High-order gas-kinetic scheme for numerical simulations of wind turbine with nacelle and tower using ALM and IBM

This paper presents a novel high-order gas-kinetic scheme integrated with the actuator line model and immersed boundary method on GPUs to accurately simulate three-dimensional wind turbine flows, including nacelle and tower effects, while validating its ability to capture complex wake interactions and turbulent statistics against experimental data.

The Gist

Imagine you are trying to predict how a giant wind turbine will behave in a storm. You want to know exactly how much power it will generate, how much stress the blades will feel, and how the "wind shadow" (the wake) behind it will mess with other turbines nearby.

Doing this on a computer is like trying to simulate a hurricane in a bathtub. The air is moving fast, swirling in tiny, chaotic eddies, and the turbine has complex parts like a tower and a heavy engine housing (the nacelle).

This paper introduces a new, super-smart computer program to solve this problem. Here is the breakdown using simple analogies:

1. The Engine: The "High-Order Gas-Kinetic Scheme" (GKS)

Think of the air around the turbine not as a smooth fluid, but as a swarm of billions of tiny, invisible billiard balls bouncing around.

• Old Way (Low-Order): Imagine trying to predict the path of these balls by taking a blurry, low-resolution photo. You might see the general direction, but you miss the tiny, chaotic spins and collisions that create turbulence.
• New Way (High-Order GKS): This new method takes a 4K, high-definition video of every single ball. It tracks them with incredible precision. Because it sees the tiny details, it can predict exactly how the air will swirl, break apart, and mix. It's like upgrading from a sketch to a photorealistic movie.

2. The Problem: The Tower and the Blades

Wind turbines have three main parts:

1. The Blades: They spin fast and are long and thin.
2. The Nacelle: The heavy box at the top holding the gears.
3. The Tower: The pole holding it all up.

Simulating these is tricky because they are shaped differently.

• The Blades (The Actuator Line Model): Instead of trying to draw every tiny curve of the spinning blades (which would take forever), the computer treats the blades like a "magic line" of force. It's like drawing a glowing wire in the air that pushes the wind just like a real blade would.
• The Tower & Nacelle (The Immersed Boundary Method): These are solid, blocky objects. The computer uses a "ghost wall" technique. It places invisible dots all over the surface of the tower and nacelle. When the wind hits these dots, the computer says, "Stop! You can't go through here," and pushes the air around them.

3. The Superpower: GPUs (Graphics Cards)

Simulating this level of detail requires a massive amount of brainpower.

• The Analogy: If one computer processor is like a single student trying to solve a math problem, this paper uses GPUs (the chips in gaming computers) as a whole army of students working together.
• They split the sky into thousands of tiny pieces and assign each piece to a different "student" (GPU). They talk to each other instantly, allowing the simulation to run fast enough to be useful for real engineering.

4. What They Tested

The team ran three types of tests to prove their new method works:

1. The Smooth River (Turbulent Channel Flow): They simulated wind flowing between two walls to make sure their "high-definition" math was accurate. It passed with flying colors.
2. The Stick in the Water (Circular Cylinder): They simulated wind blowing past a round pole (like a tower). They checked if the "vortex street" (the swirling wake behind the pole) looked real. It did.
3. The Real Deal (Wind Turbines):
   • Test A (NREL 5MW): They simulated a massive, standard wind turbine. They found that their "high-definition" method could see the tiny, chaotic swirls in the wind shadow much better than older, "blurry" methods. This is crucial because those tiny swirls determine how much energy the next turbine in line will get.
   • Test B (NTNU "Blind Test"): This was the ultimate challenge. They simulated a turbine with a tower and nacelle included.
     • The Discovery: When the blade spins past the tower, the wind gets messed up. The tower creates its own little whirlpool. When the blade's own whirlpool hits the tower's whirlpool, they crash into each other, creating turbulence much earlier than expected.
     • The Result: The new method predicted that the power output would wiggle up and down slightly every time a blade passed the tower. Older methods missed this because they ignored the tower or used low-resolution math.

The Bottom Line

This paper is about building a super-accurate, high-speed wind tunnel inside a computer.

By combining a high-definition math engine (High-Order GKS) with smart tricks for blades (ALM) and towers (IBM), and running it on a super-fast army of graphics cards, the researchers can now predict exactly how wind turbines will behave in the real world. This helps engineers design better, more efficient wind farms that can survive storms and generate more clean energy.

Technical Summary

Here is a detailed technical summary of the paper "High-order gas-kinetic scheme for numerical simulations of wind turbine with nacelle and tower using ALM and IBM."

1. Problem Statement

The increasing scale of modern wind turbines (exceeding 200m in diameter) presents significant challenges for Computational Fluid Dynamics (CFD). These large rotors operate at high Reynolds numbers ($\sim 10^7$), generating complex, multi-scale turbulent wakes.

