Electroluminescence (EL) imaging is widely used to detect defects in photovoltaic (PV) modules, and machine learning methods have been applied to enable large-scale analysis of EL images. However, existing methods cannot assign multiple labels to the same pixel, limiting their ability to capture overlapping degradation features. We present a multi-channel U-Net architecture for pixel-level multi-label segmentation of EL images. The model outputs independent probability maps for cracks, busbars, dark areas, and non-cell regions, enabling accurate co-classification of interacting features such as cracks crossing busbars. The model achieved an accuracy of 98% and has been shown to generalize to unseen datasets. This framework offers a scalable, extensible tool for automated PV module inspection, improving defect quantification and lifetime prediction in large-scale PV systems.

Individual photovoltaic (PV) module health monitoring can be a daunting task for operation and maintenance of solar farms. Modules can be inspected through luminescence, thermal imaging, and current-voltage (I-V) curve analyzes for identification of damage and power loss. I-V curves provide easily interpretable data to determine module health as they directly provide electrical performance metrics. However, in order to obtain these curves, modules must be disconnected from the array and either removed to a solar simulator or characterized in situ with corrections for module temperature, the incident solar spectrum, and intensity. Luminescence or thermal images of a module are relatively easy to acquire in situ. Electroluminescence (EL) images highlight physical defects in the modules but do not provide easily interpretable features to correlate with electrical performance. This work presents a SWin transformer network to predict I-V curves for PV modules from their corresponding EL images. The predicted I-V curves allow the accurate prediction of the maximum power point (MPP), short-circuit current I-sc, and open-circuit voltage V-oc with a mean error less of than 1%. Comparing single diode model (SDM) parameters extracted from the predicted curves to those extracted from the true curves, the series resistance R-s demonstrates a mean error of 5.19%, and the photocurrent I a mean error of 0.197%. The shunt resistance R-sh and dark current I-o parameters are predicted with larger errors because of their sensitivity to small changes in the I-V curve.

PV cell cracks can cause various levels of power loss depending on the type and severity of crack, cell characteristics, and external stressors. We present a statistical framework for predicting power loss based on these factors, based on experimental and finite element analyses. Using x-ray topography, we measure translation of external thermomechanical stress into internal deformation and strain on cracked cells. Experiments are replicated in finite element models to validate prediction of cell mechanical and electrical behaviors. Together these approaches inform the probabilistic estimation of PV cell power loss associated with cell crack geometry and external stressors, enabling fast, prognostic evaluation of PV modules with existing cell cracks using only electroluminescence imaging and environmental operating conditions.

Due to the extreme thermomechanical stresses imposed on photovoltaic (PV) modules during manufacture, shipping, installation, regular operation, and extreme weather, fracture of Si cells may be inevitable. As the industry moves thinner glass and larger format modules, and as the frequency of extreme weather events continues to increase, the problem of cell cracks is likely here to stay. Therefore, when cracks are observed, it is important to understand their immediate effects on PV performance. In this work we are using a Variational Autoencoder (VAE) trained with preprocessed EL images. The results of the training, latent space, is analyzed to detect clusters of certain defects. IV-curves of the same solar cells used for the EL images are used to develop a correlation model between the latent space and the performance loss.

The risk of cell cracks continues to be an issue for photovoltaic systems worldwide. Damage caused by extreme weather and mishandling during transportation and installation can lead to underperforming panels and eventual power degradation to the system. To address cell-crack-induced degradation directly, we have formulated a carbon nanotube additive for commercial screen-printable silver pastes. This work highlights the beginning-of-life increase in module power, efficiency, and fill factor by adding carbon nanotubes to commercial silver paste. We conclude this work by applying digital image correlation to quantify the movement of cell fragments and full-sized module certification via a modified IEC 63209 accelerated stress test, which demonstrates that the CNT-reinforced gridlines significantly reduce cell cracks.

