及以前】基于多模态融合的方法
- [|Journal of Materials Science Research，中科院未收录] [|红外和可见光图像] [|] [|] [|多晶光伏太阳能电池中的划痕和分流] [|] [|] [|] [|] [Fusion of Thermal and Visible Acquisitions for Evaluating Production-borne Scratches and Shunts in Photo-Voltaic PV Cells](https://www.ccsenet.org/journal/index.php/jmsr/article/view/20250) **[报告|红外和可见光图像融合]**
- [|IEEE Transactions on Industrial Informatics，中科院1区] [|电热图（ET）和电致发光（EL）] [|] [|？] [|划痕、断裂网线、表面杂质、隐藏裂缝] [|光伏电池] [|] [|] [|] [Electromagnetic Induction Heating and Image Fusion of Silicon Photovoltaic Cell Electrothermography and Electroluminescence](https://ieeexplore.ieee.org/abstract/document/8736340) **[报告|提出了一种基于L1范数的图像融合规则，用于融合ET和EL图像的稀疏向量，实现了两种波长检测数据的集成和互补]**
- [|IEEE PES Asia-Pacific Power and Energy Engineering Conference (APPEEC)] [|] [|] [|] [|] [|] [|] [|] [|宣称在自建数据集上达到99.49] [Research on Fault Detection of PV Array Based on Data Fusion and Fuzzy Mathematics](https://ieeexplore.ieee.org/abstract/document/5749018) **[报告|多传感器数据融合]**
- [|Ad Hoc Networks 15'，SCI升级版 计算机科学3区，CCF C] [|非图像：传感器+太阳辐照度] [|] [|] [|] [|] [|] [|一种异常检测算法，通过计算可疑数据的平均值对光伏模块的问题进行分类和定位] [|] [Intelligent photovoltaic monitoring based on solar irradiance big data and wireless sensor networks](https://www.sciencedirect.com/science/article/pii/S1570870515001390) **[报告|为了检测监测大数据中的光伏（PV）模块问题，首先开发了一种两类数据融合方法，将WSNs传感器节点的监测数据整合起来。然后设计并通过监测中心现有的太阳辐照度大数据训练了一种创新的半监督支持向量机（SVM）分类器。]**
- [|] [|非图像：瞬时电压、电流和温度] [|] [|] [|] [|] [|] [|] [|] [Research on Applying Multi-Sensor Data Fusion in Photovoltaic Array Monitoring System](https://www.semanticscholar.org/paper/Research-on-Applying-Multi-Sensor-Data-Fusion-in-Jin-qiao/38f757ca2d7e4b307bb2cc6d8e533077c3236031) **[报告|设计了一种基于无线传感器网络和多传感器数据融合的光伏阵列监控系统。传感器节点捕捉光伏电站的瞬时电压、电流和温度。然后系统负责对这些数据进行初级融合，并通过基于ZigBee的无线传感器网络将初级数据包传输到中心节点。中心节点负责对所有初级数据包进行高级融合并传输高级数据服务器通过数据包残值解析算法，分析各个光伏阵列的状态，判断光伏阵列是否正常工作]**
## 【2023】光伏组件的故障监测方法综述
- [|Energies，中科院4区] [|] [|] [|] [|] [|] [|] [|] [|] [Artificial Intelligence in Photovoltaic Fault Identification and Diagnosis: A Systematic Review](https://www.mdpi.com/1996-1073/16/21/7417) **[报告|介绍了人工智能（AI）技术在解决PV系统故障检测问题方面的最新进展：结合传统方法和AI方法似乎可以获得更好的结果。此外，集成学习和堆叠分类器在诊断故障方面尤其有效。然而，还存在一些需要克服的挑战。公开可用于测试这些方法的数据短缺。（->数据短缺需要以多模态互补）在实际环境中，一些复杂的算法可能也难以应用。研究发现一些故障类型在研究中受到的关注比其他类型更多]**
- [|International Journal of Low-Carbon Technologies，中科院4区] [|] [|] [|] [|] [|] [|] [|] [|] [Detection, location, and diagnosis of different faults in large solar PV system—a review](https://academic.oup.com/ijlct/article/doi/10.1093/ijlct/ctad018/7083000?login=false) **[报告|将太阳能光伏系统中可能发生的故障分为物理、环境和电气故障，并对它们进行了进一步分类]**
