D.C. Jordan1*, R.M. Smith1, S.R. Kurtz1, N. G. Dhere2, S.A. Pethe2, A.Kaul2, S. Pulver3, A. Cronin3
1National Renewable Energy Laboratory (NREL), 1617 Cole Blvd., Golden, CO 80401
2Florida Solar Energy Center (FSEC), 1679 Clearlake Road, Cocoa, FL 32922-5703
3University of Arizona, Department of Physics, PO Box 210081, Tucson, AZ 85721
In this paper we present the results of an outdoor study of copper-indium-gallium-diselenide (CIGS) photovoltaic (PV) modules from Shell Solar. Two different module types are compared at three different locations. ST-80 modules were monitored at the Outdoor Test Facility (OTF) at the National Renewable Energy Laboratory (NREL) in Golden, Colorado. ST-40 modules, the predecessor module to the ST-80 were investigated in three different climates. These locations included the cooler climate at NREL, the hot and humid climate at the Florida Solar Energy Center (FSEC) in Cocoa, Florida and the hot and dry climate at the Tucson Electric Power PV yard in Tucson, Arizona. Annual degradation rates were determined and the impact of module development progression on the stability is assessed.
The ability to accurately predict power delivery over the course of time is key to the maturing Photovoltaics (PV) industry growth . For realistic PV lifespan estimation, the knowledge of power decline over time is essential. It is important for all involved stakeholders: utility companies, investors and researchers alike. Utility companies require this knowledge because of their obligation to deliver power reliably to their customers. Investors must estimate the internal rate of return of their investments and allocate resources accordingly. Researchers and developers need to estimate the time when a module has declined to a point where it is considered failed, and then study the mechanisms of failure, which, in turn, will lead to improved modules in the future. Indoor testing has played an important role in testing PV modules and certifying them for field deployment. However, a clear correlation between accelerated testing and outdoor performance is still being sought. In the absence of such a clear correlation outdoor, field testing remains a vital part of estimation of PV field performance and lifetime.
One difficulty with outdoor testing is that due to seasonal changes, several complete cycles are required in order to obtain reasonably accurate degradation rates. Thus, quality data needs to be maintained and monitored for several consecutive years. Another challenge of outdoor testing is that degradation rates depend on the climate at the test location. Various studies of different PV technologies in different locations have been undertaken in the past [3,4,5,6,7] but direct comparison is often difficult especially for thin films like CIGS due to transient effects. Finally, financial considerations limit the number of PV modules at each location thus making it difficult to account for possible manufacturing variations.
1. 14 Shell Solar ST-80 modules were installed as two parallel 7-module series-connected strings for a nominal power output of 1120W DC under Standard Test Conditions (STC). The modules, manufactured in 2005, were continuously monitored at the Outdoor Test Facility (OTF) at NREL from January 2006 to the present in 2009. The array was mounted at latitude tilt of 40º facing true South. The array was tied into the utility grid through a SMA Sunnyboy inverter with its own maximum-power-point-tracking algorithm. A Campbell-Scientific CR10 data acquisition system was used to collect plane-of-the-array irradiance from a Kipp & Zonen CM11 pyranometer, ambient and module temperature and power data. Additional meteorological data were obtained from NREL’s Reference Meteorological and Irradiance System (RMIS) located approximately 20m from the PV array. The data acquisition frequency was 5 s and stored as 15-min averages for PV array parameters and 1-min averages for RMIS parameters. Quarterly field I-V curves were obtained with a portable Daystar I-V tracer on a cloudless day around solar noon and corrected for temperature and irradiance in accordance with IEC 60891.
2. 28 Shell Solar ST-40 modules with 2 sub-arrays consisting of 5 parallel strings of 2 modules in series and 1 sub-array consisting of 4 parallel strings of 2 modules in series were installed for a DC output of 1120W, STC. The modules, manufactured in 1998, were continuously monitored at the same facility at NREL from December 1999 to May 2004. The array was also mounted at latitude tilt of 40º facing true South. Each sub-array was connected to an AERL 1200B peak-power tracker, which, in turn, was connected to a 0.95 Ω, 2kW fixed resistive load. Data acquisition system information is the same for the above listed ST-80 system.
