A Holistic Load Dataset of Two Multi-Family Houses Deriving from Advanced Metering Infrastructure

Abstract

For many research applications, high-resolution datasets are essential. This includes various machine learning approaches and representative load profiles for evaluating energy management strategies in the laboratory or modelling. As freely accessible data are still not sufficiently available for various investigations, this work presents a dataset of two identical multi-family houses, originating from Bielefeld, Germany, with a total of 48 households. It was collected within the funded project MELANI, considering the multi-use of photovoltaic and battery energy storage systems in multi-family houses. The authors, from the elenia Institute located at TU Braunschweig and the National Metrological Institute, were part of this project. The data was measured via an advanced metering infrastructure, the special feature of which is the direct use of household meters for control algorithms. The dataset contains load profiles (active power in watts) at a high resolution of one second for an entire year, specifically from March 2024 to February 2025. A distinction is made for providing different time series for households, electric vehicles, shared facility loads, and photovoltaic generation. The power values range up to 12kW for household and EV loads, 17kW for the shared facility load, and around 45kW for PV generation. To assess the actual data availability, a revision strategy that identifies various errors to achieve a complete second-by-second dataset is developed. Data gaps are also presented in a data availability diagram. The average data availability across all time series is 95.74%. The revision strategy includes script-based revision through persistence, the use of parallel data acquisition, and interpolation. The annual consumption of households varies between 576 and 4460 kWh. A comparison with the German standard load profile suggests that the household load profiles collected are sufficiently representative. In addition to the household loads, their respective electric vehicle loads, as well as the shared facility load and photovoltaic generation, are revised and provided. The dataset is made available for download via Zenodo.