- [|International Conference on Green Energy, Computing and Intelligent Technology (GEn-CITy 2023)] [|] [|] [|] [|] [|] [|] [|] [|] [Comparative review on recent electrical and non-electrical methods utilized for fault detection and diagnosis in PV systems](https://digital-library.theiet.org/content/conferences/10.1049/icp.2023.1780) **[报告|比较了最近关于基于电气和非电气方法的PV故障检测和诊断技术的研究成果。详细讨论了检测准确性、计算时间和考虑的故障类型等性能指标]**
- [|Energies，中科院4区] [|电致发光图像] [|] [|] [|] [|] [|] [|] [|] [A Review on Defect Detection of Electroluminescence-Based Photovoltaic Cell Surface Images Using Computer Vision](https://www.mdpi.com/1996-1073/16/10/4012) **[报告|使用卷积神经网络（CNN）架构实现基于电致发光（EL）的光伏电池自动缺陷检测]**
## 【2022】光伏组件的故障监测方法综述
- [|IEEE Conference and Expo Transportation Electrification Asia-Pacific (ITEC Asia-Pacific)] [|] [|] [|] [|] [|] [|] [|] [|] [Review on Fault Characterization and Diagnosis Technique in Photovoltaic Systems](https://ieeexplore.ieee.org/abstract/document/9941956) **[报告|不同的故障检测方法：视觉诊断方法： 只能诊断一些已经产生严重后果的故障。成像解决方案： 使用不同类型的摄像头传感器来识别光伏模块中的热点或老化。电气检测方法： 讨论了离线和在线两种基于电气的技术。其中，I-V曲线测量方法是常用的离线技术。MPPT为在线方法中最简单和最准确的诊断方法： 总体而言，基于最大功率点跟踪（MPPT）的诊断方法是所有在线方法中最简单和最准确的，可以嵌入商业逆变器中。]**
- [|Energies，中科院4区] [|] [|] [|] [|] [|] [|] [|] [|] [Artificial Neural Network for Fault Diagnosis of Solar Photovoltaic Systems: A Survey](https://www.mdpi.com/1996-1073/15/22/8693) **[报告|回顾了人工神经网络在PV系统故障诊断方面的最新研究进展。讨论了不同广泛应用的人工神经网络模型，包括多层感知器（MLP）、概率神经网络（PNN）、径向基函数网络（RBF）、卷积神经网络（CNN）和稀疏自编码器（SAE）]**
## 【2021】光伏组件的故障监测方法综述
- [|RENEWABLE & SUSTAINABLE ENERGY REVIEWS，中科院1区] [|] [|] [|] [|] [|] [|] [|] [|] [Potential measurement techniques for photovoltaic module failure diagnosis: A review](https://www.sciencedirect.com/science/article/pii/S1364032121008108) **[报告|PV模块降级的主要来源包括变色、剥离、腐蚀和破裂。这可以通过红外（IR）热成像和I-V曲线测量直接定量化，另外还可以考虑环境参数，如温度、湿度和紫外辐射]**
- [|RENEWABLE & SUSTAINABLE ENERGY REVIEWS，中科院1区] [|] [|] [|] [|] [|] [|] [|] [|] [Application of Artificial Neural Networks to photovoltaic fault detection and diagnosis: A review](https://www.sciencedirect.com/science/article/pii/S136403212030798X) **[报告|综述了ANN和混合ANN模型在光伏故障检测和诊断（FDD）方面应用]**
- [|Machines，中科院3区] [|] [|] [|] [|] [|] [|] [|] [|] [Review of Artificial Intelligence-Based Failure Detection and Diagnosis Methods for Solar Photovoltaic Systems](https://www.mdpi.com/2075-1702/9/12/328) **[报告|基于人工智能的方法的主要特点以及光伏故障检测和诊断应用的有效性]**
- [|IEEE Access，中科院3区] [|] [|] [|] [|] [|] [|] [|] [|] [Deep Learning-Based Fault Diagnosis of Photovoltaic Systems: A Comprehensive Review and Enhancement Prospects](https://ieeexplore.ieee.org/abstract/document/9530686) **[报告|基于深度学习的FDD被划分为五类：基于卷积神经网络（CNN）的FDD、基于循环神经网络（RNN）的FDD、基于堆叠自动编码器网络（SAEN）的FDD、基于深度信念网络（DBN）的FDD以及基于深度迁移学习（DTL）的FDD，其中指出了它们的主要优势和缺点]**
- [|International Conference on Applied System Innovation (ICASI)] [|] [|] [|] [|] [|] [|] [|] [|] [Fault identification and diagnosis methods for photovoltaic system: A review](https://ieeexplore.ieee.org/abstract/document/9568414) **[报告|测量的准确性与监测/故障检测系统的效力相关；如果光伏逆变器提供不准确的测量数据，可能需要其他测量设备，如传感器]**
