This paper is a technical overview of DeepMind and Google's recent work on reinforcement learning for controlling commercial cooling systems. Building on expertise that began with cooling Google's data centers more efficiently, we recently conducted live experiments on two real-world facilities in partnership with Trane Technologies, a building management system provider. These live experiments had a variety of challenges in areas such as evaluation, learning from offline data, and constraint satisfaction. Our paper describes these challenges in the hope that awareness of them will benefit future applied RL work. We also describe the way we adapted our RL system to deal with these challenges, resulting in energy savings of approximately 9% and 13% respectively at the two live experiment sites.
@article{arxiv.2211.07357,
title = {Controlling Commercial Cooling Systems Using Reinforcement Learning},
author = {Jerry Luo and Cosmin Paduraru and Octavian Voicu and Yuri Chervonyi and Scott Munns and Jerry Li and Crystal Qian and Praneet Dutta and Jared Quincy Davis and Ningjia Wu and Xingwei Yang and Chu-Ming Chang and Ted Li and Rob Rose and Mingyan Fan and Hootan Nakhost and Tinglin Liu and Brian Kirkman and Frank Altamura and Lee Cline and Patrick Tonker and Joel Gouker and Dave Uden and Warren Buddy Bryan and Jason Law and Deeni Fatiha and Neil Satra and Juliet Rothenberg and Mandeep Waraich and Molly Carlin and Satish Tallapaka and Sims Witherspoon and David Parish and Peter Dolan and Chenyu Zhao and Daniel J. Mankowitz},
journal= {arXiv preprint arXiv:2211.07357},
year = {2022}
}