PhD awarded with distinction
Defended my thesis "Scalable Reinforcement Learning for Optimal Electric Vehicle Charging" at TU Delft and was awarded the PhD cum laude. Many thanks to my promotors, Dr. Pedro P. Vergara and Prof. Peter Palensky.
Stavros Orfanoudakis
Postdoctoral Researcher · Intelligent Electrical Power Grids, TU Delft
I am a Postdoctoral Researcher in the Intelligent Electrical Power Grids group at TU Delft, working on reinforcement learning, graph neural networks, large language models, and physics-informed machine learning for power and energy systems, with a focus on scalable and trustworthy AI for real-world operation. Previously, I was a Research Scholar at ETH Zürich's Power Systems Laboratory and a Visiting Researcher at MIT's Center for Transportation & Logistics.
Research interests
- Reinforcement Learning
- Physics-Informed ML
- LLMs & GNNs for Power Systems
- Energy Transition & Smart Grids
- Multi-Agent Systems
PhD · Electrical Engineering, Mathematics & Computer Science
MSc · Electrical & Computer Engineering
BSc & MEng · Electrical & Computer Engineering
“I envision intelligent AI systems as a key force in the energy transition: autonomous, resilient, and trustworthy in real-world operation.”
News
Recent Milestones
Best Student Paper Award at SETN 2026
Our paper "SAVGO: Learning State–Action Value Geometry with Cosine Similarity for Continuous Control", with Pedro P. Vergara, received the award at the 14th Hellenic Conference on AI in Chania, Crete.
MAR 2026
Awarded Best PowerWeb Paper Award 2025 and nominated for Best Energy Paper for our Graph-RL EV coordination work.
FEB 2026
Started as Research Scholar at the Power Systems Laboratory, ETH Zürich, working on LLMs and physics-informed ML for power systems.
DEC 2025
"Scalable RL for Large-Scale EV Coordination Using GNNs" named Editor's Choice 2025 at Nature Communications Engineering.
NOV 2025
Visiting Researcher at MIT Center for Transportation & Logistics.
NOV 2025
Granted the NCCR Automation Fellowship for "Physics-Informed LLMs for Sequential Decision-Making in Power Systems".
SEP 2025
Invited talk at the Linux Foundation Energy Summit Europe, Aachen.
FEB 2025
Featured in a TU Delft user story: "Simulating a Thousand Charging Stations on DelftBlue".
APR 2023
Started my PhD at TU Delft, Intelligent Electrical Power Grids group.
OCT 2021
Our autonomous broker TUC-TAC finished 2nd in the Power Trading Agent Competition (PowerTAC) 2021: TU Crete news · official results.
NOV 2020
TUC-TAC won 1st place in the international PowerTAC 2020 automated electricity-trading competition: official results · our EUMAS paper.
Selected publications
Research Highlights
"Scalable Reinforcement Learning for Large-Scale Coordination of Electric Vehicles Using Graph Neural Networks"
"A Graph Neural Network-Enhanced Decision Transformer for Efficient Optimization in Dynamic Smart Charging Environments"
"Physics-Informed Reinforcement Learning for Large-Scale EV Smart Charging Considering Distribution Network Voltage Constraints"
"EV2Gym: A Flexible V2G Simulator for EV Smart Charging Research and Benchmarking"
Archive
Publications
Journal articles, peer-reviewed conference papers, and preprints under review.
Under review
Preprints · 3
Journal publications
Peer-reviewed · 10
2026
"Physics-Informed Reinforcement Learning for Large-Scale EV Smart Charging Considering Distribution Network Voltage Constraints"
"A Graph Neural Network-Enhanced Decision Transformer for Efficient Optimization in Dynamic Smart Charging Environments"
"Physics-Informed Neural Network with Adaptive Activation for Power Flow"
"Open-Source Algorithms for Maximizing V2G Flexibility Based on Model Predictive Control"
2025
"High-Temporal-Resolution Dataset of Uni-, Bidirectional, and Dynamic Electric Vehicle Charging Profiles"
"Scalable Reinforcement Learning for Large-Scale Coordination of Electric Vehicles Using Graph Neural Networks"
2024
"EV2Gym: A Flexible V2G Simulator for EV Smart Charging Research and Benchmarking"
"A Comprehensive Analysis of Agent Factorization and Learning Algorithms in Multiagent Systems"
"PowerFlowNet: Power Flow Approximation Using Message Passing Graph Neural Networks"
2022
"Novel Meta-Learning Techniques for the Multiclass Image Classification Problem"
Conference publications
Peer-reviewed · 12
2026
"Flow Matching Policy Optimization with Mirror Descent and Entropy Constraints"
"SAVGO: Learning State–Action Value Geometry with Cosine Similarity for Continuous Control"
"Community-to-Vehicle: Integrating Electric Vehicles into Energy Communities -- A Swiss Case Study"
"Can A.I. Revolutionize EV Dispatch?"
