| 17 Sep 2025
The Lattice Boltzmann Method for Wind Farm Simulations: A Review
Abstract. High fidelity numerical simulations are currently a key method to gain insight into flows in very large wind farms. However, these simulations are extremely costly in terms of computational resources and the progress in computational efficiency has been outrun by the growing size of wind farms and the need for simulations. The adoption of massively parallel hardware, namely graphics processing units (GPUs), by the wind energy community has begun but the numerical structure of the Navier-Stokes equations hinders an efficient use of such hardware. That is one of the reasons, the lattice Boltzmann method has gained increasing attention in recent years. By construction, this method for simulating the Navier-Stokes equations is perfectly parallelizable and well suited for massively parallel hardware. However, as with every new method, the foundation of the method is not widely known by the wind energy community and often met with doubt. This review paper collects the various methods necessary for a potential GPU-resident wind farm flow solver based on the lattice Boltzmann method. Furthermore, it discusses various aspects of the application of the lattice Boltzmann method to wind farm flows and related flow regimes. It also identifies gaps in the current literature and aims to direct future research on the lattice Boltzmann method for wind farm flows.
The paper gives a review on the usage of the lattice Boltzmann method in wind farm simulation. Instead of presenting data, the various aspects of wind farm simulation, large scale LES, boundary layer functions, actuator line-models, stratified atmospheric boundary layers are discussed in textual form. Actual data is only shown for performance with different GPUs.
I belief that the paper fulfills the standards of this forum and can be published with minor revisions, albeit more quantitative discussion based on detailed data would certainly be preferable. The discussions also appear to be skewed towards a specific lattice Boltzmann model and it is not entirely clear whether this is due to the popularity of the cumulant model in the wind engineering field in general or the popularity of that models with the authors in particular.
Some specific points listed below require some attention:
l. 82: “paragraph follows the excellent description by Kruger et al.”, not saying Kruger is not excellent but scientific publications profit from avoiding judgements of this kind.
Line 141 mention Strang splitting and/or integration along characteristics to explain 2nd order convergence.
l. 144 “First, each velocity set leads to a speed of sound” how does the velocity set lead to a speed of sound?
l. 210 depending on the geometry of the boundary, the system of equations might either be over- or under-specified.
l.219 bounce forward only works for straight boundaries
l. 239 sponge layers do not primarily dampen waves. They eliminate eddies that would CAUSE waves when they interact with pressure boundary conditions. Non-reflective BCs are not an alternative to sponge layers or vice versa as both act on different problems.
2.2 Simulating wind turbines and farms in the LBM -> confusing-> Simulating wind turbines and wind farms in the LBM
l. 660: why does the simulation of urban flows reduce the need for wall models?
l.739 method method
l. 893: actuator disk has not yet been applied: An actuator disk was used in the Ph.D. thesis of Schoenherr for a ship propeller.