Research Project eKI4DS

Dynamic stability aspects are often only dealt with in passing. Currently, the static properties of a grid are examined by means of load flow analysis, followed by the selection of a few worst cases for dynamic stability analysis, whereby not all possible scenarios can be considered due to the high computational effort involved. However, the increasing volatility of renewable energies and the fast dynamics of power electronic systems make it necessary to analyze more scenarios in detail and evaluate them systematically in order to meet the dynamic requirements.

Explainable AI methods, such as machine learning, make it possible to learn new correlations from empirical or synthetic data and make predictions for unknown situations. It is crucial to understand the reasons behind the predictions in order to create trust in the results. In many areas, AI has the potential to match or exceed the performance of human experts. Various architectures have already been tested for use in power grids, with Graph Neural Networks (GNNs) being emphasized as they can directly take into account the network structure of power grids. In this project, the focus is specifically on the application of AI to analyze the stability of power grids.

The eKI4DS project aims to develop, validate and demonstrate methods of explainable artificial intelligence (AI) to analyze the dynamic security of power grids efficiently and transparently. These approaches enable optimized use and planning of grids, avoiding unnecessary and costly expansion projects.

Dynamic security is of central importance, as electricity grids are constantly exposed to planned and unplanned disruptions. A key task of grid planning and operation is to ensure that the grid remains stable even in the event of disruptions. It is crucial that the grid maintains a stable operating point after a disturbance and that all technical limits of the grid actors are respected to prevent a cascade of faults and possible system collapse.

Currently, systems are often operated conservatively, which means that capacities are not fully utilized to provide buffers for possible disruptions. Timely assessment of the dynamic security of the ever-changing grid status is necessary in order to take proactive measures against security and stability risks. This would allow the existing infrastructure to be used more efficiently.