Browsing by Author "Pratt, Lawrence E"
Now showing 1 - 20 of 28
Results Per Page
Sort Options
Item A benchmark dataset for defect detection and classification in electroluminescence images of PV modules using semantic segmentation(2023-12) Pratt, Lawrence E; Matheus, Jana; Klein, RElectroluminescence (EL) images enable defect detection in solar photovoltaic (PV) modules that are otherwise invisible to the naked eye, much the same way an x-ray enables a doctor to detect cracks and fractures in bones. This paper presents a benchmark dataset and results for automatic detection and classification using deep learning models trained on 24 defects and features in EL images of crystalline silicon solar cells. The dataset consists of 593 cell images with ground truth masks corresponding to the pixel-level labels for each feature and defect. Four deep learning models (U-Net_12, U-Net_25, PSPNet, and DeepLabv3+) were trained using equal class weights, inverse class weights, and custom class weights for a total of twelve sets of predictions for each of 50 test images. The model performance was quantified based on the median intersection over union (mIoU) and median recall (mRcl) for a subset of the most common defects (cracks, inactive areas, and gridline defects) and features (ribbon interconnects and cell spacing) in the dataset. The mIoU measured higher for the two features compared to the three defects across all models which correlates with the size of the large features compared to the small defects that each class occupies in the images. The DeepLabv3+ with custom class weights scores the highest in terms of mIoU for the selected defects in this dataset. While the mIoU for cracks is low (25%) even for the DeepLabv3+, the recall is high (86%), and the resulting prediction masks reliably locate the defects in complex images with both large and small objects. Therefore, the model proves useful in the context of detecting cracks and other defects in EL images. The unique contributions from this work include the benchmark dataset with corresponding ground truth masks for multi-class semantic segmentation in EL images of solar PV cells and the performance metrics from four semantic segmentation models trained using three sets of class weights.Item Characterization of solar photovoltaic modules in outdoor and indoor conditions(CIGRE Southern Africa National Committee, 2019-10) Mkasi, Hlaluku W; May, Siyasanga I; Basappa Ayanna, Manjunath; Pratt, Lawrence E; Roro, Kittessa TThis paper summarizes the results of solar photovoltaic (PV) module electrical characterization under outdoor and indoor conditions in the Republic of South Africa. Under outdoor conditions, the bi-facial PV modules on fixed racks and dual-axis trackers yielded more energy than mono-facial modules on the same mounting configuration due to the active backside. The indoor current-voltage (IV) measurements conducted on an A+A+A+ sun simulator matched within ± 3% of the manufacturer’s nameplate ratings on a sample of crystalline PV modules, with some exceptions. The measured electrical characteristics from the outdoor system was corrected to standard test conditions (STC) and compared with indoor measurements at STC. The corrected maximum power from the outdoor system measured roughly 5% below maximum power as measured on the indoor system.Item Defect detection and quantification in electroluminescence images of solar PV modules using U-net semantic segmentation(2021-11) Pratt, Lawrence E; Govender, Devashen; Klein, RElectroluminescence (EL) images enable defect detection in solar photovoltaic (PV) modules that are otherwise invisible to the naked eye, much the same way an x-ray enables a doctor to detect cracks and fractures in bones. The prevalence of multiple defects, e.g. micro cracks, inactive regions, gridline defects, and material defects, in PV module can be quantified with an EL image. Modern, deep learning techniques for computer vision can be applied to extract the useful information contained in the images on entire batches of PV modules. Defect detection and quantification in EL images can improve the efficiency and the reliability of PV modules both at the factory by identifying potential process issues and at the PV plant by identifying and reducing the number of faulty modules installed. In this work, we train and test a semantic segmentation model based on the u-net architecture for EL image analysis of PV modules made from mono-crystalline and multi-crystalline silicon wafer-based solar cells. This work is focused on developing and testing a deep learning method for computer vision that is independent of the equipment used to generate the EL images, independent of the wafer-based module design, and independent of the image quality.Item The effect of soiling on