Research

Conventional power systems typically come with a small number of large-scale generation units. Future power systems, on the contrary, are typically associated with a large number of decentralized small-scale units with uncertain infeed. Contrary to conventional power systems, their structure is steadily changing due to new units, e.g., rooftop photovoltaic generators, being installed or old ones being decommissioned. Moreover, their complexity is significantly higher due to a large number of small-scale units. Therefore, existing identification and control methods for power grids are likely to run into limitations in the future.

Based on the huge amount of measurement data that is being collected in power systems, data-based approaches from control and system identification appear as a obvious choice to overcome this barrier. These allow complex systems to be modeled and promise rapid adaptation to changing network topologies. Hereby we want to identify bottlenecks and increase the share of renewable feed-in by adapting control strategies.

Multimodal energy systems comprise different energy carriers, such as electrical, thermal and hydrogen. The components considered in such grids comprise electric ones, such as renewable generators, storage units and electrical loads, and thermal ones, such as thermal storage units or thermal loads, but also units that provide both, heat and electricity, such as combined heat and power plants. Coupling different sectors allows for a more environmentally friendly operation where the efficiency can be increased by orchestrating the units in ways that reduce the wastage of available energy from uncertain renewable generators.

Given the decentralized nature of such energy systems, control schemes that can be implemented in distributed ways appear as a natural choice. Specifically, we use methods from distributed control and optimization to enable multimodal climate-friendly energy networks.

Microgrids are small power systems that can function either connected to or isolated from a larger power grid. They consist of various components such as storage units, both renewable and conventional generators, as well as consumers. A central task in this field involves determining the optimal operation of microgrids with a high share of renewable energy sources. This entails devising strategies for controlling the energy storage units and maximizing the use of uncertain renewable sources, all while ensuring safe and reliable operation.

To address this task, we have developed several model predictive control (MPC) approaches for managing microgrid operations. These approaches differ in how they handle the uncertainties associated with both the load demand and the renewable energy sources:

1. Certainty equivalence MPC relies on a single forecast and fully trusts it for decision-making.
2. Minimax MPC adopts time-varying forecast intervals, accommodating variations in predictions.
3. Risk-neutral stochastic MPC relies on a probability distribution from the forecast and fully trusts it.
4. Risk-averse MPC, in contrast, does not completely rely on the forecast probability distribution.

In our research, we conduct comprehensive comparisons of such different approaches to determine the most effective methods for managing uncertain generation in future power systems.

One strategy to manage the operation of complex future energy systems with a significant share of small-scale renewable energy sources involves dividing the overall grid into smaller microgrids (MGs). Each MG appears as a distinct entity to the external world and can both provide and consume energy. Allowing power exchange between these entities generally improves the overall performance of the system. Specifically, interconnecting the MGs typically comes with advantages such as smoothing effects coming from geographical dispersion of renewable generators, leading to a potential increase in usable renewable generation.

Our research focus in this field are strategies that enabling optimal trading among a network of interconnected microgrids while preserving the autonomy of each individual MG. Within this focus, various distributed MG schemes have been developed. In an early work, a hierarchical distributed model predictive control (MPC) approach was introduced. This approach involves a central coordinator responsible for managing the transmission of power between individual MGs. Through a distributed optimization scheme, power setpoints are determined by sequentially solving and communicating the outcomes of local problems and a problem at the central coordinator. In a more recent work, this scheme was extended by incorporating a condition that safeguards the self-interests of each MG. Additionally, the central coordinator could be eliminated, and the algorithm implemented in a fully distributed manner, relying solely on peer-to-peer communication between neighbouring MGs. Currently, we are working on stochastic MPC schemes that account for uncertain loads and available renewable infeed.

In future power systems, many rotating generators, e.g., coal power plants, will be replaced by inverter-interfaced renewable sources. Unfortunately, not all ancillary services, such as frequency control or a desired short circuit behavior, can be satisfactorily provided by these renewable sources. Therefore, the use of storage units which provide ancillary services has been advocated.

Our research in this field focuses on some of these ancillary services. These are provided, among other things, by appropriate voltage and current controls of the plants, as well as primary and secondary control schemes. So far, our work included the development of short circuit strategies for decentralized grid-forming storage units on this topic. Moreover, we investigated decentralized control of grid-forming units which included wide-ranging theoretical studies and practical tests with physical inverter hardware in the MW-range. Additionally, we worked on different distributed secondary control schemes.