The PV-Degradation project is an applied research project funded by the Cyprus Research Promotion Foundation under grant number ΤΕΧΝΟΛΟΓΙΑ/ΕΝΕΡΓ/0311(ΒΙΕ)/12. The goal is to develop a protocol for evaluating degradation of thin-film and crystalline Silicon photovoltaics (PV) by combining performance monitoring from PV systems, individual PV modules and indoor characterization.
At the University of Cyprus we monitor and assess the performance of PV systems operating in the field. Measurements are collected from both grid-connected systems and from current-voltage (IV) characterization of modules on the plane of the array. Indoor characterization is performed using electroluminescence, dark IV and testing at Standard Test Conditions (STC) using a solar simulator.
PV-Degradation is a project that aims to analyse these measurements in order to formulate a robust protocol for the accurate determination of degradation. The protocol will be benchmarked in different regions and validated through round-robin procedures.
The project started in June 2013, and is scheduled to run for 24 months. It is co-financed by the European Regional Development Fund and by the Republic of Cyprus with grant number ΤΕΧΝΟΛΟΓΙΑ/ΕΝΕΡΓ/0311(ΒΙΕ)/12.
This paper provides a review of methodologies for measuring the degradation rate, RD, of photovoltaic (PV) technologies, as reported in the literature. As presented in this paper, each method yields different results with varying uncertainty depending on the measuring equipment, the data qualification and filtering criteria, the performance metric and the statistical method of estimation of the trend. This imposes the risk of overestimating or underestimating the true degradation rate and, subsequently, the effective lifetime of a PV module/array/system and proves the need for defining a standardized methodology. Through a literature search, four major statistical analysis methods were recognized for calculating degradation rates: (1) Linear Regression (LR), (2) Classical Seasonal Decomposition (CSD), (3) AutoRegressive Integrated Moving Average (ARIMA) and, (4) LOcally wEighted Scatterplot Smoothing (LOESS), with LR being the most common. These analyses were applied on the following performance metrics: (1) electrical parameters from IV curves recorded under outdoor or simulated indoor conditions and corrected to STC, (2) regression models such as the Photovoltaics for Utility Scale Applications (PVUSA) and Sandia models, (3) normalized ratings such as Performance Ratio, RP, and PMPP/GI and, (4) scaled ratings such as PMPP/Pmax, PAC/Pmax and kWh/kWp. The degradation rate results have shown that the IV method produced the lowest RD and LR produced results with large variation and the largest uncertainty. The ARIMA and LOESS methods, albeit less popular, produced results with low variation and uncertainty and with good agreement between them. Most importantly, this review showed that the RD is not only technology and site dependent, but also methodology dependent.
In this work, the seasonality and performance loss rates of eleven grid-connected photovoltaic (PV) systems of different technologies were evaluated through seasonal adjustment. The classical seasonal decomposition (CSD) and X-12-ARIMA statistical techniques were applied on monthly DC performance ratio, RP, time series, constructed from field measurements over the systems' first five years of operation. The results have shown that the RP of crystalline silicon (c-Si) technologies was higher during winter. This was also the case for the coppereindium gallium-diselenide (CIGS) and cadmium telluride (CdTe) technologies but with lower seasonal amplitude. The amorphous silicon (a-Si) technology exhibited a different seasonal profile, with high RP during summer and autumn and low during winter. In addition, the trends extracted from the application of CSD and X-12-ARIMA on three-year, four-year and five-year RP time series were used to estimate linear performance loss rates. A comparison between standard linear regression (LR), CSD and X-12-ARIMA has shown that CSD and X-12-ARIMA resulted in higher rates overall for c-Si, 1.07 and 0.93%/year respectively, but with significantly less uncertainty than LR. Lastly, it was shown that X-12-ARIMA provided statistical inference in the presence of outliers and produced model residuals that were uncorrelated, in contrast to CSD.
This paper presents the results of an extensive testing campaign for validating the time series analysis approach to the estimation of the linear field performance loss rates (PLR) of grid-connected photovoltaic (PV) systems of different technologies operating side-by-side at the PV Technology test site of the University of Cyprus since June 2006. Fifteen-minute average measurements of the array power at the maximum power point, PA, were used to construct time series of the performance ratio, RP, of each array. The time series were analysed with regARIMA and classical seasonal decomposition (CSD) in order to extract the trend. Then, linear regression (LR) was used to calculate the slope. To validate the results, all arrays were disassembled and every module was tested at Standard Test Conditions (STC) in a class A+A+A+ solar simulator, in order to calculate the nominal array degradation rate. Comparison of both methods has shown good agreement between the time series analysis approach and the indoor testing approach, for PV arrays with no identified failures through electroluminescence (EL). On the contrary, for modules with identified failures through EL, the nominal array degradation rate was higher in comparison to the field PLR. Differences between the two methods have been shown to be due to cracked cells, hotspots and spectral response mismatch. Lastly, the comparison has shown that amongst the time series analysis methods, regARIMA produced statistically significant PLR with low uncertainty and the best agreement with the nominal array degradation rates.
