Data-driven urban energy saving optimization: a path to carbon reduction in smart cities

Traditional urban management models often suffer from data silos and delayed responses, hindering effective energy conservation under the “dual carbon” strategy. To address this, we propose a multi-source data-driven closed-loop optimization mechanism, validated using operational data from Shenzhen. By integrating NB-IoT and sensor data across buildings, transportation, and industry, the system establishes a “perception-determination-execution-feedback” loop. Crucially, a dual-mode architecture comprising a main closed-loop and an emergency response loop enables rapid dynamic regulation during high-load scenarios like holidays and extreme weather. Pilot results in Nanshan and Longgang districts demonstrate significant improvements: building energy consumption decreased by 12.7%, industrial carbon emissions by 11.8%, and traffic management latency was compressed to under 3 s, achieving a maximum overall energy saving of 14.1%. Robustness experiments confirmed stable performance across normal, peak, and emergency traffic conditions. This study offers a replicable, data-driven pathway for real-time decision-making and green governance in megacities.