Advancing building facade solar potential assessment through AIoT, GIS, and meteorology synergy

Abstract

The assessment of building solar potential plays a pivotal role in Building Integrated Photovoltaics (BIPV) and urban energy systems. While current evaluations predominantly focus on rooftop solar resources, a comprehensive analysis of building facade BIPV potential is often lacking. This study presents an innovative methodology that harnesses state-of-the-art Artificial Intelligence of Things (AIoT) techniques, Geographic Information Systems (GIS), and Meteorology to develop a model for accurately estimating spatial–temporal building facade BIPV potential considering 3 Dimension (3D) shading effect. Here, we introduce a zero-shot Deep Learning framework for detailed parsing of facade elements, utilizing cutting-edge techniques in Large-scale Segment Anything Model (SAM), Grounding DINO (Detection Transformer with improved denoising anchor boxes), and Stable Diffusion. Considering urban morphology, 3D shading impacts, and multi-source weather data enables a meticulous estimation of solar potential for each facade element. The experimental findings, gathered from a range of buildings across four countries and an entire street in Japan, highlight the effectiveness and applicability of our approach in conducting comprehensive analyses of facade solar potential. These results underscore the critical importance of integrating shadow effects and detailed facade elements to ensure accurate estimations of PV potential.