The OFFIS project, funded by the German Federal Ministry for Economic Affairs and Climate Action under the grant number 03EI4013C, ran from 1 May 2020 to 31 December 2023. It was carried out by the OFFIS Institute for Information Technology together with the Zentrum für Technomathematik at the University of Bremen, the Optimization working group at TU Ilmenau, and associated partners IAV GmbH and EWE Netz GmbH. The authors Tobias Grimm, Rebeca Ramirez, Emilie Frost, Mohammad Arhum, Jirapa Kamsamrong, Pauline Helene Rühmann, Martin Tröschel, and Sebastian Lehnhoff are responsible for the report’s content.

The project addressed the increasing prevalence of neighbourhood grids that generate their own renewable electricity. These grids must simultaneously satisfy local self‑supply and contribute surplus power to the larger distribution network, creating a bi‑level, multi‑objective optimisation problem. To solve this, the team developed a multi‑agent system in which each actor—whether a neighbourhood or the distribution network—is represented by an autonomous agent that negotiates a joint operating point. The core of the negotiation mechanism is an extension of the COHDA algorithm, now capable of handling multiple objectives. The extended algorithm is available as open‑source software on GitHub (mango‑library/negotiation/multiobjective_cohda) and can be integrated into future research projects.

The technical work comprised several components. First, a simulation framework was built using mosaik to represent the physical network topology and the agents’ interactions. Data sets describing realistic summer and winter load and generation profiles were incorporated. The agent system was designed to operate both on the neighbourhood level and on the distribution‑network level, allowing for hierarchical coordination. Communication topologies were evaluated in terms of message volume and convergence speed. In the summer scenario, the agents achieved a balanced power exchange that met local demand while injecting surplus into the grid within a few negotiation rounds. In the winter scenario, the system adapted to lower renewable output by increasing local consumption and reducing grid injections, again converging rapidly. Although the project did not complete a full comparison with alternative approaches from the partner institutions, the results demonstrate that the multi‑agent, multi‑objective COHDA extension can solve the bi‑level optimisation efficiently and reliably.

Performance metrics reported in the evaluation include the number of negotiation rounds required for convergence (typically fewer than ten), the total communication overhead (on the order of a few kilobytes per agent per round), and the achieved objective values: minimised local curtailment and maximised grid support. These figures confirm that the approach scales to realistic network sizes and can operate in near real‑time, making it suitable for integration into operational tools for distribution system operators.

In addition to the software artefacts, the project produced a peer‑reviewed publication titled “Integration of smart neighbour grids into the German distribution grid: A Perspective,” which discusses the policy relevance of neighbourhood grids for the German energy transition. The project’s findings were presented at workshops and conferences, fostering knowledge exchange between academia and industry. The open‑source release of the COHDA extension and the simulation environment provides a foundation for subsequent studies on distributed energy resources, grid stability, and optimisation under uncertainty. The OFFIS project thus contributes both a practical toolset and a conceptual framework for enabling neighbourhood grids to support the broader electricity system while maintaining local self‑sufficiency.