Resistive loading of photovoltaic modules and ... - Wiley Online Library

PROGRESS IN PHOTOVOLTAICS: RESEARCH AND APPLICATIONS Prog. Photovolt: Res. Appl. 2006; 14:567–575 Published online 2 May 2006 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/pip.693

Application

Resistive Loading of Photovoltaic Modules and Arrays for Long-Term ExposureTesting C. R. Osterwald*,y , J. Adelstein, J. A. del Cueto, W. Sekulic, D. Trudell, P. McNutt, R. Hansen, S. Rummel, A. Anderberg and T. Moriarty National Renewable Energy Laboratory, 1617 Cole Blvd., Golden, CO, USA

This paper investigates the feasibility of using fixed resistors instead of active maximum power tracking as electrical loads for long-term exposure testing of photovoltaic modules and arrays. The method of investigation was to compare resistive versus active loading on two modules for which historic current-voltage data over time were available. Also, a small amorphous silicon array was installed with a resistive load and the performance has been monitored versus time. The major conclusion of this work is that fixed resistive loading is an inexpensive and viable means of loading photovoltaic devices for exposure testing if the resistance value used is close to the ratio of the voltage to the current at the maximum power point under Standard Test Conditions. Copyright # 2006 John Wiley & Sons, Ltd. key words: photovoltaics; modules; arrays; loading; exposure; testing

INTRODUCTION

O

ne method of monitoring photovoltaic (PV) system performance uses regressions to Performance Test Conditions (PTC), which are defined as 1000 W/m2 plane-of-array global irradiance, 20 C ambient temperature, and 1 m/s wind speed (the PTC regressions have also been called ‘PVUSA ratings’ in the past, as they were originally developed for the Photovoltaics for Utility Scale Applications project in the early 1990’s). We employ the PTC method to monitor the output of a number of small, grid-connected PV systems installed at our outdoor test site at the National Renewable Energy Laboratory (NREL) on a monthly basis. It has been shown that identical degradation rates can be measured both with the PTC method and with careful temperature and irradiance corrections of performance data to standard conditions.1 The advantages of the PTC method are that PV module and array temperature measurements are not needed, and that the magnitude of the power obtained at PTC is much closer to actual operational power values (as compared with those at 25 C cell temperature and 1000 W/m2 total irradiance, i.e., Standard Test Conditions or STC). In 2003, NREL initiated a subcontracted program to install high-voltage ( 600 V) strings of thin-film modules in hot, humid climate locations, and study possible problems caused by moisture ingress.2 To minimize costs and complexity for the subcontractors, the program specifications called for the series-connected module

* Correspondence to: Carl R. Osterwald, National Renewable Energy Laboratory, 1617 Cole Blvd., Golden, CO, USA. y E-mail: [email protected]

Copyright # 2006 John Wiley & Sons, Ltd.

Received 14 July 2005 Revised 14 November 2005

568

C. R. OSTERWARD ET AL.

Figure 1. Schematic diagram of PV module loading

strings to be loaded with fixed-value resistors, and instrumented with sufficient data-acquisition capability to allow performance monitoring with PTC regressions. Thus, PTC regressions were intended to be the primary means of detecting any performance degradation of the test modules. As a result, the motivation for the work reported in this paper was to verify that it is technically feasible to employ the PTC method on resistively loaded PV modules and arrays. This verification has been done in two ways. First, a small test system was installed at the NREL site with resistive loading and appropriate instrumentation, and second, historical outdoor current versus voltage (I–V ) measurements have been post-processed to simulate resistive loading.

RESISTIVE LOADING Exposure testing of PV modules can be used to determine degradation rates of performance3, or for discovery of failure mechanisms.4,5 Regardless of the purpose, the test samples are loaded electrically during the exposure testing, as illustrated in Figure 1, where the test module is connected to a resistance that represents the electrical load. Note that if the test module is shorted, RLOAD is nearly zero, and if the module output terminals are left unconnected, the load resistance can be considered infinite. However, if the objective of the exposure testing is to monitor performance versus time in an environment close to that of actual use, these two simple load conditions (i.e., shorted or open circuited) are not adequate and, in fact, can cause problems by themselves. For example, it is known that shorted is the most stressful load condition for crystalline-Si modules and can result in destructive hot-spot heating as solder joints degrade.6 It has

