Avoiding false inter-zonal meshing in the clustering of a large-scale German power grid

The ongoing transformation towards a sector-coupled energy system based on renewable generation leads to more complex grid-based energy system models. It is a crucial task to reduce the models’ complexities in order to keep optimisation problems tractable while at the same time generating appropriate results. This work adds a Dijkstra’s algorithm to a commonly used k-means Clustering method to reduce the spatial complexity of a German transmission and sub-transmission grid model. The novel approach leads to more accurate results while reaching faster calculation times. In particular, it successfully avoids false inter-zonal meshing. Consequently, the more accurate modelling of inter-cluster connections results in up to 41% higher grid expansion needs and significantly changed spatial allocations of network and storage expansion. Where geographical or political borders (e.g. the former inner-German border) have led to scarcely interconnected grid topologies (especially on the sub-transmission grid level), the impact is particularly high. The presented work follows open-source and open-data principles. • Adding a Dijkstra’s algorithm to a k-means Clustering avoids false inter-zonal meshing. • Higher grid expansion needs are seen when applying the presented clustering method. • Topological changes in reduced grids affect the distribution of grid and storage expansion. • Avoiding false meshing is essential regarding geographical or political borders. • The work complies with open source and open data principles.