LF Energy Summit 2024

Our goal is to find topological actions that solve congestion problems and lower the redispatch volume in the grid using a self-developed gpu-native approach and results on the real transmission grid with open-source packages (like PowSyBl, Pandapower, Jax) As topology actions are basically free, this could reduce the redispatch costs (in Germany around 3-4 billion € per year) incurred by various transmission system operators. There are three main, difficult challenges to solve: (i) the number of combinatorial possibilities, (ii) non-linearity and (iii) consistent and available data. In the first part we will present results on a scalable gpu-native load flow solver, on the topology optimization and challenges to handle real grid data with PowSyBl and pandapower. Next to it, we would also like to share first insight from the open-source research project GNN4GC – Graph Neural Network 4 Grid Control. In the second part we would like pose some questions and have an interactive sessions with people in the audience on challenges and approaches we face.