Reinforcement Learning and Rule-Based Peer-to-Peer Pricing in Residential PV-BES Communities

Research compares rule-based and reinforcement learning-based pricing mechanisms for peer-to-peer (P2P) electricity trading within residential communities utilizing photovoltaic (PV) systems. Initial findings indicate that traditional rule-based methods, such as bill-sharing and supply-demand-ratio pricing, surpass deep Q-network (DQN) based reinforcement learning approaches in communities relying solely on PV. However, the introduction of battery energy storage (BES) significantly alters performance dynamics.

Why it matters

This research provides critical insights into the efficacy of different pricing models for localized energy trading, which is vital for the strategic development and adoption of distributed energy resources. Understanding the performance of these mechanisms, particularly in the context of emerging technologies like battery storage, will influence investment decisions and operational strategies for future energy grids and community microgrids.

Key insights

• Rule-based pricing mechanisms (bill-sharing, mid-market rate, supply-demand-ratio) serve as benchmarks for comparison.
• Reinforcement learning (RL) is implemented via a Deep Q-Network (DQN) for pricing strategies.
• Performance evaluation focuses on community savings, financial indicators, and operational metrics.
• In PV-only community configurations, rule-based benchmarks demonstrate superior performance over the best RL policies.
• The presence of battery energy storage is noted to impact the comparative performance of these pricing mechanisms.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2609.01680