Figure 2. I–V curve of a 60 W crystalline-Si module at STC, and the corresponding power versus load resistance curve. This module has a fill factor of 699% and an RSTD value of 425


Prog. Photovolt: Res. Appl. 2006; 14:567–575

PHOTOVOLTAIC MODULES AND ARRAYS LOADING FOR EXPOSURE TESTING

569

Figure 3. Outdoor I–V curves for the same crystalline-Si module in Figure 2 measured on a single clear day (September 20, 2002) over an irradiance range of 70–1100 W/m2. Also shown are three resistive load lines including RSTD ¼ 425 , which is the ratio of VMAX to IMAX at STC

also been suggested that leaving certain thin-film modules open-circuited could cause damage due to higher electric fields in thin active layers.7 Therefore, it is desirable to employ a load condition that approximates actual use conditions for exposure testing. An ideal solution would be an active device that effectively adjusts RLOAD so that the test samples are maintained at the maximum power point (PMAX) regardless of the irradiance and module temperature. An obvious way to accomplish this would be to use inverters, which need PMAX tracking capabilities of the PV input to maximize efficiency.8 Unfortunately, inverters have a number of disadvantages for exposure testing, including high unit cost and greatly increased test system complexity and costs through the need for an AC line

Figure 4. Power versus irradiance determined from the September 2002 PERT data for the same crystalline-Si module used in Figures 2 and 3, while loaded at 225 and the maximum power point Copyright # 2006 John Wiley & Sons, Ltd.


570


Figure 5. Power versus irradiance determined from the September 2002 PERT data for the same crystalline-Si module used in Figures 2 and 3, while loaded at 425 and the maximum power point

connection. A stand-alone inverter does not require an AC line connection, but these are generally more expensive because they must be self-commutating. As a result, using inverters as active loads is not an attractive solution, especially if individual modules are to be tested. Computer-controlled electronic loads can also be used for exposure testing, but require software to perform maximum power tracking.9 Electronic loads can also be expensive—on the order of $3 US per Watt versus $1 US per Watt of output power for a grid-connected inverter. Fixed resistive loads are an obvious alternative, and can greatly reduce the test system costs and complexity. The disadvantage is that there is no way to implement PMAX tracking. Due to this shortcoming, the objective of this paper is to examine the implications of fixed resistive loads for exposure testing of PV modules and arrays. More precisely, how does the use of a resistive load affect the output power of a PV system?

Figure 6. Power versus irradiance determined from the September 2002 PERT data for the same crystalline-Si module used in Figures 2 and 3, while loaded at 625 and the maximum power point Copyright # 2006 John Wiley & Sons, Ltd.



571

Figure 7. Monthly PTC regressions using I–V curve data from the crystalline-Si module shown in Figure 2 for several load conditions. The power data were derived from I–V curve measurements obtained with the PERT (see text), which were then combined with irradiance, ambient temperature, and wind speed data. Also shown are linear regression fits for each load condition, which are summarized in Table I; the dashed lines are associated with the unfilled symbols. The module power data were filtered by excluding all points where the irradiance was less than 800 W/m2. Note that the seasonal variations in the PTC regressions are temperature related (especially visible for the maximum power loading); the output power of crystalline-Si modules is higher in the cold winter months

