A Comparative Study of Reordering Techniques for Consecutive Numbering in Power Grids

Document Type : Research Article

Author

Department of Electrical Engineering, Golestan University, Golestan , Iran

Abstract

The structure and ordering of nodes in power grids significantly affect the performance of algorithms used for simulation, optimization, and visualization. While synthetic generation of power grids has gained momentum in both transmission and distribution domains, limited research has addressed how node indexing impacts matrix sparsity patterns and computational performance. This paper presents a comprehensive study of node reordering techniques aimed at achieving consecutive numbering in complex power networks. Four categories of methods are explored: heuristic approaches, optimization-based strategies, graph-theoretic algorithms, and artificial intelligence (AI)-driven models. We revisit existing heuristics and propose a novel metaheuristic optimization method that maximizes diagonal density across multiple bandwidth levels. For radial distribution grids, we develop a graph-based method using depth-first search (DFS) that aligns node numbering with physical structure. In addition, we evaluate the feasibility of using convolutional and feedforward neural networks to learn reordering patterns from data. Despite training on thousands of synthetic graphs, standard AI models fail to produce valid permutations, highlighting the need for permutation-aware architectures. Extensive experiments on synthetic transmission and distribution networks confirm that method selection should align with network topology. Optimization excels in meshed systems, while graph-based DFS is ideal for tree-like networks. A comparative analysis is provided to guide future applications and research in this space.

Keywords

Main Subjects


Volume 4, Issue 2
Issue in progress
July 2025
  • Receive Date: 02 July 2025
  • Revise Date: 21 November 2025
  • Accept Date: 29 December 2025