the PV performance ratio for different PV systems(SASEC 2018, 2018-06) Mienie, S; Basappa Ayanna, Manjunath; Pratt, Lawrence E; Roro, Kittessa T; Diale, MSoiling is an important and unpredictable factor that affects PV plant performance. Soiling is dependent on so many factors making it hard to quantify and accurately incorporate in PV prediction models and levelized cost of electricity (LCOE). In this experiment, a comparison was done to establish the effect of soiling for 2 different PV plants installed at the Council of Scientific and Industrial Research (CSIR) campus in Pretoria. The soiling rate for the performance ratio (PR) was 1.2 % per week on the 558 kWp DC single-axis tracker system and 0.4 % per week on the 202.3 kWp DC dual-axis plant over the period from June 2017 to August 2017. Soiling has a smaller effect on dual-axis systems as compared to single-axis systems due to the soiling accumulation mechanics and PV system elevation above ground level.Item Effect of using global horizontal or plane of array irradiance for monitoring sun tracking solar photovoltaic plants performance(Southern African Solar Energy Conference (SASEC), 2018-06) Basappa Ayanna, Manjunath; Pratt, Lawrence E; Mienie, S; Roro, Kittessa TThe performance ratio (PR) is a metric commonly used for assessing the performance of a photovoltaic (PV) plants worldwide. The standard PR calculation method (PRPoA) used in solar PV industry as per the International Electro-technical Commission (IEC) standard uses the solar irradiance in the Plane of Array (PoA) as reference. The CSIR PR calculation method (PRGHI) used Global Horizontal Irradiance (GHI) as a reference for the first phase of PV plants realized as part of the Energy Autonomous Campus program. The PRGHI holds the EPC contractor liable for any underperformance due to poor layout or tracker operations. The PRGHI is variable in the short term and highly variable across the seasons compared with the PRPoA. The temperature corrected PRGHI and PRPoA are calculated to minimize any biases due to temperature which may arise from different weather conditions. The PRPoA is less variable in the short term and across the seasons, so any loss or gain in performance is more easily identified for both the single axis and dual axis tracker systems. The seasonal effect on PRGHI is larger for the dual axis tracker system compared to single axis tracker PV system. The increasing PRGHI between summer and winter solstice for the dual axis tracker system is due to a decrease in the reference irradiance caused by the higher angle of incidence of the sun on the GHI reference sensor and is not due to an increase in PV electrical output. The predicted and actual PRGHI and PRPoA are compared for the single and dual axis tracker systems for a period of one year without temperature correction. The absolute delta between the predicted and actual PR for the single axis tracker was -10% for PRGHI and -4.5% for PRPoA. The absolute delta for the dual axis tracker was -4.5% for PRGHI and +0.4% for PRPoA. The large absolute PRGHI delta between the actual and predicted PRs for the single axis tracker is investigated and this paper focuses only on the tracker performance. The actual tracker tilt angle performed optimally during the summer solstice period but suboptimally during the early morning and late afternoon between the summer and winter equinoxes. Future work will characterize the effect of the sub-optimal tracker performance in terms of PoA irradiance and energy production.Item Electroluminescence imaging of photovoltaic modules(2019-10) Malefafana, MT; Pratt, Lawrence E; Basappa Ayanna, Manjunath; Roro, Kittessa TElectroluminescence (EL) imaging is a non-destructive characterization technique used widely across the solar PV industry. Similar to an x-ray image, the EL image allows the analyst to peer inside the PV module and see features that are otherwise impossible to view with the naked eye. Those features include grain boundaries, interconnects, defects in the silicon, cracked cells and inactive cell areas. The integrity of the solar cells is critical because the cells are the generators that convert solar radiation into direct current electrical energy. Therefore, EL is used at a number of points in the supply chain to minimize the negative impact from cracked cells and inactive areas. EL images are collected at the factory during manufacturing processes, quality assurance batch testing, routine operations and maintenance as well as during failure analysis of fielded modules in real world conditions. In this paper, the basic techniques used to capture EL images at the CSIR are presented. Some of the typical features that can be seen on multi-crystalline silicon and mono-crystalline silicon photovoltaic (PV) modules, as well