Reproduce “Reinforcement Learning and Rule-Based Peer-to-Peer Pricing in Residential PV-BES Communities” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Pablo Benalcazar, Maciej Kalka, Wilian Guamán, Jacek Kamiński
An annual simulation of a 20-household residential PV community compares three transparent peer-to-peer pricing rules with Deep Q-Network pricing under photovoltaic-only and battery-storage settings. Rule-based, budget-balanced mechanisms deliver €829.98 in PV-only community savings, while the learnable supply–demand-ratio policy reaches €734.23 without storage and €978.52 with storage, outperforming the tested multiplier design but not the rule-based benchmarks in the PV-only case.
Key insights
- Budget-balanced rule-based pricing preserves the value of all physically matchable trades, whereas separate RL buying and selling quotes create an internal spread that reduces community savings.
- The learnable SDR formulation is better suited to adaptive pricing than the tested multiplier action space: it lowers the annual spread from €210.46 to €92.49 relative to the fixed SDR control while maintaining nearly the same traded energy.
- Storage extends local energy availability and improves RL outcomes, but fixed battery dispatch and unequal household benefits mean higher aggregate savings do not necessarily imply fairer distribution.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.