⚡ electrical engineering

Airflow and Energy Optimization in the turbine-aided data center using Entransy approaches

This study numerically investigates a turbine-aided cooling solution for data centers using Computational Fluid Dynamics and Entransy-based analysis to optimize airflow distribution, enhance energy efficiency, and improve heat dissipation through the strategic manipulation of impinging turbulent flows and rotor positioning.

Original authors: Zhihang Song, Qiang Yue

Published 2026-10-01

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Zhihang Song, Qiang Yue

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Data centers are the vast, humming warehouses of the modern world, housing the servers that power our digital lives. Inside these rooms, thousands of computers generate immense heat, requiring a constant, powerful flow of cool air to prevent them from melting down. The standard method involves raising the floor to create a hidden plenum, a hollow space where air conditioning units push cold air upward through perforated tiles. This air travels into the cold aisles between rows of server racks, absorbs the heat, and returns to the top of the room to be cooled again. However, this system is often inefficient. The air sometimes moves too fast, crashing into the floor and creating turbulent, wasted energy, or it fails to reach the hottest spots, leaving servers to overheat while other areas remain unnecessarily cold. The challenge for engineers is to tame this chaotic flow, ensuring the cooling air moves exactly where it is needed without wasting power.

A team of researchers set out to solve this problem by introducing a new element into the underfloor space: a small wind turbine. Rather than viewing the rushing air as a problem to be fought, they treated it as a resource to be harnessed. They built a detailed computer simulation of a typical data center room, measuring 8.75 meters long and 6.4 meters wide, filled with two rows of server racks. In their model, they placed a turbine rotor in the path of the cooling air as it descended from the CRAC unit's bottom into the underfloor plenum. The goal was to see if this turbine could do two things at once: slow down the air just enough to stop it from crashing into the floor and wasting energy, and simultaneously capture some of that moving air's power to generate electricity. They tested different positions for the turbine and adjusted how much resistance the turbine offered to the airflow, essentially tuning how much the air had to push against the blades.

The results of their simulations revealed that the turbine acts as a gentle regulator for the air. Without the turbine, the high-speed air from the cooling unit would slam into the floor, creating a chaotic splash of pressure that wasted momentum. When the turbine was added, it absorbed some of that force, smoothing out the flow and reducing the pressure hitting the floor. This was particularly effective when the turbine was positioned higher up in the underfloor space and when it was set to offer a moderate amount of resistance to the air. The researchers found that this arrangement not only saved energy by reducing turbulence but also allowed the turbine to harvest a small amount of power from the airflow itself. The most efficient setup for generating power occurred when the turbine was placed 0.73 meters from the slab and the airflow resistance was tuned to a specific level, creating a sweet spot where the air moved steadily through the blades.

Beyond just saving energy, the turbine improved the actual cooling of the servers. In the simulation, the presence of the turbine helped distribute the cold air more evenly across the floor tiles. This led to a more uniform temperature in the aisles, reducing the amount of hot air that would otherwise get sucked back into the front of the servers. The team observed a temperature drop of about 2 degrees Celsius at the server inlets compared to a room without the turbine. This means the servers could run cooler and more efficiently, potentially extending their lifespan and reducing the load on the main air conditioning units. The researchers used a specific method of analysis, known as the Entransy approach, to measure how well the heat was moving through the system. This method helped them quantify the improvements in airflow organization, confirming that the turbine helped organize the chaotic air into a more orderly and effective stream.

To ensure their findings were robust, the researchers used a statistical technique called Response Surface Methodology. This allowed them to map out the relationship between the turbine's position, its resistance to the air, and the overall performance of the cooling system. They discovered that the best results came from a specific combination of factors: a moderate level of air resistance and a turbine height of about 0.21 meters from the slab for optimal cooling efficiency, or 0.73 meters for maximum power generation. The computer model predicted these outcomes with high precision, showing a very small margin of error when compared to their detailed calculations. The study suggests that adding a simple turbine to the underfloor space of a data center is a viable way to improve both energy efficiency and cooling performance. It turns a wasted force—the rushing air—into a useful tool, offering a new way to manage the complex thermal environment of modern computing facilities.

