A real-world energy management data set from a smart company building for optimization and machine learning

We present a real-world data set obtained from monitoring a smart company building over the course of six years. The data set describes the energy consumption of various sites within the building, energy production via a photovoltaic system and a combined-heat-and-power plant, and the detailed operation of the heating and cooling system. The data set further contains measurements from an on-site weather station for the same time period. The data set covers periods of normal operation before the onset of the Covid-19-pandemic, periods of reduced operation during, and after, the pandemic. We describe the recording, processing, and curation strategy to generate the data set. The data set enables the application of a wide range of methods in the domain of energy management, including optimization, modelling, and machine learning to optimize building operations and reduce costs and carbon emissions.

Methods
During the recording time span, a multitude of issues occurred which affected the collected data, like measurement outages, maintenance and device replacements.

In order to produce a consistent and research-grade data set, these issues need to be addressed and corrected. We apply a cleaning and post-processing pipeline to the data, which consists of seven steps:

Specification and detection of issues with rule-based detection mechanism
Data harmonization to ensure consistency in naming and sign convection
Application of issue correction
Time alignment of all measurements
Resampling into equidistantly sampled time series (1 min, 15 min, 1 h)
Calculation of missing dependent measurements
Export the time series in gzip-compressed CSV files

Furthermore, based on the corrected and resampled time series, we provide a reduced dataset. It consists of a less complex representation of the building energy consumption, production of both electricity, heating and cooling, as well as weather measurements.