Energy Efficiency Through Waste-Heat Recovery: Hybrid Data-Centre Cooling in District Heating Applications

Overview

Growing demand for computing resources increases electricity use and cooling needs in data centres, while creating decarbonisation opportunities through integrating waste heat into district heating (DH). The study investigates using waste heat from the hybrid cooling system in third (3G), fourth (4G) and fifth (5G) generation DH systems with TRNSYS dynamic simulations. The control logic assigns 70% of the thermal load to DLC and 30% to CRCC; free cooling engages for DLC at outdoor temperature at or below 25 °C and for CRCC at or below 10 °C. For 3G DH a high-temperature water-to-water heat pump is needed.

For 5G and 4G DH integration, PUE stays below 1.6 and ERF above 0.85, while 3G DH requires a higher temperature lift and gives PUE of about 2.1 with ERF around 0.78. Primary energy at the data-centre boundary falls by about 88% (5G DH), 86% (4G DH) and 68% (3G DH) relative to the reference system, with similar CO2 reductions. Under Croatian conditions 3G DH needs high-temperature heat pumps to reach a 75/60 °C winter regime and a 65/50 °C summer DHW regime.

With a heat-purchase price 10% below the variable DH tariff, payback is under one year for 5G and 4G DH but about 14 years for 3G DH. Compared with the existing hybrid cooling system without heat recovery, useful heat delivered to the DH network increases by approximately 25% (5G DH), 35% (4G DH) and 62% (3G DH).

The conclusion is that low-temperature integration into 5G and 4G DH, directly through heat exchangers, achieves the most favourable combination of data-centre indicators and system performance, enabling extensive use of free cooling and reduced chiller operation. Limits: a single real HPC facility under Croatian conditions; results do not transfer directly to other DH regimes and climates.

Key points

1. The hybrid cooling system has a DLC loop at 25-30 °C and a CRCC (chilled-water rack cooler) loop at 10-15 °C.
2. The control logic assigns 70% of the thermal load to DLC and 30% to CRCC; free cooling engages for DLC at outdoor temperature at or below 25 °C and for CRCC at or below 10 °C.
3. For 5G and 4G DH integration, PUE stays below 1.6 and ERF above 0.85, while 3G DH requires a higher temperature lift and gives PUE of about 2.1 with ERF around 0.78.
4. Primary energy at the data-centre boundary falls by about 88% (5G DH), 86% (4G DH) and 68% (3G DH) relative to the reference system, with similar CO2 reductions.
5. 3G DH under Croatian conditions needs high-temperature heat pumps to reach a 75/60 °C winter regime and a 65/50 °C summer DHW regime.
6. With a heat-purchase price 10% below the variable DH tariff, payback is under one year for 5G and 4G DH but about 14 years for 3G DH.

Using it in design

• When connecting waste heat to DH, fix the DH generation early: for 5G and 4G direct heat-exchanger connection suffices (PUE below 1.6, ERF above 0.85), while 3G needs a high-temperature heat pump (PUE about 2.1, payback about 14 years).
• In hybrid cooling, split the heat load between the DLC (25-30 °C) and chilled-water (10-15 °C) loops (70/30) and add the free-cooling thresholds (25 °C and 10 °C) to the control logic.
• Set the heat sale price relative to the DH tariff (for example 10% below) and show payback sensitivity.

Who should read it

Engineers and planners intending to connect the waste heat of HPC and high-density data centres to district heating.

Document contents

1. 1. Introduction
2. 2. Materials and Methods (2.1 System Configuration: existing hybrid cooling system, integrating DC waste heat into 5G, 4G and 3G DH; 2.2 Simulation, 2.3 Input Data and HPC Model Validation, 2.4 Instrumentation and Measurement Uncertainty, 2.5 Main Modelling Assumptions, 2.7 Energy Performance Calculation, 2.9 Calculation of Costs, 2.10 Regulatory Framework for WH Producers, 2.11 Operational Cost Components)
3. 3. Results and Discussion (3.1 Simulation Results, 3.2 PE Consumption and CO2 Emissions, 3.3 Cost Performance Results and Sensitivity Analysis, 3.4 Limitations and Future Directions)
4. 4. Conclusions

Related standards

• EN 50600-4-6 (ERF)
• ISO/IEC 30134-2 (PUE)

Read the document

Publication details

Authors

Damir Požgaj; Boris Delač; Branimir Pavković; Vedran Medica-Viola

Publisher

Applied Sciences (MDPI AG)

Year

2025

Volume

16

Issue

1

Pages

323

Document language

English

Source type

Scientific

Access

Open licence — downloadable

DOI

10.3390/app16010323

Licence

CC BY 4.0 — share with attribution

Licence evidence

www.mdpi.com

Citation