# High-resolution energy data from a sustainable industrial production area in Karlsruhe

## TL;DR

This paper introduces a high-resolution industrial energy dataset with detailed electrical measurements over seven years, supporting better energy management and machine learning applications.

## Contribution

The novel contribution is a large-scale, high-resolution industrial electricity dataset with detailed measurements and metadata for long-term analysis.

## Key findings

- The dataset includes over 74 billion data points collected at 5-second resolution from 22 industrial machines and a photovoltaic system.
- It features up to 190 measured quantities per device, including harmonic spectra and fundamental waveform characteristics.
- The dataset is enriched with external metadata like weather, electricity prices, and emission factors.

## Abstract

Understanding and optimizing industrial energy systems requires datasets that capture detailed electrical behavior at high temporal resolution over long time periods. Such data are essential for analyzing power quality, identifying operational patterns, and developing data-driven models for forecasting, control, and predictive maintenance. Yet, most existing open datasets lack the temporal granularity, measurement diversity, and machine-level detail needed to reflect the complexity of industrial environments. To address this gap, we present a large-scale, high-resolution dataset of industrial electricity measurements comprising more than 74 billion data points collected at 5-second resolution over up to seven years. The dataset includes 22 industrial machines and one photovoltaic system, with up to 190 measured quantities per device, including three-phase voltages and currents, active, reactive, and apparent power, harmonic spectra, total harmonic distortion, and fundamental waveform characteristics. In addition, the dataset is complemented by external metadata such as weather, electricity prices, and emission factors. This unique combination of long-term coverage, high sampling rate, and rich feature space enables insight into industrial energy dynamics and provides a robust foundation for advancing machine learning, digital twins, and intelligent energy management in industrial environments.