ASSUME – Deep reinforcement learning for electricity market modeling

The transformation of electricity markets associated with the transition towards high shares of renewable power generation results in the constant development of market mechanisms. However, changing the current market design does affect all other markets and their participants in not necessarily foreseeable ways, as the last changes in the German reserve markets demonstrated. This raises the need for tools and simulation models to investigate and understand such complex market interplay and predict possible adverse effects and misuse of market power. The ASSUME project aims to develop an open-source, agent-based simulation toolbox for electricity markets using deep reinforcement learning algorithms.

Objectives

During the workshop, participants will explore two case studies that showcase the versatility of the ASSUME framework:

  1. Exploring Market Dynamics with DRL:
      • Understand the power of DRL within a market simulation context.
      • Learn how to analyze bidding strategies developed by DRL agents using xAI methods.
      • Gain insights into the potential outcomes of proposed market zone splits.
  2. Modeling Demand Side Management (DSM) units with ASSUME:
      • Dive into modeling steel plants as DSM units within ASSUME and calculate their flexibility.
      • Explore how to offer this flexibility on a redispatch market for optimal resource allocation.

Agenda

  1. Introduction to the ASSUME Framework (15 minutes)
      • Overview of the project’s objectives and the simulation toolbox.
      • Highlight key features and showcase existing use cases.
      • Interactive Q&A session.
  2. Tutorial I: DRL for Market Analysis (75 minutes)
      • Introduction to DRL and explainable AI (xAI) methodologies.
      • Hands-on experience using ASSUME for a practical case study on potential outcomes of proposed market zone splits.
      • Analyze outcomes using advanced xAI techniques.
  3. Tutorial II: DSM Units in the Redispatch Market (75 minutes)
      • DSM agent implementation and flexibility calculation within ASSUME.
        o Run simulations with DSM units, comparing cost-optimal operation in the day-ahead market to flexible operation in the redispatch market.
      • Analyze results and engage in discussions.
  4. Wrap-up and Q&A Session (15 minutes)
      • Recap key workshop insights and takeaways.
      • Open floor for participants to ask questions, share feedback, and further explore workshop topics.

Requirements

  1. Hardware: A laptop or desktop computer with internet access.
  2. Software: No local software installation is required for most participants.
  3. Google Account: Participants should have a Google account to access and run tutorials in Google Colab.
  4. Alternative Setup (Optional): For those without a Google account or preferring a local environment:

Moderators

Nick Harder

Nick Harder

University of Freiburg

Nick Harder is a research associate and doctoral student at INATECH, University of Freiburg. He focuses on using deep reinforcement learning to model electricity markets and understand participant behavior. Nick coordinates the ASSUME project, developing tools to analyze electricity markets using advanced machine learning. His expertise combines sustainable systems engineering with AI applications in energy markets.

Kim K. Miskiw

Kim K. Miskiw

KIT

Kim K. Miskiw is a research associate and doctoral student at KIT’s Chair for Information and Market Engineering. Her research focuses on deep reinforcement learning in electricity markets, agent-based modeling, and stochastic optimization. She previously worked at KIT’s Institute for Industrial Production. Kim holds a Master’s in Industrial Engineering from KIT, with a thesis on optimized bidding strategies in electricity markets.

Manish Khanra

Manish Khanra

Fraunhofer ISI

Manish Khanra is a research associate at Fraunhofer ISI, specializing in renewable energy integration and decarbonizing Hard-to-Abate sectors. He holds an M.Sc. in Renewable Energy and Energy Efficiency. Manish has researched district heating optimization and works on agent-based power market simulations, sector coupling, and analyzing economic and technical aspects of renewable energy integration.