Authors: Soumyendu Sarkar, Avisek Naug, Ricardo Luna, Antonio Guillen, Vineet Gundecha, Sahand Ghorbanpour, Sajad Mousavi, Dejan Markovikj, Ashwin Ramesh Babu

Published on: March 21, 2024

Impact Score: 8.6

Arxiv code: Arxiv:2403.14092

Summary

• What is new: A multi-agent Reinforcement Learning framework, DC-CFR, optimizes data center operations for carbon footprint reduction, energy consumption, and cost in real-time.
• Why this is important: The rising energy consumption of machine learning workloads necessitates sustainable, low carbon emission data centers.
• What the research proposes: The proposed DC-CFR MARL framework dynamically optimizes cooling, load shifting, and energy storage in data centers based on renewable energy availability and other factors.
• Results: DC-CFR achieved significant improvements over the ASHRAE controller, reducing carbon emissions by 14.5%, energy usage by 14.4%, and energy cost by 13.7%.

Technical Details

Technological frameworks used: Data Center Carbon Footprint Reduction (DC-CFR) multi-agent Reinforcement Learning (MARL)

Models used: Reinforcement Learning agents

Data used: Real-world dynamic weather data, grid carbon intensity

Potential Impact

Data center operators, cloud computing companies, sustainable technology solutions market

Want to implement this idea in a business?

We have generated a startup concept here: GreenSync Data Solutions.