Reference Designs for Data Center Heat Reuse, Revision 1.0

Overview

The document opens with a business section on how heat reuse can be integrated into data centre operation. A stakeholder map defines interfaces from the data centre through heat transmission to the end user. A seven-level maturity model (0-6) sets the demarcation point: level 0 is own use and level 1 is a tap-off from chilled water or a heat exchanger (roughly 30 C supply water); heat-pump levels deliver 30-70 C or 30-100 C and above, and the more mature levels move risk to the data centre operator and make it a heat retailer. The document concentrates on levels 0 and 1.

The technical core is the heat recovery station. Essential hardware is a liquid-to-liquid heat exchanger with plumbing, metering, valves and actuators. At levels 0-1 the demarcation point is the wall of the heat reuse room and all essential equipment is owned and operated by the data centre; if the host needs a higher temperature it provides the boost. The space table assumes a full-capacity AHRI-certified plate-and-frame exchanger, a maximum 10 psi drop, PG25 on both sides, a maximum 2.13 m/s pipe velocity, a low delta T of 4.4 C and a high delta T of 16.7 C. The heat reuse loop counts as a secondary heat rejection path, and dry coolers downstream are still sized for the full IT load.

Scenario families are defined by the ITE cooling method: air cooling (return generally not above 30 C), moderate-temperature liquid cooling (generally not above 40 C) and high-temperature liquid cooling (above 40 C, in some cases reaching 65 C). Retrofits split into cases where heat is already transferred to a liquid at some point and where it is not; the second is harder because it needs an added air-to-liquid loop and fan system changes. The document warns that replacing N+1 dry coolers with a heat reuse station can reduce redundancy.

The demand-matching section covers seasonal and intraday variation, heat storage (sensible, latent, thermochemical), heat pumps, boilers, heat-to-cooling and heat-to-power. Impacts on PUE, WUE and CUE and the ERF and ERE metrics follow: when heat is used without physical transformation all three improve, and when a high-energy transformation is needed they may worsen. This is given as a qualitative rule of thumb, with measured impact by climate region left to future work. Appendices contain six block diagrams, two-phase DLC heat reuse and additional heat pump considerations.

Key points

1. Maturity model levels 0-6: levels 0-1 (own use or tap-off via heat exchanger) deliver roughly 30 C supply water; heat-pump levels deliver 30-70 C or 30-100 C and above.
2. For levels 0-1 the demarcation point is the wall of the heat reuse room, with all essential equipment owned by the data centre; temperature boosting is the host's responsibility.
3. Space table assumptions: full-capacity AHRI-certified plate-and-frame HX, 10 psi max pressure drop, PG25 on both sides, 2.13 m/s max pipe velocity, low delta T 4.4 C and high delta T 16.7 C.
4. Data centre heat can exceed 45 C in some cases but generally reaches 25-40 C; by scenario family return is up to ~30 C for air cooling, up to ~40 C for moderate liquid cooling and above 40 C up to 65 C for high-temperature liquid cooling.
5. Heat reuse is a secondary heat rejection path; downstream dry coolers are sized for the full IT load and the system falls back to standard heat rejection if heat reuse fails.
6. Thermal storage is space-intensive: bridging 1 MW for 30 minutes needs about 43 m3 of water and 1 MWh needs about 80 m3.
7. Heat pump Carnot COP depends on the temperature lift; for an 80 C target, 40 C supply gives a theoretical COP of 8.8 and 50 C gives 11.8; a conservative space allowance is at least 10 m2 per MW of heat pump design capacity.
8. Heat to cooling: adsorption chillers start at 45 C and exceed COP 0.4 above 55 C; single-effect absorption starts at 65 C and can exceed COP 0.75 above 80 C; the most suitable hot-water range is 55-85 C.
9. ORC heat-to-power converts less than 4% of the heat to electricity at these low temperatures.
10. Reused waste heat (EREUSE, kWh) is measured per EN 50600-4-6, counting heat used outside the data centre boundary that displaces other energy; ERE = (1 - ERF) x PUE.

Using it in design

• Provide a heat-recovery-station block in the site layout with space sized from capacity (plate HX footprint, 10 m2 per MW for heat pumps, about 43 m3 per MW per 30 minutes of storage) and a demarcation boundary to the heat host.
• Have the heat-reuse calculator compute Carnot and realistic COP from data centre supply and host temperatures and recommend a boosting technology by temperature band.
• Meter EREUSE at the boundary per EN 50600-4-6 to compute ERF, subtracting any part used for cooling the data centre.
• Warn if converting N+1 dry coolers to a heat reuse station reduces heat-rejection redundancy.
• In new hydronic loops include taps and valved connections for future heat reuse even when it is not planned now; the paper notes this costs far less than retrofitting a running system.
• Consider splitting the heat exchanger into several units so some heat can still be delivered during maintenance, and put industry-standard fluid quality requirements in the contract with the host.

Who should read it

Mechanical engineers, architects and decision makers designing data centres that can export heat.

Document contents

1. Executive summary
2. Compliance with OCP Tenets
3. Introduction
4. Business considerations
5. Stakeholders
6. Data center heat reuse maturity model
7. Design considerations
8. General design of the heat recovery station
9. Components and connections of the heat recovery station
10. Fluid monitoring with heat host
11. Location of the heat recovery station
12. Expected power, network, cooling connections, security level in the heat reuse station
13. Measurement and reporting of excess heat
14. Expected redundancy/resiliency in the devices of the heat reuse room
15. Loops design, cascading
16. Data centers retrofits for heat reuse
17. Matching of the heating demand
18. Seasonality and intraday variations
19. Heat storage
20. Boosting or transformation mechanisms (heat pumps, boilers, heat to cooling, heat to power)
21. Metrics
22. Conclusion
23. Appendix I - Reference designs, block diagrams
24. Appendix II - Two-phase DLC heat reuse
25. Appendix III - Heat pump heat reuse additional considerations

Related standards

• EN 50600-4-6
• ISO/IEC 30134-6
• AHRI Standard 400
• OCP Data Centers Heat Reuse 101

Read the document

Publication details

Authors

Open Compute Project Foundation

Publisher

Open Compute Project Foundation

Year

2025

Document number

Rev 1.0

Document language

English

Source type

Industry body

Access

Open licence — downloadable

Licence

CC BY-SA 4.0 — share with attribution under the same licence

Licence evidence

www.opencompute.org

Citation