Technical Summary: Airflow and Energy Optimization in the Turbine-Aided Data Center Using Entransy Approaches

Problem Statement
Modern data centers face significant challenges regarding the maldistribution of cooling turbulence and associated energy wastage. Specifically, the over-provisioned cooling mass flow from Computer Room Air Conditioning (CRAC) units often leads to inefficient flow boundary conditions and unpredictable heat flow transportation at the rack level. While raised-floor configurations are standard, they frequently suffer from hot spots and energy losses due to poor airflow organization. Existing evaluation metrics, such as Power Usage Efficiency (PUE) or heat indices (RHI/SHI), often provide qualitative assessments of bypass and recirculation but lack the quantitative physical magnitude required to optimize complex, multivariate cooling physics. Furthermore, the potential of harvesting energy from the over-provisioned cooling flow itself remains under-explored in the context of data center underfloor environments.

Methodology
The study employs a multi-faceted numerical approach to investigate a newly proposed "turbine-aided" cooling infrastructure within a single-CRAC, raised-floor data center environment.

Computational Fluid Dynamics (CFD): The physical domain (8.75m × 6.4m × 5.15m) containing two rows of server racks and a 0.914m high underfloor plenum was modeled using FLOTHERM software. The simulations utilized the LVEL K-Epsilon Turbulence model with Reynolds-averaged Navier-Stokes (RANS) calculations. A grid independence study confirmed that a mesh of 356,402 cells provided a balance between precision and efficiency.
Turbine-Aided Modeling: A compact turbine model was introduced in the underfloor plenum to interact with the downward CRAC flow. The turbine was characterized by a porous-jump boundary condition representing rotor resistance. Key parameters included the axial induction factor ( $a$ ) and the turbine-to-slab distance ( $h$ ). The model calculated power harvest ( $P_t$ ), thrust coefficients ( $C_T$ ), and pressure drops based on actuator disc theory and Glauert curves.
Entransy-Based Analysis: To evaluate heat transferability and energy dissipation more rigorously than traditional indices, the study applied the Entransy approach. This involved calculating Entransy dissipation at the room level ( $\Delta J_{E,rm}$ ), rack level ( $\Delta J_{E,rk}$ ), and specifically the "under-organized" loss ( $\Delta J_{E,u0}$ ), which quantifies the inefficiency caused by cold/hot aisle air mixing and non-uniform flow.
Response Surface Methodology (RSM): A Central Composite Design (CCD) was implemented to optimize the interaction between the axial induction factor ( $a$ ) and the rotor positioning ( $h$ ). A second-order polynomial model was fitted to the Entransy loss data to identify optimal operating conditions.

Key Results

Flow Impingement Reduction: The introduction of the turbine-aided structure effectively reduced the impinging flow pressure upon the plenum slab boundary. Increasing the axial induction factor ( $a$ ) and adjusting the rotor height ( $h$ ) significantly attenuated the momentum of the CRAC-provisioned mass flow, particularly in the centerline region.
Energy Harvesting: The turbine-harvested power was found to be dependent on both the axial induction factor and the rotor position. Maximum power was achieved when $0.6 < a < 0.7$ with a rotor height of $h=0.73$ m. However, at very high induction factors ( $a=0.8$ ) and large distances, the flow tended to pass by the rotor rather than through it due to an impediment process over the rotor's resistance surface, which reduced energy harvest.
Thermal Performance: In the optimal configuration ( $a=0.5, h=0.55$ m), the turbine-aided system demonstrated a 2°C reduction in server inlet temperatures compared to the baseline without the turbine. This improvement was attributed to more uniform z-directional tile flow momentum and a reduction in hot air recirculation at the rack tops.
Optimization Accuracy: The Entransy-based RSM optimization identified a minimum Entransy loss ( $\Delta J_{E,u0}$ ) of 136,523 W·K at $h=0.21$ m and $a=0.5$ . This result showed a high degree of agreement (within 0.6% relative error) with direct CFD calculations, validating the RSM approach for this application.

Significance and Claims
The paper claims that the proposed turbine-aided infrastructure offers a dual benefit: it improves the energy efficiency of the data center by harvesting momentum from the cooling flow and enhances thermal management by organizing the airflow more effectively. By treating the CRAC-provisioned mass flow axial induction and rotor resistance positioning as primary factorial components, the study reveals that these interactions can significantly reduce flow momentum waste at the plenum bottom.

The authors emphasize the value of the Entransy-based approach as a more robust tool for evaluating heat flow transportation compared to empirical indices, as it provides a quantitative measure of "under-organized" energy losses. The successful integration of Entransy theory with Response Surface Methodology is presented as a valid scientific method for visualizing and enhancing the interdependencies between multivariate heat and mass transfer mechanisms. Ultimately, the research suggests that turbine-aided solutions, when optimized via these methods, can offer creative insights for future advanced cooling and energy enhancements in data centers.

More like this