Real-world energy data of 200 feeders from low-voltage grids with metadata in Germany over two years

The last mile of the distribution grid is crucial for a successful energy transition, as more low-carbon technology like photovoltaic systems, heat pumps, and electric vehicle chargers connect to the low-voltage grid. Despite considerable challenges in operation and planning, researchers often lack access to suitable low-voltage grid data. To address this, we present the FeederBW dataset with data recorded by the German distribution system operator Netze BW. It offers real-world energy data from 200 low-voltage feeders over two years (2023-2025) with weather information and detailed metadata, including changes in low-carbon technology installations. The dataset includes feeder-specific details such as the number of housing units, installed power of low-carbon technology, and aggregated industrial energy data. Furthermore, high photovoltaic feed-in and one-minute temporal resolution makes the dataset unique. FeederBW supports various applications, including machine learning for load forecasting, conducting non-intrusive load monitoring, generating synthetic data, and analyzing the interplay between weather, feeder measurements, and metadata. The dataset reveals insightful patterns and clearly reflects the growing impact of low-carbon technology on low-voltage grids.

💡 Research Summary

The paper introduces the FeederBW dataset, a comprehensive, real‑world collection of low‑voltage (LV) feeder measurements from the German distribution system operator Netze BW. Data were recorded from 200 distinct LV feeders over a two‑year period (2023‑2025) with a one‑minute temporal resolution. For each feeder, the dataset contains per‑phase RMS current, active and reactive power, power factor, and a single voltage measurement (shared among feeders at the same substation). The measurement hardware is the SMIGHT Grid 2 sensor‑gateway system, which harvests energy from the feeder itself and therefore requires a minimum total current of 32 A for stable operation. Current accuracy is ±3 % in the 5‑400 A range; voltage accuracy is also ±3 % in the 85‑264 V range.

Beyond raw electrical quantities, FeederBW provides rich metadata derived from Netze BW’s customer management system. For each feeder the metadata includes the number of connected residential units, the installed capacity of low‑carbon technologies (photovoltaic (PV) systems, heat pumps, electric‑vehicle (EV) chargers), and average hourly consumption figures for industrial and commercial customers. These metadata entries are event‑driven, updated whenever a new installation is registered or consumption statistics are revised, allowing researchers to track the evolution of distributed generation and demand response assets over time.

Weather information is integrated at the zip‑code level using the German Weather Service’s ICON‑D2 numerical weather prediction model. Hourly variables such as temperature, precipitation, wind speed and direction, global horizontal irradiance, diffuse and direct radiation, humidity, and pressure are supplied. This weather layer enables detailed studies of the weather dependence of PV generation, heat‑pump operation, and EV charging demand.

The authors position FeederBW against existing public LV datasets, highlighting its superior scale (200 feeders vs. typically <10), higher temporal resolution (1 min, second only to a 5‑second dataset), inclusion of both load and generation data, and the presence of extensive customer‑side metadata and weather data. While the dataset does not contain explicit grid topology, it can still complement topology‑based studies by providing realistic time‑series inputs.

Data stewardship follows FAIR principles: each data object receives a DOI via Zenodo, ensuring findability, accessibility, interoperability, and long‑term preservation. The dataset is distributed in common formats (CSV, Parquet) with detailed licensing and provenance documentation. Recommended train‑validation‑test splits are provided to facilitate reproducible machine‑learning experiments.

Potential applications are numerous. The high‑resolution feeder measurements support load and generation forecasting models at the feeder level, which are essential for distribution system operators’ state estimation and congestion management. The combination of aggregate feeder data with detailed metadata enables non‑intrusive load monitoring (NILM) research, allowing the decomposition of feeder‑level consumption into constituent device categories. Researchers can also generate synthetic feeder data for grid simulation studies, improve data quality through imputation or anomaly detection techniques, and explore the interplay between weather, low‑carbon technology penetration, and power‑factor dynamics.

In summary, FeederBW fills a critical gap in publicly available LV grid data by delivering a large‑scale, high‑resolution, and richly annotated dataset that captures the evolving landscape of decentralized generation and electrified heating/transport. Its release is expected to accelerate reproducible research in load forecasting, grid analytics, and the development of data‑driven tools for the energy transition.