'''Your interests/skills:''' Deep Learning, Graph Neural Networks, Electricity Networks

'''Details:'''<<BR>>
Electricity grids are increasingly heavily used due to the rise of power-consuming technologies like electric vehicles and heat pumps, which are needed for the energy transition towards a carbon-free energy supply.
To guarantee the safety and stability of the grid, grid operators must know the grid's state at every moment in time.
Only few measuring stations exist in the grid, and therefore the voltage must be estimated at grid nodes without measurements to infer the full state of the grid.
Graph Neural Networks (GNNs) are suited to work on heterogeneous graph structures like an electrical grid.
This was already tested at Fraunhofer ISE.
In this thesis, you are going to extend and improve the GNN model by Fraunhofer ISE.
The performance of GNNs with different layer types, hyperparameter settings and training data will be tested.