as bi-facial modules are presented. Two case studies have been analyzed as part of the failure analysis. One shows the impact that a handling mishap can have on the integrity of the solar cells. The second case study shows the use of EL, infrared thermography and the electrical characterization to investigate a module with burned back sheet.Item The field performance of a 558 kWp ground mounted single-axis PV system in Pretoria, South Africa(2017-12) Roro, Kittessa T; Basappa Ayanna, Manjunath; Pratt, Lawrence E; Lekoloane, Gaoshitwe A; Klein, Peter; Koopman, Stephen JAccording to a recent Green Tech Media (GTM) publication [Moskowitz, 2017], 23% of all ground-mounted photovoltaic (PV) systems are installed on trackers. Trackers are especially beneficial for commercial and utility scale applications, particularly in sunniest parts of the world where the direct normal irradiance (DNI) component of sunlight is relatively high. However, very few studies have been reported on the field performance of single axis tracking systems installed and operated in the Southern Africa region. In this study, therefore, the real-world performance of a 558 kWp ground mounted single-axis PV system at the Council for Scientific and Industrial Research (CSIR) Pretoria campus over a period of 21 months was analyzed. The plant performance was evaluated in terms of the standard performance ratio based on the irradiance in the plane of array and in terms of the performance ratio based on the Global Horizontal Irradiance. It is anticipated that the result of the study will provide useful inputs in assisting system designers, researchers and operators of solar PV plants in South Africa in understanding the optimal operation of solar tracking systems.Item Field performance of bifacal modules in South Africa(SASEC, 2019-11) Basappa Ayanna, Manjunath; Mkasi, Hlaluku W; Pratt, Lawrence E; Roro, Kittessa TThis study presents the initial results of field performance analysis conducted on bi-facial modules installed on a fixed tilt rack and dual axis tracker system in Pretoria, South Africa. Bifacial modules produce better energy yield compared to monofacial modules depending on the geographical location, albedo of the back surface, height of a module from the surface and tilt angle. Currently multiple standards are under development to predict the bifacial gains and the field performance data is required to validate the models. This paper presents the bifacial gains over the monofacials without a high albedo back surface in a fixed tilt system, intercomparison of the grass cover, unpainted and white painted pavement and overall bifacial gain over the monofacials in dual axis tracker system. The bifaciality factor and efficiency curve of two bifacial types measured on indoor sun simulator is presented.Item Financial benefits of integrating east–west fixed, single and dual axes tracking PV systems in the supply mix of a South African Research Campus(2018-08) Tazvinga, Henerica; Moholisa, Tsolo E; Basappa Ayanna, Manjunath; Pratt, Lawrence E; Hlatshwayo, Mluleki C; Mushwana, CrescentThe use of Photovoltaic (PV) systems in distribution networks has increased in the past few years. The benefits of PV systems connected to the grid include reducing the carbon footprint, improving energy access, and security as well as the management of demand. The cost of PV systems has dropped rapidly in recent years, and the trend is expected to continue in future making this clean technology more attractive economically. This paper presents three different PV systems (single axis tracker, dual axis tracker and fixed rooftop) connected to the grid. Data measured over a period of a year are used in the analysis to quantify the benefits of such systems for a South African commercial entity. The results show considerable benefits attributable to the use of various PV system configurations. The outcomes of this work illustrate that the various configurations examined could result in significant benefits for industrial and commercial entities in cost reductions and addressing environmental issues.Item Forecast accuracy for solar resource in Tshwane using solcast data services(2024-11) Pratt, Lawrence E; Basappa Ayanna, Manjunath; Mkasi, Hlaluku WThis paper compares forecasted global horizontal irradiance (GHI) data from SolCast to GHI recorded at the CSIR meteorological station in Tshwane, South Africa. The study compares the SolCast forecast to the CSIR measured values and computes accuracy metrics for +1-hour, solar noon, +1-day, and +7-day ahead forecasts. The +1-hour ahead and solar noon forecasts are based on the average value for a single hour, while the +1-day and +7-day ahead are based on the insolation over 24 hours. Based on the limited data collected for this study, the