Data Centers as Prosumers: Waste Heat Recovery for District Heating

Welcome to issue 4 of our heatbeat Research Newsletter. After last month's review of the existing 5th generation district heating and cooling networks in Europe, this issue is dedicated to the role of data centers as prosumers in district energy systems. The selected paper reviews data centers as prosumers with a focus on both integrating renewable energy sources and capturing data centers' waste heat for district heating. For this newsletter, we focus mainly on the potentials of data centers as sources of waste heat in district heating networks, but also recommend the other aspects of the original paper.

The Short version

In this issue, we present our take on the paper "A review of data centers as prosumers in district energy systems: Renewable energy integration and waste heat reuse for district heating" by Pei Huang et al. The paper is a collaboration between the Department of Energy and Built Environment at Dalarna University and EcoDataCenter in Sweden.

The paper aims at promoting a view of global optimization for data center operations that considers all the involved energy flows including both the data center's energy supply and the utilization of the resulting waste heat. This holistic approach understands data centers as prosumers with high electricity consumption, but also producing large quantities of waste heat while cooling the IT facilities with a potential of utilization for heating applications.

To support this holistic take on data centers as prosumers, the authors review data center cooling technologies, the integration of renewable energy sources in the electricity supply, advanced control strategies, and options to recover the produced waste heat for utilization in district heating networks. They find that simple technical solutions for using data center waste heat in district heating have so far proven to be the most economic options. Furthermore, they conclude that there is still a lack of global optimization of operating data centers as energy prosumers.

Our full summary

The paper "A review of data centers as prosumers in district energy systems: Renewable energy integration and waste heat reuse for district heating" by Pei Huang et al. investigates the current state of the art regarding the role of data centers as prosumers in district energy systems. The paper is a collaboration between the Department of Energy and Built Environment at Dalarna University and EcoDataCenter in Sweden.

The authors illustrate the general importance of data centers in energy systems by citing that data centers consume a share of 3 % of the global electricity supply while accounting for 4 % of global greenhouse gas emissions. And for the future, these numbers are expected to increase significantly. As this has large implications both from an economic and an ecological perspective, the authors draw from a large body of research regarding optimization of energy flows to and from data centers for their review.

As a foundation for their review, the authors start by giving a comprehensive overview of data centers and available cooling technologies. They show that different types of computing equipment within the data center have different operating temperatures and lead to different heat densities. For the temperatures, they distinguish between standard servers and high-performance computing (HPC) servers. For the heat densities, the paper shows that newer data centers have significantly increased their heat dissipation rates to up to 10 kW/m². For a safe operation of the data center, this heat needs to be removed for the servers to operate within their recommended ranges of ambient temperature and humidity.

The cooling technologies available for removing heat from the data center are grouped into 3 categories:

• Air-cooled systems: The most common cooling technology where air flows through the server racks to transfer heat out of the data center
• Water-cooled systems: Liquid coolant is in direct contact with the computing equipment and can remove heat at higher temperatures
• Two-phase cooled systems: Liquid coolant in contact with computing equipment evaporates with even higher cooling performance by utilizing latent heat

For the most common air-cooled systems, the paper identifies the server rack exhausts as the optimal location to capture waste heat in order to reach temperatures of 35 - 45 °C. To quantify how much of the data center's energy consumption can be recovered as waste heat, the paper mentions the Energy Reuse Factor defined as the reused energy divided by the data center's electricity consumption. As reference values for the Energy Reuse Factor, the paper cites values of 25 - 45 % when heat is recovered from the air return, 55 % when recovered within the cooling system's chiller and estimations identifying a potential of up to 68 %. When more innovative cooling options are used, the paper mentions that water-cooled systems can reach temperatures of around 60 °C and two-phase cooling systems even up to 80 °C.

Regarding the control strategies for data center operation, the paper shows that moving server loads to times of high availability of renewables can reduce greenhouse gas emissions. Yet, the paper also describes that currently, such control concepts are mainly focused on the supply side without optimizing for waste heat potentials, in part because common metrics to measure data center efficiency do not consider this from a global perspective of the data center as prosumer.

For the integration of the recovered waste heat into district heating, the paper suggests different options involving a central heat pump at the data center to raise the temperature of the waste heat to be fed into the network while simultaneously cooling the data center. In order to decouple the waste heat supply and the network demand, the authors suggest adding a second ground sourced heat pump which can help to shift and store heating and cooling energy with the help of 1 or 2 borefields. While adding the borefields opens up storage capacity and thus can increase waste heat utilization, the authors also mention the increased costs of such a system, so that the simpler direct connection without borefields is identified as a more economical solution.

All these options evaluated in the paper are focused on integrating the waste heat into a conventional district heating network at relatively high temperatures, requiring a central heat pump. In contrast, low temperature networks and the 5th generation district heating and cooling networks as discussed in our previous newsletter issue 3 open up further interesting potentials to integrate data center waste heat directly without a central heat pump.

Of course, for such applications the trade-offs between operating one central heat pump with the network at higher temperatures and decentralized heat pumps in a low temperature network need to be evaluated carefully. At heatbeat, we see such evaluations as promising applications for our simulation methods to compare different system setups for the optimal utilization of potential waste heat sources like data centers in district heating to make best use of their role as prosumers in the energy system.

Further Information

The original article can be found at https://doi.org/10.1016/j.apenergy.2019.114109 . Within the scope of our newsletter, we were only able to focus on the aspects directly related to district heating. Yet, the paper includes many more interesting details on the integration of renewables and advanced control strategies for data centers. Therefore, we highly recommend the paper in its entirety.

In addition, the research group in Dalarna University is now developing a simulation platform, based on agent based modelling and Geographic Information System (GIS) technique, to exam the techno-economic performance of the 5GDHC technology and explore the feasible applications in Baltic-Nordic regions. It will adopt an integrated multi-disciplinary approach by combining technological database development, digital-GIS resource mapping, techno-economic simulation, business model optimization and policy-orientated feasibility study for dissemination/replication. Such research is funded by Nordic Energy Research. For more information on this project visit https://www.du.se/en/research/research-projects2/?code=HDA2021-00001 .

The next issue of our newsletter will be released on March 3, 2021.