SIMULATOR-BASED REINFORCEMENT LEARNING FOR DATA CENTER COOLING OPTIMIZATION

Summary from the TLDR Web Dev newsletter

Meta has successfully used reinforcement learning to optimize energy and water consumption in its data centers. Its approach uses historical data and a physics-based simulator to train a model that determines the optimal airflow setpoint for various weather conditions.