normalised root mean squared deviation (nRMSD) is 28.9% for the +1-hour ahead forecast at the CSIR for forecasts made in the morning. This forecast error from the morning forecasts is nearly double the nRMSD reported in a study conducted by the SolCast team on forecasts across 33 sites worldwide for forecasts made at all hours of the day (16.6%). The high forecast error is likely due to the relatively low irradiance value forecasted and measured in the morning hours of the winter season captured in the study dataset. By contrast, the nRMSD for forecasts made for solar noon of the same day was 11.3%, which falls within the range of values reported by the SolCast team. The normalised mean bias deviations (nMBD) for the daily insolation for +1-day and the +7-day ahead forecasts were both 5.2% at the CSIR.Item Global irradiance on photovoltaic array(2018-06) Mkasi, Hlaluku W; Basappa Ayanna, Manjunath; Pratt, Lawrence E; Roro, Kittessa TThe amount of available solar irradiance is the most important parameter to determine solar energy generation for any given location. Satellite-based irradiance data is often used as a basis for predicting the energy generation for any location, as ground-based irradiance data is only available for specific sites. This paper presents a comparison of ground-based measured irradiance and satellite-based predicted irradiance and the corresponding modelled alternating current (AC) energy generation from a photovoltaic power plant. The 558 kW DC single-axis tracked photovoltaic (PV) plant located at the Council for Scientific and Industrial Research (CSIR) Pretoria campus is used for the case study. This paper briefly describes the techniques used to determine the Plane of Array (POA) irradiance from the Global Horizontal Irradiance (GHI) as well as the theory of operation for ground-based GHI sensors. The annual ground-based measured GHI at two locations with the CSIR campus were 3.1% and 4.9% lower compared to PVGIS Typical Meteorological Year (TMY) (1998-2011) and 1.9% and 3.6% lower compared with Meteonorm TMY (1960-2000) data, respectively. System Advisor Model (SAM) was used to model the AC energy output from the PV system. A strong correlation between the POA irradiance and alternating current (AC) power generation was calculated (R-squared = 0.987). The annual AC energy generation based on ground-based GHI measurements predicted 3.6% and 5.9% lower compared to PVGIS and 2.4% and 4.7% lower compared to Meteonorm TMY data, respectively.Item Impact of measured spectrum variation on solar photovoltaic efficiencies worldwide(2022-08) Kinsey, GS; Riedel-Lyngskær, NC; Miguel, A; Boyd, M; Braga, M; Shou, C; Cordero, RR; Duck, BC; Maweza, Elijah L; Pratt, Lawrence EIn photovoltaic power ratings, a single solar spectrum, AM1.5, is the de facto standard for record laboratory efficiencies, commercial module specifications, and performance ratios of solar power plants. More detailed energy analysis that accounts for local spectral irradiance, along with temperature and broadband irradiance, reduces forecast errors to expand the grid utility of solar energy. Here, ground-level measurements of spectral irradiance collected worldwide have been pooled to provide a sampling of geographic, seasonal, and diurnal variation. Applied to nine solar cell types, the resulting divergence in solar cell efficiencies illustrates that a single spectrum is insufficient for comparisons of cells with different spectral responses. Cells with two or more junctions tend to have efficiencies below that under the standard spectrum. Silicon exhibits the least spectral sensitivity: relative weekly site variation ranges from 1% in Lima, Peru to 14% in Edmonton, Canada.Item The importance of PV module reliability research in South Africa(SASEC 2018, 2018-06) Pratt, Lawrence E; Roro, Kittessa T; Basappa Ayanna, ManjunathThis paper will introduce the new C450 international standard for PV module quality assurance programs. Quality assurance testing has been on the rise in the PV industry following the rise in development of large scale PV plants, sometimes under the name of “reliability” test programs or “bankability” studies. The test programs are largely based on published International Electrotechnical Commission (IEC) standards for PV module design certification and safety testing. The Council for Scientific and Industrial Research’s Energy Centre (CSIR-EC) in support from its Technology Localization and Implementation Unit (TLIU) is constructing a solar PV module reliability research and testing laboratory at its Pretoria campus to support the growing South African renewable energy market, mainly focussing PV module reliability and quality assurance testing. PV module reliability issues impact degradation rates, future energy production of PV