CALIBRATION OF SPECTRORADIOMETERS FOR OUTDOOR DIRECT SOLAR SPECTRAL IRRADIANCE MEASUREMENTS

Transcription

1 CALIBRATION OF SPECTRORADIOMETERS FOR OUTDOOR DIRECT SOLAR SPECTRAL IRRADIANCE MEASUREMENTS Paraskeva Vasiliki 1*, Matthew Norton 1,2, Maria Hadjipanayi 1, George E. Georghiou 1 1 Photovoltaic Technology, ECE, University of Cyprus, 75 Kallipoleos St., Nicosia, 1678, Cyprus 2 European Commision, DG JRC, Institute for Energy and Transport, Renewable Energy Unit, TP 450, Ispra (VA), Italy 1* corresponding author: Vasiliki Paraskeva (phone , Fax , vparas01@ucy.ac.cy) (Matthew Norton 1,2 phone , Fax , msh.norton@gmail.com) (Maria Hadjipanayi 1 phone , Fax , hadjipanayi.maria@ucy.ac.cy) (George E.Georghiou 1 phone , Fax , geg@ucy.ac.cy) ABSTRACT: Spectral s are an important aspect of environmental monitoring for photovoltaics, especially concentrator photovoltaics (CPV) that use multi-junction cells with spectrally dependent output. There are only a few reliable sources of spectrally resolved direct normal irradiance (DNI) data in the world since the equipment is expensive and calibration is hard to maintain. Efforts have been made to produce a high quality data set of spectra with reduced uncertainties. An evaluation of the of both indoor and outdoor spectrum s and their origins is presented in the paper. A traceable system was used to calibrate two sets of spectroradiometers and repeated s have been carried out to find the uncertainties of repeatability, positional sensitivity and lamp stability. The two systems were then used outdoors alongside broadband solar irradiance devices, to determine the differences of the spectral s between the systems. Finally, a collimator was attached to one of the spectroradiometer systems and s were taken using both collimated and non-collimated systems to establish the impact of the collimator on spectral s. Keywords: multi-junction solar cells, calibration, spectral s, uncertainties 1 INTRODUCTION The arrival of new PV technologies (such as multijunction thin film and concentrators) on the market has highlighted the need for better spectrally resolved s of solar irradiation. These new PV devices consist of several junctions with narrower spectral responses to the solar irradiance than the common crystalline silicon devices and for that reason they are more sensitive to the spectral variation of the solar irradiance. Establishing the long term energy yield and rating of concentrator photovoltaic systems and modules is an area of active development, with the IEC standard under development to provide methods for an energy rating [1]. Predicting the energy yield or determining an energy rating for such devices is consequently more complicated, as the device outputs need to be determined for a more complex set of environmental conditions [2,3]. In order to extract such information, it is necessary to have high quality environmental data with low uncertainties. Spectral s play an important role in the understanding of the performance of multi-junction devices and accurate of the spectrum is required. The solar spectrum arriving on a terrestrial PV device varies with the time of day, the season, the weather conditions and also with the position of the receiver [4,5]. In order to evaluate the energy yield potential of multi-junction devices at a particular location, high quality spectral s should be taken routinely. Moreover, many factors introduce uncertainties during the of solar irradiance affecting the procedure. In an attempt to produce a high quality data set of spectra with minimum uncertainties, the sources of first need to be established and quantified. Recently, increased effort by the scientific community has been made to improve knowledge and experience in the outdoor of solar irradiance spectra, best highlighted by the annually organised spectroradiometer inter-comparisons [6,7]. This work supports the abovementioned effort by investigating problems with s that are applicable to a wide range of devices. Such data is essential for understanding and predicting the impact of spectral irradiance upon concentrator photovoltaic technologies that use multiple-junction cells. An investigation of the uncertainties of spectral s and their origins is presented in this paper. Understanding the origin and nature of the uncertainties will provide an insight into the factors affecting the uncertainties encountered during outdoor spectrum s and will thus assist in finding ways of reducing them. 2 APPROACH 2.1 Experimental apparatus In an attempt to build a reliable spectral system, two sets of calibrated spectroradiometers have been used. These two systems are of the same type and almost identical. They consist of two units, one for performing s in the visible and ultraviolet region of the spectrum (using a silicon based CCD) and the second one for s in the near infrared region. The two spectra are merged with dedicated software so that the final spectra irradiance is in the range nm. The irradiance to be measured is then directed into the units through the use of fibre-optic cables that have a cosine corrector, or diffuser, at their entrance aperture. The outdoor system uses a 7 m long fibre-optic cable to direct light from a mounting point on an accurate solar tracker into a climate-controlled cabinet and into the

