Optimization and Machine Learning

In the Optimization and Machine Learning department, we apply and develop numerical optimization methods and algorithms for the design and operation of complex energy systems. Important challenges in optimization of such systems are nonlinearity (in particular nonconvexity), dynamics, multiple scales, uncertainty and discrete decisions. Our expertise covers (deterministic) global optimization, stochastic programming, bilevel optimization, and dynamic optimization. A particular focus of our work is the combination of numerical optimization and machine learning (ML). Tailoring optimization methods allows us to solve design and operation problems with artificial neural networks embedded into hybrid system models. To tackle large scale problems, we develop methods for ML-based surrogate modeling and algorithms for parallel and high performance computing architectures

Control Solutions

The Control Solutions department focuses on the development of methods and algorithms for the control of power systems and their components. The focus is on application-oriented methods that can operate in real time, are scalable and are suitable for the operation of decentralized systems. In order to be able to test corresponding control solutions, the department designs and operates a "Hardware-in-the-loop" laboratory. The laboratory is based on a holistic real-time simulation environment in which the innovative solutions are tested by simulating the multi-physical infrastructures and the communication networks under almost realistic conditions.

High Performance Computing

The High-Performance Computing department pursues the goal of further developing the simulation and optimization-based methods used for the planning and operation of energy systems with regard to numerical mathematics and high-performance computing.

In this way, the most efficient use of the IT resources used in the respective application is aimed for. This includes commodity workstations, server clusters and networks as well as supercomputers, specialized high-speed networks and hardware accelerators. Central aspects here are, among others, the runtime analysis of existing simulation and optimization software, the subsequent improvement of existing algorithms, the design of new algorithms with a focus on modern IT resources, and their implementation as well as validation

Networks and Complex Systems

The department networks and complex systems develops mathematical methods and tools for a stable operation of the electric power grids. One focus of our work is the stability and robustness against outages and equipment failures. We use simulations and concepts of mathematical graph theory to understand what makes a power grid vulnerable and how we can design stable and robust networks. A second focus is power system control and the interactions with other parts of the energy system. Using method from machine learning, we quantify and explain how external influences and perturbations affect the operation of the power system.

Energy Grids

The Energy Grids department develops and applies models for the simulation and optimization of multi-physical energy infrastructures. We focus on electricity, gas and heating grids as well as their information and communication structures. Thereby, we also develop integrated energy market models to derive the distribution and transmission tasks for the energy infrastructures under consideration. The main subject of research are the stationary and dynamic characteristics of current and future energy systems, which are characterized by high shares of renewable energies and increasing sector coupling. One focus on the further and new development of models is to be able to use solution methods that enable the use of server clusters and high-performance computers. In addition, we develop scalable methods with which the increasingly complex control tasks in energy networks can be solved faster than in real time.

Building and Districts

The “Buildings and Districts” department develops and demonstrates innovative low-carbon and energy efficient solutions for buildings and (city) districts. In future energy systems, the energy supply for electricity, heating and cooling will be characterized by smaller, decentralized energy sources. The coordination of the multiple energy producers and consumers is a major challenge. For this purpose, the department develops e.g. an information and communication platform (ICT platform) and agent-based control approaches. The developed solutions are thoroughly tested in a near to real environment, e.g. in the "Living Lab Energy Campus (LLEC) Smart area (“Reallabor”) at Forschungszentrum Jülich. To integrate users, the department develops web-based dashboards that provide various stakeholders with information, e.g. on the current energy consumption. Another focus of the department is the development of methods and model libraries for the dynamic modeling of heating networks and building energy systems.

Industrial Energy Systems

Research in the Industrial Energy Systems department focuses on economic and low-CO2 solutions for industrial sites integrating electricity, heat, cooling, and chemical energy carriers. The shift from a consumer-driven to a producer-driven energy system is both an opportunity and a challenge for energy-intensive industries. By means of intelligent system integration (digitalization) and flexible operation (demand side management) industrial processes can actively support grid stability and reduce their energy cost. Furthermore, application of Power-to-X technologies allows to decarbonize production processes. These advantages are in conflict with the high demands on the security of supply and the intense international competition energy-intensive producers face. We develop models and methods to optimally design and operate future industrial energy systems by means of multi-objective optimization.

Last Modified: 02.02.2024