plants, and ultimately the Levelized Cost of Energy (LCOE) paid over the lifetime of a PV system. The LCOE from the CSIR’s 558kW DC single-axis tracker plant would increase from 0.83 Rand/kWh to 0.90 Rand/kWh if the PV module degradation increases to 1.6% per year from the predicted 0.8% per year. The CSIR reliability laboratory supports the decisions made by project developers, installers, contractors, owners and investors that will ultimately lead to more reliable, lower PV generation costs for South African consumers by reducing the risk associated with PV module quality and reliability.Item Inter laboratory comparison of indoor performance tests on crystalline silicon solar PV modules(2019-11) Pratt, Lawrence E; Basappa Ayanna, Manjunath; May, Siyasanga I; Roro, Kittessa T; McCleland, JC; Zikhali, QThe paper presents results from an inter laboratory comparison (ILC) of crystalline silicon PV module performance measurements from three different institutions within South Africa. The analysis quantifies the differences in power, current, and voltage measurements and compares power to the stated uncertainties for the measurements. The paper also explains the basics of PV module performance measurements, the link to PV module nameplate ratings, and describes the sources of uncertainty in the measurements. The results of the 2019 ILC I-V measurements among three South African institutions show differences of +/- 2.0% or less for maximum power measurements of crystalline PV modules, relative to the CSIR results. These differences are within the stated uncertainty for the measurements but greater than the +/-0.5% differences among international reference labs reported in a 2017.Item Modelling photovoltaic power output using machine Learning techniques(2022-08) May, Siyasanga I; Pratt, Lawrence E; Roro, Kittessa T; Bokoro, PThis work focuses on modelling the power output of multiple PV technologies installed at the outdoor test facility of the Council for Scientific and Industrial Research (CSIR) in Pretoria. Extreme Gradient Boosting (XGB) and Artificial Neural Network (ANN) machine learning models are trained with historic time-series datasets (measured meteorological and PV electrical parameters) to model the historical output power of the photovoltaic (PV) system. To facilitate the training, sub-hourly measured data from January to November 2019 were averaged at hourly intervals. For testing, sub-hourly data from January 2020 to March 2020 was divided into clear, moderate, and cloudy skies monthly records. Outliers were identified and removed from the data during pre-processing. The short circuit current (Isc) and PV output have shown a very strong correlation (r 2 =0.93) since PV output is heavily influenced by array irradiance and current generation. PV output strongly correlated with plane of array irradiance and albedo (r 2 =0.83,0.69), and with module temperature (r 2 =0.70), depending on the module type. To quantify model accuracy, root mean squared error (RMSE) was used. ANN outperforms XGB by a wide margin based on the RMSE values. ANN produced the lowest RMSE values with 4. 1W to XGB record high 17. 5W during moderate skies. The majority of the observed maximum RMSE values came from XGB modelling. The trained models will be used to predict PV output power using only forecasted weather data as inputs in future work.Item Power output predictions of photovoltaic system using machine learning(2021-11) May, Siyasanga I; Pratt, Lawrence E; Roro, Kittessa T; Bokoro, PThis work focuses on developing prediction models for the power output of multiple PV technologies installed at the outdoor test facility on the Pretoria campus of the Council for Scientific and Industrial Research. Random Forest (RF) and Adaboost machine learning models are trained with historic time-series data sets (measured meteorological and PV electrical parameters) to predict historical output power of the photovoltaic (PV) system. Sub-hourly measured data from January 2019 until November 2019 was averaged to hourly intervals for training and testing. The data undergo a pre-processing step where outliers are identified and removed. A very strong correlation (r2 ~ 0.99) was calculated between Isc and PV output because PV output is largely determined by the plane of array irradiance and the resulting current generation. A strong correlation between PV output and plane of array (0.89 < r2 < 0.99) and between PV output and module temperature (0.62 < r2 < 0.72) are also calculated, depending on the module type. The models are then trained on the datasets and the accuracy is quantified based on the root mean squared error (RMSE) between the actual measured PV output and the predicted PV output of different PV technologies. RF generally outperformed the Adaboost