Cyber-physical integrated resource scheduling for distribution networks based on cloud-edge-end collaborative federated consensus learning

The widespread integration of high-penetration renewable energy and massive distributed resources has introduced significant challenges to the economic efficiency, operational stability, and real-time responsiveness of distribution networks. Traditional centralized optimization struggles with renewable intermittency and limited communication resources, whereas existing distributed consensus methods often converge slowly and are susceptible to local optima. Therefore, this paper proposes an cyber–physical integrated resource scheduling architecture for distribution network energy management. For day-ahead scheduling, a cloud–edge federated learning algorithm is adopted to enhance prediction accuracy and economic performance. For intraday scheduling, a delay-fairness-aware matching mechanism and edge-end collaborative hierarchical consensus framework are introduced to optimize multi-agent interaction object selection and realize dynamic consensus update, substantially accelerating consensus convergence. Simulation results demonstrate that the proposed method outperforms conventional algorithms in prediction accuracy, economic efficiency, photovoltaic (PV) consumption enhancement, and convergence behavior, providing an effective solution for coordinated optimization in distribution networks with high penetration of renewable energy.