Glass breakage has become a major reliability issue in utility-scale photovoltaic power plants over the last similar to 5 years, as module sizes have trended larger and glass thickness has decreased to decrease module cost per watt. Module failures have been reported as both 'spontaneous' and resulting from extreme weather such as hail and high windspeeds. Here we explore the costs associated with more robust module types, including smaller form factors and thicker glass and frames. We consider not only module costs but also system level effects for components, shipping, and installation. Our analysis shows that 'hail-resistant' thicker glass and thicker frames are cost-viable strategies for improving glass reliability. Reverting to smaller modules, however, prohibitively increases system-level costs.

Photovoltaic (PV) module materials and technologies continue to evolve as module manufacturers and buyers try to minimize costs, maximize performance, and speed deployment. Both silicon and thin film modules are converging toward similar similar to 3m(2) glass-glass designs with thinner glass sheets to increase power output while reducing module weight, and both types are increasingly mounted on single-axis trackers. At the same time, an increasing number of PV sites have been reporting spontaneous glass breakage in early life systems deployed with these "big, floppy modules." In this article, we identify the concurrent module changes that may be contributing to increased early failure, explain the trends, and discuss their reliability implications. We suggest that larger, thinner glass sheets along with variations in heat treatment and quality may be contributing to glass vulnerability. We note that trends toward weaker or back-mounted frames may also be contributing to module failures, especially for "extra-extra-large" modules mounted on trackers. Combinations of these trends may have pushed modules to a threshold at which increasing early failures are causing the front edge of the "bathtub curve" to re-emerge. Current qualification testing appears to be ineffective for catching these early failures in new module designs, and module buyers do not have enough reliability information-or cannot prioritize such information-during module procurement. Additional research is needed to identify the field conditions leading to glass breakage and if there is one or multiple limiting flaws in new module designs causing glass breakage. Early failures may be mitigated by returning to more robust designs or ensuring better module testing and quality assurance.

Photovoltaic (PV) systems have become a cornerstone of renewable energy strategies, particularly due to the significant reduction in solar power costs over the past decade. However, the long-term reliability of PV installations presents a persistent challenge, requiring the development of advanced monitoring and predictive maintenance strategies. A wide range of data types is used to evaluate the health of PV systems, including environmental conditions, electrical performance, and inspection imagery. These data enable methodologies such as machine learning (ML) models for lifetime prediction and computer vision techniques for defect detection. However, the acquisition of high-quality and comprehensive data is difficult, particularly in terms of long-term consistency and data variety. Publicly available data sets serve as valuable resources for addressing these challenges, but they often suffer from fragmentation and are difficult to access. This paper presents a comprehensive review of existing open-source data sets related to PV degradation, analyzing their features, functionalities, and potential applications. We categorize these data sets based on the specific aspects of PV system information they cover, such as environmental conditions, operational monitoring, image inspection and module materials, and propose relevant tools and ML models for processing them. In addition, we propose practices for future data collection and usage, while also discussing potential directions in data-driven research. Our aim is to enhance data utilization and publication among researchers and industry professionals, promoting a deeper understanding of the role of data in enhancing the performance and durability of PV systems.

The advent of bifacial PV systems drives new requirements for irradiance measurement at PV projects for monitoring and assessment purposes. While there are several approaches, there is still no uniform guidance for what irradiance parameters to measure and for the optimal selection and placement of irradiance sensors at bifacial arrays. Standards are emerging to address these topics but are not yet available. In this paper we review approaches to bifacial irradiance monitoring which are being discussed in the research literature and pursued in early systems, to provide a preliminary guide and framework for developers planning bifacial projects.

Hail poses a significant threat to photovoltaic (PV) systems due to the potential for both cell and glass cracking. This work experimentally investigates hail-related failures in Glass/Backsheet and Glass/Glass PV modules with varying ice ball diameters and velocities. Post-impact Electroluminescence (EL) imaging revealed the damage extent and location, while high-speed Digital Image Correlation (DIC) measured the out-of-plane module displacements. The findings indicate that impacts of 20 J or less result in negligible damage to the modules tested. The thinner glass in Glass/Glass modules cracked at lower impact energies (similar to 25 J) than Glass/Backsheet modules (similar to 40 J). Furthermore, both module types showed cell and glass cracking at lower energies when impacted at the module's edges compared to central impacts. At the time of presentation, we will use DIC to determine if out-of-plane displacements are responsible for the impact location discrepancy and provide more insights into the mechanical response of hail impacted modules. This study provides essential insights into the correlation between impact energy, impact location, displacements, and resulting damage. The findings may inform critical decisions regarding module type, site selection, and module design to contribute to more reliable PV systems.

