Review of methods to account for the solar spectral influence on photovoltaic device performance

The accuracy of photovoltaic (PV) performance forecasts is essential for improving grid penetration, fault detection, and financing of new installations. Failing to account for the spectral influence on PV performance can lead to weekly errors of up to 14% even for relatively stable technologies such as polycrystalline silicon. There exist a range models, known as spectral correction functions (SCFs), to account for the spectral influence on PV performance forecasts. These SCFs use different methods to characterise both the shift in PV performance due to the spectrum, and the solar spectrum itself. This review analyses the merits and limitations of seven commonly used spectral characterisation indices — five proxy variables (air mass, clearness index, precipitable water, aerosol, diffuse solar radiation ratio) and two variables extracted from the spectral distribution (average photon energy, depth of a water absorption band). The same analytical approach is adopted to review a further four indices (mismatch factor and its variants, (weighted) useful fraction, normalised short-circuit current) that are commonly used to characterise the variation in PV performance due to the solar spectrum. A review of ten SCFs that are based on these indices is undertaken to analyse the current state of the art of spectral correction modelling. The results of the review show that whereas some proxy-variable methods offer a simple and convenient way to account for the spectral influence in PV performance forecasts, they are surpassed in terms of accuracy by SCFs based on parameters derived directly from the spectrum, such as the average photon energy and the depth of spectral absorption bands. A decision-making framework is proposed to guide PV performance modellers in their choice of spectral correction model. The framework considers system specifications, climate, data availability, etc. The results of this work may be applied in, for example, software packages for PV performance forecasting to enable more accurate case-specific power forecasts. In future work, a standardised comparison of all SCFs and their respective indices is necessary to quantify the differences between a wider range of models than is currently available in the literature and substantiate the proposed framework.