Earth Notes: On Data Centre Heat Recovery (2024)

Updated 2024-08-15.
Notes for summer school group challenge "Energy and heat recovery for Mega Data Centres" #PhD #heating #research
Notes for Network+ heating & Cooling Summer School to .

Group Challenge

Summer school attendees were informed by email on :

The group challenge over the 2.5 days will focus on "Energy and heat recovery for Mega Data Centres"

It is recommended that you:

  • Research existing data centres and their heat recovery techniques
  • Understand the progression to Hyperscale datacentres and the potential that this brings.
  • Research grades of heat and the different methods used to recover them.
  • Research other waste heat recovery techniques within other industries (e.g. major manufacturing such as steel production, chemical processing plants etc.).

Existing data centre heat recovery

Waste heat recoveries in data centers: A review: data centres ... are influential energy consumers and carbon emitters in building or even global energy sectors (around 3% of global energy consumption [in 2019, 4% GHGs])... [yuan2023recoveries]

Data center heat reuse: How to make the most of excess heat: examples of heat reuse include district/network home space heating, agriculture, fish harms, wood pellet drying.

Other possible uses include DHW (Domestic Hot Water), absorption or adsorption cooling, electricity generation via Organic Rankine Cycle (ORC) and Seebeck-effect devices and desalination, and thermal storage may assist with overall recovery [yuan2023recoveries].

Hyperscale data centres

Accoding to IBM: What is a hyperscale data center?:

A hyperscale data center is a massive data center that provides extreme scalability capabilities and is engineered for large-scale workloads with an optimized network infrastructure, streamlined network connectivity and minimized latency.

...

A hyperscale data center differs primarily from traditional data centers by virtue of its sheer size. A hyperscale data center requires a physical site large enough to house all associated equipment—including at least 5,000 servers and quite possibly miles of connection equipment. As such, hyperscale data centers can easily encompass millions of square feet of space.

The first (2006) was Google's Dalles (OR) at 1.3Msqft; the current (2024) largest, in China, is 10.7Msqft.

The largest hyperscale cloud providers are AWS (~32% market share), Microsoft Azure and Google Cloud Platform.

Power requirements from MW to GW; above hyperscale definition implies >100MW.

US data centre power requirements are predicted to double by 2030 to 35GW. (Similar to current GB grid mean consumption.)

Grades of heat and their recovery

Data centre heat is all fairly low grade because it is low temperature, in turn because of the relatively low ceiling operation temperature for IT equipment. Waste heat temperatures from ~20°C to 60°C (liquid cooling system) are available from Table 1 [yuan2023recoveries]; the higher the better. (Heat grade can also be raised with heat pumps.)

Thermopedia: Waste Heat Recovery: Waste heat can be considered as either low grade (<100°C), medium grade (100°C--400°C) or high grade (>400°C).

About half of industrial process heat is >400°C.

[su2021multigrade] puts the thresholds at 230°C and 650°C, and puts fuel cell, cement and steel industries into the high-grade category; glass, electricity generation and automobile transportation are in medium; a further six are in the low-grade category including food and petrochemicals. This paper describes a multitude of recovery theoretical and deployed mechanisms from sewage (evaporation condenser) to cement (flue gas).

Non-data-centre waste heat recovery techniques

ScienceDirect topic: Waste Heat Recovery System.

It is often the case that within one manufacturing process plant, eg chemical engineering, there are places that need heating and others that need cooling. With care these can be coupled ("process integration") to minimise external energy demands and waste emissions.

One of my banking clients needed no heating for the human occupants in winter because it was able to reuse waste heat from the servers in the same building; the building did require cooling in summer however.

At home I have MHRV (Mechanical Heat Recovery Ventilation) that recovers about half the heat from exhaust air in the bathroom and kitchen. Notable is when very humid air from the bathroom has the moisture's latent heat recovered and water drips outside.

District heating and cooling schemes can share thermal flows between buildings close by.

Winning solution

The team that I was in won the group challenge (ie planning for better use of 'hyperscale' data centre waste heat), mainly by making use of a 60°C stream from liquid cooling for domestic space heat and DHW, ~10k+ homes in heating climates for each 100MW of data centre power demand, tacking fuel poverty and earning a social licence to operate.

Ideal Job

We were also asked to bring along a copy of a job advert for your ideal job (either now or in 5 years' time).

Mine would be something like:

Fractional CTO/CSO/CEO of start-up with climate impact, preferably in the energy efficiency space, maybe distributed demand-side response at scale, eg millions of participating end-points or more.

References

(Count: 5)