EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Hybrid Multi‐Agent Reinforcement Learning and Consensus Control for Dynamic V2G‐Enabled Voltage Regulation in Active Distribution Networks

Mojtaba Ajoudani, Seyed Reza Mosayyebi, Ramazan Teimouri Yansari

IET Generation, Transmission & Distribution · 2026-01 · 20 (1)

DOI: 10.1049/gtd2.70413 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IET Generation, Transmission & Distribution, online erschienen am 27.08.2026. Verfasst von Mojtaba Ajoudani und 2 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 47 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Active low‐voltage distribution networks with high penetration of photovoltaic (PV) systems and vehicle‐to‐grid (V2G)‐enabled electric vehicles (EVs) face significant voltage regulation challenges due to renewable intermittency, stochastic EV behaviour, and dynamic plug‐and‐play participation of mobile storage resources. Conventional voltage‐control approaches are often reactive, model‐dependent, or unable to accommodate time‐varying agent populations in real time. This paper proposes a Hybrid Multi‐Agent Reinforcement Learning and Consensus (HMC) framework for decentralised voltage regulation in PV‐rich active distribution networks. The proposed architecture integrates three coordinated layers: (i) fast local droop control for immediate stabilisation, (ii) multi‐agent proximal policy optimisation (MAPPO) under centralised training and decentralised execution (CTDE) for anticipatory decision‐making, and (iii) a modified ratio‐consensus mechanism with dynamic graph management for fair coordination and seamless real‐time integration of transient V2G resources. The control problem is formulated as a partially observable Markov decision process (POMDP) considering stochastic PV/load variations and EV user constraints. The proposed framework is validated on a modified IEEE 33‐bus distribution feeder using realistic stochastic datasets and compared with distributed consensus control (DCC), model predictive control (MPC), alternating direction method of multipliers (ADMM), and standalone deep reinforcement learning baselines. Simulation results demonstrate that the proposed framework reduces voltage violations by up to 82.9%, decreases network losses by 18.0%, and lowers BESS cycling by 36.8% while maintaining 94.7% EV user satisfaction. Furthermore, the framework remains robust under forecast uncertainty and communication impairments, including ±30% forecast errors, 500 ms communication delays, and 20% packet loss, demonstrating its practical suitability for future V2G‐enabled active distribution networks.

Zitieren

APA

Ajoudani, M., Mosayyebi, S. R., & Yansari, R. T. (2026). Hybrid Multi‐Agent Reinforcement Learning and Consensus Control for Dynamic V2G‐Enabled Voltage Regulation in Active Distribution Networks. IET Generation, Transmission & Distribution, 20(1). https://doi.org/10.1049/gtd2.70413

Vancouver

Ajoudani M, Mosayyebi SR, Yansari RT. Hybrid Multi‐Agent Reinforcement Learning and Consensus Control for Dynamic V2G‐Enabled Voltage Regulation in Active Distribution Networks. IET Generation, Transmission & Distribution. 2026;20(1). doi:10.1049/gtd2.70413

BibTeX
@article{Ajoudani2026,
  author = {Ajoudani, Mojtaba and Mosayyebi, Seyed Reza and Yansari, Ramazan Teimouri},
  title = {Hybrid Multi‐Agent Reinforcement Learning and Consensus Control for Dynamic V2G‐Enabled Voltage Regulation in Active Distribution Networks},
  journal = {IET Generation, Transmission & Distribution},
  year = {2026},
  volume = {20},
  number = {1},
  doi = {10.1049/gtd2.70413}
}
Kostenlos starten →