Answers are grounded in this paper and cite their sources.
In this paper, we develop a deep reinforcement learning (DRL) framework for transmission interface power flow adjustment, which aims to regulate the aggregate power flow over critical groups of transmission lines through generation redispatch. The proposed framework is built upon a Factorizable State Attribution (FSA) mechanism, which decomposes the power system state graph into four electrical factor graphs corresponding to active power, reactive power, voltage magnitude, and phase angle. By incorporating transmission interface information, FSA learns interface-aware attribution weights and constructs task-adaptive state representations for different interface adjustment tasks. The attributed representation is then used by a Dueling DQN policy to select corrective redispatch actions. Experiments on IEEE 118-bus, Chinese 300-bus, and European 9241-bus systems show that FSA improves adjustment success rates over representative DRL baselines and achieves much faster inference than OPF, while providing interpretability through the learned factor attribution map.
No takes yet. Share an insight, caveat, or question.
Zhang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: