Reinforcement learning
Data Center Cooling Optimization
A reinforcement-learning agent that learns continuous cooling control for a multi-zone data centre — holding every zone in bounds while cutting power. A custom Gymnasium environment models the thermal dynamics; a Soft Actor-Critic agent learns the policy.
A closer look.
- A custom Gymnasium environment models thermal dynamics across multiple cooling zones.
- A Soft Actor-Critic agent learns a continuous control policy with Stable-Baselines3.
- The policy aims to keep every zone within temperature bounds while reducing power use.