WindPRO training

WAsP model limitations

Introduction to CFD
 2

INDEX
1. WAsP model

2. Limitations of WAsP

3. Fundamentals of CFD

1. Conservation law
2. Governing equations
3. Need f models -> turbulence models
4. Numerical solving approach
5. Mesh examples
6. CFD full RANS
The WAsP model
The WAsP model

• WAsP is a micro-scale linear function model

• Separated handling (superposition of); terrain +

roughness + atmospheric stability

• Accurate predictions only for close to neutrally-

stable atmospheric conditions, for wind flows over
low and smooth hills (attached flows).

• The reference and predicted sites must be subject

to the same weather regime (and the reference
data reliable, of course).
The WAsP model

• Input data is annual/long-term mean

flow: Weibull, wind rose… -> TAB

• Process in two steps:

• 1. Removes local flow effects from

measurements to produce a
“regional wind atlas”

• 2. reapplies local flow effects at

each WTG position
The WAsP model

• WAsP assumes that there is a unique

speed-up factor between the two sites
(reference and predicted) for each wind
direction sector which is determined by
the roughness field and local terrain
heights at both sites.

• It understand the terrain as a

perturbation of a flat reference.

• The speed-up factor is assumed to be

independent of climatic conditions.
Limitations of WAsP
WAsP limitations
Summary

• Complex terrain

• Complex roughness

• Atmospheric conditions far for neutral or

neutrally-stable.

• Always check excessive speed-ups

WAsP limitations
When to use it?
Fundamentals of CFD
Fundamentals of CFD
Conservation law
• The variation of the total amount of a quantity inside a given domain is equal to the balance
between the amount of that quantity entering and leaving the considered domain, plus the
contributions from eventual sources generating that quantity.
• Fluid flows are described totally by the conservation of the following three quantities:
➢Mass
➢Momentum
➢Energy

• If all variables which are contained on those derived equations from these conservation laws,
were known, fluid dynamics would be defined, but as it was shown later, additional equations are
needed to close the problem, and they will change depending on turbulence approximation level.
Fundamentals of CFD. Governing equations
Governing equations. Mass conservation (continuity)
• Independent of the nature of the fluid.
• Independent of the forces acting on the fluid.
• It expresses the empirical fact that in a fluid system, mass cannot disappear from the system, nor
be created.

Integral conservative

0 0
Independent on forces.
Assumption: absence of
Differential conservative
external mass sources.

Differential non-conservative
Fundamentals of CFD. Governing equations
Governing equations. Momentum conservation

Quantity: Momentum. Product of the mass and velocity.

• Convective flux of momentum

•No momentum diffusion (impossible

in a fluid at rest)

Term (I)
 6
Fundamentals of CFD. Governing equations
Governing equations. Momentum conservation

•The internal forces cancel at all internal points due to the

action-reaction rule and only the surface points remain for the
internal forces contributions.

•External forces such as those ones acting directly on the

volumetric mass of the fluid elements. Examples: gravitational,
electric, magnetic forces.

•Surfaces forces: Acting on the surface of the fluid elements.

Two main sources: pressure and shear/normal stress (viscous).
 10
Fundamentals of CFD. Governing equations
Governing equations. Energy conservation

Term (III) = Term (IV)

Integral conservative

Differential conservative

Differential conservative
(enthalpy form)
Fundamentals of CFD. Governing equations
Governing equations. Summary. Navier-Stokes equations
Fundamentals of CFD. Governing equations
Governing equations. Summary. Navier-Stokes Equations

Being, viscous shear stress:

Fundamentals of CFD. Governing equations
Governing equations. Other considerations

• Navier-Stokes equations configures a coupled system of nonlinear partial differential equations, and
hence are very difficult to solver analytically. To date, there is no general closed-form analytical
solution to these equations that could be obtained exactly and mathematically.

• There are five equations but six unknown flow field variables ρ, p, u, v, w, e. So, supplementary
equations are needed.

• In aerodynamics, it is generally reasonable to assume the gas as perfect gas, so standard equation
of state could be applied, but it introduces another unknown variable, temperature.

• Additional thermodynamic relation between state variables must be included to close the set of
equations and flow filed unknown variables.

• Note: for highly compressible flows, thermal conductivity and viscosity varies with temperature more
than negligible so, additional expressions are usually included, such as Sutherland’s law for viscosity
or experimental ones for k.
Fundamentals of CFD. Governing equations
Governing equations. Summary.

Exact solution exists but

Governing equations: currently is not possible to
Navier-Stokes be solved strictly from
equations analytical point of view.

Need of numerical solution

Need of of the problem
models
Why? Discretization of control
volume and equations.

Algebraic system approach is Partial derivative equations

solved through numerical calculus coupled equations are
methodologies reduced up to a algebraic
system of equations.
Fundamentals of CFD. Approximation Levels
Need of models. Why? Turbulence

• These complexities in fluid dynamics pose considerable challenges for CFD, in particular for turbulent
flows.

• High complexity turbulent flows needs high size of the meshes which discretize the control volume,
which have very high computational time to be solved, above all for industrial purposes. That is WHY
different turbulence models are necessary.

