A hybrid multi-agent deep actor-critic learning and particle swarm optimization algorithm for active voltage control in smart grids with renewable energies
The increasing penetration of renewable and distributed energy resources presents significant challenges for voltage regulation in modern power distribution systems. Traditional optimization-based controllers often fail to deliver reliable performance under real-time constraints and high system unce...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-09-01
|
| Series: | e-Prime: Advances in Electrical Engineering, Electronics and Energy |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2772671125001822 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The increasing penetration of renewable and distributed energy resources presents significant challenges for voltage regulation in modern power distribution systems. Traditional optimization-based controllers often fail to deliver reliable performance under real-time constraints and high system uncertainty. To address these issues, recent research has increasingly focused on Multi-Agent Reinforcement Learning (MARL) algorithms to coordinate control units across distributed grid regions. While MARL offers strong potential, its performance can be limited by over-generalization and insufficient adaptability to diverse operating conditions. In this study, we propose a novel hybrid framework that integrates MARL with Particle Swarm Optimization (PSO) to address these challenges. Our method combines MARL’s ability to learn from experience with PSO’s efficient search capabilities to enhance voltage control across distributed networks. The proposed algorithm is evaluated on the MAPDN platform under 33-bus and 322-bus scenarios using four MARL variants. Experimental results demonstrate that the hybrid method achieves up to 10x reduction in power loss and consistently maintains a perfect control rate of 1.0, significantly outperforming standalone MARL approaches. The framework is scalable, adaptable to various MARL models and metaheuristic algorithms, and offers promising implications for data-driven voltage regulation in renewable-heavy smart grids. |
|---|---|
| ISSN: | 2772-6711 |