Developing automated methods to estimate spectrally resolved direct normal irradiance for solar energy applications

Abstract

We describe four schemes designed to estimate spectrally resolved direct normal irradiance (DNI) for multi-junction concentrator photovoltaic systems applications. The schemes have increasing levels of complexity in terms of aerosol and circumsolar irradiance (CSI) treatment, ranging from a climatological aerosol classification with no account of CSI, to an approach which includes explicit aerosol typing and type dependent CSI contribution. When tested against ground-based broadband and spectral measurements at five sites spanning a range of aerosol conditions, the most sophisticated scheme yields an average bias of þ 0:068%, well within photometer calibration uncertainties. The average spread of error is 2:5%. These statistics are markedly better than the climatological approach, which carries an average bias of 1:76% and a spread of 4%. They also improve on an intermediate approach which uses Angstrom€ exponents to estimate the spectral variation in aerosol optical depth across the solar energy relevant wavelength domain. This approach results in systematic under and over-estimations of DNI at short and long wavelengths respectively. Incorporating spectral CSI particularly benefits sites which experience a significant amount of coarse aerosol. All approaches we describe use freely available reanalyses and software tools, and can be easily applied to alternative aerosol measurements, including those from satellite.