• Energy Research

Solar photovoltaic (PV) modules are susceptible to manufacturing defects, mishandling problems or extreme weather events that can limit energy production or cause early device failure. Trained professionals use electroluminescence (EL) images to identify defects in modules, however, field surveys or inline image acquisition can generate millions of EL images, which are infeasible to analyze by rote inspection. We develop a rapid automatic computer vision pipeline (∼0.5 seconds/module) to analyze EL images and identify defects including cracks, intra-cell defects, oxygen-induced defects, and solder disconnections. Defect identification is achieved with a machine learning model (Random Forest, ResNet models and YOLO) trained on 762 manually-labeled EL images of PV modules. We compare model performance on an imbalanced real-world validation set containing 134 EL images and determine that ResNet18 and YOLO are the optimal models; we next evaluated these models on a dedicated testing set (129 module images) with resulting macro F1 scores of 0.83 (ResNet18) and 0.78 (YOLO). Using a field EL survey of a PV power plant damaged in a vegetation fire, we analyze 18,954 EL images (2.4 million cells) and inspect the spatial distribution of defects on the solar modules. The results find increased frequency of ‘crack’, ‘solder’ and ‘intra-cell’ defects on the edges of the solar module closest to the ground after fire. We also find an abnormal increase of striation rings on cells which were assumed to be caused mainly in fabrication process. Our methods are published as open-source software. It can also be used to identify other kinds of defects or process different types of solar cells with minor modification on models by transfer learning.

Solar photovoltaic (PV) modules are susceptible to manufacturing defects, mishandling problems or extreme weather events that can limit energy production or cause early device failure. Trained professionals use electroluminescence (EL) images to identify defects in modules, however, field surveys or inline image acquisition can generate millions of EL images, which are infeasible to analyze by rote inspection. We develop a rapid automatic computer vision pipeline (∼0.5 seconds/module) to analyze EL images and identify defects including cracks, intra-cell defects, oxygen-induced defects, and solder disconnections. Defect identification is achieved with a machine learning model (Random Forest, ResNet models and YOLO) trained on 762 manually-labeled EL images of PV modules. We compare model performance on an imbalanced real-world validation set containing 134 EL images and determine that ResNet18 and YOLO are the optimal models; we next evaluated these models on a dedicated testing set (129 module images) with resulting macro F1 scores of 0.83 (ResNet18) and 0.78 (YOLO). Using a field EL survey of a PV power plant damaged in a vegetation fire, we analyze 18,954 EL images (2.4 million cells) and inspect the spatial distribution of defects on the solar modules. The results find increased frequency of ‘crack’, ‘solder’ and ‘intra-cell’ defects on the edges of the solar module closest to the ground after fire. We also find an abnormal increase of striation rings on cells which were assumed to be caused mainly in fabrication process. Our methods are published as open-source software. It can also be used to identify other kinds of defects or process different types of solar cells with minor modification on models by transfer learning.

This study compared module power loss for 36 modules that endured various accelerated aging test sequences before installation outdoors on a 10-kWp array in Birmingham, AL, USA for 1.72 to 2.72 years. Twelve modules endured standard IEC 61215 aging tests and 24 endured Qualification Plus (Qual Plus). Modules in each group were further split into two test sequences with different exposures. Electrical parameter variations were analyzed as a function of aging test and field exposure history. Fill factor loss was determined to be the cause of observed decreases in power output during accelerated aging tests, while decreases in both open circuit voltage and fill factor dominated the power loss during subsequent on-sun testing. Quantified cell crack features were extracted via computer vision tools from electroluminescence images and correlated with power loss. Results illustrate that standard aging tests led to negligible cracks, while Qual Plus test sequences yielded more severe cracks. While correlating results from qualification tests with in-field performance degradation parameters remains a challenge, this study provides new insights on specific environmental stressors and crack features that may play a role in power loss. Insights on accelerated aging protocols are discussed.

This study compared module power loss for 36 modules that endured various accelerated aging test sequences before installation outdoors on a 10-kWp array in Birmingham, AL, USA for 1.72 to 2.72 years. Twelve modules endured standard IEC 61215 aging tests and 24 endured Qualification Plus (Qual Plus). Modules in each group were further split into two test sequences with different exposures. Electrical parameter variations were analyzed as a function of aging test and field exposure history. Fill factor loss was determined to be the cause of observed decreases in power output during accelerated aging tests, while decreases in both open circuit voltage and fill factor dominated the power loss during subsequent on-sun testing. Quantified cell crack features were extracted via computer vision tools from electroluminescence images and correlated with power loss. Results illustrate that standard aging tests led to negligible cracks, while Qual Plus test sequences yielded more severe cracks. While correlating results from qualification tests with in-field performance degradation parameters remains a challenge, this study provides new insights on specific environmental stressors and crack features that may play a role in power loss. Insights on accelerated aging protocols are discussed.

To enable health monitoring and fault diagnosis of PV modules using current-voltage characteristics (I–V curves), it is generally necessary to correct the I–V curves measured under different environmental conditions to the standard condition. The most common correction methods are those from IEC 60891: 2021 standard. However, these methods can introduce significant errors when dealing with degraded PV modules due to the inability to account for changes in resistance. To address this, we propose an improved I–V curve procedure, denoted Pdynamic, which considers different types of degradation by dynamically deriving the correction coefficients from the measured I–V curves. To evaluate the performance, we simulate I–V curves across a wide range of irradiance and temperature for the healthy and degraded module, where the degradation involves increased series resistance, decreased shunt resistance, or both. The results reveal that Pdynamic can produce corrected I–V curves closer to the reference ones than Procedures 1, 2, and 4 of the IEC 60891:2021 standard. Moreover, Pdynamic exhibits resilience to both seasonal fluctuations and varying levels of degradation. These results highlight Pdynamic as a promising and robust I–V curve correction method, particularly for degraded PV modules. A Python-based open-source tool for this procedure is also available at https://github.com/DuraMAT/IVcorrection.