This is the authors' abstract. We don't add key points for this paper.
The high penetration of distributed energy resources in decarbonized low-voltage networks facilitates peer-to-peer (P2P) energy trading for profit sharing. To preserve privacy and satisfy low-voltage network constraints, this paper proposes a novel multi-agent reinforcement learning approach with message-passing mechanisms for the P2P energy market. The message-passing mechanism is designed to transmit key features derived from surrogate models, which are parametric models that map the system state to physically feasible costs and continuous double auction market rewards. This strategy enables prosumers to maximize total reward while adhering to network constraints by relying only on local observations and message passing from surrogate models, thereby preserving the privacy of prosumers, network operators, and market operators. To accommodate unforeseen topology changes, the Moore-Penrose pseudoinverse updates low-voltage network surrogate parameters for more accurate operational state estimation. A counterfactual baseline is developed for prosumer agents to explore better bidding policies and sustain participation incentives in the P2P energy market. The numerical results demonstrate the superiority of the proposed approach in terms of episodic reward and privacy preservation compared to other competitors.
Answers are grounded in this paper and cite their sources.
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: