📄 中文摘要
随着建筑运营数据的日益丰富，强化学习（RL）被广泛应用于从数据中直接学习控制策略，以应对大规模建筑集群的复杂性和不确定性。然而，现有的仿真环境大多侧重于建筑端性能指标，缺乏对电网层面影响的系统评估，同时其实验工作流程仍然依赖于手动配置和大量编程专业知识。因此，提出了AutoB2G，一个自动化的建筑-电网协同仿真框架，该框架能够仅基于自然语言任务描述完成整个仿真工作流程。该框架扩展了CityLearn V2，以支持建筑与电网之间的协同仿真。
AutoB2G：一种基于大型语言模型的自动化建筑-电网协同仿真框架
📄 English Summary
AutoB2G: A Large Language Model-Driven Agentic Framework For Automated Building-Grid Co-Simulation
The increasing availability of building operational data has led to the application of reinforcement learning (RL) for learning control policies directly from data, addressing the complexity and uncertainty of large-scale building clusters. However, most existing simulation environments focus on building-side performance metrics and lack systematic evaluation of grid-level impacts, while their experimental workflows still heavily rely on manual configuration and substantial programming expertise. To address these issues, AutoB2G is proposed as an automated building-grid co-simulation framework that completes the entire simulation workflow based solely on natural language task descriptions. This framework extends CityLearn V2 to support Building-to-Grid (B2G) interactions.