A Reproducible Open-Data Pipeline for Synthetic Low-Voltage Grid Generation in Germany

Abstract:
The transition to decentralized energy systems requires detailed modeling of low-voltage (LV) distribution grids, but this data is often scarce or confidential. Synthetic grid generation tools are now common in literature. However, the reproduction and verification of the results from these tools remains quite difficult due to opaque and bespoke data pipelines. This paper presents a comprehensive open-data-based preprocessing pipeline for generating input data for synthetic LV grid generation in Germany. We leverage openly available geospatial datasets, including 3D building models (LoD2), census demographics (Zensus 2022), and official geodata from cadastral authorities, including street networks with addresses, and integrate them into a harmonized Infrastructure Database (InfDB). The pipeline employs a reproducible, scalable process using SQL in a dockerized PostgreSQL environment to allocate census population data to building geometries, classify building types, and construct a street graph for network routing consistent with available cadastral information. This also implies more accuracy in peak load estimations and asset sizing in the synthetic LV grid generation. We demonstrate this using an improved version of our pylovo tool on a Munich case study with robust household-to-building and building-to-network allocation. The approach is scalable to large geographic areas, including nationwide implementation, and is implemented transparently to be verified and reproduced by the community.