regression. Both regression models achieved minimal RMSE on predictions for the thin film module technologies with maximum RMSE of 0.2 W for Adaboost and 1.2 W for the Random Forest. In future work, the trained models will be used to forecast future electricity production from PV plants using only forecasted weather data as inputs.Item Power quality analysis of a grid tied 3.3 kWp rooftop solar photovoltaic system with battery storage(2019-06) May, Siyasanga I; Basappa Ayanna, Manjunath; Pratt, Lawrence E; Mienie, S; Roro, Kittessa TThis paper outlines the initial results for power quality analysis (PQA) of a single phase 3.3kWp photovoltaic (PV) system with battery storage connected to a low-voltage system at the Council for Scientific and Industrial Research (CSIR), Pretoria campus. The purpose of this study is to understand the harmonic distortion behaviours for three different sky conditions: clear, moderate and cloudy condition. The measured current harmonic distortions are categorized to different order intervals as per NRS 097-2-1:2017. The categorized maximum current harmonic distortions of each hour interval within a day are compared against the requirements stipulated in NRS 097-2-1:2017. The measured current harmonic distortions are inclusive of existing harmonics in the building, hence it is difficult to quantify the harmonics generated from the solar PV. The future plans include measurement of these harmonics in the building without PV to determine the baseline and then calculate the harmonic distortions originating from the PV system for different sky conditions.Item Power quality analysis of low voltage grid-tied network with PV system(SASEC, 2019-11) Mnisi, MI; Bokoro, P; Roro, Kittessa T; Basappa Ayanna, Manjunath; Pratt, Lawrence EPower quality analysis (PQA) in power systems has become important to consider since there has been an increase in distributed energy resources installed on the existing grid, particularly in the area of renewable energy generation. South Africa is primarily depending on generating electricity from fossil fuel, as a result, the shortage of fossil fuel in the country has influenced the increase in electricity tariff. On the other hand, market price for installing photovoltaic (PV) system is decreasing, hence low voltage (LV) customers are beginning to utilise the advantage. Concerns have been raised about the power quality of grid-tied PV system because PV systems incorporate devices such as inverters for power conversion. Inverters have been regarded as a possible source of power quality problems according to many previous studies published. PQA of a low voltage South African network connected with PV system will be addressed in the paper. This paper analyses the level of harmonic distortions measured in the low voltage grid network the presence of PV system and compares the measured Total Harmonics Distortion on current (THDI) with the NRS 097 standard to verify if the network with PV system meets the requirements.Item PV module reliability scorecard - Round 1(2021-11) Pratt, Lawrence E; Basappa Ayanna, Manjunath; May, Siyasanga I; Mkasi, Hlaluku W; Maweza, Elijah L; Roro, Kittessa TSolar PV system component reliability is key to longterm success for the PV industry, given the 25-year lifespan of PV generators. PV modules are one key component needed to deliver on the lifetime production of PV systems, as they are the engines that convert the fuel from the sun into the electrons for use here on Earth. This research presents the results from the first round of reliability testing for commercially available PV modules based on accelerated stress tests conducted at the CSIR Energy Centre. Four module types with a specific bill of materials (BOM) were subjected to a series of accelerated stress tests as defined by the C450 international standard. Pre- and post-stress characterizations were conducted to quantify the changes in safety and performance due to the accelerated stress. PV module power decreased by less than 3% for all modules subjected to mechanical loads, humidity-freeze, thermal cycling, and PID. However, the degradation rates among the four BOMs differed significantly (p<0.001) following the thermal cycling sequence. PV module power decreased by as much as 9% following the damp heat stress prior to the final stabilization. The work was conducted to demonstrate the value of accelerated stress testing now available in South African to support the PV industry. The round one results highlight potential differences in the long-term performance of PV module BOMs in the field, supporting the value of accelerated stress testing to de-risk investments in PV generators.Item Solar PV indoor and outdoor testing(2018-10) Pratt, Lawrence EPresentation on the solar PV indoor and outdoor testing.