Proximal Policy Optimization-Based Reinforcement Learning Control of Single-Stage Multiport Inverter in Islanded Microgrids
IEEE Transactions on Industrial Electronics · 2026-04 · 73 (4) · 5663-5674
DOI: 10.1109/tie.2025.3634402 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Industrial Electronics, 2026-04. Verfasst von Yujia Kuang und 4 weiteren Autor:innen. Die Arbeit wurde bisher 4-mal zitiert (Crossref) und nennt 40 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Basic and Applied Basic Research Foundation of Guangdong Province, Science and Technology Planning Project of Guangdong Province.
Abstract (Originalsprache)
The single-stage multiport inverter (SSMPI) is a high-efficiency and cost-effective solution for islanded microgrids with hybrid energy source integration owing to the elimination of intermediate dc–dc converters. However, achieving simultaneous ac voltage control and dc port power distribution remains challenging due to inherent nonlinearities in power distribution and coupling between control objectives. Given this, a proximal policy optimization (PPO)-based reinforcement learning (RL) control scheme is proposed to achieve both ac voltage control and power distribution in SSMPI. In the proposed method, nonlinear features and coupling issues of the SSMPI are formulated as a Markov decision process (MDP) comprising the state, the action, and a reward function that encodes the constraints and objective priorities. A deep actor-critic network is designed to efficiently capture the nonlinear mapping between the environment and control variables. Based on this formulation, offline training enables the agent to learn the optimal policy, while sensitivity analysis of reward weights ensures stable and balanced control performance. Experiment results demonstrate that the proposed scheme effectively controls ac voltage and dc power despite nonlinearities and coupling, reducing power ripples by 52.22% and 88.97%, and ac voltage THD by 27.34% and 83.44% on average over previous methods.
Förderung
- National Natural Science Foundation of China
- Basic and Applied Basic Research Foundation of Guangdong Province
- Science and Technology Planning Project of Guangdong Province
- Shenzhen Science and Technology Program
Zitieren
APA
Kuang, Y., Zhou, D., Shen, Z., Zou, J., & Liu, X. (2026). Proximal Policy Optimization-Based Reinforcement Learning Control of Single-Stage Multiport Inverter in Islanded Microgrids. IEEE Transactions on Industrial Electronics, 73(4), 5663-5674. https://doi.org/10.1109/tie.2025.3634402
Vancouver
Kuang Y, Zhou D, Shen Z, Zou J, Liu X. Proximal Policy Optimization-Based Reinforcement Learning Control of Single-Stage Multiport Inverter in Islanded Microgrids. IEEE Transactions on Industrial Electronics. 2026;73(4):5663-5674. doi:10.1109/tie.2025.3634402
BibTeX
@article{Kuang2026,
author = {Kuang, Yujia and Zhou, Dehong and Shen, Zewei and Zou, Jianxiao and Liu, Xin},
title = {Proximal Policy Optimization-Based Reinforcement Learning Control of Single-Stage Multiport Inverter in Islanded Microgrids},
journal = {IEEE Transactions on Industrial Electronics},
year = {2026},
volume = {73},
number = {4},
pages = {5663--5674},
doi = {10.1109/tie.2025.3634402}
}
EUVIMED