Keeping the servers in a data center cool requires a massive amount of power, especially when it comes to the large facilities operated by tech giants such as Google LLC. Now the company is trying to harness artificial intelligence to cut the electricity bill.

Google today revealed that it has built a specialized AI to automatically manage the sophisticated cooling equipment in its data centers. According to the company, the system’s ability to fine-tune cooling automatically has already led to a dramatic increase in power efficiency.

The project builds on work that Google’s DeepMind AI division first published in 2016. The group had developed a system that could collect operational data about cooling equipment and offer engineers recommendations on how to optimize power use. Google’s new AI can take over the task entirely, but it operates under human supervision for safety reasons.

The system works by creating a snapshot of the cooling equipment in a data center every five minutes. It takes into account thousands of different metrics ranging from facility temperatures to more nitty-gritty details such as the speed at which the heat pumps are operating. Based on this information, Google’s AI decides what actions can be taken to optimize electricity consumption.

“Optimal actions computed by the AI are vetted against an internal list of safety constraints defined by our data center operators,” Google wrote. “Once the instructions are sent from the cloud to the physical data center, the local control system verifies the instructions against its own set of constraints. This redundant check ensures that the system remains within local constraints and operators retain full control of the operating boundaries.”

According to Google, there is a total of eight different mechanisms in place to ensure the AI works as intended. In case something goes wrong, the system simply falls back to the predefined automation rules that are otherwise used to manage cooling systems.

Google said the AI is delivering average energy savings of 30 percent in the data centers where it has been deployed. That’s less than the 40 percent maximum reductions provided by the recommendation system developed back in 2016, but Google explained that the software is deliberately constrained so to balance power efficiency with reliability. Plus, the search giant sees the AI increasing power savings over time as it collects more data.

“We’re excited that our direct AI control system is operating safely and dependably, while consistently delivering energy savings,” Google wrote. “However, data centers are just the beginning. In the long term, we think there’s potential to apply this technology in other industrial settings, and help tackle climate change on an even grander scale.”

Image: Google