|
David Faiman
Ben-Gurion National Solar Energy Center,
Jacob Blaustein Institutes for Desert Research,
Ben-Gurion University of the Negev
Sede Boqer Campus, 84990 Israel
E-Mail: faiman@bgu.ac.il
Fax: +972-8-659-6736
INTRODUCTION
The power output of photovoltaic cells depends on the intensity of the incoming light, its spectral content and the cell temperature. In order to be able to predict the performance of a pv system, therefore, it is of paramount importance to be able to quantify cell performance in a reproducible manner. The standard laboratory technique for this purpose is to employ a solar simulator and a calibrated reference cell. Such a setup enables module performance to be assessed under constant, standard, illumination and temperature conditions. However, this technique has three inherent weaknesses.
First, it is difficult to synthesize a light spectrum, from electric lamps, which closely approximates that of natural sunshine. On the one hand the interference filters needed to remove unwanted peaks, particularly in the near IR, tend to be expensive and not very stable with time. Furthermore, the relatively high electric power densities needed to produce a uniform luminous intensity of 1000 Wm-2 over the entire area of a test module do not help to promote filter stability.
Secondly, in addition to filter instability, the lamps themselves undergo ageing. This has led to the development of so-called "flash" test techniques in which a minimum of the lamp's total serviceable life time is required for each test. But flash tests can introduce spectral variations as the lamp warms up and further uncertainties about the degree to which steady state conditions may or may not exist in the module while its I-V curve is being measured.
Lastly, although calibrated reference cells may be constructed for testing modules employing crystalline or polycrystalline silicon cells, there are materials (such as amorphous silicon) for which a stable reference cell can not be fabricated. The obvious solution, of employing a black-body pyranometer as reference standard, is not feasible for flash tests owing to the finite time (typically several seconds) such a pyranometer takes to reach steady state itself.
The following discussion will demonstrate the degree to which the natural sun conditions - at our desert test laboratory - can be employed as a reliable and reproducible standard for testing a wide variety of module types.
ACHIEVING NATURAL STC TEST CONDITIONS
Standard Test Conditions (STC) are specified as being 1000 W m-2 insolation of spectral type AM1.5 (defined by convention) and 25o C cell temperature.
(a) Insolation conditions
In order to measure I-V curves in our outdoor test laboratory the test module is placed on a stand that enables it to be orientated at normal incidence to the incoming solar beam direction. Measurements are made at and around solar noon on cloudless days. Fig.1 displays the measured global insolation on a sun-tracking plane on a typical clear day at Sede Boqer (June 26, 1993). The insolation is measured with a thermopile pyranometer that is periodically compared, using the normal incidence method, with a secondary standard the calibration of which was established at the World Meteorological Organization in Davos, Switzerland. We prefer the use of this kind of radiation sensor to a calibrated reference cell because, on the one hand, its spectral sensitivity is relatively neutral, i.e. it does not favor any particular kind of module. Secondly, the available solar resource at any given site is usually measured with this kind of instrument. Hence the module efficiency as determined with a thermopile pyranometer provides a better estimate of how much available solar energy may be converted to electricity
than does an efficiency figure that was derived from a reference cell.
Figure 1: Global irradiance measured on a sun-tracking plane at Sede Boqer, June 26, 1993
From Fig.1 one sees that, in spite of some slight cloudiness in the late afternoon on this particular day, the irradiance remained constant to a few parts per mille for more than two hours around solar noon. On clear days there is, accordingly, ample time to measure many I-V curves without fear that the irradiance will change either during a measurement (which lasts for several seconds) or from one measurement to the next (which may be separated by several minutes). Indeed, in summer, conditions are often sufficiently stable to enable sequences of measurements to be continued from one day to the next if, for example, many modules are to be compared.
(b) Spectral conditions
At Sede Boqer (Latitude = 30.8o N) the noon-time sun has a zenith angle qz that varies from about 8o in summer to about 54o in winter. These figures correspond to the angular conditions (i.e. arcsec qz) of AM1.01 to AM 1.70, respectively. We have measured the spectral content of the natural global irradiance incident on the plane of a test module set at normal incidence to the solar beam direction. Such measurements are performed with a Li-Cor spectroradiometer on days when modules are tested, i.e. on cloudless days. Fig.2 shows a typical scan (small crosses) taken at 2 pm on March 6, 1995. At this time of day, on this date, the solar zenith angle corresponds to AM 1.5 at the latitude of our laboratory. The scanning interval of the spectroradiometer was set at 1 nm intervals in wavelength over the range 300-1100 nm. Also shown in Fig.2 are the IEC standard intensities (circles) for AM 1.5 over this range of wavelengths.
