Traditional physics-based simulation approaches for urban facade photovoltaic potential assessment remain computationally intractable for metropolitan-scale deployment, requiring weeks to months of processing time that effectively paralyzes evidence-based urban energy policy development. This computational barrier has prevented the transition from theoretical renewable energy potential to operational decarbonization planning tools despite building facades representing the primary scalable pathway for distributed solar generation in space-constrained urban environments. To overcome this fundamental barrier, we introduce E2AY-Net, an end-to-end deep learning framework that transforms urban facade PV assessment from slow multi-stage simulation into instantaneous spatially-resolved energy yield generation. E2AY-Net integrates three specialized encoding pathways: convolutional neural networks capturing hierarchical urban morphological features across spatial scales from individual facades to city-scale configurations, Transformer architectures processing arbitrary-length meteorological sequences spanning hours to years with attention mechanisms preserving long-range temporal dependencies, and multilayer perceptrons accommodating diverse photovoltaic module specifications. Validated in the hyper-dense urban environment of Hong Kong, the framework achieves comprehensive annual assessment of the complete urban domain in 33.79 s with 678,955 computational acceleration, while maintaining engineering-grade accuracy with 5.56% mean relative error and 84.6% of building surfaces achieving predictions within 10% tolerance. The anytime capability enables flexible assessment across arbitrary planning horizons from short-term feasibility studies to comprehensive annual evaluations through processing variable-length meteorological sequences in single forward passes without architectural modification or pipeline re-execution. Strategic deployment analysis reveals that targeted installation on the highest-performing 25% of facades, concentrated within merely 9.2% of total available facade area, achieves 3200 GWh annual generation potential with 1500 kt CO2 emission reduction capacity. This work establishes a practical breakthrough enabling the transition from computationally intractable urban energy assessment to real-time interactive planning tools, fundamentally transforming urban building envelopes into accessible distributed energy infrastructure for evidence-based decarbonization policy development across space-constrained metropolitan environments worldwide.
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号