Spectral irradiance distribution analysis: The impact on photovoltaic performance combined with the impact of clouded skies and a comparison to model outputs

Abstract

Five-minute measurements of both the average photon energy and photovoltaic performance are used in this project for the analysis of the impact of variations in the average photon energy on photovoltaic performance. Four days of measurements have been analyzed, these four days comprise a clear sky day, a day with alternating high and low irradiance peaks (super irradiance) and two clouded days. The impact of this variation is clearly shown in the comparison between the average photon energy versus irradiance plot and the efficiency versus irradiance plot of the clear day data. These plot show that the relatively higher average photon energy relates to a relatively lower efficiency. This effect confirmed by the overall statistic taken over all four days whereby the higher than average average photon energy and lower than average efficiency has the highest share in the measurements. The difference between the two investigated module types (amorphous and (multi)crystalline) is proven in the efficiency versus irradiance plot. The behaviour at the relatively higher average photon energy is similar in both cases. In the case of the relatively lower average photon energy the efficiency of the amorphous modules remains at the same level as at average average photon energies. The impact of clouds on the spectral irradiance distribution is investigated in this project. The comparison is made between the average photon energy of the measured data and the extraterrestrial average photon energy. In this way a description of the whole atmospheric impact is made. This shows that the clouded sky impact on the average photon energy on the earth’s surface is that it increases the average photon energy. To conclude this project a comparison between modelled and measured average photon energy is made. This shows that the modelled average photon energy is underestimating the measured data on the clouded days. It also shows that the variability on the super irradiance day is lost in the modelling output. For model results on the clear day the values line up closely to the measured values.