In my role as a Senior Scientist for Machine Learning at Fraunhofer IEE and as the head of the BMBF-funded junior research group RL4CES at the University of Kassel, I focus on leveraging machine learning to optimize processes in the energy sector. My academic background — a degree in computer science with a focus on theoretical computer science and a PhD in machine learning — forms the foundation for my specialization in advanced learning methods such as self-supervised learning, few-shot learning, and reinforcement learning.

As the leader of the RL4CES research group, I explore the application of deep reinforcement learning (DRL) to optimize complex decision-making processes in the energy industry. Our goal is to develop intelligent algorithms that autonomously adapt to new situations, making grid control and energy trading more efficient, secure, and cost-effective.

Traditional optimization methods are reaching their limits due to the growing complexity of the energy system, as they are often not applicable in real-time. Existing rule-based systems have limited responsiveness to unexpected changes. In contrast, DRL enables the creation of dynamic and adaptive solutions that enhance grid stability and optimize intraday energy trading.

Through close collaboration with companies in the energy sector, we ensure the practical implementation of our research results. We place particular emphasis on explainable AI methods to build trust in automated decision-making processes and promote their widespread industrial adoption. With RL4CES, we actively contribute to advancing the energy transition through innovative AI technologies.

Due to the increasing complexity of the energy system, deep learning models are increasingly being used to control and monitor it. But what happens when crises, severe weather events or political decisions have a lasting impact on the dynamics of the energy system, e.g. through sharply rising or falling prices? Conventional deep learning models then must be retrained based on the new data. What if you could apply what you have already learnt to the new situation instead? This transfer from one problem to another similar one, is the basic idea behind the research areas of transfer and meta-learning, which I am investigating in detail in my doctorate using the use case of energy trading. In addition to this strongly application-driven perspective, I would like to use my mathematical expertise to create a theoretical basis for future research in this area.

I studied mathematics with a focus on stochastics, statistics and optimization at the Technical University of Dresden and later focused on machine learning, especially during my semesters abroad in Dublin (Ireland) and Gothenburg (Sweden). Since my master thesis (2021) I am now here at the Fraunhofer Institute IEE in Kassel. Since then, I have been developing deep reinforcement learning agents for short-term energy trading.

Areas of Interest:Deep Reinforcement Learning, Transfer- und Meta-Learning, Energiehandel, Mathematische Grundlagen im Reinforcement Learning

I am a researcher at Fraunhofer IEE and joined the RL4CES group in the Department of Intelligent Embedded Systems at the University of Kassel in 2022. My research goal is to advance the autonomy of energy systems, in particular by integrating neural networks in the areas of grid control and electricity trading. Therefore, I am currently writing my PhD thesis on these topics, with a special focus on the application of Deep Reinforcement Learning (DRL). In terms of the underlying methods, I mainly cover the application of DRL, but also apply supervised and deep learning models to ensure sufficient training.

Areas of Interest:Deep Reinforcement Learning, Grid Control, Electricity Trading, Machine Learning, Graph Neural Networks, Time Series

I have been working at Fraunhofer IEE since 2022 and am doing my doctorate on Graph Neural Networks at the University of Kassel. I completed my Bachelor's and Master's degrees in Applied Computer Science at the University of Göttingen, where I specialised in machine learning. In my PhD, I am working on the development of graph neural networks that help to operate power grids more efficiently and improve the integration of renewable energies. Modelling power grids as graphs offers the possibility of mapping relationships and interactions between grid components. In contrast to conventional neural networks, graph neural networks can use this structure directly to learn complex relationships. I therefore consider them to be a particularly promising method for optimising power grids.

I am a PhD candidate and research associate at Fraunhofer IEE and the University of Kassel, and I have been part of the RL4CES team since 2023. I completed my Bachelor's and Master's degrees in Applied Computer Science at the University of Göttingen, focusing on explainable and interpretable machine learning models.

My research revolves around the application of Deep Reinforcement Learning (DRL) for controlling transmission networks. A key aspect of this work is managing large action spaces when optimizing power grids. To address this, I develop methods for reducing action sets using evolutionary algorithms and for representing action spaces. Additionally, I explore the integration of Graph Neural Networks (GNNs) into DRL models to incorporate the grid's topological structure directly into the decision-making process, enabling more robust and efficient control strategies.

Another focus of my work is enhancing the transparency and explainability of DRL models for grid operations. Since conventional DRL agents often act as black-box models, I design methods to make their decision-making processes more comprehensible. This aims to build trust among grid operators and decision-makers in AI-based control systems and improve their practical applicability.

Areas of Interest: Deep Reinforcement Learning, Graph Neural Networks, Explainable AI, Grid Control, Renewable Energy

I am currently a PhD student at the University of Kassel and Fraunhofer IEE, working on the topic of Trustworthy Artificial Intelligence as part of the RL4CES team. I have a Master's degree in Mathematics from the University of Augsburg, where I specialized in Statistics, Probability Theory and Optimization. Since 2020 I am working as a research associate at Fraunhofer IEE on the application of machine learning in power systems. My research focuses on improving the adversarial robustness and interpretability of Deep Reinforcement Learning (DRL) algorithms for grid operations.

Traditional DRL agents for grid operations are based on highly complex black-box models whose actions and processes are opaque and difficult for humans to understand. This makes it challenging to monitor the systems and opens up new opportunities for cyber-attacks. In the worst case, an attacker can manipulate a DRL agent so that its actions cause a blackout. To address this problem, I am developing interpretable DRL agents for grid operations and investigating their robustness against attacks. I am also studying different defense strategies to protect DRL agents from adversarial attacks.

Areas of Interest: Deep Learning, Reinforcement Learning, Explainable AI, Adversarial Machine Learning, Renewable Energy

I am a PhD candidate at the University of Kassel and have been part of the RL4CES team since November 2024. I completed my Bachelor's and Master's degrees in Mathematics at the University of Bayreuth, with a focus on algebra and statistical learning theory. This laid the foundation for a strong theoretical understanding of machine learning. In my research, I explore the application of Deep Reinforcement Learning (DRL) for planning in networks, particularly power grids. The key question is: where might power supply be at risk, and how can we proactively prevent it? I investigate models characterized by strategic and forward-looking reasoning and apply them to energy systems. The goal of my research is to develop DRL algorithms that can realistically assess problems in grid operations and respond with targeted and understandable actions. To make the predictions of these models more interpretable and user-friendly, I explain the DRL model’s outputs using simpler, more transparent models — a technique known as model distillation.

Interessengebiete: Deep Reinforcement Learning, Graph Neural Networks, Explainable AI, Grid Control, Mathematical Foundations of Reinforcement Learning, Model Distillation