Snow is a significant challenge for photovoltaic (PV) systems at northern latitudes, where the pace of deployment is rapid but snow-related power losses can exceed 30% of annual production. Accurate snow-related power loss estimation methods for utility-scale sites can support snow mitigation strategies, inform resource planning and validate predictive snow-loss models. This study builds on our previous work on inverter-based detection of snow, and its implications for utility-scale power production, by validating the accuracy of our snow-loss method across different PV sites and system designs and highlighting its value in bringing greater visibility to PV plant operations in winter. Our estimation method is both novel and scalable, requiring only standard monitoring data to correlate snow-related losses with meteorological data. As demonstrated here, our validation method involved three main steps: 1) estimation of performance losses for multiple systems by comparing measured inverter data to modeled data; 2) application of a detection framework to identify which performance losses are snow-related; and 3) comparison of snow-related losses among three utility-scale sites differing in tilt angle. Results show that utility-scale systems at higher tilt angles consistently shed snow more quickly/completely than their lower-tilt counterparts. Further, monthly and seasonal snow losses are inversely and non-linearly correlated with tilt angle when normalized for cumulative snowfall. These results are consistent with the findings of previous studies and support the broad applicability of this method to fixed-tilt utility-scale PV systems around the world that routinely experience snow-related performance losses.

Concentrating solar power (CSP) plants with integrated thermal energy storage (TES) have successfully been coupled with photovoltaics (PV) + chemical battery energy storage (BES) in recent commercial-scale projects to balance system cost and diurnal power availability. Sandia National Laboratories has been tasked with designing an advanced solar energy system to power Kirtland Air Force Base (KAFB) where Sandia is co-located in Albuquerque, NM, USA. This design process requires optimization of individual components and capacities of the hybrid system. Preliminary modeling efforts have shown that a hybrid CSP+TES/PV+BES in Albuquerque, NM is sufficient for net-zero power generation for Sandia/KAFB for the next decade. However, the ability to meet the load in real-time (and minimize energy export) requires balance of generation and storage assets. Our results also show that excess PV used to charge TES improves resilience and overall renewables-to-load for the system. Here we will present the results of a parametric study varying the land use proportions of CSP and PV, and TES and BES capacities. We evaluate the effects of these variables on energy generation, real-time load satisfaction, site resilience to grid outages, and LCOE, to determine viable hybrid solar energy designs and their cost implications.

Cell cracks reduce power production from PV systems worldwide. The difficult-to-detect and gradual reduction in power production caused by cell cracks has yet to be resolved by module and cell manufacturers, and the ever-growing demand for solar energy continues to climb. To address cell-crack-induced degradation directly, we have formulated a carbon nanotube additive for commercial screen-printable silver pastes. In this work, we highlight the increase in cell efficiency by adding carbon nanotubes and the mitigated power loss from accelerated mini-module fracture testing. We conclude this work with full-sized module certification via a modified IEC 63209 stress testing, digital image correlation to quantify the movement of cell fragments, and field testing.

