AI is changing the demands placed on data centre power systems at the same time as grid requirements are becoming more complex. GreenScale has published a new open-source document to help operators and project teams navigate both.
Grid & AI Load Compatibility: System Design Requirements for AI Load, Grid Code & Fault Ride-Through has been developed by Liam Newcombe, SVP Product, Energy Strategy & Innovation at GreenScale, with contributions from vendors, partners and independent experts and is being made openly available for the industry to use, test and build upon.
What the requirements address
The document sets out practical design considerations for data centres to:
- Handle new variable AI IT power demand profiles
- Meet new grid connection policies and local electricity network requirements
- Comply with new large demand load fault ride-through requirements
- Support new obligations such as demand response
- Adapt to future changes in grid codes and tenant load profiles
- Open for the industry to use and build upon
The guidance promotes a base system design approach to address a range of foreseeable operating conditions. This includes identification of the load- and site-specific design constraints and how these impact design, equipment selection, and specification. For example, the specification of fault ride-through-capable UPS or mechanical equipment capable of operating across the full voltage and frequency ranges required.
GreenScale is releasing the document under a Creative Commons Zero (CC0) licence allowing companies and individuals to use and build upon it in accordance with the licence terms.
GreenScale is inviting data centre operators, designers, engineering consultants, utilities, technology providers and individual experts to review and reuse the document, and contribute to the continuing industry conversation.
Quotes
“The deployment of AI requires data centres to move from being simple customers of the electricity grid to participants in the grid. Data centres as a class are now large enough to impact grid stability, which means facilities need to adapt, incorporating capabilities such as fault ride-through to ensure this impact is positive, resulting in a resilient, stable grid. The task is no longer simply to protect the IT equipment from the grid, but to meet the combined requirements of the IT equipment, data centre infrastructure, and electricity grid.”
“This document is published as a work in progress. It provides an initial approach that we hope others in the industry will test, challenge and improve. By publishing it open source, we hope to provide a useful starting point for others, sharing our learnings both direct and from our contributors.”
– Liam Newcombe, SVP Product, Energy Strategy and Innovation at GreenScale