Fast dynamic reconfiguration for large-scale distribution networks via decision-oriented spatiotemporal integrated reduction

Dynamic reconfiguration (DR) is essential for the secure and economic operation of active distribution networks. However, DR is commonly formulated as a computationally demanding mixed-integer nonlinear programming (MINLP) problem. Existing studies often improve tractability through period division and spatial key switch selection. These approaches are usually driven by load-profile similarity and do not fully account for topological characteristics or decision consistency across time steps. Static key switch selection also cannot adapt to time-varying optimal topologies, which can lead to suboptimal reconfiguration decisions. To address these limitations, this paper proposes a decision-oriented spatiotemporal dimensionality reduction method. First, a multi-objective reconfiguration model is formulated to minimize network loss and the load balancing index. Second, an integrated reduction framework is developed. By quantifying multi-time-step decision interoperability and incorporating switch operation similarity, the proposed method jointly determines the period division and period-wise critical switches. This design reduces the solution space in both temporal and spatial dimensions while preserving decision quality. The method is validated on standard 16-node and 72-node systems and on a real-world 395-node network in Southern China. Simulation results show that the proposed approach achieves a favorable trade-off between solution quality and computational efficiency compared with traditional reduction methods.