Photovoltaic modules undergoing laboratory hail tests were observed using high speed video to analyze the key characteristics of impact-induced glass fracture, including crack onset time, initiation location relative to the impact site, and propagation trends. Fifteen commercially representative glass- lass thin-film modules were recorded at 300,000 frames per second during hail impacts which happened to cause glass fracture. Images were processed to identify the time between impact and first plausible glass crack appearance (average 126 mu s, standard deviation 59 mu s) along with the time to a confirmed crack (average 158 mu s, standard deviation 77 mu s), during the ice ball impacts which had a median kinetic energy of 47 J delivered by 55 mm diameter balls. Limiting factors for identifying glass crack timings were ice ball fragmentation obscuring the impact site and indistinct initial crack appearance, which were inherent to the images and not improved with processing. Computational simulations corresponding to each impact event showed that glass stresses were still localized to the impact site during times with definitively identifiable fracture, and even impacts which did not induce failure created local stress magnitudes exceeding stress levels associated with static glass fracture. These observations confirm that impact-induced glass failure is a time- and rate-dependent phenomena. Results from this study provide baseline metrics for developing a glass fracture criterion to predict module damage during hail impact events, which in turn allows for analysis of design features that may affect damage susceptibility.

Due to software fragmentation, PV system modeling teams can be limited to language specific packages, preventing cross-sectional analysis of different modeling techniques and workflows. To this end, PVplr, a popular PV performance modeling R software package, has been ported to the Python programming language. To verify and test the robustness of the port, NSRDB data has been used to simulated PV installations at native resolution ( 2 million Sites), with a variety of degradation rates, degradation patterns, and modules. Performance Ratios were calculated using the ported functions from pvplr-python and compared against Rdtools YoY values. Due to the complicated nature of degradation, a new metric has been proposed to quantify the performance loss of a system. The cumulative production loss, is the total amount of energy lost due to the degrading performance of the system. Cumulative production loss alleviates the problems with fitting linear functions to non-linear degradation. Cumulative Production loss was shown to better estimate the total loss revenue for degradation patterns. XbX + UTC was found to most accurately predict the total lost revenue in simulated systems.

System health monitoring is an essential task in the operation and maintenance of any photovoltaic (PV) system. Typically, electroluminescence (EL), thermal imaging, and current-voltage (IV) curve analyses are used to analyze PV modules with each providing unique insights into system health. While it is relatively easy to acquire an EL or thermal image of a panel in-situ, acquisition of IV curves requires electrical disconnection of the panel from the array and either removal to a solar simulator or characterization and correction for the incident solar spectrum and intensity. In this work we show that, with the use of a transfer-learned Swin transformer model, we can predict accurate IV curves from EL images. Extracting single diode equation parameters from the predicted IV curves yielded an error less than 1%+/- 1% for the maximum power point (MPP), short-circuit current I-sc, open-circuit voltage V-oc and photocurrent I. The series resistance Rs and number of series cells nN(s)V(th) denoted as N were predicted with errors of similar to 5%+/- 7% and similar to 3%+/- 2%, respectively. Prediction of the shunt resistance R-sh and dark current I-o yielded larger errors, likely due to sensitivity to small changes in the IV curve.

Precise and comprehensive records of system performance is imperative for the efficient monitoring and prediction of Photovoltaic (PV) power generation. Nonetheless, the inevitable failure of real-world sensors and monitoring devices results in information loss. Additionally, the high variability in production data can obfuscate erroneous measurements as nominal values. The occurrence of such “missingness” significantly impacts the stability and precision of performance estimation. By leveraging the inherent value dependencies present in PV production data, graph data driven Digital Twin models can be created for individual sites that capture the high frequency spatial weather patterns that determine the precise performance for that specific instance in time. Through varying the graph structure, the same model architecture can target both outlier detection, and imputation of missing values. st-GAE, an existing spatio-temporal graph autoencoder, was used to detect and impute outliers for a collection of 98 inverters with 5 minute interval data for a period of two years. Outlier detection was compared against existing maintenance logs which were available for all of the systems. stGAE was shown to correctly identify 90% of maintenance events and was able to reconstruct those missing values with a MAE of less than 1.5W.

The impact of snow losses on the performance of solar modules is unneglectable. Northern latitudes are expanding their solar energy portfolio and need reliable prediction methods and guidelines on how to reduce these losses. In this work, we compare the performance of landscape and portrait orientation for state of the art half-cut solar cell modules during winter months. We show how to adapt the Marion snow loss model to account for module orientation to improve performance modeling for PV systems affected by snow.