Solar cell efficiency divergence due to operating spectrum variation

Abstract

A limitation in the performance rating of solar cells and modules is that they are tested using a single value for the solar spectrum. To map the performance expected under the varying spectra found in operating conditions, solar cell efficiencies have been evaluated using solar spectra generated by the National Solar Radiation Database, applied to confirmed record-efficiency cell parameters. Nine solar cell types (single-junction and multijunction) are evaluated using spectra at more than forty locations, spanning 76° of latitude and 150° of longitude, at hourly intervals over a year. Relative to the standard test efficiency, increases in annual operating efficiency are seen in cadmium telluride and single-junction perovskite designs, while efficiency decreases are observed in two-terminal multijunction structures. Though silicon exhibits the least variation, its-3% to +2% range in relative efficiency is equivalent to ~20° C of temperature variation. This divergence in operating efficiencies indicates that evaluation using a single spectrum is not a sufficient basis for comparison, or prediction of energy yield in operation. Application of an additional "operating spectrum," to complement the standard test spectrum, is proposed. Fig. 1. Source locations for spectra in this study. Each marker represents about 4000 hourly spectra, for wavelengths from 280 nm to 4000 nm. In six locations, hourly spectra were generated for direct normal irradiance (DNI) using inputs from the Typical Meteorological Year 3 database (NREL, 2010) to generate spectra using the SMARTS model (Gueymard, 1995). Map courtesy of NREL's NSRDB.

Impact of measured spectrum variation on solar photovoltaic efficiencies worldwide

“…Cells with a wider spectral response (Figure 2) are more tolerant to the increased spectral variation that arises from sun tracking. In Figure 6 and Appendix F, increasing the sun tracking is seen to benefit single-junction cells with a wider spectral response range (silicon, CIGS), but is a detriment to cells with a narrower spectral response (cadmium telluride, perovskites) 23,39,40 . For cells with more than one junction, the result is more mixed, as losses due to current mismatch between the junctions trade against the wider spectral response of the overall stack.…”

“…Consistent with the ratings for temperature and broadband irradiance 16 , rating under more than one condition would enable interpolation and extrapolation to other spectral conditions. Figure 4 suggests that AM0, the standard spectrum in space applications 58 , could be re-purposed to bracket terrestrial spectrum variation 23 . Fortunately, characterization under two spectra does not require two measurements: if the spectral response of a cell is known, efficiency under a single condition can be translated to that under a second condition using a calculation known as "spectral mismatch correction 83 .…”

“…While synthetic spectra have demonstrated the ability to predict solar energy generation under a narrow set of field conditions [68][69][70] , direct, ground-level measurements such as those presented here are needed to fine-tune the models and verify that the synthetic spectra remain accurate across the range of atmospheric conditions under which solar energy is generated worldwide. Synthetic spectra have previously been applied to analysis of the nine cell types in [ 23 ]. Differences between results from synthetic 66,67 and measured spectral irradiance are shown in Appendix E. Spectral irradiance sensors can be mounted in orientations with varying degrees of solar tracking (Figure 1).…”

“…Variation of mere tenths of a percent now imply billions of euros gained or lost over the decades of operation -before the cost of carbon is included 22 . As detailed in [ 23 ], for a single 100-MW PV plant that displaces fossil-fuel generation, a 2% variation in operating performance is equivalent to about €150,000 in annual revenue and ~1,500 tons of avoided CO 2 , or nine one-way flights between New York and Paris each year. As of 2021, the world has installed PV capacity equivalent to eight thousand such plants 24 .…”

“…In the past two decades, as new solar cell designs using novel materials have proliferated 25 , evidence for a need to incorporate spectral variation [4][5][6] has grown harder to ignore. However, given the limited amount of measured spectral irradiance data available, previous studies on its impact have had to rely on relatively small data sets, or synthetic spectra 5,23,39,57 .…”