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

Description

Abstract (English)

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.

Other

Usage notes

We encourage scientists conducting ML experiments with the dataset to follow a consistent split of the data into training and test to enhance comparability. We recognize that deviations may be appropriate depending on the research question or methodology. For locational splits, we recommend to use feeders F1 - F160 for training and F161 - F200 for testing. When conducting a cross validation, we recommend using a five-fold-cross validation with the folds F1 - F40 (test feeders in the first iteration), F41 - F80 (test feeders in the second iteration), F81 - F120 (test feeders in the third iteration), F121 - F160 (test feeders in the fourth iteration) and F161 - F200 (test feeders in the fifth iteration). The locational splits correspond to a random distribution. 

For temporal splits, we recommend to use the first year (April 1, 2023 to March 31, 2024) for training and the second year (April 1, 2024 to March 31, 2025) for testing. In addition, we encourage users to apply cross validation for time-series forecasting. We do not recommend folds for the time-series cross validation because they can differ based on your research question.