PRO-LAB

PRO-LAB addresses the rising energy demands of computationally intensive AI applications, which can consume up to 20 times more electricity under load than in idle mode. With the existing Green-IT-Housing-Center and photovoltaic installations serving as infrastructure, the project provides an ideal foundation for developing methods to operate AI training processes as energy-efficiently as possible. This approach combines technological innovation with ecological responsibility and strengthens regional competitiveness through forward-looking research.

Schematische Funktionsweise der intelligenten Aufgabenverteilung

The project aims to develop an intelligent, energy-efficient control system for AI systems that integrates energy data, building data from cooling and ventilation systems, and weather information. Based on this data, algorithms will be created to dynamically and efficiently distribute AI workloads, particularly when renewable energy supply exceeds demand. To achieve this, a modern data and computing infrastructure will be established and analyzed within the context of the interdisciplinary energy research laboratory. The ultimate goal is the development of intelligent task distribution to prevent peak loads. This solution is intended to serve as a foundation for reducing energy consumption, lowering CO₂ emissions, and enhancing the efficiency of computing infrastructure, thereby contributing to the climate goals of the federal state of Bremen.

Data will be centrally collected, processed, and analyzed in real time to optimally plan AI training processes. The system will consider not only energy demand but also the thermal load on the computing infrastructure. The developed approaches will be tested in a laboratory environment and validated using measured data.