Direct beam and hemispherical terrestrial solar spectral distributions derived from broadband hourly solar radiation data

Multiple junction and thin film photovoltaic (PV) technologies respond differently to varying terrestrial spectral distributions of solar energy. PV device and system designers are concerned with the impact of spectral variation on PV specific technologies. Spectral distribution data are generally very rare, expensive, and difficult to obtain. We modified an existing empirical spectral conversion model to convert hourly broadband global (total hemispherical) horizontal and direct normal solar radiation to representative spectral distributions. Hourly average total hemispherical and direct normal beam solar radiation, such as provided in Typical Meteorological Year (TMY) data, are spectral model input data. Default or prescribed atmospheric aerosols and water vapor are possible inputs. Individual hourly and monthly and annual average spectral distributions are computed for a specified tilted surface. The spectral range is from 300 nm to 1800 nm. The model is a modified version of the Nann and Riordan SEDES2 model. Measured hemispherical spectral distributions for a wide variety of conditions at the Solar Radiation Research Laboratory at the National Renewable Energy Laboratory, Golden, Co. and Florida Solar Energy Center (Cocoa, FL) show that reasonable spectral accuracy of about ±10% is obtainable with exceptions for weather events such as snow. Differing cloud climatology and variable albedo and aerosol optical depth atmospheric conditions can lead to spectral model differences of 30–40%.

► Spectral distribution data are needed to assess photovoltaic material performance.
► We describe a model to convert hourly global and direct solar to spectral data.
► Model is designed primarily for data from Typical Meteorological Year (TMY) files.
► Any hourly global, direct solar irradiance, aerosol, and water vapor data may be used.
► Clear and overcast sites have best accuracy (10%), partly cloudy sites worst (40%).