Cloud-Edge coordinated economic scheduling for integrated Source-Grid-Load-Storage system considering demand response

• A scheduling model for integrated source-grid-load-storage systems with refined industrial load models is established. • A cloud-edge coordinated scheduling architecture based on a customized physics-embedded neural network is proposed. • A privacy-preserving cloud-edge collaborative scheduling algorithm is developed and comprehensively validated. When coordinating adjustable loads in integrated source-grid-load-storage systems to participate in demand response programs, it is often challenging to simultaneously capture the complex dynamic regulation characteristics of industrial loads and overcome the bottlenecks of production data confidentiality. To address these issues, this paper proposes a cloud-edge coordinated economic scheduling method based on a physics-embedded neural network architecture. By structurally integrating the physical dynamics and operational constraints of heterogeneous devices into the neural network, this approach enables the efficient utilization of the system’s regulatory flexibility while ensuring data privacy. Specifically, this study encompasses three main contributions. First, a comprehensive scheduling model for the integrated source-grid-load-storage system is established, incorporating refined regulation models for two typical industrial loads. Second, a customized physics-embedded neural network is constructed, wherein the physical dynamics and operational constraints of heterogeneous devices are structurally encoded into specialized neurons. Third, a privacy-preserving cloud-edge collaborative scheduling algorithm is developed, in which distributed backpropagation and a gradient correction mechanism are constructed to jointly derive economic scheduling strategies while strictly preserving data privacy. Finally, multi-scenario simulations under varying production tasks, along with comprehensive comparisons and ablation analysis, are conducted to validate the effectiveness and practical advantages of the proposed method.