Graph Optimization-Based Automatic Partitioning Method for Electromagnetic Transient Simulation of AC-DC Distribution Networks

Abstract

The large-scale integration of renewable generation and power-electronic devices has led distribution networks to exhibit AC-DC hybrid configurations and multi-node common-bus structures, significantly increasing the scale and coupling complexity of electromagnetic transient (EMT) simulation. Traditional branch-cut partitioning methods provide limited parallel speedup under such topologies and cannot fully exploit multi-core computing resources. To address this issue, this paper proposes a graph optimization-based automatic partitioning method to generate optimal partition schemes for node-tearing-based parallel simulation. A node-cut optimization model is formulated under node-assignment and boundary constraints, with the dual objectives of minimizing subdomain size and the number of boundary nodes, and solved using the ε-constraint method to automatically obtain partition results. According to the optimized partition results, the admittance matrix is reordered and reconstructed by relocating torn nodes to the boundary, forming a bordered block-diagonal (BBD) structure that enables parallel subdomain solutions with serial boundary coordination. Case studies demonstrate that, while maintaining nearly identical numerical accuracy, the proposed method significantly reduces the solution time of node-voltage equations, achieving a speedup of nearly three times. It is therefore well suited for parallel EMT simulation of AC-DC distribution networks with multi-node common-bus structures.