Background

General Background

With the rapid advancement of mobile communication and artificial intelligence technologies, data centers have become critical infrastructure for data processing and are now among the primary drivers of global electricity demand growth. In 2022, the total electricity consumption of data centers worldwide reached 460 TWh, accounting for approximately 2% of global electricity demand, and this figure is projected to exceed 800 TWh by 2026. Given their substantial energy consumption and growing share of total energy use, improving the energy efficiency of data centers is of paramount importance, particularly under the prerequisite of ensuring operational safety and reliability.

From an internal perspective, the major energy-consuming components of a data center include IT (Information Technology) servers, cooling systems, power distribution systems, and auxiliary supporting equipment. Power Usage Effectiveness (PUE), defined as the ratio of the total energy consumption of a data center to that of IT equipment, is the most widely adopted metric for evaluating data center energy efficiency. A PUE value closer to unity indicates higher overall efficiency. Since auxiliary systems primarily function to provide a stable and secure operating environment for IT equipment, PUE is largely used to assess the performance of cooling and power distribution systems.

Among these, improving the energy efficiency of cooling systems represents the most effective and impactful approach to reducing PUE, offering significant potential for energy savings. Current research on data center heat rejection processes can be broadly categorized into three main directions. First, enhancing the performance of internal heat transfer processes, such as improving chip-level heat dissipation, so that the heat rejection process more closely approaches the operating temperature of the chips. Second, utilizing lower-temperature heat sinks to increase the use of natural cooling sources, for example, by employing indirect evaporative cooling technologies to achieve cooling below the ambient wet-bulb temperature. Third, recovering and utilizing waste heat from data centers for applications such as heating, cooling, and power generation; although this approach does not directly reduce the energy consumption of the heat rejection process, it improves the overall energy utilization efficiency of the data center.

Current Situations of Data Center Heat Exhaust Technologies

Chip Cooling Technologies

According to the type of cooling medium in direct contact with the chip, cooling strategies are generally classified into air cooling and liquid cooling. Air cooling remains the most mature, widely adopted, and cost-effective solution in current data center applications.

In recent years, the rapid development of technologies such as generative artificial intelligence, 5G, and video-based social platforms has led to a substantial increase in the integration density of electronic components within chips. Consequently, the heat flux density of chips has risen significantly. Under conditions where the effective heat dissipation area and allowable operating temperature of chips remain largely unchanged, the increasing heat flux density necessitates a substantial enhancement of the convective heat transfer coefficient between the chip and the cooling medium. In this context, conventional air cooling is no longer sufficient to meet the thermal management requirements of high-performance chips.

Liquid cooling, characterized by significantly higher convective heat transfer coefficients between the coolant and the chip, is capable of accommodating much higher heat flux densities. As a result, it has become the most extensively studied and rapidly developing chip cooling technology over the past decade. Liquid cooling technologies can be further categorized into cold plate cooling, microchannel cooling, embedded cooling, immersion cooling, pool boiling, jet impingement, spray cooling, and heat pipe cooling. Current research on liquid cooling primarily focuses on the optimization of cooling media and heat transfer structures.

Low-Temperature Cooling Source Technologies

The natural environment provides abundant cooling resources, and their effective utilization is one of the primary approaches to improving the energy efficiency of data center cooling systems. Among these, ambient air is the most readily accessible natural cooling source; however, the achievable cooling temperature varies depending on the type of air-side cooling technology employed.

When air is used directly for sensible heat exchange, the lowest attainable cooling temperature is limited by the dry-bulb temperature. In contrast, evaporative cooling technologies can further reduce the cooling temperature to the wet-bulb temperature, or even approach the dew-point temperature under certain conditions. Based on the type of cooling medium utilized, evaporative cooling technologies can be broadly classified into air-side and water-side systems.

Waste Heat Recovery Technologies

The recovery and utilization of waste heat generated by data centers can significantly improve overall energy utilization efficiency and has become a major research focus in recent years. Extensive efforts have been made to explore various pathways for data center waste heat utilization, including applications in district heating and domestic hot water supply, auxiliary power generation, absorption and adsorption cooling, organic Rankine cycles (ORC), thermoelectric conversion, biomass conversion, as well as seawater desalination and wastewater treatment.

Among these approaches, the direct use of data center waste heat for district heating or hot water supply is widely regarded as the most economically viable and practically promising option. Data center waste heat recovery systems for heating applications not only enhance energy utilization efficiency but also contribute to clean heating solutions, demonstrating considerable application prospects and development potential. However, from a technical perspective, the temporal mismatch between data center heat rejection and end-user heat demand remains a key constraint limiting the effective utilization of waste heat.

