Determinants of the distribution of utility-scale photovoltaic power facilities across the globe

Abstract

Photovoltaic power (PV) is the fastest-growing source of renewable electricity. Making reliable scenarios of PV deployment requires information on what drives the spatial distribution of PV facilities. Here we empirically derive the determinants of the distribution of utility-scale PV facilities across six continents, using a mixed effects logistic regression modelling approach relating the occurrence of over 10 000 PV facilities to a set of potential determinants as well as accounting for country and spatially correlated random effects. Our regression models explain the distribution of PV facilities with high accuracy, with travel times to settlements and irradiation as the main determinants. In contrast, our results suggest that land cover types are not strong determinants of the PV distribution, except for Asia and Africa where the PV distribution is related to the presence of agriculture, short natural vegetation and bare land. For Europe and Asia a considerable part of the variance in PV distribution is explained by inter-country differences in factors not included in our fixed determinants. Relevant determinants identified in our study are in line with the main assumptions made in cost of electricity (COE) maps used in the IMAGE integrated assessment model (IAM). However, we found correlations (Spearman ρ ) of −0.18–0.54 between our PV probability maps and IMAGE’s COE maps. These may partly be explained by conceptual differences between our empirically-derived probability maps and the COE maps, but we also recommend using higher-resolution maps of PV potential and COE computations such as used in IAMs.

Keywords

• utility-scale PV
• regression models
• integrated assessment models
• determinants of distribution
• PV potential