As a first step toward answering this question, consider Figure 2, which is the I–V curve of a 60 W crystallineSi module at Standard Test Conditions (STC, 1000 W/m2 global irradiance, 25 C). Also plotted on this graph is the output power as a function of load resistance; these are equal to V I and V I, respectively. This curve has a maximum at a load resistance of 425 , which corresponds to the maximum power point at STC and is designated RSTD. Therefore, RSTD is defined as the ratio of the voltage at PMAX (VMAX) to the current at PMAX (IMAX), at STC. Note that as RLOAD is increased from zero, the output power rises almost linearly up to PMAX, and then decreases at a slower rate as RLOAD approaches infinity. Next, a series of I–V curves for the same module used in Figure 2 were measured outdoors every 15 min on a clear day; several of these are plotted together in Figure 3, along with three resistive load lines. These I–V curves were measured with the Performance and Energy Ratings Testbed (PERT) at the NREL test site in Golden, Colorado; this measurement system is described in the paper by del Cueto.10 Looking at the intersection of the 425 ( ¼ RSTD) load line with the I–V curves, it can be seen that at low irradiances the module operates with a nearly constant current. Higher irradiances operate near, and then eventually past, the knee of the I–V curve. Because of the logarithmic dependence of voltage on irradiance, if the load line intersects the I–V curve beyond the knee of the curve, output power is essentially independent of irradiance. An excellent way to view the effects of resistive loading is to plot output power versus irradiance. Because the PERT periodically measures I–V curves, it is possible to extract not only the maximum power point, but also the output power at a given load resistance. This was done by fitting the I–V data, which were converted to power versus resistance, to fourth-order polynomials and evaluating the polynomials at the desired RLOAD. Using an

Table I. Linear-regression fit results for the crystalline-Si module PTC regressions shown in Figure 7 Load condition: Initial power (W): Degradation rate (%/year): Correlation coefficient r2:

255

493 091 052


425

595

596 087 020

514 065 014

PMAX 614 097 037


572


Figure 8. I–V curve of a 40 W thin-film module at STC, and the corresponding power versus load resistance curve. This module has a fill factor of 496% and an RSTD value of 76

entire month of PERT I–V curve data from September 2002 (this month includes the data presented in Figure 3), Figures 4–6 show power versus irradiance for three values of RLOAD plotted against the PMAX data (the three RLOAD values are the same as the load lines plotted in Figure 3). The PMAX loading is nearly linear with irradiance, except for a small roll-off caused by higher temperatures at higher irradiance. With a resistive load of 225 (¼ 053 RSTD), the dependence is nonlinear up to 800 W/m2 and considerably lower than PMAX (Figure 4). In Figure 5 with RLOAD ¼ RSTD, there is still a large divergence at low irradiance, but at 800 W/m2 and higher it appears to track the PMAX curve. If the load resistance is raised to 625 ( ¼ 147 RSTD) as in Fig. 6, a similar shape is seen except that above 800 W/m2 the output power is independent of irradiance.

PERFORMANCE MONITORING EXAMPLES Using all the historical PERT data for the same crystalline-Si module shown in Figure 2, and the I–V curve processing software described above, we were then able to calculate power versus time at PMAX and at several resistive load values. Applying the PTC regressions on a monthly basis to these data resulted in Figure 7, and parameters obtained with the linear fits are listed in Table I. Note that the RSTD load case (425 ) shows an overall power level close to the PMAX tracking case, whereas the high- and low-resistance cases are both considerably lower. Yearly seasonal variations are clearly visible in the PMAX data, but are much less noticeable

Table II. Linear-regression fit results for the thin-film module PTC regressions shown in Figure 9 Load condition: Initial power (W): Degradation rate (%/year): Correlation coefficient r2:

46

76

115

PMAX

358 223 082

456 365 093

412 345 088

460 374 091




573

Figure 9. Monthly PTC regressions using I–V curve data from the thin-film module shown in Figure 8 for several load conditions

with resistive loads. The low-resistance (255 ) case shows the highest linear correlation coefficient, which is indicative of the operating point being predominately on the flat portion of the I–V curve. Note that the highresistance (595 ) case has the smallest correlation coefficient, which could be the result of the output power being independent of the irradiance above 800 W/m2, as was seen in Figure 6. Next, I–V curve data from a thin-film module was processed in a similar fashion. The STC I–V data for this module are shown in Figure 8; when compared with the Si module in Figure 2, this module has a significantly lower fill factor and the peak in the resistance versus power curve is much broader. It is interesting to note that the PMAX tracking and RSTD loads for this module are in good agreement; the power versus resistance peak is a likely explanation for this behavior. Finally, the thin-film module has much smaller random variations in the

Figure 10. Power versus irradiance determined from the September 2002 PERT data for the same thin-film module used in Figure 8, while loaded at 767 and the maximum power point Copyright # 2006 John Wiley & Sons, Ltd.


