In earlier times, energy supply was strictly hierarchical: a power plant produced electricity and distributed it centrally via several substations to hundreds of thousands of households. Today, energy distribution is organized in a much more decentralized manner. Homeowners produce their own electricity with photovoltaic systems on their roofs, for example. They feed the surplus into the common grid. Electricity distribution has thus become a complex matter with many players. AI-based algorithms decide when, where and how much electricity is delivered to the consumer at what time of day.

The research area Energy of OFFIS - Institute of Computer Science from Oldenburg is dedicated to the realization of intelligent energy systems. Eight research groups with different focal points are working, among other things, on energy-efficient concepts for the power supply of tomorrow.

AI controls decentralized energy storage

Battery storage systems are an important component here. The flexibility of such energy depots makes it possible to compensate for short-term fluctuations of weather-dependent wind power and solar plants. In order to be able to use the currently still comparatively expensive storage systems in an economically and technically optimal way, OFFIS is investigating the possibilities for a multi-purpose use of battery storage. For example, parked e-cars are to serve as local energy storage systems, from which electricity can also be tapped when needed - depending on the calculation of the AI.

The Oldenburg scientists are also concerned with the question of how AI can optimize the forecasting of grid conditions in the distribution grid. External data - for example, from the current traffic situation and the weather forecast - are to be used to automatically determine the best possible network utilization.

Resilient power grids against cyber attacks

However, the digitalization of the energy supply also harbors new threat scenarios with great potential for damage to society. For example, cyber attacks paralyzed parts of Ukraine's entire power grid in 2015 and 2016. The consequences were blackouts lasting for hours. Since these incidents at the latest, it has become clear that power grids have become important targets of malicious attacks. The question that therefore arises: How can vulnerabilities be reliably found and closed?

AI offers a solution here as well. Traditional methods for analyzing operational security quickly reach their limits. The Oldenburg researchers are currently working on software that simulates attacks on the power grid and takes immediate countermeasures. In the process, the AI system plays against itself like a chess player, developing ever newer and more sophisticated defensive measures through the attack simulations.

Electricity must be available everywhere and at all times. At the same time, it should be cheap, environmentally friendly and produced as regionally as possible. To ensure all of this, intelligent energy generation and supply is needed. Scientists at Oldenburg's OFFIS - Institute for Computer Science are researching how decentralized power grids can be better controlled using artificial intelligence methods.