Figure 2: Natural solar spectrum measured at Sede Boqer at 2 pm, March 6, 1995 (crosses). IEC standard AM1.5 values (circles) are superimposed.
Particularly striking in Fig.2 is the fact that no re-scaling has been performed: At the time of day when the solar zenith angle corresponded to AM1.5 the measured intensity at most wavelengths was found to be extremely close to the corresponding international standard intensities. This is especially noticeable at the various peaks and troughs in the figure. If we exclude the 300-400 nm UV region, for which this instrument has poor accuracy, and assess the remaining 50 IEC points of reference our measured spectrum exhibits a mean bias error of only 3%, the two worst points being 12% low (at 757.5 nm - interpolated from measurements at 757 and 758 nm) and 32% high (at 930 nm). Since, moreover, both of these points lie on narrow, rapidly changing, parts of the spectrum (respectively, O2 and H2O absorption bands) their significance is not high.
Similar scans have been performed at other times of the year and at various times on either side of AM 1.5 angular conditions, in order to determine the extent to which the spectrum of Fig.2 remains stable. At other times of day it is necessary to rescale the intensity and perform integrals over various wavelength intervals but at such times we do not perform module tests. For our present purposes it suffices to state that on clear days at all months of the year the noontime natural insolation at Sede Boqer is far closer to the IEC AM1.5 spectrum than any solar simulator that has yet been built.
Some indirect evidence for this fact may be seen by measuring Isc for a reference module, at noon time, on clear days at various times of the year. This parameter is relatively insensitive to temperature uncertainties (discussed below). Therefore, provided we normalize the measurements to 1000 W m-2, any variations in measured Isc must be due to spectral effects. Fig.3 shows twelve such measurements of Isc that were made during 1994. One sees that, except for a single measurement that was about 2% above average, spectral variations are typically at the 1% level throughout the entire year provided measurements are performed on clear days.
Figure 3: Isc measurements, normalized to 1000 Wm-2, performed on a reference module
at Sede Boqer on 12 clear days throughout the year 1994. Error bars shown are a nominal 1%
(c) Temperature conditions
Our standard method for assessing cell temperature while an I-V curve measurement is in progress is to follow the output of a thermocouple taped to the rear side of the module. This offers the advantage of representing the actual in-use temperature of the module but suffers from the disadvantage that momentary fluctuations in wind speed may change the temperature by typically ± 2o C during an experimental run lasting several minutes. For more accurate out-
Figure 4: Typical Voc vs Tmod curve obtained outdoors with insulated module backing
door work, specifically when measuring temperature coefficients of the various cell parameters, we insulate the rear side of the module with polystyrene sheeting in order to minimize the effect of wind. This results in the module operating at somewhat higher temperatures than would normally be the case at any given insolation level but it does result in curves that are very smooth as a function of temperature. Fig.4 shows an example of a typical temperature coefficient curve obtained in this manner. The module under test had first been cooled in an air-conditioned room, carried outdoors with insulated panels front and back, placed in the test stand and then exposed to solar radiation. Its I-V curve and temperature were subsequently measured every few minutes - quite frequently at the beginning, as the module temperature rose rapidly, but less frequently as the module temperature neared its steady state.
STC conditions may either be read directly from curves such as Fig.4 or the fitted slopes can be used for making adjustments to measurements obtained in the field.
PUTTING IT ALL TOGETHER
One of our experimental projects (which studies module performance degradation) involves modules being periodically tested on one clear day each month. These modules include a reference sample that is kept indoors except when tests are performed. Its parameters are therefore expected to be stable compared to those of modules that are in use under conditions of constant exposure. Fig.5 shows the measured values of Pmax, reduced to STC, for the reference module at various times of the year. It is important to note that the measurements in Fig.5 are for an uninsulated module and, accordingly, subject to a temperature uncertainty of about ± 2o C, as stated above. This temperature uncertainty translates into an approximately 2% uncertainty in the measured peak power of the module under test.
Figure 5: STC-adjusted results of Pmax measurements on an uninsulated module on clear days
at Sede Boqer during 1993-1994. Error bars shown are a nominal 2%.
From Fig.5 it is evident that, in our outdoor tests at Sede Boqer, module efficiency can be determined to an accuracy of approximately ± 2% at any time of the year provided measurements are made in the noontime period on clear days. It is interesting to observe that this is approximately the degree of accuracy with which solar insolation can be measured using a well-maintained pyranometer.
