Journal of Applied Research in Electrical Engineering

Journal of Applied Research in Electrical Engineering

Distribution Network Reconfiguration and Radiality Detection using Deep Reinforcement Learning

Document Type : Research article

Authors
Department of Electrical Engineering, Faculty of Engineering, Ferdowsi University of Mashhad, Mashhad, Iran
Abstract
Distribution network reconfiguration (DNR) is one of the most effective methods for reducing losses and damages in distribution networks, resulting in the least investment for power companies. One of the most critical constraints of the distribution network is that it remains radial during operation. In this paper, a novel method is presented for detecting the radiality of distribution networks. Then, this novel method is combined with the Double Deep Q Network algorithm, which is used to determine the optimal structure of the IEEE 33- and 69-bus distribution networks. This approach is implemented to minimize power losses and voltage deviation. Additionally, distributed generation sources, such as wind turbines and photovoltaic cells under fixed and uncertain capacity with varying load conditions, are utilized to investigate their impact on this approach.

Highlights

• Proposing a novel method for validating the radiality of distribution networks. 
• Employing deep reinforcement learning as a model-free approach to solve the 
distribution network reconfiguration problem. 
• Investigating the impact of sensitivity analysis in hyperparameters of deep 
reinforcement learning agents. 
• Validation of the proposed method on well-known IEEE distribution 
networks, such as 33- and 69-bus. 
• Validating the proposed algorithm under uncertainties arising from varying 
load conditions and distributed generation units.

Keywords
Subjects

Volume 4, Issue 2
Summer and Autumn 2025
July 2025
Pages 165-177

  • Receive Date 12 October 2025
  • Revise Date 31 January 2026
  • Accept Date 25 February 2026