IEK-10 focuses on the development of models and algorithms for simulation and optimization of decentralized, integrated energy systems. Such systems are generally described by high-fidelity simulation models, thus eluding optimization-based design and control. We aim to bridge this gap by using machine learning based reduced-order models (ROMs). The resulting models will facilitate energy-optimal, cost-efficient and safe design and operation of future energy systems.
Background
In the field of Scientific Machine Learning modern machine learning approaches are deployed to speed up simulation and optimization with complex models while maintaining their high accuracy. To this end, data-driven ROMs replace or complement physics-based models.
Scientific Machine Learning most commonly resorts to artificial neural network (ANN) based models, which are built and trained by so called automatic differentiation (AD) frameworks. These AD frameworks leverage the fact that all code and components describing the model are differentiable and thus accessible by the training algorithms of machine learning. The AD frameworks can be further extended to train ANNs in a special setting or with specific desired properties. The goal of this thesis is to develop a method for training computationally efficient surrogate models in the field of energy systems.
Your profile
- Student (w/m/d) in the field of Computer Science, CES, Engineering, Applied Mathematics and related fields
- Strong mathematical background
- You are able to code in Python or Julia, ideally you also know C++
- You have prior theoretical and/or practical knowledge in machine learning
- Highly motivated and capable of working independently
We offer
- A highly motivated team in one of Europe’s biggest research facilities
- An excellent scientific and technical infrastructure
- Working environment with clear focus on optimization and machine learning
- Supervision by experts in relevant fields
Contact:
Danimir Doncevic, M.Sc.
Forschungszentrum Jülich
Institute of Energy and Climate Research – Energy Systems Engineering (IEK-10)
Email: d.doncevic@fz-juelich.de