Probabilistic switch state identification based on scarce grid measurements: a case study on German distribution grids

Abstract:
The integration of decentralized energy resources (DERs) such as photovoltaic installations, electric vehicles, and heat pumps in distribution grids can lead to voltage stability issues and grid congestion. To effectively manage these chal-lenges, distribution system operators (DSOs) require accurate models of their distribution grid topologies. However, DSOs often lack transparency of the actual switch configurations and operational topologies of their grids, especially in medium and low voltage networks where switches are operated manually, and documentation is scarce. This paper pro-poses a probabilistic two-stage approach to identify the most likely switch configuration from a set of plausible candidates. In the first stage, a Markov Chain Monte Carlo (MCMC) method is used to efficiently generate a set of candidate topol-ogies. In the second stage, the candidate topologies are further evaluated using available load and voltage measurements, using a Bayesian two-step framework. The proposed method is validated on two real-world distribution grid datasets from a German DSO, including a 273-bus medium voltage network and a 3042-bus low voltage grid. By providing DSOs with the most likely topology candidates, the method can guide their efforts to improve grid transparency and support the integration of DERs.