Work

Efficient reconfiguration of flexible distribution networks: A physics-informed interactive branch-and-bound framework

Year: 2026
Type: article
Abstract: Flexible distribution networks (FDNs) are characterized by more complex and frequent topology reconfiguration, which is enabled by extensive controllable switches. However, the numerous discrete switches create a combinatorial explosion, challenging conventional branch-and-bound (B&B) methods for fast and reliable decision-making. In this regard, this paper proposes a physics-informed deep learning framework to accelerate B&B for FDN reconfiguration, which interacts within the B&B branching loop to strictly preserve global optimality. First, a shortest power supply path-based switch variable evaluation mechanism is proposed to quantify the impact of variable changes on network power distribution. Then, a hybrid graph learning architecture-based branching variable prediction model is developed to predict the optimal branching variable by jointly capturing variable physical attributes, historical branching trajectories, and global search tree features. Based on this, an interactive iterative framework embedded within the B&B process is established to dynamically optimize branching decisions, achieving reliable and fast solving while preserving solution space completeness. Simulation results demonstrate that the proposed method significantly reduces the size of the B&B search tree and computational time.
Language: English
Open Access status: gold
Estimated APC paid: $3,920
OpenAlex ID: W7204100759
FWCI:
Cites: 30
Cited by:
Related to: 10