The CFD Approach

WindSim and Meteodyn WT are commercial software packages that use the computational fluid dynamics technique (CFD) to produce simulations of the Atmospheric Boundary Layer (ABL) for wind power applications. These simulations solve the time-averaged equations of mass and momentum conservation (Navier-Stokes equations). Both WindSim and Meteodyn WT are based on the 3D CFD tool Phoenics [2] and incorporate the MIGAL solver. MIGAL is an iterative linear equations solver which applies an algorithm that simultaneously updates the wind speed and pressure across the whole computational domain (“coupled resolution”). This method demands more storage capacity, but increases the speed and robustness of the package when compared to the standard SIMPLE algorithm. MIGAL is developed by MFRDC and has been fully validated on number of academic case studies [1, 2].

The CFD approach requires significantly more computational resource than a classical WAsP analysis, as the calculations are significantly more complex. A flow domain is created and defined by a set of boundary conditions which control the way that the air flows in and out of the domain. A 3D grid is created within the domain and the Navier-Stokes equations are solved at each discrete point on the grid. Flow must be setup to travel through the domain parallel to the sides.

Due to this construction, the model is subject to discretisation errors and can only evaluate wind from a single direction at a time. A separate computation is undertaken for each direction sector. The predicted wind regime can then be synthesised from the results of the single direction runs.

A GH CFD analysis is carried out on a 10 degree direction basis. The turbine locations are at least 8 km away from the edge of the computational domain for each calculation. The horizontal resolution of the Meteodyn WT grid domain is 25 m with an “expansion coefficient” of 1.1 and the vertical resolution is 4 m, which are the defaults recommended by the software developers. The “expansion coefficient” is used to reduce the horizontal resolution of the grid in the far field of the points of interest. The grid used by GH currently consists of up to 6 million cells. For all calculations GH assumes a neutrally stable boundary layer within the CFD code. Should wind speed and temperature data be available from taller masts with multiple sensors at multiple heights, the stability level of the site can be better defined. For each direction sector, it is ensured that the wind speed predictions at all nodes in the computational grid vary by less than 3% in the final iteration of the calculation.

In any CFD analysis it is important to establish “grid independent” results. Grid independence is achieved when further refinement or expansion of the grid does not result in significant changes to the converged CFD solutions. There is often a trade-off between complete grid independency, resolution and number of calculation points for a given computational resource. Therefore in order to strike this balance, the CFD calculations are repeated from 2 to 5 times depending on the terrain complexity and the average of the 2 to 5 runs taken as the final CFD results. The different runs have the centre of the grid domain displaced by 1 or 2 km. By this method any grid dependency is significantly mitigated. Furthermore, as detailed above, GH undertake calculations for every 10 degree direction sector and re-average results to derive 30 degree direction sector speed-ups from the masts to the turbine locations, which are used to predict the wind speed at each turbine location. This method further reduces grid dependency and noise in the results.

[1] Ferry M., 2000, “The MIGAL solver”, Proc. Of the Phoenics Users Int. Conf., Luxembourg, 2000

[2] Ferry M., 2002, “New features of the MIGAL solver”, Proc. Of the Phoenics Users Int. Conf.,Moscow, Sept. 2002