Generative Adversarial Networks for Climate-Sensitive Urban Morphology: An Integration of Pix2Pix and the Cycle Generative Adversarial Network

Land, Journal Year: 2025, Volume and Issue: 14(3), P. 578 - 578

Published: March 10, 2025

Urban heat island (UHI) effects pose significant challenges to sustainable urban development, necessitating innovative modeling techniques optimize morphology for thermal resilience. This study integrates the Pix2Pix and CycleGAN architectures generate high-fidelity models aligned with local climate zones (LCZs), enhancing their applicability studies. research focuses on eight major Chinese coastal cities, leveraging a robust dataset of 4712 samples train generative models. Quantitative evaluations demonstrated that integration substantially improved structural fidelity realism in synthesis, achieving peak Structural Similarity Index Measure (SSIM) 0.918 coefficient determination (R2) 0.987. The total adversarial loss training stabilized at 0.19 after 811 iterations, ensuring high convergence structure generation. Additionally, CycleGAN-enhanced outputs exhibited 35% reduction relative error compared Pix2Pix-generated images, significantly improving edge preservation feature accuracy. By incorporating LCZ data, proposed framework successfully bridges climate-responsive planning, enabling adaptive design strategies mitigating UHI effects. enhance generation, while classification produce forms align specific climatological conditions. Compared model trained by coupled alone, approach offers planners more precise tool designing optimizing layouts mitigate effects, improve energy efficiency,

Language: Английский

Strategic management of solar generation for solar electric vehicle charging in microgrids using deep reinforcement learning

Journal of Renewable and Sustainable Energy, Journal Year: 2025, Volume and Issue: 17(3)

Published: May 1, 2025

The integration of solar electric vehicles (SEVs) into microgrids, particularly those enriched with photovoltaic (PV) systems, presents unique challenges due to the inherent variability in energy and dynamic consumption patterns SEVs. This study aims address these complexities by developing an advanced operational framework that enhances management flows within leveraging capabilities modern artificial intelligence. Utilizing a deep double Q-network (DDQN), this research introduces sophisticated method dynamically adapt fluctuations generation SEV demands, ensuring efficiency, sustainability, grid stability. methodology encompasses detailed mathematical modeling generation, consumption, storage dynamics, integrated environmental economic constraints simulate realistic microgrid scenarios. DDQN is employed optimize distribution strategies real-time, based on predictive analytics responsive control mechanisms. approach not only copes stochastic nature renewable sources usage but also capitalizes aspects improve overall performance. paper contributes novel management, for systems incorporating SEVs PV generation. By optimizing interplay between power availability charging requirements, provides strategic insights can guide infrastructure investments tactics, promoting more efficient economically viable systems. proposed models are expected significantly advance field paving way development smarter, resilient urban environments.

Language: Английский