Modelling the spatial distribution of installed solar photovoltaic capacity

Abstract

The global transition toward decarbonized energy systems has intensified the need for accurate, spatially resolved data on installed solar photovoltaic (PV) capacity. However, inconsistencies in reporting, limited geographic granularity, and varying measurement standards pose challenges for planning and assessment. This thesis addresses these challenges by developing spatial models to estimate, benchmark, and forecast installed PV capacity across global, regional, and subregional scales.

Structured as a thesis by publication, the work comprises three core studies. The first develops a global model of installed PV capacity at the national level, identifying key geographic and socioeconomic drivers. The second estimates regional capacity across 36 European countries, including those lacking official regional data. The third focuses on Great Britain, modelling subregional capacity. Collectively, the models disaggregate national capacity, benchmark deployment, and forecast where future capacity is likely to be installed—supporting efforts to monitor generation, reduce connection delays, plan grid expansion, and address land-use conflicts by identifying areas where solar development may compete with other uses, including agriculture.

The models exhibit strong performance across spatial scales. The Global Model estimates annual capacity additions with a global error of 9.7%. The European Model estimates cumulative capacity at the NUTS 2 level and achieves a national error of 19.5% when applied across all countries. In countries with available regional data—including the UK, Italy, Spain, Belgium, Germany, and France—the error falls to 2.5%. The GB Model achieves a national MAPE of 5.4% at the NUTS 3 level.

Across spatial scales, a shift in deployment drivers emerges. National capacity is shaped by socioeconomic factors, while regional and subregional deployment is driven by land-use characteristics, with artificial surfaces and agricultural areas as strong predictors. While solar irradiation is often assumed critical, the models show that structural and socioeconomic conditions are more influential, particularly in developed markets.