48 ST-40 modules were divided into 2 strings, each string was held at negative and positive potential with respect to ground for a nominal maximum DC power output of 960 W, STC. The modules, manufactured in 2003, were monitored April 2005 to September 2007 at FSEC. While installation of the modules had been completed at the end of 2003, the very active Atlantic hurricane season of 2004 led to a number of disruptions in 2004 such that continuous monitoring did not commence until 2005. In addition, the modules were sent back to NREL in 2005 for I-V characterization. The modules were mounted at a latitude tilt of ca. 29º facing true South. The array was a stand-alone system with DC power delivered to a series of fixed 175 Ω precision load resistors. The data acquisition system consisted of two AM 16/32 Multiplexers and a Campbell-Scientific CR10 datalogger to collect all voltage and current data, 2 pyranometers readings and other meteorological data. The data acquisition frequency was 15 s and stored as 15-min averages for all data. Similarly to NREL, quarterly field I-V curves were obtained with a portable Daystar I-V tracer and corrected for temperature and irradiance in accordance with IEC 60891.
38 Shell Solar ST-40 modules were installed for a nominal power output of 1520W DC, STC. The modules, manufactured in 2003, were continuously monitored at the Tucson Electric Power PV yard, from September 2003 to the present in 2009. In 2008, Tucson Electric Power authorized University of Arizona researchers to conduct performance measurements on the PV yard. The array was mounted at latitude tilt of 32º facing true South. It was tied into the utility grid through a Xantrex TR inverter with its own maximum-power-point tracking. AC power was monitored with a utility-grade AC current meter. Computerized data acquisition of DC power and meteorological data was initiated in October 2008. For meteorological data prior to this date, data were obtained from the Arizona Meteorological Network station, approximately 10 km distant from the PV yard. The data acquisition frequency was 5 min for the AC power data and was stored as 15-min averages. The sampling frequency for the meteorological data was hourly such that the data had to be interpolated to 15-min intervals.
Degradation Rate Measurements
Power data were normalized to Performance Test Conditions (PTC)  by using the PVUSA multiple regression of equation 1. Monthly data were then charted as a time series over several seasonal cycles to study degradation rates. [11,12]
P is the calculated DC power, E is the irradiance, T the ambient temperature, ws the wind speed and a1 to a4 are regression coefficients. Data at low irradiance levels, i.e. below 700 W/m2 were eliminated from the analysis for 2 reasons: First of all, extrapolation from low-irradiance levels to PTC conditions, increases the model uncertainty. Second of all, it has been shown that at low-light conditions the performance of maximum-power-point tracking and resistively loaded systems deviate substantially. However, at high-irradiance levels with an appropriately calibrated and selected load resistor, the resistive load approximates the maximum power, although with more scatter in the data. The degradation rate is then determined by using a linear standard least square fit to the time series. The analytical uncertainty quoted below is determined from the Standard Errors of the linear fit using Gaussian error propagation.
Figure 2 shows the DC power regression at PTC conditions for the ST-80 array located at NREL. The degradation rate RD is determined to be (+0.4±0.5) %/year. The degradation rate determined from maximum power points of field I-V data curves is determined to be (-0.2±0.4)%/year and agrees with the PTC regression within the statistical uncertainty. In addition, the nominal power rating for the array, corrected for PTC conditions, is shown as a dashed line. It can be seen that the ST-80 modules show no significant degradation after more than 4 years of field exposure in Colorado and still perform to the manufacturer’s specifications.
Figure 3 shows the PTC regression analysis for the ST-40 modules located at NREL. The degradation rate is determined to be (-3.5±0.4)%/year. It can be seen that the array at the beginning performed according to the manufacturer’s specifications. However, over the course of several seasonal cycles the performance decreases markedly. Although there is a substantial gap in manufacturing year the later ST-80 modules exhibit much better long-term stability.
The PTC regression analysis for the modules deployed in Florida is shown in Fig.4. Due to the above-mentioned weather-related interruptions, it can be seen that the FSEC monitoring time span is shorter than the periods at NREL. The 7-month interruption in data collection from September 2005 to approximately May-06 while the modules were awaiting I-V characterization at NREL is indicated by the broken horizontal axis of Fig. 3. The modules were stored in a cool, dark room during this time period and were not exposed to further field degradation. Degradation rates were determined to be (-4.1±2.2)%/year and (-3.8±2.3)%/year for the positive and negative string, respectively. The difference in degradation rate between the positive and negative string is within statistical uncertainty. In addition, the degradation rate from maximum-power points of field I-V curves was determined to be (-3.1±1.6)%/year and agrees well with the rate determined from PTC regression method. Furthermore, the nameplate rating adjusted for PTC conditions is also shown in Fig. 4 by the dashed line. It can be seen that degradation rate between the ST-40 modules of Colorado and Florida are similar. The uncertainty of the Florida data is larger possibly due to (a) a much shorter monitoring time period and (b) the absence of maximum-power tracking.