"Safe Reinforcement Learning for V2G-Enabled Electric Vehicle Aggregators"
2024
"A Dynamic Prediction Tool for Vehicle-to-Grid Operation and Planning"
"Energy Storage Systems Planning in the Electric Distribution System Considering the Grow of PV Penetration"
"Reinforcement Learning for Optimized EV Charging Through Power Setpoint Tracking"
2023
"A Novel Aggregation Framework for the Efficient Integration of Distributed Energy Resources in the Smart Grid"
2022
"The Performance Impact of Combining Agent Factorization with Different Learning Algorithms for Multiagent Coordination"
"Intelligent Robotic System for Urban Waste Recycling"
2021
"Aiming for Half Gets You to the Top: Winning PowerTAC 2020"
Open source & collaborations
Projects
A curated portfolio of open-source research software and collaborative initiatives in intelligent energy systems, reinforcement learning, and EV charging optimization.
434
97
EV2Gym
Research-grade Python environment for large-scale EV smart charging and V2G simulation. Provides an OpenAI Gym interface for reproducible reinforcement-learning experiments in Vehicle-to-Grid settings.
EV-GNN
Python package for graph-based reinforcement learning in EV charging coordination. Built to support scalable experiments and reproduce the Nature Communications Engineering study.
DT4EVs
Python package for offline EV charging optimization using decision transformers, designed for data-driven scheduling and benchmark evaluation in dynamic charging environments.
PowerFlowNet
Python package for graph-neural-network-based power-flow approximation, enabling fast surrogate modeling for grid analysis and optimization workflows.
DRIVE2X
EU Horizon consortium advancing vehicle electrification through bidirectional smart charging. My TU Delft PhD work contributed large-scale coordination methods for real-world deployment scenarios.
Supervision & teaching
Teaching
Lectures, tutorials, and co-supervision of MSc theses across ETH Zürich, MIT, TU Delft, and the Technical University of Crete.
Lectures & tutorials
Courses Taught
Introduction to Deep Reinforcement Learning
Guest Lecturer · AI Minor Program
Invited lecture covering policy/value methods, exploration strategies, and practical implementation guidance.
Tutorial on Reinforcement Learning
Guest Lecturer · InnoCyPES Colloquium
Hands-on tutorial session focused on modeling sequential decision-making problems in energy systems.
Artificial Intelligence
Teaching Assistant · Tutorials & Project Supervision
Supported tutorials, graded coursework, and supervised student projects in core AI topics.
Multiagent Systems
Teaching Assistant · Tutorials & Project Supervision
Delivered tutorial support and project mentoring on coordination, planning, and learning in multiagent settings.
MSc thesis supervision
Students
Short-Term Power Imbalance Forecasting: Accuracy vs. Interpretability
Global Vessel Routing Using Reinforcement Learning and Graph Neural Networks
Bibi van den Berg
Quantifying the Economic V2G Benefits: Cost-Optimal V2G Integration in Residential Settings
Ruben Eland
Improving EV Aggregators' Workplace Charging: A Safe Reinforcement Learning Approach
Yunus Emre Yılmaz
EV Charging Strategies through Power Setpoint Tracking: A Reinforcement Learning Approach
Antonios Mastorakis
Development of a Competitive Autonomous Agent for Smart-Grid Energy Markets
Student project supervision
Projects
Model-Based Reinforcement Learning for Energy Management
Predict-Then-Optimize: Data-Driven Electric Vehicle Charging Control for Day-Ahead Operations