Scalable fairness shaping with LLM-guided multi-agent reinforcement learning for peer-to-peer electricity markets

Peer-to-peer (P2P) energy trading is becoming central to modern distribution systems as rooftop PV and home energy management systems become pervasive, yet most existing market and reinforcement learning(RL) designs emphasize efficiency or private profit and offer little real-time guidance to ensure equitable outcomes under uncertainty. To address this gap, we propose FairMarket-RL, a fairness-aware multiagent RL framework in which a large language model (LLM) critic reads a compact summary of each cleared auction and returns three normalized slot-level fairness scores—Fairness-to-Grid (FTG), Fairness-Between-Sellers (FBS), and Fairness-of-Pricing (FPP) that shape bidding policies within a continuous double auction under partial observability and discrete price–quantity actions; these scores are integrated into the reward via ramped coefficients and tunable scaling so fairness guidance complements, rather than overwhelms, economic incentives. The environment models realistic residential load and PV profiles and enforces hard constraints on prices, physical feasibility, and policy-update stability. By scalable fairness shaping we mean that the same LLM-guided reward design and policy class can be trained on a small pilot community and then transferred, without architectural changes, to larger communities and longer horizons while preserving both fairness and economic performance. Across a progression of experiments from a small pilot to a larger simulated community and a mixed-asset real-world dataset, the framework shifts exchanges toward local P2P trades, lowers consumer costs relative to grid-only procurement, sustains strong fairness across participants, and preserves utility viability. Sensitivity analyses over solar availability and aggregate demand further indicate robust performance, suggesting a scalable, LLM-guided pathway to decentralized electricity markets that are economically efficient, socially equitable, and technically sound. • LLM-guided MARL framework enhances fairness in P2P electricity markets. • LLM critic provides three slot-level fairness scores to shape agent bidding. • Realistic load, PV profiles, and physical constraints modeled in the auction. • Scalable design transfers from small pilots to larger communities unchanged. • Experiments show more P2P trades, lower costs, and robust fairness outcomes.