AI-driven microgrids and renewable integration are crucial for sustainable, efficient development of data centres, writes Barny Evans, Director of Sustainability and ESG at Turley.
Data centres are crucial to the AI revolution. To make the most of them, however, requires a sophisticated approach. AI itself should be crucial in planning how to power data centres effectively at the heart of the AI revolution. Used correctly, it would allow us to develop and support the transition to net zero CO2 emissions more efficiently and cost-effectively.
Data centres account for about 1% of the UK’s electricity demand. Globally, they use more electricity than the whole of the UK – and that’s only going to grow a lot more with AI. According to the National Grid, data centre demand will increase six-fold in the next decade, comparable to the demand of the whole of London. This growth is creating challenges as the demand for spare grid capacity competes with our needs for housing and other uses.
Heat
One suggestion to improve the situation is to utilise the waste heat that data centres produce. While this idea has merit, it presents some practical challenges. Data centres often have different infrastructure requirements and security needs, compared to residential or leisure facilities, which can make co-location difficult.
Additionally, capturing and transporting heat is both costly and challenging, and data centres tend to release the majority of their heat during the summer when the demand for it is at its lowest. There are opportunities for symbiosis, however, if it is planned into the start of a wider site or project, such as industrial laundries with a consistent demand for heat, though these applications may be limited in scale.
Power
You will often hear that data centres are using 100% renewable energy, but this is normally on a net basis; that is the data centre operator buys as much renewable energy as they use in a year. The crucial thing for the future is that renewable energy must be used as it is generated. We can make increasing amounts of solar and wind energy, at decreasing costs, but we need to balance that supply with demand in real time.
The really big opportunity, therefore, is how data centre power demand can flex and interact with other uses on sites, using AI and microgrids. Co-locating data centre sites with logistics and manufacturing is natural and allows the opportunity for them to share electricity from solar generation, adjusting demand as needed.
If you are planning a new mixed-use site, it could contain data centres, warehousing, manufacturing, massive solar roofs, maybe a solar or wind farm, and lots of electric vehicle charging. A traditional approach might require each of these uses to determine its own maximum grid capacity requirements, potentially making development expensive, inefficient, and slower as we await grid updates.
A microgrid takes a different approach, operating a network as if all these uses were part of one energy system able to share energy and respond to each other by adjusting their power usage. If there is lots of solar power, the data centre may ramp up its processing, the EV charging may turn-up, the manufacturing facilities can increase production, and so on, to soak it up, rather than exporting it. If there is low wind and power generation then the EV charging can be minimised, the data centre can postpone some processing, etc. Data centres can be surprisingly flexible with their demand. One report suggests that up to 40% of demand can be shifted.
Combining this in a microgrid with other uses and generators would be hugely beneficial; the entire site can reduce the size of the grid connection, reduce energy bills, support the transition to a net zero network and free up capacity for housing and other needs.
As we continue to push the boundaries of artificial intelligence, we must also continue to use human intelligence to use energy optimally and significantly enhance the AI revolution.
