摘要
针对多区域电网协同调度面临的经济性、稳定性与隐私保护需求,提出一种基于联邦学习的分布式协同调度方法。首先,构建横向联邦架构,建立涵盖成本、稳定运行与隐私保护的多目标优化模型。其次,引入异步通信机制以降低通信同步开销。通过形式化建模明确调度目标和约束条件,利用联邦学习实现跨区域数据的隐私保护与协同优化,并结合压缩感知等技术提升通信的高效性和可靠性。实验结果显示,该方法可使发电成本平均降低12.5%,频率偏差减少60%,数据隐私性显著增强,同时通信效率得到有效改善。研究成果为多区域电网的高效安全协同调度提供了可行路径。
To address the economic efficiency,stability,and privacy protection requirements in multi-regional power grid coordinated scheduling,a distributed coordinated scheduling method based on federated learning is proposed.First,a horizontal federated architecture is constructed,establishing a multi-objective optimization model that covers cost,stable operation,and privacy protection.Second,an asynchronous communication mechanism is introduced to reduce communication synchronization overhead.Through formal modeling,scheduling objectives and constraints are clarified.Federated learning is employed to achieve privacy-preserving cross-regional data sharing and collaborative optimization,while techniques such as compressed sensing are integrated to enhance communication efficiency and reliability.Experimental results demonstrate that this method can reduce generation costs by an average of 12.5%,decrease frequency deviations by 60%,significantly enhance data privacy,and effectively improve communication efficiency.The research provides a feasible approach for efficient and secure coordinated scheduling in multi-regional power grids.
作者
王书欢
刘涛
李悦
WANG Shuhuan;LIU Tao;LI Yue(State Grid Zhoukou Power Supply Company,Zhoukou,Henan 466000,China)
关键词
联邦学习
多区域电网
协同调度
federated learning
multi-regional power grid
coordinated scheduling