The results for the PTC regression obtained for the Arizona data is shown in Fig. 5. The degradation rate was determined to be (-1.5±0.6) %/year which is smaller than the Colorado or the Florida data. However, the uncertainty is slightly larger than the ST-80 data set in Colorado despite maximum-power tracking. This may be due to that (a) inverter efficiency is convoluted in the measurement because only AC power was measured and (b) meteorological data had to be obtained from a nearby site and needed to be interpolated. The results for the PTC regression obtained for the Arizona data are consistent with the observation that the energy yields from the ST-40 system decreased by (-1.6±0.4) %/year relative to eight other systems at the same test yard in Arizona. This is the mean and standard deviation of the relative degradation rate for the ST-40 modules compared to a group that includes three systems with polycrystalline silicon modules, two systems with multijunction silicon modules, two systems with monocrystalline silicon modules, and one system with amorphous silicon modules. While these comparison systems may be degrading as well, our analysis of the AC power output from these systems over three years confirms that the grid-tied ST-40 modules in Arizona degraded more rapidly than the comparison systems.
The main challenge of this paper was the different methodologies used in the data collection, as outlined in the experimental section. The PV community will face a similar challenge over the next few years as more and more PV system will be installed world-wide.
Table 1 summarizes the determined degradation rates from the various locations. Shell Solar ST-40 modules showed a significant degradation rate in all 3 locations, Colorado, Florida and Arizona while ST-80 modules showed remarkable stability over several years in Colorado. The ST-40 modules in Florida appear to be degrading at the highest rate though a conclusive statement is difficult due to the large uncertainty. The detrimental impact of temperature and humidity on PV reliability has been well documented. Thus, it would not be surprising to find a significant degradation rate difference between the cool and dry climate of Colorado and the hot and humid climate of Florida. The Colorado ST-40 data show an annual decline almost as much as the Florida data, apparently refuting the impact of climate differences on degradation. However, it has to be kept in mind that the Colorado modules, though of the same type, were manufactured 5 years earlier compared to the Arizona and Florida locations. It is a reasonable assumption that the earlier modules at Colorado represented an earlier revision than the later modules at Arizona and Florida exacerbating the degradation rate
Typically, NREL has observed on average annual degradation rates of (-1.0 ± 1.0)%/year across several dozen array systems. Therefore, the ST-80 modules appear remarkably stable to this overall average while the ST-40s underperform. Although the impact of climate cannot be excluded it seems likely that module improvement from ST-40 to ST-80 resulted in better stability of the latter one. The ST-80 module featured several manufacturing improvements, amongst them improvements to the packaging which might have made them less susceptible to climatic conditions such as moisture ingress. The design of the ST-40 uses ethylene vinyl acetate (EVA) to laminate circuit plates to a cover glass sheet and use a Tedlar/polyester/Al/Tedlar (TPAT) backsheet. The EVA completely surrounds the circuit plate and forms a hermetic seal. For the ST-80 module the design was modified towards a glass-on-glass module using a low moisture permeable edge seal which seems a likely reason for its observed stability. 
While the apparent stability is interesting to note it needs to be verified by (1) extending the monitoring period and/or (2) testing the same set of modules in a harsher hot and humid climate.
We presented results of an outdoor study of copper-indium-gallium-diselenide (CIGS) photovoltaic (PV) modules from Shell Solar. Two different module types were compared at three different locations. It was found that ST-80 modules showed marked stability over 4 years while ST-40 modules showed significant declines in all 3 locations.
The stability of the ST-80 modules is interesting to note, however it needs to be verified by extending the monitoring period and thus determining the degradation rate more accurately. Alternatively, the modules could be tested in a hot and humid climate known to exacerbate degradation rates. The apparent stability also emphasizes the need for long-term outdoor testing where significant degradation may not be confidently measurable until after several years of monitoring. It is essential to carry out these long-term studies to accurately determine module degradation rates and thus forecast PV module output under actual operating conditions. Moreover, the resulting understanding by continuously monitoring the interaction between module output of various designs and meteorological parameters in different locations will lead to improved next-generation modules.