TESTING COMMERCIAL PV MODULES
At the Ben-Gurion National Solar Energy Center we maintain a static test bed upon which a variety of commercial PV modules are exposed to the sun while feeding resistive loads. These modules are removed 4 times a year and tested in the above indicated manner. The purpose of these regular tests are (a) to correctly characterize the modules for potential use under Middle East climatic conditions, (b) to identify module types that are stable for long periods of time. Table 1 shows some typical results.
|
Module |
Pmax(Simulator) [W] |
Pmax(Outdoors) [W] |
Difference [%] |
|
Germany 1 |
53.7 |
46.6 |
- 13 |
|
Japan 1 |
68.5 |
62.8 |
- 8 |
|
Japan 2 |
80.2 |
78.6 |
- 2 |
|
Russia 1 |
54.2 |
44.3 |
- 18 |
|
Spain 1 |
52.6 |
48.8 |
- 7 |
|
USA 1 |
47.2 |
43.4 |
- 8 |
Table 1: Comparison of various manufacturer's STC peak power simulator ratings, for specific modules, compared to outdoor measurements at Sede Boqer.
In Table 1 the column headed Pmax(Simulator) refers to measurements provided by the manufacturer for the specific module under test. Manufacturer's tests are usually performed with a solar simulator because local weather conditions are not sufficiently stable to ensure adequate quality control. However, because the light spectrum of solar simulators varies from one situation to the next and, in all cases, is different from the true solar AM1.5 spectrum, substantial differences are obtained when these modules are tested under real sun conditions. In the case of one module shown in Table 1 the manufacturers have overstated the power output by more than 20%.
Figure 6 : Measured spectrum of a, filtered metal halide lamp, solar simulator (crosses) and IEC Air Mass 1.5 standard spectrum (circles)
It must be emphasized that these differences are not deliberate deception by the manufacturers but, instead, result principally owing to spectral differences between commercial solar simulators and real sun conditions. In order to illustrate this point Fig.6 shows the spectrum of a commercial solar simulator compared to the IEC standard AM1.5 spectrum. The difference between Fig.6 and the natural solar spectrum we have at Sede Boqer (Fig.2) is striking.
SUMMARY AND CONCLUSIONS
(1) Noontime outdoor lighting conditions on clear days at Sede Boqer closely correspond to the IEC definition of the AM1.5 solar spectrum at all times of the year.
(2) At such times, climatic conditions are sufficiently stable to enable the determination of Pmax to a precision of ± 2%. This is comparable to the accuracy of a well-maintained pyranometer and superior to the accuracy obtainable from solar simulators.
(3) PV modules characterized by commercial solar simulators seriously overestimate module performance. This can lead to expensive mis-understandings.
(4) The Ben-Gurion National Solar Energy Center offers a uniquely accurate characterization service to the international community of PV manufacturers and research laboratories.
PRICES AND CONDITIONS
The Ben-Gurion National Solar Energy Center currently (2000) performs two kinds of test on commercial PV modules: The Short Test and the Long Test.
The Short Test consists of two parts.
Part 1 - Determination of the module temperature coefficients.
Part 2 - Determination of the module I-V curve at STC.
The Long Test consists of two parts.
Part 1 - The entire Short Test.
Part 2 - The second part of the Short Test, repeated each month for an entire year.
The client should note that these tests are not intended to be regarded as "qualifying tests" in that no pass/fail or grade is awarded the modules. Instead, the client receives a confidential report containing all measurement data, graphs, least-squares curve fits and our conclusions regarding the STC values of the I-V curve parameters Voc, Isc, Vpp, Ipp, Pmax and FF. In the case of the Long Test the report includes our conclusions regarding any module degradation that may have been noted during the year of tests.
The cost of the Short Test is $5,000. The cost of the Long Test is $10,000.
These costs include all handling charges incurred in Israel (customs clearance and module transportation from the airport to the Ben-Gurion National Solar Energy Center's desert location) and the cost of sending a test report to the client by international air express. Not included is the cost of returning the module to the client should this be required.
Pre-payment should be made to "Ben-Gurion University (National Solar Energy Center)".
Modules should be insured and sent air-freight, addressed to "The Ben-Gurion National Solar Energy Center, Ben-Gurion University, Israel". The accompanying documents should clearly state that the contents are "test samples with no commercial value".
For further information, please contact Prof. David Faiman
Phone: +972-8-659-6933, Fax: +972-8-659-6736, E-mail: faiman@bgu.ac.il
|