Reconstruction of solar spectral resource using limited spectral sampling for concentrating photovoltaic systems

One of the challenges associated with forecasting and evaluating concentrating photovoltaic system (CPV) performance in diverse locations is the lack of high-quality spectral solar resource data. Various local atmospheric conditions such as air mass, aerosols, and atmospheric gases affect daily CPV module operation. A multi-channel filter radiometer (MFCR) can be used to quantify these effects at relatively low cost. The proposed method of selectively sampling the solar spectrum at specific wavelength channels to spectrally reconstruct incident irradiance is described and extensively analyzed. Field spectroradiometer (FSR) measurements at the University of Ottawa's CPV testing facility (45.42 degrees N, 75.68 degrees W) are fed into our model to mimic the outputs from the MCFR. The analysis is performed over a two year period (2011-2012), using 46,564 spectra. A recommendation is made to use four aerosols channels at 420, 500, 780, and 1050 nm, one ozone channel at 610 nm and one water vapour channel at 940 nm, all of which can be measured with ubiquitous Si photodiodes. A simulation of this MFCR channel configuration produces an RMS error under 1.5% over 96% of the 350-1830 nm range, when compared with the FSR, for the 2012 data set in Ottawa.