Distributionally Fair Peer-to-Peer Electricity Trading

TL;DR

This work tackles fairness in semi-decentralized P2P electricity trading by grouping peers according to energy poverty and minimizing the maximum distributional disparity across groups using the Wasserstein distance $\oldsymbol{\mathcal{W}}$. The proposed Distributionally Fair P2P model reformulates the trading problem to minimize $D_{max}$ while preserving near-optimal community profits under a sacrifice parameter $\epsilon$, and employs an alternating optimization algorithm to handle the resulting bilinear non-convexity. Case studies on the IEEE 33-bus system show that the method can reduce inter-group unfairness by up to 70.1% during peak hours and that fairness improves further with a non-profit community PV plant; congestion and tariff structure also influence outcomes. While promising, the approach may converge to local optima, motivating future work on MILP reformulations for global optimality and on alternative grouping criteria to broaden applicability.

Abstract

Peer-to-peer energy trading platforms enable direct electricity exchanges between peers who belong to the same energy community. In a semi-decentralized system, a community manager adheres to grid restrictions while optimizing social welfare. However, with no further supervision, some peers can be discriminated against from participating in the electricity trades. To solve this issue, this paper proposes an optimization-based mechanism to enable distributionally fair peer-to-peer electricity trading. For the implementation of our mechanism, peers are grouped by energy poverty level. The proposed model aims to redistribute the electricity trades to minimize the maximum Wasserstein distance among the transaction distributions linked to the groups while limiting the sacrifice level with a predefined parameter. We demonstrate the effectiveness of our proposal using the IEEE 33-bus distribution grid, simulating an energy community with 1600 peers. Results indicate that up to 70.1% of unfairness can be eliminated by using our proposed model, even achieving a full elimination when including a non-profit community photovoltaic plant.

Figures (6)

• Figure 1: Community's distribution system.
• Figure 2: Households' production and consumption profiles.
• Figure 3: Electricity tariffs on 15/10/2022 (left) and 08/07/2024 (right).
• Figure 4: Cummulative PV production, consumption, and surplus.
• Figure 5: Trade distributions in reference and fair ($\varepsilon=100\%$) scenarios (08/07/2024 - I).