Accurate parameter extraction for photovoltaic models is critical for optimizing system efficiency and ensuring operational stability. However, conventional deterministic and metaheuristic methods often suffer from premature convergence due to the strong nonlinearity and multimodality of the objective function. To address these challenges, this paper proposes a framework capable of adaptive strategy selection, termed large language model-based swarm intelligence. By embedding Meta LLaMA-3.1-405B-Instruct as the high-level strategic reasoning core, the proposed method integrates a triple stagnation detection mechanism that dynamically switches between optimization algorithms to prevent local optima entrapment. Furthermore, a physics-informed validation module is incorporated to enforce boundary constraints and ensure the physical feasibility of the identified parameters. Comparative experiments demonstrate that large language model-based swarm intelligence yields superior accuracy over state-of-the-art metaheuristics, achieving root mean square error reductions of 84.6% for the single diode model, 84.3% for the double diode model, and 82.1% for the triple diode model.
Solar photovoltaic deployment is pivotal for achieving carbon neutrality by harnessing renewable energy in urban environments. Accurately assessing the role of urban morphology in shaping solar potential distribution is critical for policy design, particularly in metropolitan areas where vertical surfaces constitute a significant share of directly sun-exposed urban areas. However, conventional models relying on fixed albedo values oversimplify multi-reflective dynamics, underestimating indirect radiation contributions and introducing systemic errors in solar potential estimation. This simplification further obscures the interplay between urban morphology and solar redistribution, leading to an incomplete understanding of morphological impacts. In this work, we propose a machine learning-based framework that integrates material-specific albedo with 3D urban geometry to quantify morphological importance. Simulation results identify the mean façade albedo as the dominant annual-scale driver (22.7% relative importance), followed by building volume (21.6%) and façade area (7.5%). Seasonal and diurnal analyses reveal temporal dependencies: volumetric factors govern solar gains by affecting shadow effects during temporal transitions, while density and aggregated neighborhood reflectance synergistically enhance multi-reflections in these periods. These findings underscore the necessity of balancing urban morphology with material albedos to maximize solar harvesting in urban planning.