The authors would like to thank Tucson Electric Power and Bill Henry for its authorization to use their PV yard for performance and degradation measurements. In addition, the rest of NREL’s Reliability group and Jaap van der Burgt of ADVANCIS GmbH for helpful discussions and interactions are gratefully acknowledged. . This work was supported by the U.S. Department of Energy under Contract No. DOE-AC36-08GO28308 with the National Renewable Energy Laboratory.
* Correspondence to: Dirk Jordan, National Renewable Energy Laboratory, 1617 Cole Boulevard, Golden, CO 80401, USA, firstname.lastname@example.org
1 Kurtz S, Granata J, Quintana M. Photovoltaic-Reliability R&D toward a Solar-Powered World. Proceedings of the SPIE, San Diego, 2009, 7412.
2 Osterwald CR, Adelstein J, del Cueto JA, Kroposki B, Trudell D, Moriarty T. Comparison of Degradation Rates of Individual Modules Held at Maximum Power. Proceedings of the 4th IEEE World Conference on Photovoltaic Energy Conversion, Hawaii, 2006.
3 Vazquez M, Rey-Stolle I, Photovoltaic Module Reliability Model Based on Field Degradation Studies. Progress in Photovoltaics: Research and Applications 2008; 16: 419.
4 Marion B, Adelstein J, Boyle K, Hayden H, Hammond B, Fletcher T, Canada B, Narang D, Kimber A, Mitchell L, Rich G, Townsend T. Performance Parameters for Grid-Connected PV Systems. Proceedings of the 31stIEEE PV Specialist Conference, Orlando, 2005, 1601.
5 Kimber A, Dierauf T, Mitchell L, Whitaker C, Townsend T, Newmiller J, King D, Granata J, Emery K, Osterwald C, Myers D, Marion B, Pligavko A, Panchula A, Levitsky T, Forbess J, Talmud F. Improved test method to verify the power rating of a photovoltaic (PV) project. Proceedings of the 34th IEEE PV Specialist Conference, Philadelphia, 2009.
6 Vignola F, Krumsick J, Mavromatakis F, Walwyn R. Measuring Degradation of Photovoltaic module performance in the field. ASES National Solar Conference 2009, Buffalo.
7 Dhere NG, Kaul A, Khatri SD, Pethe SA, Kumar B. Outdoor Monitoring and High Voltage Bias Testing of Thin Film PV Modules. 2006.
8 Del Cueto JA, Pruett J, Cunningham D. Stabilization of High-Efficiency CdTe Photovoltaic Modules in Controlled Indoor Light Soaking. National Center for Photovoltaics and Solar Program Review Meeting, Denver, 2003.
9 Standard Test Conditions (STC): Irradiance = 1000 W/m2, Tmodule=25ºC and Spectrum = 1.5 AirMass.
10 Performance Test Conditions (PTC): Irradiance = 1000 W/m2, Tambient = 20ºC, Wind Speed = 1 m/s
11 Jennings C. PV Module Performance at PG&E. Proceedings of the 20th PV Specialist Conference, Las Vegas, 1988, 1225.
12 Whitaker CM, Townsend TU, Wenger HJ, Iliceto A, Chimento G, Paletta F. Effects of Irradiance and other factors on PV Temperature Coefficients. Proceedings of the 22nd IEEE PV Specialist Conference, Las Vegas, 1991, 608.
13] Whitaker CM, Townsend TU, Newmiller JD, King DL, Boyson WE, Kratochvil JA, Collier DE, Osborn DE. Application and validation of a new PV performance characterization method. Proceedings of the 26th IEEE PV Specialist Conference, Anaheim, 1997, 1253.
14 Osterwald CR, Adelstein J, del Cueto JA, Sekulic W, Trudell D, McNutt P, Hansen R, Rummel S, Anderberg A, Moriarty T. Resistive Loading of Photovoltaic Modules and Arrays for Long-Term Exposure Testing. Progress in Photovoltaics: Research and Application 2006; 14: 567–575.
15 McMahon T. Accelerated testing and failure of thin-film PV modules. Progress in Photovoltaics: Research and Application 2004; 12: 235.
16 Wieting R, Gay R, Nguyen H, Palm J, Rischmiller C, Seapan A, Tarrant D, Willett D. CIS Thin Film Manufacturing at Shell Sola: Practical Techniques in Volume Manufacturing. Proceedings of the 31st IEEE PV Specialist Conference, Orlando, 2005, 177.