A deep learning approach to photovoltaic cell defect classification

Cracks in solar cells are one of the problems responsible for low efficiency among photovoltaic (PV) modules and these cracks are hard to avoid. Cracks usually can occur during transportation, handling or manufacturing process. Micro cracks degrade the power output in photovoltaic modules due to the inactive cell area. Micro-cracks are invisible to detect by naked eye and hence the Electroluminescence (EL) imaging were introduced to analyse the cracks in PV module. The electroluminescence is the most useful method to detect the cracks in the solar cell. A si-CCD camera was used in the dark condition to capture the EL image under forward bias conditions. In this study, mono-crystalline and multicrystalline were taken from the outdoor condition which were exposed for almost eight years and were analysed using electroluminescence imaging. From the electroluminescence imaging, the cracks were measured. The cracks were analysed by pattern and both panels were compared. From the study it was found that multicrystalline photovoltaic module shows the more crack compared with monocrystalline photovoltaic module. The crack in the individual solar cell and their relative efficiency in the two different types of crystalline modules have been also been presented.

Defect detection and quantification in electroluminescence images of solar PV modules using U-net semantic segmentation

Electroluminescence (EL) imaging is one of the most common characterisation techniques for photovoltaic cells and modules. EL images contain information on both the open-circuit voltage (Voc) and series resistance (Rs) of the device. However, separating the two effects and identifying features related to each parameter can be challenging. In this study, a novel approach for decomposing EL images into Voc and Rs maps using a convolutional neural network architecture is presented. A deep learning model was first trained on paired EL and photoluminescence images that were generated using a simulation tool. Results obtained using the validation set show that the trained model is able to accurately differentiate between features related to Voc and Rs in EL images, thus replacing the need for multiple types of measurements. The proposed method presents a unique approach to analyse EL images, unlocking new capabilities that have the potential to advance solar cell characterisation.

Automated defect identification in electroluminescence images of solar modules

Efficient deep feature extraction and classification for identifying defective photovoltaic module cells in Electroluminescence images

Electroluminescence (EL) imaging is a photovoltaic (PV) module characterization technique, which provides high accuracy in detecting defects and faults, such as cracks, broken cells interconnections, shunts, among many others; furthermore, the EL technique is used extensively due to a high level of detail and direct relationship to injected carrier density. However, this technique is commonly practiced only indoors—or outdoors from dusk to dawn—because the crystalline silicon luminescence signal is several orders of magnitude lower than sunlight. This limits the potential of such a powerful technique to be used in utility scale inspections, and therefore, the interest in the development of electrical biasing tools to make outdoor EL imaging truly fast and efficient. With the focus of quickly acquiring EL images in daylight, we present in this article a drone-based system capable of acquiring EL images at a frame rate of 120 frames per second. In a single second during high irradiance conditions, this system can capture enough EL and background image pairs to create an EL PV module image that has sufficient diagnostic information to identify faults associated with power loss. The final EL images shown in this work reached representative quality SNRAVG of 4.6, obtained with algorithms developed in previous works. These drone-based EL images were acquired with global horizontal solar irradiance close to one sun in the plane of the array.

This review paper presents a comprehensive analysis of electroluminescence (EL) imaging techniques for photovoltaic (PV) module diagnostics, focusing on advancements from conventional indoor imaging to outdoor and daylight EL imaging. It examines key challenges, including ambient light interference and environmental variability, and highlights innovations such as infrared-sensitive indium gallium arsenide (InGaAs) cameras, optical filtering, and periodic current modulation to enhance defect detection. The review also explores the role of artificial intelligence (AI)-driven methodologies, including deep learning and generative adversarial networks (GANs), in automating defect classification and performance assessment. Additionally, the emergence of drone-based EL imaging has facilitated large-scale PV inspections with improved efficiency. By synthesizing recent advancements, this paper underscores the critical role of EL imaging in ensuring PV module reliability, optimizing performance, and supporting the long-term sustainability of solar energy systems.

Thermal and electroluminescence (EL) imaging techniques are widely accepted as powerful tools for analyzing solar cells. We have identified and characterized various defects in photovoltaic devices with sub-micron spatial resolution using a novel thermoreflectance imaging technique that can simultaneously obtain thermal and EL images with a mega-pixel silicon-based CCD. Linear and non-linear shunt defects are investigated as well as electroluminescent breakdown regions at reverse biases as low as -5V. Pre-breakdown sites with electroluminescence are observed. The wavelength flexibility of thermoreflectance imaging is explored and thermal images of sub-micrometer defects are obtained through glass that would typically be opaque for infrared light. Image sequences show a 10μs thermal transient response of a 15μm defect in a polysilicon solar cell. Nanosecond reverse bias voltage pulses are used to detect breakdown regions in thin-film a-Si solar cells with EL.

