Assessment of Large-Scale Solar Photovoltaic Potential on Building Roofs and Facades Using Geo-Aware Graph Attention Networks

Assessing building photovoltaic (PV) potential is crucial for urban energy planning and achieving net-zero energy building goals. However, it still faces numerous limitations and challenges at the urban scale. Traditional solar irradiation simulation methods are computationally inefficient and difficult to apply to large-scale, high-resolution urban building analysis. Machine learning methods for predicting solar irradiation often inadequately consider the interrelationships among buildings. Moreover, most studies have overlooked the significance of building facades in PV power generation potential, with only a few small-scale studies simultaneously considering both rooftop and facade PV potential. This study proposes a novel geo-aware graph attention network (GA-GAT) model to accurately and efficiently predict solar irradiation on large-scale urban buildings. The model thoroughly considers the complex interactions among neighboring buildings and assesses the potential for buildings to achieve net-zero energy and net-zero electricity based on the predictions. The study focuses on approximately 45000 buildings in Manhattan, New York City, analyzing solar irradiation on rooftops and facades with a spatial resolution of 1 meter. Through this city-wide, high-resolution solar irradiation model, the study further explores the potential for Manhattan to achieve net-zero energy (NZEB) and net-zero electricity buildings (NZEB-e). The results show that achieving NZEB remains challenging. However, if both roof and facade are considered, NEZB-e has the potential to be achieved in many buildings. This research provides a scientific basis and decision support for urban PV planning and NZEB policy formulation.