574


Figure 11. PTC regressions versus time for an a-Si system consisting of eight series-connected modules with a fixed load resistance of 225 , which is equivalent to 104 RSTD

PTC regressions compared to the crystalline-Si module (Table II; Figure 9). While considering both of these observations, compare Figure 5 with Figure 10, which show power versus irradiance for the PMAX tracking and RSTD loads for both modules. Above 800 W/m2 the PMAX tracking and RSTD for the thin-film module track each other very closely with much less scatter. Again, the tracking behavior is probably due to the broad resistance versus power peak, and the reduced scatter is possibly due to smaller temperature coefficients. As a test of the resistive loading concept, a thin-film system consisting of eight amorphous silicon (a-Si) modules connected in series was installed at our site prior to the beginning of the hot and humid exposure subcontracts. The system is instrumented with appropriate data acquisition and has a load resistance of 225 . Based on outdoor I–V measurements of the entire string, this load resistance is very close to RSTD. Two years of data have been collected, and the PTC regressions are presented in Figure 11. A rapid initial degradation typical of a-Si devices within the first 2 months of operation is apparent with the drop from 80 W, and the output power has been declining since that time. Two peaks that correspond to a-Si annealing during summer hightemperature operation are visible. The overall degradation rate has been about 7% per year.

CONCLUSIONS Using I–V curve data recorded at 15-min intervals over a 10-year period, a direct comparison between maximum-power-point tracking and fixed resistive loads for outdoor exposure testing of PV modules has been made. The results of the comparison show that using fixed resistive loads can closely reproduce changes in module performance over time that are observed with maximum-power-point tracking loads. A resistively loaded a-Si array of eight modules has been monitored for over 2 years, and the performance versus time of this system shows features that are commonly observed in a-Si modules and arrays. From this work, we conclude that fixed resistive loads are a viable option for monitoring long-term PV performance, provided the load resistance value used is close to RSTD. The closeness to RSTD needed depends on the fill factor of the test sample.

REFERENCES 1. Marion B, Adelstein J. Long-term performance of the SERF PV systems. Proceedings of the NCPV and Solar Programme Review Meeting 2003, 24–26 March 2003, Denver; 199–201, NREL/CD-520-33586. Copyright # 2006 John Wiley & Sons, Ltd.



575

2. Dhere NG, Hadagali VV, Shirolikar JS. Outdoor monitoring and high voltage bias testing of thin film PV modules. Proceedings of the DOE Solar Energy Technologies Program Review Meeting, 25–28 October 2004, Denver; 348–349, DOE/GO-102005-2067. 3. Osterwald CR, Benner JP, Pruett J, Anderberg A, Rummel S, Ottoson L. Degradation in weathered crystalline-silicon PV modules apparently caused by UV radiation, Proceedings of 3rd World Conference on Photovoltaic Energy Conversion (WCPEC-3), 11–18 May 2003, Osaka; 2911–2915. 4. Quintana MA, King DL, McMahon TJ, Osterwald CR. Commonly observed degradation in field-aged photovoltaic modules. Conference Record of the Twenty-Ninth IEEE Photovoltaic Specialists Conference—2002, 19–24 May 2002, New Orleans; 1436–1439. 5. Osterwald CR, McMahon TJ, del Cueto JA. Electrochemical corrosion of SnO2:F transparent conducting layers in thinfilm photovoltaic modules. Solar Energy Materials and Solar Cells 2003; 79; 21–33. 6. Wohlgemuth J. Private communication, 2001. 7. McMahon TJ. Private communication, 2003. 8. Ross JN. Practical Handbook of Photovoltaics, Chap. IIIb-1: System Electronics. Elsevier: Oxford, 2003; 565–588, ISBN 1-85617-390-9. 9. del Cueto JA, Osterwald C, Pruett J. Advanced indoor module light-soaking facility. Proceedings of the DOE Solar Energy Technologies Program Review Meeting, 25–28 October 2004, Denver, 417–418, DOE/GO-102005-2067. 10. del Cueto JA. Model for the thermal characteristics of flat-plate photovoltaic modules deployed at fixed tilt. Conference Record of the Twenty-Eighth IEEE Photovoltaic Specialists Conference—2000, 15–22 September 2000, Anchorage; 1441–1445.

