Multi-microgrid collaborative adaptive optimization scheduling method based on attention-based deep reinforcement learning

Yun Shen

doi:10.1016/j.ijepes.2025.111494

Return

Multi-microgrid collaborative adaptive optimization scheduling method based on attention-based deep reinforcement learning

2026-01-01

DOI：10.1016/j.ijepes.2025.111494

Original

Save

View PDF

Abstract

En 中文

Microgrids, as a component of flexible resource optimization and regulation, provide important support for enhancing the safe and economic operation of distribution networks. Existing collaborative optimization studies predominantly employ static models with fixed microgrid numbers, without considering the impact of changes in microgrid grid-connected and off-grid operation states on collaborative performance. Therefore, this paper proposes an attention-enhanced reinforcement learning strategy for dynamic optimization of distribution network under microgrid operational state switching. Firstly, the operational domain quantifying power interaction between the distribution network and microgrids is evaluated to achieve efficient control and collaborative optimization by characterizing microgrids' external power characteristics while visualizing the dispatchable region of tie-line power for optimal scheduling. In the day-ahead regulation stage, a day-ahead scheduling model considering voltage violation risks is established to reduce the operational risks associated with the transition from microgrid grid-connected to off-grid. For intra-day regulation, an improved value network incorporating an attention mechanism is adopted, and autonomous coordination capabilities are obtained through dynamic allocation of attention weights by agents, enabling stable regulation under random disturbances in microgrid grid-connected and off-grid operations. Finally, the feasibility of the proposed method is verified through numerical simulation.

Keywords:

Microgrid

Distribution network

Off-grid operation

Multi-agent reinforcement learning

Attention mechanism

Optimal scheduling

AI Summary

Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.