Efficient Management of Day-Ahead Energy Markets via Multi-Agent Reinforcement Learning - a Hybrid Model Case Study

Matan Levy, Sarah Keren, Itay Segev, Alexander Tuisov

Keywords: Reinforcement Learning, Energy Markets, Energy Systems, Multi-Agent Reinforcement Learning

TL;DR: This study examines the optimization of day-ahead hybrid electricity markets using multi-agent reinforcement learning.

Abstract: This study examines the optimization of day-ahead hybrid electricity markets. The shift from centralized systems to public-private models introduces many challenges, including the introduction of independent market players and renewable energy sources (RESs). A formal model of market participants’ behavior is developed, and a multi-agent reinforcement learning (MARL) framework is proposed to optimize system operator strategies, incorporating dynamic pricing and dispatch scheduling to reduce operational costs, ensure stability, and align market incentives. A new and adaptable simulation environment, compatible with state-of-the-art methods, is presented. Evaluations in increasingly complex settings demonstrate the efficacy of our framework in managing the complexities of modern electricity markets.

Submission Number: 5