ASSUME: An agent-based simulation framework for exploring electricity market dynamics with reinforcement learning

Abstract

Electricity markets are undergoing transformative changes driven by integrating renewable energy and emerging technologies, and evolving market conditions such as shifting demand patterns, regulatory reforms, and increased price volatility. To address the complexity of electricity markets and their interactions, we present ASSUME, an open-source agent-based simulation framework that incorporates multi-agent deep reinforcement learning for modeling adaptive market participants. ASSUME offers a modular architecture for representing generator and demand-side agents, bidding strategies, and diverse market configurations. ASSUME has been proven effective in multiple research studies, demonstrating its ability to analyze complex bids, demand-side flexibility, and other market scenarios. By incorporating adaptive strategies through deep reinforcement learning, ASSUME supports dynamic strategy exploration, enabling a deeper understanding of electricity market behaviors. With its flexible architecture, documentation, tutorials, and broad accessibility, ASSUME ensures usability across different user groups, minimizing technical overhead and freeing up human resources for deeper insights into operational, economic, and policy-related challenges in this critical sector.