2 spectroradiometers. The other set of spectroradiometers uses a 1m long fibre-optic cable. To investigate the effect of collimation on the spectrum, a custom-made collimator was fabricated at the Joint Research Centre (JRC) with specifications designed to match those typically found in pyrheliometers (specifically a slope angle of 1 degree and an acceptance angle of ±2.5 degrees). The first spectroradiometer set was calibrated at the JRC under a standard halogen lamp with a NIST-traceable calibration file. 2.2 Experimental procedure Since it was not possible to perform a direct calibration of the outdoor system, it was necessary to first calibrate the set of spectroradiometers used for indoor at the JRC (System 1), and then to transfer this calibration to the outdoor set (System 2). The original calibration of System 1 was transferred to System 2 by performing a side-by-side of a halogen lamp in the laboratory. In this procedure a number of s were performed indoors to determine the uncertainties that may arise from the system, positional sensitivity and lamp stability. After calibration, both spectroradiometers were installed outdoors, and side-by-side s were taken again to determine the uncertainties and their origins outdoors. Then a collimator was attached to one of the spectroradiometers and s were taken using both collimated and non-collimated systems. Outdoor s of global and direct normal irradiance were also taken with a pyranometer and a pyrheliometer respectively and compared with s taken with both calibrated spectroradiometers. The uncertainties introduced at each stage have been calculated. 3 RESULTS AND DISCUSSION 3.1 Indoor calibration Initially indoor s of a standard halogen lamp were performed to calibrate System 1. The results of this procedure are shown in Figure 1. The discontinuity in the middle of the graph indicates the crossover point between the UV-VIS and NIR spectrometers used to take the s. This discontinuity has been attributed to a combination of noise levels and sensitivity differences (to temperature and perhaps also orientation) and a corresponding higher multiplication factor in the calibration files of this region of the spectrum. Based on the standard deviation calculated as shown, it is clearly observed that the worstcase scenario for is obtained in the spectral region between nm. This range is very important since it belongs to the spectral region of all types of photovoltaics. Lower noise is seen in wavelengths above 1000 nm, where the NIR detector is used. The analysis focused on the following main sources: 1. The equipment (spectroradiometers, lamp, power supplies, fibre optics), which were deemed to affect the repeatability. 2. The positional reproducibility. The transfer of the calibration from one system to another relies on the ability to accurately determine the irradiation falling on a specific plane and to locate both the spectroradiometer input optics on the same plane. The inability to do this perfectly introduces a source of. 3. The original calibration. System 1 was calibrated initially at JRC. The analysis was performed at the JRC as part of the internal quality system. The relevant documents report a standard level of calibration of ±1.2%. Figure 1: Comparison of repeated spectrum s and the associated standard deviation of system 1. Higher deviation in the visible range is clearly observed. The estimated values of uncertainties from the indoor analysis are given in Table I. As seen from Table I, the largest part of the arises from the initial calibration procedure while the uncertainties originating from lamp stability are negligible. The combined standard calculated indoors is 1.72%. Repeated s taken from System 2 during calibration indoors is shown in Figure 2. Higher deviation is obseved in the visible range as before. The estimated values of uncertainties in that case is shown in Table II. The initial calibration in that case is the combined standard of System 1. Table I: List of indoor calibration uncertainties of System 1 Source of Value ± Prob. Divisor Standard Initial calibration Accuracy of repositioning of the fiber optic diffuser 1.23% Normal % 0.29% Normal % 2.07% Rectangular %