Importance and Emergency

Despite substantial progress, further research on data center cooling and waste heat recovery is still required for several reasons.

1. Lack of unified target parameters for liquid cooling and air cooling system.
2. Lack of unified system analysis structure and method of Data center cooling systems.
3. Lack of real cases investigation of Data Center Cooling Systems.
4. Lack of understanding of the performance characteristics of systems utilizing different cooling sources.
5. The scope of application and suitable conditions for existing evaporative cooling technologies remain unclear.
6. Waste heat recovery from data centers continues to face several challenges, including limited application scale, low utilization efficiency, high initial investment, and long payback periods.

Thus, in light of the rapid growth in data center energy consumption and the urgent need for improved energy efficiency, several key issues must be systematically investigated and ultimately addressed.

Key Problems Needed to be Solved

• Establishment of unified performance evaluation metrics: it is necessary to develop a comprehensive set of evaluation indicators capable of systematically assessing both data center cooling performance and waste heat recovery effectiveness. These metrics should encompass Power Usage Effectiveness (PUE), Water Usage Effectiveness (WUE), inlet air or liquid temperature, target control temperatures, as well as integrated optimization of chip computational power consumption and cooling energy efficiency.
• Development of a real-world data center cooling system case study dataset: a comprehensive dataset should be established based on statistical analysis of operational data from existing data centers. This effort aims to identify key challenges in cooling system design, airflow management, and chip-level liquid cooling. To ensure consistency and reliability, a well-defined survey framework and data collection protocol are required.
• Determination of the applicability and operating conditions of evaporative cooling technologies: the performance of evaporative cooling technologies, such as direct evaporative cooling (DEC) and indirect evaporative cooling (IEC), varies under different climatic conditions in terms of supply water temperature, fan energy consumption, and water usage. Therefore, their applicability ranges and operational constraints in data center cooling systems across different regions require systematic qualitative and quantitative investigation.
• Research on liquid cooling technologies: at present, liquid cooling represents the primary solution for thermal management of high heat flux chips and is also the most rapidly developing cooling technology. Further research is required in areas such as the optimization of heat transfer performance in cold plate liquid cooling systems, the distribution and flow dynamics of coolant, and issues related to coolant-induced corrosion and material compatibility.
• Research on data center waste heat district heating: District heating based on data center waste heat is considered the most economically viable and promising utilization pathway. However, its current deployment remains limited in scale, with low waste heat utilization efficiency, high capital investment, and long payback periods. This technology faces three fundamental challenges: a mismatch between waste heat temperature levels and user heat demand temperatures, a temporal mismatch in heating demand, and a spatial mismatch in geographical distribution. Electric heat pumps and thermal energy storage systems are essential components of data center waste heat heating systems. Heat pumps can upgrade the temperature level (quality) of waste heat, while thermal storage systems enable load shifting and peak shaving, thereby improving overall waste heat utilization efficiency. Large-scale and long-distance waste heat recovery for district heating represents a key future development direction. However, further research is still required in system configuration and the optimization of operational parameters.

Objectives

Thus, the proposed Annex project intends solve the above key problems, the objectives are:

1. Data Center cooling system: Set up the unified target parameters for liquid cooling and air cooling system, including temperature parameters and electricity consumption indicators; Develop a unified system analysis structure and method, as to choose suitable cooling system for different systems.
2. Liquid cooling system: Cooperation research on optimization Study on the heat transfer performance of cold-plate liquid cooling heat exchangers; Discussion on issues in coolant distribution and flow dynamics; Give common methods for challenges related to coolant corrosion.
3. Waste heat Recovery: Research on heat pump methods to realize waste heat recovery; Combined with heat storage technologies for peak shaving in heat supply; Discuss the combination of data center cooling and waste heat recovery, to ensure the safely running of heat exhaust, then for heating system using the waste heat.
4. Application and water consumption of Evaporative cooling technologies: Research on air-side indirect evaporative cooling and water-side direct or indirect evaporative cooling for data center cooling system; Research on the scope of application and suitable conditions for existing evaporative cooling technologies.
5. Computing and Power Synergy for data centers: investigate computing resource allocation and scheduling strategies for improving energy efficiency in data centers; study the relationship between cooling system temperature levels and chip energy consumption.
6. Non-technical challenges, governance, policy recommendations: analyze regulatory frameworks, planning regulations, economic feasibility, and stakeholder responsibilities related to data center cooling and waste heat utilization; develop practical guidelines for practitioners, policymakers, municipal authorities, district heating companies, data center operators, and technology suppliers.