Search

Energy losses of photovoltaic (PV) plants because of soiling are a problem in all regions, including Germany. Soft soiling, caused by a uniform dust film, is shading a PV module and is reducing yield depending on the thickness of the debris layer. Hard soiling, caused by agglomerations of dust or dirt, only covers parts of a PV module, and it causes local shading. A spot check of modules from different PV power plants in Germany revealed a power reduction because of soft- and hard soiling of up to 6%. Determination of soiling type and amount is a prerequisite for decisions about cleaning and for optimizing yield. In this article, we show that luminescence imaging can be used to detect and characterize soiling and quantify losses. For this purpose, we compare luminescence images before and after cleaning and we show that soiling becomes detectable and power losses quantifiable from difference images. We also show that the impact of hard soiling can be quantified from a single photoluminescence (PL) image. This technique may enable a fully automated quantification of power losses because of hard soiling in PV- modules and strings in the future.

Feasibility of the identification of backsheet (BS) composition of field-aged Si photovoltaic (PV) modules based on the analysis of characteristic patterns of UV-excited fluorescence (UVF) of ethylene vinyl acetate copolymer encapsulant is shown. The approach includes BS identification by vibrational spectroscopies and the formation of a database of BS-related characteristic UVF patterns. A multivariate principal component analysis of the UVF images of unknown samples combined with the UVF pattern database allows the BS type to be identified only from UVF images. The geometric place of particular samples in the principal component plots is indicative of the general degradation status of PV modules. The proposed method can be applied for high-throughput noncontact characterization of solar PV plants.

Degradation of backsheets (BSs) of commercial silicon PV modules is currently recognized as a source of reduced module performance and module failure. Monitoring of the BS state in the field is possible by using non-destructive and highly informative near-infrared absorption (NIRA) spectroscopy. Application of NIRA for the analysis of multi-layer polyethylene terephtalate (PET) based BSs, which dominate the PV module market, is challenging due to a large variety of possible BS configurations that show only small differences in NIRA spectra. In the present work, a spectroscopic tool for the structural identification of PET-based BSs is introduced. The method is based on a principal component analysis of a database of 250 representative NIRA spectra of BSs of different types. It allows a BS with an unknown structure to be assigned to one of 12 different types based solely on its NIRA spectrum. The identification was successfully validated on a test collection of 45 selected BSs and shown to be feasible for the field deployment. Further automation of NIRA measurements and spectral analysis are expected to elevate the proposed tool to the level of a non-intrusive high-throughput field analysis of the BS composition and state in operating PV module grids.

Increasing deployment of photovoltaic (PV) plants requires methods for automatic detection of faulty PV modules in modalities, such as infrared (IR) images. Recently, deep learning has become popular for this. However, related works typically sample train and test data from the same distribution ignoring the presence of domain shift between data of different PV plants. Instead, we frame fault detection as more realistic unsupervised domain adaptation problem where we train on labeled data of one source PV plant and make predictions on another target plant. We train a ResNet‐34 convolutional neural network with a supervised contrastive loss, on top of which we employ a k‐nearest neighbor classifier to detect anomalies. Our method achieves a satisfactory area under the receiver operating characteristic (AUROC) of 73.3% to 96.6% on nine combinations of four source and target datasets with 2.92 million IR images of which 8.5% are anomalous. It even outperforms a binary cross‐entropy classifier in some cases. With a fixed decision threshold, this results in 79.4% and 77.1% correctly classified normal and anomalous images, respectively. Most misclassified anomalies are of low severity, such as hot diodes and small hot spots. Our method is insensitive to hyperparameter settings, converges quickly, and reliably detects unknown types of anomalies making it well suited for practice. Possible uses are in automatic PV plant inspection systems or to streamline manual labeling of IR datasets by filtering out normal images. Furthermore, our work serves the community with a more realistic view on PV module fault detection using unsupervised domain adaptation to develop more performant methods with favorable generalization capabilities.

Reliable and durable solar power plants require PV modules with high-grade polymer encapsulants and backsheets (BSs). For performance analyses of PV installations fast, reliable and non-destructive methods for determining composition and degradation state of polymer components need to be developed. Here, we show that the structure of some common polymer BSs can be determined in the field in real time by analyzing near-infrared transmission (NIRT) spectra collected under illumination with natural sunlight. The potential of this “Solar-NIRT” method was probed by field measurements on a multi-MW PV power plant where four major BS types were identified by multispectral cross-sectional Raman imaging. Additionally, degradation of a particular BS type was found to result in distinct changes in NIRT spectra allowing the degraded BSs to be classified as a separate type. Principal component analysis (PCA) applied to a collection of 62 Solar-NIRT spectra allowed to create a map of five clusters, each corresponding to a particular BS type. The feasibility of using the PCA cluster map for the identification of unknown samples was shown on a test set of 13 different BSs. The Solar-NIRT is relatively fast, non-invasive, selective, can be upgraded to a non-contact regime making it a promising tool for high-throughput characterization.