3 Figure 2: Repeated spectrum s and the standard deviation of system 2 taken indoors. Table II: List of indoor calibration uncertainties of System 2 Source of Value ± Prob. Divisor Standard Initial calibration Accuracy of repositioning of the fiber optic diffuser 1.72% Normal % 0.23% Normal % 3.59% Rectangular % The combined in this case is 2.70% and it is larger than in the case of calibration of System 1 as it is expected. The transfer of calibration from one system to another introduces larger in spectral s. 3.2 Outdoor operation After calibration, the two sets of spectroradiometers were installed side-by-side on a solar tracker that followed the sun to within an error of 0.1º. The tracking error is very small but still introduces in the spectrum s. A typical pair of scans of the global normal spectrum is shown in Figure 3. Measurements showed that the difference in the spectrum varied from 8% to 10% depending on the solar irradiance at the moment of capture. At lower solar irradiance levels the difference was around 8% while at high solar irradiance the difference could reach 10%. Overestimation of the values of spectral irradiance was observed in system 2. Since both spectroradiometers were calibrated, the difference could be attributed to background noise levels of the devices. Since the initial calibration was performed at much lower irradiance levels it was believed that the low signal to noise ratio could account for discrepancies when measuring at higher intensities. For that reason the difference in spectrum between both spectroradiometers at higher irradiance levels is greater whereas at lower irradiances the difference gets smaller. The units were then calibrated indoors at higher irradiance levels by moving the halogen lamp closer to spectroradiometer units. The sources of uncertainties remained the same in that case. To confirm the correct operation of the spectroradiometers in an outdoor application the spectroradiometers were again set-up outdoors as in the previous case. Measurements were taken on two different days in order to investigate the stability of the system and investigate the uncertainties introduced by repositioning of the fibre-optic cable (see Figure 4). Figure 4: Comparison of global normal irradiance spectrum taken with both units. The calibration of the second unit was performed at higher irradiance levels. The first s demonstrated a 4.4% difference in global irradiance between both spectroradiometers. System 2 presented higher values in the visible range ( nm). Measurements were repeated next day and the difference in solar irradiance between both systems was around 3.5%. Variation of measured spectrum for the first day between both units is present in Figure 5. Figure 3: Comparison of global normal irradiance spectrum taken with both units. Figure 5: Difference between outputs of both units at each wavelength.

4 This procedure shows that irradiance levels during calibration have a significant impact on the correct calibration of the systems. This suggests there may be a degree of non-linear behavior with irradiance intensity, and therefore the calibration of spectroradiometers should be performed close to the irradiance levels under which they will operate. Integration of the spectrum over the entire range results in an average difference of 40 W, which is relatively small and indicates better agreement between data. However, there still remained an unacceptable variation between the two systems, and it was suspected that one significant factor could be the influence of the 7 m long fibre-optic cable. In an attempt to eliminate the difference between both units and to investigate the effect of the fibre optic cabling we performed the calibration outdoors in-situ. In so doing, it would be possible to reduce the effect of twisting and bending of the fibre optic cable since the whole system would be calibrated with the cable in its normal operating position, apart from a slight change in bending due to tracking over the day. To perform calibration at the actual site (on the solar tracker) calibration was performed during a clear day. Then an outdoor inter-comparison between units was performed during the day (see Figure 6). Data showed very small differences between the units (around 2%) and this represents the best result obtained. Table III: List of outdoor uncertainties Source of Value ± Prob. Divisor Standard Initial calibration 1.72% Normal % 1.85% Normal % The combined standard in the case of outdoor calibration is 2.52% and it is slightly lower than in the case of indoor calibration indicating better stability of the system. Deviation between both spectroradiometer units at different times of the day is displayed on Figure 7. The average difference between both units is below ±0.1 W/m 2.nm indicating reasonable agreement between s. Very small variation of the over time was presented in the infrared region and indicates stability of the units and systematic effects. However in the visible region the variation over time is larger indicating other sources of. This is attributed to changes in the sensitivity of the CCD particularly to variations in ambient temperature and perhabs also total light intensity. Figure 7: Difference between outputs of both units at each wavelength over time showing the stability of the. Figure 6: Comparison of irradiance taken with both units. The calibration was performed at the actual site. Multiple s have been carried out at different times of the day. The outdoor inter-comparison showed 1.9% difference between units in the morning, 2% difference during midday and 0.36% in the afternoon where the irradiance levels are low. The s show only a small difference over the entire range, mostly in the range covered by the UV-Visible spectroradiometer. Also the estimated values of uncertainties from the outdoor analysis are given in Table III. In the case of outdoor calibration in-situ the due to repositioning is absent since the fibreoptic is not removed during calibration and actual procedures. Comparison of the integrated irradiance of both units with pyranometer s indicated differences around 6.4% between data. The comparison cannot be performed directly since the pyranometer has the ability to measure irradiance in wavelengths far beyond 1700 nm. Nonetheless, it is useful to compare relative differences between the sensors because the irradiance over 1700 nm is fairly constant over the conditions of interest. This comparison showed that performing the calibration at the actual site where the impact of the fibre-optic cabling is minimised, appeared to be the best way to reduce errors during the calibration procedure. In all the above cases the sensor integration time was chosen to minimise noise as well as saturation of the. The influence of ambient temperature was suggested as a daily trend in the difference between spectroradiometer and pyranometer s. However, this difference was less than 2% over the range of temperatures so far recorded, and therefore did not contribute significantly to the of the outdoor s.

