Abstract
In a photovoltaic system, the prediction of long-term performance may require a statistical knowledge of solar spectral distribution for some types of PV cells. However, the measuring items of current PV plant monitoring does not contain any spectral observation. To attain it in an easy manner the author discusses the capabilities of ‘skipped sampling’. It can be classified to; (1) sampling with equidistant wavelength interval, and (2) sampling with non-equidistant wavelength boundary. The former is further classified to discrete sampling and average sampling. All the cases have been systematically studied for 5 types of PV cells by numerical methods. The analysis adopted plural different sampling phases to assure possible maximum estimation errors.
Major conclusions are; (1) Average sampling allows larger sampling interval approximately 3 to 5 times as much as discrete sampling, (2) it is demonstrated that non-equidistant sampling having appropriate wavelength boundary can decrease the required number of spectral bands compared with equidistant one, and (3) a proposed 6 band spectroradiometer gives precisions within -4 to +2.5% for efficiency evaluation and -2 to +1% for its seasonal deviation evaluation. It can be adopted as a practically excellent solution for system performance evaluation.