- [|Engineering Review，中科院未收录] [|] [|] [|] [|] [|] [|] [|] [|] [A comprehensive review of diagnosis technics for faults occurring in the photovoltaic generator](https://er.riteh.hr/index.php/ER/article/view/1714) **[报告|总结了许多与光伏发电机降级形式、故障类型和诊断技术]**
- [|International Journal of Green Energy，中科院4区] [|] [|] [|] [|] [|] [|] [|] [|] [Fault diagnosis of visual faults in photovoltaic modules: A Review](https://www.tandfonline.com/doi/full/10.1080/15435075.2020.1825443) **[报告|可见故障、电气故障和其他故障]**
## 光伏组件的故障监测数据集
- [] [|红外] [|？] [|？] [|11种类型的故障] [|光伏模块] [|10000+10000] [InfraredSolarModules|Infrared solar module dataset for anomaly detection] [](https://ai4earthscience.github.io/iclr-2020-workshop/papers/ai4earth22.pdf) [](https://github.com/RaptorMaps/InfraredSolarModules) [](https://www.kaggle.com/code/marcosgabriel/dataset-intro-infrared-solar-modules) **[报告|该数据集包含太阳能场中不同异常情况的真实红外图像]**
- [|IEEE Transactions on Industrial Informatics，中科院1区] [|电致发光] [|生产制造] [|？] [|10] [|光伏电池] [|36,543] [|] [|宣称在自建数据集上达到99.49] [PHOTOVOLTAIC CELL ANOMALY DETECTION DATASET] [](https://ieeexplore.ieee.org/abstract/document/9744494/algorithms?tabFilter=dataset#algorithms) [](https://ieee-dataport.org/documents/photovoltaic-cell-anomaly-detection-dataset) **[报告|提供公开数据集。第一个为光伏太阳能电池异常检测提供框架真值的公共数据集（目标检测任务）]** 重点
- [|mdpi data，中科院未收录] [|可见光图像] [|] [|？] [|动态重配置和最大功率点跟踪] [|光伏电池] [|5211] [|采用基于图像的方法来估算光伏板的电特性] [|] [Dataset for Detecting the Electrical Behavior of Photovoltaic Panels from RGB Images] [](https://www.mdpi.com/2306-5729/7/6/82) [](https://ieee-dataport.org/documents/photovoltaic-cell-anomaly-detection-dataset)) **[报告|介绍了一个数据集，将RGB图像与不同辐照度和阴影条件下的光伏板电数据进行关联；此外，数据集还提供了补充的天气数据和其他图像特征，以支持估算模型的训练。具体而言，该数据集被设计用于支持基于图像的电数据估算模型的设计，以替代大量的传感器]**
- [|Remote Sensing，中科院2区] [|红外] [|航拍图像] [|？] [|] [|] [|] [|] [|] [RID—Roof Information Dataset for Computer Vision-Based Photovoltaic Potential Assessment] [](https://www.mdpi.com/2072-4292/14/10/2299) [](https://github.com/TUMFTM/RID) **[报告|引入了两个新的多类别数据集，分别用于屋顶分割和屋顶结构的语义分割，重点是从航拍图像中提取屋顶信息，尤其是在光伏潜力分析方面的应用。虽然基于2D航拍图像的深度学习光伏潜力分析包含了结构信息，但缺乏坡度和阴影信息。因此，有必要研究从航拍图像学习这些信息或将2D和3D数据合并以进一步提高潜力估计准确性。除了光伏潜力分析，丰富的3D模型还可以推动其他研究和应用，如城市噪声扩散建模、屋顶隔热评估或建筑能源需求估算等方面的发展。]**
- [|35th EU PVSEC 2018] [|电致发光图像] [|实验室环境] [|？] [|裂纹、断裂互连、PID 和电池质量] [|太阳能电池] [|2,426] [|] [|] [elpv-dataset] [](https://userarea.eupvsec.org/proceedings/35th-EU-PVSEC-2018/5CV.3.15/) [](https://github.com/zae-bayern/elpv-dataset) **[报告|从高分辨率电致发光 (EL) 图像中提取的由 2,426 个太阳能电池组成的数据集用于自动缺陷概率识别。收集的图像包含单晶和多晶太阳能模块中功能正常和有缺陷的太阳能电池，其退化程度不同。这些图像由专家标记，专家根据每张图像中缺陷的可能性对太阳能电池进行分类。标记图像可用于开发计算机视觉和机器学习方法，以自动检测不同的缺陷，如裂纹、断裂互连、PID 和电池质量，并用于预测功率效率损失。]**