5 3.3 Impact of collimator Finally a collimator was attached on System 2 outdoors and s were taken using both collimated and non-collimated systems to extract the impact of the collimator on spectral s. Outdoor data showed reduction of the spectrum in the presence of collimator since the collimator cuts out the diffused light, which is mostly of higher energy. Multiple s were performed in the presence of the collimator and the corresponding standard deviation was calculated. Higher deviation was present again in the visible region. In an attempt to observe the effect of the dust in the correct operation of the spectroradiometers and to the uncertainties of the system, s have been undertaken before and after the cleaning of the collimator. Measurements after cleaning are almost identical with the s taken beforehand. The combined standard in the case of the collimated spectroradiometer is therefore 3.66%. Comparison of the output spectra of the system with collimator and pyrheliometer has also been performed for short time periods in order to indicate the correct operation of the units as well as calibration drifts that maybe appear in short-term operation. The spectrum taken with the outdoor unit have been integrated and then compared with the pyrheliometer output. Initially the difference in irradiance between the two units is around 40 W/m 2.nm which corresponds to a difference of 4.4% which is acceptable. After 2 months of continuous operation of the units comparison was performed again. The difference in irradiance between both systems was found to be 45 W/m 2..nm demonstrating a small drift between the different sensors. The short-term degradation of the operation of the units suggest that periodic validation of the units is required. Further longer-term study is underway in order to investigate long-term drifts and environmental effects on. 4 CONCLUSIONS Figure 8: Repeated spectrum s and the corresponding standard deviation (a) before cleaning and (b) after cleaning. The list of uncertainties in the presence of collimator is shown in Table IV. Table IV: List of uncertainties in the presence of collimator. Source of Value ± Prob. Divisor Standard Initial calibration 2.5% Normal 1 2.5% 1.75% Normal % Accuracy of repositioning of the fiber optic diffuser 3.5% Rectangular % This work provides a starting point for understanding the uncertainties present during spectrum s and establishes a reliable system of spectral s with known uncertainties. A set of spectroradiometers calibrated at JRC (system 1) have been used to calibrate a second set of spectroradiometrs (system 2). Several s of spectral irradiance have been captured with system 2 indoors and outdoors under different conditions and compared with the original calibrated system 1. Results showed that there is good repeatability indoors under controlled conditions. Outdoor tests have demonstrated the impact of various factors such as fibre-optic bending and non-linearities on the proper operation of the spectroradiometers. These results showed that calibration at high irradiance levels is required. Non-linearities have been shown to cause significant influence on the correct operation of the spectroradiometers and have to be eliminated. Also calibration on the actual site has been shown to minimise calibration errors since it likely removes errors introduced by twisting/bending of the fibre optic cable. The results of the calibrated set of spectroradiometer used outdoors are in agreement with broadband devices such as pyranometers and pyrheliometers showing the correct operation of the systems. A comparison of the integrated spectroradiometer data against pyrheliometer and pyranometer s differ around 4.4% for DNI and 6.4% for GNI. A short-term study of the spectroradiometers showed a very small calibration drift on the operation of the units demonstrating that periodic recalibration of the system is needed. Further longer-term study is underway in order to investigate long-term drifts and environmental effects on. Acknowledgement This work has been co-financed by the European Regional Development Fund and by Republic of Cyprus in the framework of the project Spectrally Tuned Solar Cells for Improved Energy Harvesting with grant

6 number ΤΕΧΝΟΛΟΓΙΑ/ΕΝΕΡΓ/0311(ΒΙΕ)/13. 5 REFERENCES [1] Concentrator Photovoltaic (CPV) Module and Assemply Performance, Testing Standard Conditions, (2012) [2] M. Muller, B. Marion, J. Rodriguez, S. Kurtz, Proceedings 7th International Conference on Concentrating PV Systems (2011) 336. [3] R. Galleano, A. Gambardella, T. Huld, Proceedings 26th European Photovoltaic Solar Energy Conference (2011) 182. [4] S. Nann, C. Riordan, Proceedings 21st IEEE Photovoltaic Specialist Conference 2 (1990) [5] F. Fabero, F. Chenlo, Proceedings 22nd IEEE Photovoltaic Specialist Conference (1991) 812. [6] R. Galleano, W. Zaaiman, P. Morabito, A. Minuto, A. Spena, S. Bartocci, R. Fucci, G. Leanza, D. Pavanello, A. Virtuani, D. Fasanaro, M. Catena, M. Norton, Proceedings 8th International Conference on Concentrating PV Systems (2012) 139. [7] M. Krawczynski, M.B. Strobel, R. Gottschalg, Proceedings 24th European Photovoltaic Solar Energy Conference (2009) 3406.