Delft AI Energy Lab
AI for sustainable, reliable and effective energy systems
Energy systems are the backbone of our modern society. It is of great importance that these systems are sustainable, reliable and effective now and in the future. There is strong expertise in this field on the TU Delft campus. The Delft AI Energy Lab investigates how new AI-based methods can contribute to the management of dynamic energy systems.
Therefore we combine groundbreaking machine learning with the reliable theory of the physical energy system. For example, it is possible with the AI technique 'neural networks' to model differential equations describing dynamics in areas such as fluid dynamics, and for predicting extreme, rare events. Delft AI Energy Lab investigates these promising methods for applicability for monitoring the 'health' of parts of energy systems, and for the early detection of threats.
The Delft AI Energy Lab is part of the TU Delft AI Labs programme.
Advanced AI-based mathematical models together with scalable algorithms offer reliable diagnosis and predictive tools for modern energy systems.
Making optimization algorithms computationally efficient matters in many applications including power systems
Describing experiment methodology.
The team
Directors
Jochen Cremer
Faculty of EEMCS
Lab director
Peyman Mohajerin Esfahani
Faculty of ME
Lab director
PhD candidates
Olayiwola Arowolo
PhD candidate
Mohammad Boveiri
PhD candidate
Demetris Chrysostomou
PhD candidate
Nikolina Čović
Visiting PhD researcher
Mert Karaçelebi
PhD candidate
Shabnam Khodakaramzadeh
PhD candidate
Basel Morsy
PhD candidate
Lucas Narbondo
PhD candidate
Ali Rajaei
PhD candidate
Haiwei Xie
PhD candidate
Viktor Zobernig
PhD candidate
Education
Courses
- EE4C12 Machine Learning for Electrical Engineering, 5 ECTS, EEMCS, Electrical Engineering MSc Program"?
- MOOC Digitalisation of Intelligent and Integrated Energy Systems
- SC42150 Statistical Signal Processing, 3 ECTS, ME, Systems & Control MSc program
- SC42110 Dynamic Programming & Stochastic Control, 5 ECTS, ME, Systems & Control MSc program
Resources
Master projects
Openings
- AC Optimal Data Generation (ACODG) for Power System Security
- Graph Neural Networks for Security-Constrained Optimal Power Flow
- Estimating the Flexibility Evolution in Active Distribution Grids
- Neural Ordinary Differential Equations for Power System Dynamics
- Optimal PMU Placement for Flexibility and State Estimation
- Non-intrusive Load Monitoring of the Electricity Consumption
- Coordinated Control of Virtual Power Plants for Frequency Stability
- Offering Strategies of Virtual Power Plants in Ancillary Service Markets Based on Stochastic Programming
See also DAI Energy lab MSc Cataloge.pdf
Ongoing
Applied AI projects
- Data-Driven Adaptive Dynamic Equivalents of Active Distribution and Transmission Networks, Alex Neagu, TU Delft
- End-to-end learning for N-k SC-OPF, Bastien Giraud, TU Delft, NTNU, Norway
- Neural Ordinary Differential Equations for Frequency Dynamics, Nila Krishnakumar, TU Delft
- Reinforcement learning for transmission network topology control, Geert Jan Meppelink, TU Delft, NTNU, Norway
- Offering Strategies of Virtual Power Plants in Ancillary Service Markets Based on Stochastic Programming, Torben Zeller, RWTH Aachen University, Germany
- Market Mechanism Design for Virtual Inertia, Johnny Zheng
- Energy Community Management with Reinforcement Learning, Catarina Santos Neves, Instituto Superior Técnico, Portugal
Fundamental AI projects
- Non-intrusive load monitoring for residential houses
- Scalable dictionary learning to learn high-level features
- Sample-efficient learning for automated theorem proving in higher-order logic, Honours Programme Bachelor Project Proposal 2022-2024
- Learning Drivers’ Preferences in Delivery Route Planning through Inverse Optimization
- Fault Detection and Isolation of Nonlinear Systems with Optimized Model Mismatch
- Robustness in Fault Diagnosis applied in the Lateral Control of Automated Vehicles
Associated projects
- MSc Project proposal: Correlations for the hydrodynamics of spouted bubbling fluidized beds uzing CFD, ANN's & ecperiments
Finished
- Multivariable Anomaly Detection Framework for Multi-sensor Network, TU Delft, 2022
- Transferring Domain Knowledge to Data-driven Controller, TU Delft, 2022
- Multi-Modal End-to-End Learning for Real-Time Monitoring of Sustainable Energy Systems, TU Delft, 2022
- Deep Statistical Solver for Distribution System State Estimation, Benjamin Habib, TU Delf, Stedin, 2022
- Meter placement for state estimation in distribution networks, Sattama Datta, TU Delft, Alliander, 2022
- End-to-End Learning for Sustainable Energy Scheduling, Dariush Wahdany, TU Delft, RWTH Aachen University, 2021
- Multi-Agent Reinforcement Learning for Incentive-based Residential Demand Response, Jasper van Tilburg, TU Delft, 2021
- On the Road from Active Inference to Regret Minimization, TU Delft, 2021
- Conjugate Dynamic Programming, TU Delft, 2021
- Tractable Algorithms for Large Scale Mixed Integer Quadratic Programming: A Principal Component Analysis Approach, TU Delft, 2021
- Learning Parametric Mixed Integer Quadratic Programming via Inverse Optimization, TU Delft, 2021
- On Complexity of Data-driven Controls in Stochastic Environments, TU Delft, 2021
Partners
