Minimizing the Costs in Energy Consumption of a Data Center

Author

• Cesar Arcos: [email protected]

Abstract

I will build an AI to minimize energy consumption of a single server.

Problem

In 2016, DeepMind AI minimized a big part of Google’s cost by reducing Google Data Centre Cooling Bill by 40% using their DQN AI model (Deep Q-Learning). I will do something very similar. I will set up our own server environment, and i will build an AI that will be controlling the cooling/heating of the server so that it stays in an optimal range of temperatures while saving the maximum energy, therefore minimizing the costs. And just as DeepMind AI did, our goal will be to achieve at least 40% energy saving.

Result

We saved over 50% of energy

Run

• python training_earlystopping.py
• python testing.py