Digital Twin-Enhanced Forecasting for Rooftop Photovoltaic Systems in Urban Environments: A Scoping Review and Conceptual Framework

Accurate forecasting of rooftop photovoltaic (PV) output is essential for managing urban energy systems, yet remains hindered by weather variability, site-specific shading, and limited real-time adaptability. While digital twin (DT) technologies offer promising capabilities, combining physical models with real-time data and AI, the research landscape is fragmented and lacks synthesis. This paper conducts a scoping review to systematically analyze DT-based forecasting approaches and benchmark them against advanced non-DT methods. Findings show that DTs can improve forecasting accuracy by up to 60%, particularly under high-variability conditions, but face challenges in scalability, complexity, and data requirements. To address these, the paper proposes a novel conceptual framework integrating layered DT architecture with probabilistic AI, federated learning, and large language models. This framework provides a structured foundation for developing scalable, adaptive, and interpretable forecasting solutions tailored to smart urban PV integration.