In a power system supplied by renewable energy, all PV and wind power plants, battery storage and electric cars, smart consumers and many other units must work together to support the supply. For this, they must behave such that generation and consumption are always in balance and current and voltage everywhere in the grid do not become too large or too small. The fact that the plants can take in or release energy in quite different quantities at different times is known as flexibility. To control them all in a way that supports the goal of a secure, renewable power supply, models are needed that can represent the complexity of the entire system.
In DAI, we deal with the topic of flexibility modeling and optimization in digitized energy systems occupies on a general level as well as in relation to concrete use cases. We have developed a flexibility pipeline in order to describe the process of calculating and using flexibility from the integration of the constraints to the control of a unit. This process is illustrated in the following figure.
Flexibility pipeline to illustrate the process for calculating and utilizing flexibility.
Before calculating flexibility, the constraints that have to be taken into account are considered. On the one hand, this includes unit-specific constraints, such as possible power levels, maximum power values, start-up ramps or efficiencies. On the other hand, the constraints resulting from the application have to be considered. This concerns, for example, executed trades on the electricity market, peak shaving limits that have to be met or special time requirements from the optimization objective.
Based on these constraints, a flexibility model can be generated and the flexibility can be calculated. There are various forms of mathematical representation of flexibility. The simplest representation consists of a set of possible schedules of a unit, more complex representations often allow a higher coverage of the whole possible range of flexibility. In some use cases, it is necessary to aggregate flexibility from different units. This may be simple, complex, or not possible at all, depending on the flexibility model chosen.
Based on the calculated (and possibly aggregated) flexibility, optimization is performed with respect to one or more objectives. The optimization process for several units can be centralized as well as fully distributed and decentralized. The result of the optimization process can be transferred (if necessary after a disaggregation step) into concrete control commands for the respective units. Disaggregation here refers to the step back from the optimized aggregated flexibility to the optimized deployment plan of the individual units.
With the help of the flexibility pipeline, it is possible to precisely describe different use cases for the use of flexibility in digitized energy systems and to select the most suitable flexibility models in each case.
In order to operate electricity storage systems as economically as possible, it is meaningful to address several joint use cases (simultaneously or sequentially). In DAI, we have developed models in various projects that implement such "multi-purpose" management. One example is the Amplify flexibility model from the collaborative project with be.storaged to develop a battery storage swarm. Another example is the FRESH project, in which autonomous, electric container vehicles in the port of Hamburg are also used to provide control power.
To allow agents in a swarm of battery storage to use their assets locally and communicate unneeded flexibility to an aggregator, the Amplify flexibility model was developed. Amplify can be used to calculate battery flexibility very quickly and express it compactly. Furthermore, aggregators can use it to control battery storage based on abstract model knowledge. In order to detect conflicts that arise between use cases, Amplify is also equipped with built-in problem detection.
Amplify is published on Pypi as "amplify-model" and open source on Gitlab.com at: https://gitlab.com/offis-dai/models/amplify
In the FRESH project, a flexibility management system was developed to harness the flexibility of a fleet of battery-powered heavy-duty vehicles at the Hamburg Container Terminal Altenwerder (CTA) to stabilize the power grid. For this purpose, the fleet was integrated into a virtual power plant, and used to provide primary frequency control power (FCR). The most important requirement was not to interfere with the terminal's transport processes. To achieve this, machine learning methods were used to make predictions about the terminal's transportation needs. Then, agent-based optimization was used to determine the number of vehicles and charging stations that could be taken from the logistics operation and used for FCR within a given time period. The system was successfully tested and evaluated in a about 7-week field test at the terminal.
Within a power grid, flexibility can be used for stable grid operation. Redispatch is a mechanism that regulates this with regard to line congestions. The Int2Grids project is investigating the extent to which the flexibility of neighborhood grids can support this. A neighborhood grid is a spatially coherent grouping of generators and loads that jointly pursue self-demand-oriented optimization goals, as can be the case in smart city neighborhoods.
In order for a neighborhood grid to provide flexibility for the stable grid operation of the higher-level grid management, flexibility must be available in advance in aggregated form. In Int2Grids, flexibility is determined using a 2-step process. First, a locally optimized load profile forecast for the neighborhood grid is determined as a reference schedule. This is done using the fully distributed heuristic COHDA which is based on self-organization mechanisms. In this process, all generators and consumers participate using appropriate agents. Based on the reference schedule, the deviation potential of each unit is determined and this is provided as flexibility to the redispatch process. In the event of imminent line overloads, this flexibility can then be (partially) utilized, whereupon the operation of the units in the neighborhood grid is adjusted.
All in all, a possibility is shown here how the flexibility of locally optimized neighborhood grids can be integrated into the interconnected grid in such a way that different optimization goals can be achieved: goals of the neighborhood grid actuators as well as higher-level goals such as grid stability of the distribution grid operators.
Contact
E-Mail: benjamin.giesers(at)offis.de, Phone: +49 441 9722-747, Room: Flx-E
E-Mail: stefanie.holly(at)offis.de, Phone: +49 441 9722-732, Room: Flx-E
E-Mail: martin.troeschel(at)offis.de, Phone: +49 441 9722-150, Room: Flx-E
Dezentraler Redispatch (DEER): Schnittstellen für die Flexibilitätsbereitstellung
Duration: 2022 - 2025
National Research Data Infrastructure for the Interdisciplinary Energy System Research
Duration: 2023 - 2028
WärmewendeNordwest – Digitalisierung zur Umsetzung von Wärmewende- und Mehrwertanwendungen für Gebäude, Campus, Quartiere und Kommunen im Nordwesten
Duration: 2021 - 2025Paul Hendrik Tiemann, Marvin Nebel-Wenner, Stefanie Holly, Emilie Frost and Astrid Nieße; Applied Energy; 2025
Körner, Marc-Fabian and Nolting, Lars and Heeß, Paula and Schick, Leo and Lautenschlager, Jonathan and Zwede, Till and Ehaus, Marvin and Wiedemann, Stefanie and Babel, Matthias and Radtke, Malin; 2024
Emilie Frost and Malin Radtke and Marvin Nebel-Wenner and Frauke Oest and Sanja Stark; SoftwareX; 2024
Malte Stomberg; Martin Tröschel; ACM SIGENERGY Energy Informatics Review; October / 2024
Radtke, Malin and Lehnhoff, Sebastian; European Simulation and Modelling Multiconference (ESM '24); Oktober / 2024
Sager, Jens and Schrage, Rico; Journal of Open Source Software; October / 2024
Rico Schrage and Jens Sager and Jan Philipp Hörding and Stefanie Holly; SoftwareX; 2024
Malin Radtke; ACM SIGEnergy Energy Informatics Review; Oktober / 2024
Schrage, Rico and Tiemann, Paul Hendrik and Niesse, Astrid; SIGENERGY Energy Inform. Rev.; feb / 2023
Lesnyak, Ekaterina and Belkot, Tabea and Hurka, Johannes and Hörding, Jan Philipp and Kuhlmann, Lea and Paulau, Pavel and Schnabel, Marvin and Schönfeldt, Patrik and Middelberg, Jan; Big Data and Cognitive Computing; 2023