We present here a degradation study of a PV power plant consisting of several module pairs connecting each to a micro-inverter with monitoring. The modules comprise cell cracks, cell breakages and edge shunting. The impact of using different methodologies on the resulting relative yield loss rates is analyzed. It is demonstrated how these methods can be altered to circumvent missing irradiation sensor data generating a substitute for a relative degradation rate. The relative degradation rate of cell cracks is close to the reference with all methods and near to the detection limit which indicates little degradation for cell cracks beyond the material-inherent degradation. The modules with cell breakages and edge shunting show a higher relative yield loss of l-3 %/a depending on the method of analysis. This indicates quite a variation is possible depending on the method thus warranting caution when evaluating degradation rates or their substitutes. It was shown that cell breakages perform increasingly worse in summer leading to even higher but seasonal performance losses. Because of that, the relative yield loss can be much higher than just the relative loss in rated power due to probably averse temperature effects on defect performance. To better assess cell crack degradation, either longer observation periods or more precise methods are needed

Electroluminescence (EL) imaging is a useful modality for the inspection of photovoltaic (PV) modules. EL images provide high spatial resolution, which makes it possible to detect even finest defects on the surface of PV modules. However, the analysis of EL images is typically a manual process that is expensive, time-consuming, and requires expert knowledge of many different types of defects. In this work, we investigate two approaches for automatic detection of such defects in a single image of a PV cell. The approaches differ in their hardware requirements, which are dictated by their respective application scenarios. The more hardware-efficient approach is based on hand-crafted features that are classified in a Support Vector Machine (SVM). To obtain a strong performance, we investigate and compare various processing variants. The more hardware-demanding approach uses an end-to-end deep Convolutional Neural Network (CNN) that runs on a Graphics Processing Unit (GPU). Both approaches are trained on 1968 cells extracted from high resolution EL intensity images of mono- and polycrystalline PV modules. The CNN is more accurate, and reaches an average accuracy of 88.42%. The SVM achieves a slightly lower average accuracy of 82.44%, but can run on arbitrary hardware. Both automated approaches make continuous, highly accurate monitoring of PV cells feasible.

Polymers are essential for long-term green energy production by photovoltaics (PV). We created a unique dataset that links electrical performance data of 28,030 PV modules, combined into 1295 strings and 423 inverters with the identity of the backsheet (BS) materials of each of the tested modules. This dataset gives an insight into BS-versatility and resulting impact on the insulation resistance of strings and inverters. The BSs were analyzed using near infrared absorption spectroscopy (NIRA) plus visual inspection in the field. The insulation resistance (Riso) of selected modules and strings was measured on-site. Furthermore, reported ground impedance (GI) values were evaluated with respect to an inverter-specific threshold below which the inverters do not start operation. We differentiated seven BS-types, the three most frequent being polyamide (PA), fluorinated coating (FC), and polyvinylidene fluoride (PVDF). Although modules with PA show often BS cracking, the insulation resistance was above the threshold. Similar, the GI of inverters with modules with PVDF-BSs was high. Modules with FC-BSs feature generally low Riso–values on module-, string-, and on inverter level. Evaluating historic data reveals an increase in frequency of GI-values below the threshold for inverters connected to modules with exclusively FC-BSs. While inverters with modules having PA- or PVDF-BSs function as expected, inverters with modules having FC-BSs require maintenance to keep power yield on an acceptable level. The dataset collected in the present work highlights that the choice of BS-material of PV-modules is essential for reliable operation of PV power stations.

Cost-effective, fast, and non-destructive, on-site photovoltaic (PV) characterization methods are of interest to PV operators to determine countermeasures against defects causing power loss or against safety problems. Combining the advantages of both methods electroluminescence (EL) and thermography is photoluminescence (PL). With our PL setup, we achieved high resolution luminescence images of large area PV modules without any physical and electrical contact. Defects are recognizable with a high rate compared to indoor EL images at controlled conditions. We analyzed inactive areas, cracks, potential induced degradation, snail trails, EVA degradation, and interconnection failures and compared the PL images with different characterization methods.

