With the global transformation of the energy structure and the advancement of low-carbon goals, the role of urban indoor substations in smart grids has become increasingly important. The energy efficiency and carbon footprint of substations directly impact the sustainability of the entire urban energy system. This paper proposes a method that combines AI-driven digital twin technology to achieve precise monitoring and dynamic optimization of the carbon footprint of urban indoor substations. By constructing a digital twin model of the substation, real-time operational data is collected, and machine learning algorithms are used for predictive maintenance of equipment failures, along with optimizing energy efficiency management strategies. We developed an AI-based intelligent scheduling system to minimize energy consumption and carbon emissions while ensuring the reliable operation of the substation.