In this paper, the performance loss rates of different technology crystalline silicon (c-Si) and thin-film photovoltaic (PV) systems were estimated and compared over their first eight years of operation at the test site of the Photovoltaic Technology Laboratory, University of Cyprus (UCY) in Nicosia, Cyprus, by applying different statistical trend analysis methods on monthly Performance Ratio, RP, time series. The statistical trend analysis methods include Linear Regression (LR), Classical Seasonal Decomposition (CSD), Holt-Winters exponential smoothing (HW) and LOcally wEighted Scatterplot Smoothing (LOESS) and were applied on monthly constructed time series of RP, calculated from the fifteen-minute average array DC power at the maximum power point, PA, of each grid-connected PV system. The comparison of the estimated performance loss rates for each technology showed that the average performance loss rate of the c-Si systems was 0.75 ± 0.17 %/year. On the other hand, the average performance loss rate for the thin-film systems was 1.95 ± 0.11 %/year for all methods, with a 95 % confidence interval. The good agreement in the results between the different methods for each system also provided evidence that the performance loss rates have started to converge to a steady value. Finally, it was demonstrated that trend extraction techniques produced similar estimates between them and with very low uncertainty, even with less than five years of outdoor exposure, whereas LR was the least robust method for all technologies, since it was greatly affected by the seasonality and outliers of the time series and needed more years of data to produce reliable estimates.
The degradation rates of crystalline silicon (c-Si) and thin-film photovoltaic (PV) systems of different manufacturers and different technologies were calculated and compared for the systems’ first five years of outdoors exposure, by applying a host of different analysis methods, in order to quantify the differences between each method. These include linear regression using linear least squares (LLS), classical seasonal decomposition (CSD) and seasonal-trend decomposition by Loess (STL) on daily and monthly time series of two performance metrics, performance ratio (PR) and PR with temperature correction (PR-TC). The comparison of the resulting degradation rates for each PV group (c-Si and thin-film) showed that the monthly PR-TC-STL method provided the lowest standard deviation and a mean degradation rate of 1.12 %/year for the c-Si PV systems. On the other hand, the daily PR-TC-LLS method demonstrated the lowest standard deviation and an average degradation rate of 2.47 %/year for the thin-film PV systems. Linear regression using LLS produced the lowest degradation rates overall, but when temperature correction was applied, the calculated degradation rates were increased by 0.4 %/year. LLS also showed the lowest standard deviation for the thin-film PV systems, which was further reduced by applying temperature correction.
In this work, the performance loss rates of eleven grid-connected photovoltaic (PV) systems of different technologies were evaluated by applying linear regression (LR) and trend extraction methods to Performance Ratio, RP, time series. In particular, model-based methods such as Classical Seasonal Decomposition (CSD), Holt-Winters (HW) exponential smoothing and Autoregressive Integrated Moving Average (ARIMA), as well as non-parametric filtering methods such as LOcally wEighted Scatterplot Smoothing (LOESS) were used to extract the trend from monthly RP time series of the first five years of operation of each PV system. The results showed that applying LR on the time series produced the lowest performance loss rates for most systems, but with significant autocorrelations in the residuals, signifying statistical inaccuracy. The application of CSD and HW significantly reduced the residual autocorrelations as the seasonal component was extracted from the time series, resulting in comparable results for eight out of eleven PV systems, with a mean absolute percentage error (MAPE) of 6.22 % between the performance loss rates calculated from each method. Finally, the optimal use of multiplicative ARIMA resulted in Gaussian white noise (GWN) residuals and the most accurate statistical model of the RP time series. ARIMA produced higher performance loss rates than LR for all technologies, except the amorphous Silicon (a-Si) system. The LOESS non-parametric method produced directly comparable results to multiplicative ARIMA, with a MAPE of -2.04 % between the performance loss rates calculated from each method, whereas LR, CSD and HW showed higher deviation from ARIMA, with MAPE of 25.14 %, -13.71 % and -6.39 %, respectively.