The individual causes for power loss of photovoltaic modules are investigated for quite some time. Recently, it has been shown that the power loss of a module is, for example, related to the fraction of inactive areas. While these areas can be easily identified from electroluminescense (EL) images, this is much harder for photoluminescence (PL) images. With this work, we close the gap between power regression from EL and PL images. We apply a deep convolutional neural network to predict the module power from PL images with a mean absolute error (MAE) of 4.4 ± 4.0 % or 11.7 ± 9.5 W P . Furthermore, we depict that regression maps computed from the embeddings of the trained network can be used to compute the localized power loss. Finally, we show that these regression maps can be used to identify inactive regions in PL images as well.

Automated inspection plays an important role in monitoring large‐scale photovoltaic power plants. Commonly, electroluminescense measurements are used to identify various types of defects on solar modules, but have not been used to determine the power of a module. However, knowledge of the power at maximum power point is important as well, since drops in the power of a single module can affect the performance of an entire string. By now, this is commonly determined by measurements that require to discontact or even dismount the module, rendering a regular inspection of individual modules infeasible. In this work, we bridge the gap between electroluminescense measurements and the power determination of a module. We compile a large dataset of 719 electroluminescense measurements of modules at various stages of degradation, especially cell cracks and fractures, and the corresponding power at maximum power point. Here, we focus on inactive regions and cracks as the predominant type of defect. We set up a baseline regression model to predict the power from electroluminescense measurements with a mean absolute error (MAE) of 9.0 ± 8.4WP (4.0 ± 3.7%). Then, we show that deep learning can be used to train a model that performs significantly better (7.3 ± 6.5WP or 3.2 ± 2.7%) and propose a variant of class activation maps to obtain the per cell power loss, as predicted by the model. With this work, we aim to open a new research topic. Therefore, we publicly release the dataset, the code, and trained models to empower other researchers to compare against our results. Finally, we present a thorough evaluation of certain boundary conditions like the dataset size and an automated preprocessing pipeline for on‐site measurements showing multiple modules at once.

Degradation of backsheets (BS) and encapsulant polymer components of silicon PV modules is recognized as one of the main reasons for losses in PV plant performance and lifetime expectations. Here, we report first insights into the correlation between BS composition of PV-modules and PV power station performance by using a combination of lab- and field-imaging, as well as spectroscopic and electrical characterizations. Using field-suitable near-infrared absorption (NIRA) spectroscopy, the BS structure of 518 PV-modules, 2.5 percent of the PV-modules in a 5 MWp PV power station, was identified on-site. The variance of the BS composition was found to be heterogeneous across PV-modules of the same power class from the same manufacturer. Polymaide-based BS cause in 10 out of 100 inverters ground impedance values below 400 kOhm, which is a typical threshold for inverters connecting to the grid. For primer-based BS this low value is reached 20 times. We conclude these numbers best as possible from the available monitoring data of inverters associated with BS-types from NIRA. Challenging is the identification of degraded primer-based BSs, since visually they look healthy and undistinguishable to well performing BSs. The present results demonstrate that a deeper understanding of the relationship between bill-of-materials and performance of PV-modules is necessary to avoid/minimize inverter shut-downs, and that it can be achieved by using combinations of selected field and lab characterization methods with monitoring data., 23 pages, 9 figures

Two testing methods for photovoltaic (PV) plants, an infrared (IR)-measurement aerial system and a monitoring system on module level, are compared with respect to their capabilities for identifying irregularities of PV panels in a PV plant. For the first method, a hypothesis is tested that infrared temperature and module power can be correlated with an empirical linear correlation using the raw measurement data of the second method. A workaround is proposed how to quantify power losses for unknown PV plants showing at the same time how power losses depend on ambient conditions. Also, the data is used to improve results from the IR-measurements. The second method is explained to compare the two methods. It is shown that each method offers different advantages, helping to properly assess and schedule O&M measures.

Machine vision and applications 32(4), 84 (2021). doi:10.1007/s00138-021-01191-9, Published by Springer, Berlin; Heidelberg

Drone-based aerial thermography has become a convenient quality assessment tool for photovoltaic power plants. Often caused by defective photovoltaic modules and cells, thermal abnormalities can be localized and classified. So far, infrared images cannot be utilized for the quantitative estimation of the generated power of a solar park yet. In this context, three photovoltaic power plants with an overall nominal power of approximately 9.4 MW p are evaluated. The analysis of both the radiometric data and the monitoring data of the inverters indicate a strong correlation between defective modules and the power of the associated string.