Large-scale heat pumps: how waste heat from data centres gets into the heating network

01Waste heat that nobody uses

The cooling of a data centre removes almost all of the electricity consumed as heat – all year round and in a readily predictable way. However, the temperature level is too low for a conventional district heating network. A large-scale heat pump closes this gap: it extracts heat from the cooling water, raises it to network temperature and returns chilled water to the data centre.

02The fallacy: focusing on the machine

In early project phases, the discussion often centres on manufacturers, compressor type or refrigerant. However, the greatest influence on electricity consumption and heat price comes from two temperatures that are fixed long before the machine is selected: the source temperature from the data centre and the required flow temperature of the network.

03The principle: the temperature lift determines efficiency

The theoretical upper limit of the coefficient of performance (COP) is given by the absolute temperature of heat delivery divided by the temperature lift between source and sink. Real large-scale heat pumps achieve a fraction of this limit, known as the exergetic efficiency (Gütegrad), often about half. If the lift is halved, the coefficient of performance roughly doubles.

It follows that warm-water cooling in the data centre and low network temperatures have a double effect. They reduce the lift from both sides.

A second advantage of data centre waste heat is its constancy. Unlike outdoor air, the source is warm and frost-free all year round, and defrosting is not needed. The seasonal performance factor (SPF) is therefore usually closer to the design value than with air-to-water heat pumps. What remains decisive is how the network temperature changes over the year: weather-compensated reduction in summer noticeably improves the balance.

04A calculation example

Assume that a heat pump is to supply 1000 kW to a network with a flow temperature of 65 °C, fed from cooling water at 30 °C. With an exergetic efficiency of 0.5, this gives a coefficient of performance of around 3.5. The heat pump then needs about 290 kW of electricity; a good 710 kW comes from the waste heat.

> 80 %share of electricity costs in annual heat costs – assumed: 8000 full-load hours, 15 Rp./kWh, investment depreciated over 20 years

Because electricity dominates the heat costs, every improvement in the coefficient of performance feeds almost directly into the heat price. If the network temperature glides significantly between flow and return, multi-stage series arrangements or processes that release heat with a temperature glide are also worthwhile.

05What this means for your project

• Clarify flow and return temperatures over the whole year with the network operator, not just the design case.
• Plan data centre cooling with the warmest possible cooling water – this reduces cooling effort and temperature lift at the same time.
• Evaluate variants on the basis of the seasonal performance factor and the levelised cost of heat, not the COP on the data sheet.
• Define redundancy, peak load coverage and the safety concept for the refrigerant to EN 378 at an early stage. Information on natural refrigerants is provided in the article on the refrigerant transition.

Takeaway

In waste heat utilisation with large-scale heat pumps, network temperature is the most important cost driver: keeping the temperature lift small reduces electricity demand – and thus the heat price – for the entire service life.

SEDC AG · Article from our internal training series “Technik Praxis”, abridged for clients and users.