Modelling regional solar photovoltaic capacity in Europe: A data-driven approach for disaggregation, benchmarking, and forecasting

Abstract

The rapid expansion of solar photovoltaic (PV) technology has established it as a leading contributor to global renewable energy capacity. However, integrating solar PV into existing power grids presents significant challenges, primarily due to the variable nature of solar energy generation and the lack of accurate and complete data on installed PV capacity at the regional level. This study addresses this critical gap in capacity measurement by analysing the factors influencing regional solar PV deployment and developing models to estimate installed PV capacity across 333 regions of 36 European countries. We employed Pearson and Spearman correlation analyses to identify key geographic factors such as agricultural land area, solar irradiance and population, related to solar PV deployment. This informed the development of extreme gradient boosted parallel tree algorithm (XGBoost) models for estimating regional PV capacity. The models achieve a root mean squared capacity error (RMSE) of less than 272 MW, and explain more than 93% of the variation across 150 NUTS 2 EU regions. The models serve three primary purposes: disaggregating national PV capacity into regional figures, benchmarking inter- and intra-regional capacities, and forecasting future PV capacity distribution. The models presented in this study offer a comprehensive tool for policymakers and grid operators, enabling the design of more effective policy interventions and enhanced solar PV monitoring services. This research contributes to more sustainable and efficient energy planning in Europe.