Project 5: Intelligent control of storages for stabilizing electric grids

Description of the project

The global energy industry is in heavy transition. Classic power plants, which provide electricity in vast amounts and on schedule, are replaced by  a vast amount of small to mid-scale producers such as wind farms and solar power plants. This transformation is aimed to support the fight against global warming but also poses a huge challenge for grid operators. How to keep the balance between supply and demand if we don’t know the supply tomorrow? Storages are therefore a key tool for stabilizing the grid. The  control of these facilities itself is a (mathematical) problem.  How to decide about charging and releasing electricity based on a substantial number of stochastic input parameters?

So far, storage control problems that involve stochastic components have been solved using partial differential equations or other numerical methods (e.g. least squares Monte Carlo). They all share the same advantage: One is able to look “into” the algorithm and track its decisions. However, the complexity is on the rise. Energy storages are more and more controlled in combination with generation (wind, sun) and consumption (thermal power stations or electric vehicles). Such complex systems cannot be optimized with the aforementioned methods in reasonable time (if at all), especially considering that we might have more than one objective function. Alternative methods are heuristics such as simulation annealing, genetic algorithms or – as in many scenarios nowadays –  neural networks. Especially the latter have proven to be a powerful tool in many applications.

However, neural networks are a black box per construction and we cannot track the resulting decision, which is exactly the objective of this PhD project


First supervisor:

Prof. Dr. Stephan Schlüter, Technische Hochschule Ulm


Tandem partner:

Prof. Dr. Henning Bruhn-Fujimoto, Institut für Optimierung und OR, Universität Ulm


Consulting experts:

Prof. Dr. Volker Herbort, Technische Hochschule Ulm

Prof. Dr. Matthias Klier, Institut für Business Analytics, Universität Ulm

Prof. Dr. Karsten Urban, Institut für Numerische Mathematik, Universität Ulm