The research, published as a preprint under review at the journal Wind Energy Science by Manuel Ayala, Dennice F. Gayme, and Charles Meneveau of the Department of Mechanical Engineering at Johns Hopkins University, tackles a long-standing uncertainty in wind energy modeling. For years, scientists have simulated entire wind farms using actuator disk models, or ADMs, which represent each turbine not as a detailed machine with spinning blades but as a porous disk that extracts momentum from the passing air. The approach is computationally inexpensive, but it was unclear whether such a simplified representation could faithfully capture the fluctuations in power and thrust that real turbines experience as turbulent wind washes over them.

The alternative is the actuator line model, or ALM, which resolves each rotor blade individually as a lifting line moving through the simulated flow. ALMs capture blade-level physics, including the periodic loading that occurs each time a blade sweeps past the tower or through regions of uneven inflow. That fidelity comes at a steep price: resolving three blades on a rotating rotor demands far finer time steps and much greater computational effort than treating the rotor as a static disk. For a single turbine the extra cost is tolerable, but for a wind farm with dozens or hundreds of interacting turbines, where wakes from upstream machines batter the rotors of downstream ones, the expense can become prohibitive.

What was missing until now was a careful, statistical validation. Engineers knew that ADMs were cheaper, but nobody had systematically tested whether the power and thrust time series they produce carry the same temporal structure, spectral content, and fluctuation statistics as those from the more detailed ALM. The Johns Hopkins team set out to close that gap by comparing the two approaches directly within the same high-fidelity simulation framework, using data drawn from the Johns Hopkins Turbulence Databases wind farm simulations, known as JHTDB-Wind.

The comparison was designed to be demanding. Rather than testing the models under a single idealized wind condition, the researchers evaluated turbine-response statistics under three distinct atmospheric conditions sampled within a diurnal cycle, the 24-hour pattern of atmospheric change from stable nighttime conditions to a convective, turbulent daytime boundary layer. Atmospheric stability strongly shapes the turbulence that reaches a rotor, so a model that performs well across a full diurnal range is far more convincing than one validated only in steady, neutral conditions. The team reconstructed ADM power and thrust from disk-averaged velocity, then evaluated both models using time series, power spectra, and the distributions of fluctuations and increments, the latter capturing how sharply the output jumps over short intervals.

The verdict was strikingly positive. The actuator disk model reproduced the temporal, spectral, and statistical behavior of the actuator line model at timescales slower than the rotor frequency, meaning everything from the slow meandering of large turbulent eddies through the farm down to the gust-driven surges in output that grid operators care about. In other words, as long as the phenomenon of interest unfolds more slowly than the time it takes a blade to complete one revolution, the cheap disk representation delivers essentially the same fluctuation statistics as the expensive blade-resolving one. The blade-level detail that ALMs provide matters mainly at the fast, once-per-revolution timescale, which is precisely the range where the ADM is not intended to compete.

That finding has implications that ripple well beyond academic turbulence research. Large-eddy simulations of wind farms, in which the turbulent airflow around every turbine is resolved on a computational grid, are increasingly used to design farm layouts, optimize wake-steering control strategies, and forecast aggregate power output. If every turbine in such a simulation must be represented with a full blade-resolving ALM, the computational cost scales brutally with farm size. The new results give modelers a rigorous justification for swapping in ADMs without sacrificing the accuracy of the variability statistics that matter for grid integration and structural fatigue. A simulation that once demanded enormous supercomputer allocations can potentially be run faster, at larger scale, or with more ensemble members to capture rare events.

The study also speaks to a broader theme in computational physics: the value of knowing exactly what information a simplified model preserves and what it discards. Simplification is the lifeblood of large-scale simulation, but it is only useful when its limits are mapped. By quantifying where the ADM matches the ALM, in time series behavior, in spectral content, and in the probability distributions of fluctuations and increments, and where it does not, at the fastest rotor-period timescales, the Johns Hopkins team has drawn that map for one of the most widely used tools in wind energy. The work is deliberately framed as a brief communication, a compact result rather than an exhaustive treatise, but its conclusion is the kind that reshapes daily practice in a modeling community.

There are, of course, caveats that come with any preprint. The paper is currently under open review at Wind Energy Science, with the discussion period open until early November 2026, and its conclusions will be scrutinized by referees and the community before final publication. The validation rests on a specific simulation dataset and a particular set of atmospheric conditions within one diurnal cycle, so extending the result to other turbine designs, farm layouts, and climates will be a natural next step for the field. The authors themselves position the work as support for using ADMs as an efficient tool for predicting turbine-response variability in large wind farm simulations, a claim grounded in the statistics they examined rather than a blanket endorsement of disk models for every purpose.

Even so, the practical message is hard to overstate. As offshore wind arrays grow to hundreds of turbines and grids absorb ever-larger shares of variable renewable power, the ability to simulate fluctuation statistics affordably becomes a genuine infrastructure problem, not just an academic one. A modeling shortcut that has been trusted on intuition now rests on quantitative evidence, and that evidence suggests the humble actuator disk, a representation as old as wind turbine aerodynamics itself, remains one of the most powerful tools in the computational wind energy toolbox. For researchers planning the next generation of gigawatt-scale farm simulations, the cheapest option may also be the right one.

Subject of Research: Validation of actuator disk models against actuator line models for wind turbine power and thrust fluctuation statistics in large wind farm simulations

Article Title: Brief Communication: Actuator Disk Models reproduce Actuator Line Model power and thrust fluctuation statistics in wind farm simulations

Article References: Ayala, M., Gayme, D. F., & Meneveau, C. (2026). Brief Communication: Actuator Disk Models reproduce Actuator Line Model power and thrust fluctuation statistics in wind farm simulations. https://doi.org/10.5194/wes-2026-161

Image Credits: AI Generated

DOI: 10.5194/wes-2026-161

Keywords: wind energy, actuator disk model, actuator line model, wind farm simulation, turbulence, power fluctuations, thrust, large-eddy simulation, atmospheric boundary layer, diurnal cycle, computational modeling, grid integration