• Limitations of Current Methods: Traditional Actuator Line Models (ALM) often rely on second-order numerical schemes, which may lack the resolution required to accurately capture wake turbulence dissipation and vortex breakdown at high Reynolds numbers.
• Geometric Complexity: Most simulations simplify the turbine by omitting the nacelle and tower. However, these components significantly alter the flow field, causing local velocity deceleration, blade-tower interactions, and asymmetric wake structures, which are critical for offshore floating turbines.
• Computational Cost: Resolving these multi-scale turbulent structures with high fidelity requires computationally expensive simulations, necessitating efficient parallelization strategies.

2. Methodology

The authors propose a unified framework integrating three advanced techniques:

A. High-Order Gas-Kinetic Scheme (GKS)

• Extension: The authors extended the GKS, based on the BGK model, to simulate 3D weakly compressible isothermal flows.
• Accuracy: The scheme utilizes a two-stage fourth-order temporal discretization and a fifth-order Weighted Essentially Non-Oscillatory (WENO) spatial reconstruction.
• Physics: It captures the gas evolution process from the kinetic scale to the hydrodynamic scale, calculating both inviscid and viscous fluxes within a single framework.

B. Actuator Line Model (ALM)

• Implementation: Rotor blades are represented by a spanwise distribution of actuator points rather than resolved geometry.
• Force Calculation: Forces are calculated based on 2D airfoil theory using sampled velocities (via point-based or line-based sampling) and corrected for tip effects.
• Integration: The calculated forces are distributed to the fluid grid as a body force using a Gaussian kernel with a smearing width ($\epsilon$).

C. Immersed Boundary Method (IBM)

• Implementation: The nacelle and tower are modeled using a set of Lagrangian points on a Cartesian Eulerian mesh.
• Force Coupling: An external body force is imposed at the Lagrangian points to enforce no-slip boundary conditions. The force is re-distributed to surrounding Eulerian cells.
• Integration: Both ALM and IBM forces are added as a single external body force term ($f = f_{ALM} + f_{IBM}$) in the momentum equation of the GKS.

D. High-Performance Computing (HPC)

• The solver is implemented on Graphics Processing Units (GPUs) using CUDA and MPI.
• Parallelization: A slab decomposition strategy is used to minimize GPU-GPU communication frequency. Communication across nodes is handled via RoCE (RDMA over Converged Ethernet).

3. Key Contributions

1. First Integration: This is the first study to integrate ALM and IBM into a high-order GKS framework for wind turbine simulations.
2. Framework Extension: Successfully extended the high-order GKS from compressible flows to weakly compressible isothermal flows suitable for low-Mach number wind turbine applications.
3. GPU Acceleration: Developed a multiple-GPU enabled solver capable of handling large-scale turbulent wake simulations efficiently.
4. Comprehensive Validation: Validated the method against:
   • Turbulent channel flow (Re$\tau$=180).
   • Flow past a circular cylinder (Re=3900).
   • NREL 5 MW reference turbine (ALM only).
   • NTNU "Blind Test 1" turbine (ALM + IBM for nacelle/tower).

4. Key Results

A. Numerical Accuracy and Order of Scheme

• Validation: The high-order GKS accurately resolved turbulent channel flow and cylinder wake metrics (velocity profiles, Reynolds stresses) compared to DNS and LES reference data.
• Order Comparison: In NREL 5 MW simulations, the high-order GKS resolved vortex-dominated wake turbulence more effectively than the second-order GKS, even with the same grid resolution.
• Grid Sensitivity: While second-order schemes required locally refined grids to approach the accuracy of the high-order scheme, the high-order scheme with uniform grids provided superior resolution of transitional wake turbulence and tip vortex breakdown.

B. Smearing Kernel Width ($\epsilon$)

• The study investigated the effect of the Gaussian smearing width. A value of $\epsilon = 3\Delta x$ was identified as optimal, providing results that agreed well with reference ALM and blade-resolved solutions.

C. Impact of Nacelle and Tower (NTNU Case)

• Periodic Coefficients: Including the tower (via IBM) introduced periodic fluctuations in power ($C_P$) and thrust ($C_T$) coefficients due to blade-tower interactions, whereas simulations without the tower yielded steady coefficients.
• Wake Asymmetry: The tower vortex interacts with the tip vortex, causing an earlier transition to turbulence and an asymmetric wake distribution (velocity deficit and TKE are higher on the side where the blade rotates toward the tower).
• Quantitative Agreement: The time-averaged streamwise velocity and Turbulent Kinetic Energy (TKE) profiles predicted by the high-order GKS with IBM showed excellent agreement with NTNU experimental data.

5. Significance

• High Fidelity: The study demonstrates that high-order schemes are essential for accurately capturing the complex, multi-scale physics of wind turbine wakes, particularly the breakdown of tip vortices and the interaction with support structures.
• Realistic Modeling: By successfully integrating IBM, the method enables realistic simulations of full turbine configurations (blades + nacelle + tower), which is crucial for predicting loads and wake interactions in wind farms, especially for offshore floating platforms.
• Efficiency: The GPU-accelerated implementation makes these high-fidelity simulations computationally feasible for large-scale engineering applications.
• Future Outlook: This framework provides a robust foundation for future studies involving realistic atmospheric inflows, shear profiles, and complex wind farm layouts.