Deep learning based automatic defect identification of photovoltaic module using electroluminescence images

Accurate current–voltage (IV) characterization is essential for assessing photovoltaic (PV) module performance, yet conventional IV tracing requires physical contact and controlled conditions, limiting large-scale deployment. Electroluminescence (EL) imaging, while highly effective for detecting localized defects, remains largely qualitative and indirect in estimating actual PV module power loss. This study introduces a deep learning framework that directly predicts complete IV curves from EL images, transforming EL inspection into a quantitative, non-contact diagnostic tool. In this work, we propose a convolutional neural network (CNN) that learns the nonlinear mapping between paired EL images captured at 20% and 80% of the short-circuit current and the corresponding IV response. A total of 438 PV modules were used for model development, with performance evaluated on unseen data. The trained CNN reconstructs IV curves with high fidelity, achieving a validation accuracy of approximately 95% and low parameter deviations (<2% for key metrics such as maximum power point and fill factor). The model maintains consistent accuracy even when a single EL image is provided, supporting flexible field operation. Inference is rapid, requiring less than 0.5 s per PV module inspection, enabling real-time analysis.

Electroluminescence (EL) imaging is a powerful tool used to identify defects in photovoltaic solar cells. Typically, this type of characterization is performed in the dark using a current injection that equals short-circuit current measured at standard test conditions (STC). Restricting imaging to such a temperature range limits the information obtained about the module and cells. However, it is not trivial to develop a tool that would allow for EL imaging to be performed under a wider range of temperatures. Here we demonstrate an in situ EL imaging capability developed within an environmental chamber that allows for control of sample temperatures between –40 and 90 °C. Additionally, we demonstrate EL imaging of 4-cell mini-modules (MiMo) under front-side mechanical loading. A Raspberry Pi-connected camera with short-pass filter removed is used for EL imaging. The camera is low-cost with a small form-factor, appropriate for use in a harsh, enclosed environment. The camera is installed within a thermally isolating housing mounted within the environmental chamber. Three example cases are given for MiMos that exhibit various forms of degradation including solder fatigue and cell cracking. It is shown that by measuring at conditions above and below STC, different behaviors may be identified. In some cases, restricting characterization to STC may lead to a failure to detect damage existing in the sample.

Computer imaging of electroluminescence (EL) has been successfully applied to solar cell inspection in recent years, as EL image intensities reflect the efficiency levels and/or defects in sc-Si and mc-Si solar cells. In this paper, we propose a novel computational scheme for pseudo colorization of EL images to highlight defect regions in solar cells for human inspection. Specifically, given a template EL image and pseudo color labels on its defect regions, we impose the pseudo colors to other grayscale EL images, with respect to different defect types and image structures, by template feature clustering and pseudo color transferring. Our experiments show that the proposed approach indeed improves the readability of EL images and provides better visualization of solar cell defects.

Electroluminescence (EL) imaging of Si-based photovoltaic (PV) modules is used widely to spatially detect and characterize electrical defects, including handling and degradation-induced cracking of the component Si cells that are associated with reductions in module performance. In the present study, a commercial polycrystalline silicon PV module was subjected to accelerated lifecycle test environmental conditions and examined as a function of environmental exposure time using EL imaging. The approach followed pixel intensity distributions over each individual PV cell and confirmed a positive correlation between module conversion efficiency and results of the image analysis. Overall, an average of a 2.5% reduction in normalized EL intensity was correlated to a 0.35% reduction in actual power conversion efficiency (or a 2.3% decrease in relative efficiency). The imaging analysis technique offers a rapid, unsupervised means to assess EL data in lieu of conventional visual interpretation.

In the field of photovoltaics, electroluminescence (EL) imaging has proven useful in determining the degradation of solar panels, primarily through use of expensive equipment within a confined lab setting. This study explored the potential of conducting EL imaging using relatively inexpensive equipment. Factors that heavily influenced the quality of EL images included voltage/current settings for the purpose of forward biasing the test panel, camera settings (aperture, shutter speed, and ISO), and amount of external light (not EL) detected by the camera. Once EL images were captured, the images with the best quality were analyzed and modified through image processing and photo editing. The quality of the final images was then validated through comparison with images captured in a controlled lab with higher quality equipment. Moreover, the results of transporting a solar panel are shown to be clearly observable using low-cost EL imaging.

In this study, faults in solar panel cells were detected and classified very quickly and accurately using deep learning and electroluminescence images together. A unique and new dataset was created for this study. Monocrystalline and polycrystalline solar panel cells were used in the dataset. The dataset included intact, cracked and broken images for each solar panel cell. The dataset was preprocessed and multiplied to equalize the intact, cracked and broken numbers. Seven different deep learning architectures were used in this study. As a result of this study, 97.82% accuracy was achieved for the monocrystalline solar panel cells and 96.29% for the polycrystalline solar panel cells in the SqueezeNet architecture.