• Examples:
•One complex terrain (3D) -> ~ 300 million cells -> run 1 wind direction with unique
averaged velocity and IT condition -> 6-8 weeks for LES simulations on a ~2500 nodes
cluster.

• If models are not used, much more time or much more powerful clusters are needed to get simulations
in reasonable times. Models are crucial for industry in general.
Fundamentals of CFD. Approximation Levels
Need of models. Why? Turbulence

Most important for external

aerodynamics:

• DNS
• LES
• Hybrid RANS/LES (DES, SAS,…)
• RANS

Simpler models will be used for

initialization of models specified
above.
• Laminar flows.
• Inviscid model :
•Time-dependent Euler
equations.
•Potential flows.
Fundamentals of CFD. Approximation Levels
Need of models. Turbulence approximation levels

• A turbulent flow is composed by eddies of different size.

• Largest eddies are unstable and eventually break up into smaller eddies, and Richardson’s
so on. energy cascade of
• The energy is passed down from the large scales of the motion to smaller
scales until reaching a sufficiently small length scale such that the viscosity of eddies
the fluid can effectively dissipate the kinetic energy.
Fundamentals of CFD. Approximation Levels
Need of models. Turbulence approximation levels

• Turbulent kinetic energy spectrum vs. wavenumber of

eddies show clearly the level of accuracy of different
turbulence approaches.

• DNS approach solves directly Navier-Stokes equations

and all time and length scales of the eddies.

• LES approach filters through grid spacing what turbulent

scales are solved and which ones are modeled. Refined
grid areas solves equations, and smallest eddies below
grid spacing are modeled through subgrid models
(Smagorinsky).

• RANS approach solves averaged Navier-Stokes

equations and models turbulence through averaged
quantities representing production and dissipation of
turbulence, through different methodologies.
Fundamentals of CFD. Approximation Levels
Turbulence models. Zero-Equation Models

• At zero equation models, there are not

PDE that describes the transport of the
turbulent stresses and fluxes.
• A simple algebraic relation is used to
close the problem.
• Example: Mixing length theory.
• Approach of length over which there is
high interaction of vortices in a turbulent
flow field.

• This mixing length is

determined experimentally.
Fundamentals of CFD. Approximation Levels
Turbulence models. One-Equation Models

• Example: Spalart-Allmaras
• Turbulent kinetic energy is not calculated
through a transport equation.
• Modeled transport equation is solved for
kinematic eddy (turbulent) viscosity.
• Initially designed for aerospace
applications involving wall-bounded
flows.

Production of Destruction of
turbulent viscosity turbulent viscosity
Fundamentals of CFD. Approximation Levels
Turbulence models. Two-Equation Models

• Examples: k-ε, k-ω, k-ω-SST, k-ε Realizable

The term "realizable'' means that the model satisfies certain

mathematical constraints on the Reynolds stresses, consistent with the
physics of turbulent flows. Neither the standard k- ε model nor the
RNG k- ε model are realizable.

An immediate benefit of the realizable k- ε model is that it more accurately

predicts the spreading rate of both planar and round jets. It is also likely to
provide superior performance for flows involving rotation, boundary
layers under strong adverse pressure gradients, separation, and
recirculation.

Initial studies have shown that the realizable model provides the best
performance of all the k- ε model versions for several validations of
separated flows and flows with complex secondary flow features.
Numerical Solving Approach
Discretization schemes

• Example with easy analytical solution:

Mesh examples
Mesh examples
Mesh examples
Mesh examples
CFD full RANS
Benefits and Limitations
• Benefits:
➢ Nowadays, RANS methods have a high accuracy/time ratio
➢ They allow us to predict reasonably the location and extension of detachment regions and vertical
profiles in complex terrains ( shear, upflow, wind-veer)
➢ Although the turbulent intensity TI in absolute value could be slightly under estimated, the change of TI
with height and between wind turbine positions is well estimated

• Limitations:
➢ The higher accuracy is linked to a correct mesh generation ( including boundary conditions)
➢ Some sources of uncertainty linked to the empirical adjustment of the numerical models ( turbulent, wall
functions, canopy, bouyancy,..)
➢ Isotropic turbulent
➢ Turbulent spectrum is not captured nor synthetic wind
CFD full RANS
Boundary Conditions
• The numerical boundary conditions have a great impact of the stability and convergence of the
simulation
• The boundary conditions should be known beforehand and near to the “real” flow conditions. The
inlet boundary conditions are critical.
• As a way to minimize the impact due to the uncertainty to know the exact flow conditions at the
inlet, it is required:
• The inlet patch should be far away from the zone of interest. In that way, orography and
roughness conditions will modify the velocity and TI profile to be consistent with them.
• Flat terrain near the inlet patch to impose logarithmic velocity profiles correctly.
• The boundary condition in the sky could affect the simulation if it is near the terrain. Because of
that, the sky patch is located 5 times the difference between the minimum height and the
maximum height of the terrain
CFD full RANS
Boundary Conditions
• Terrain:
➢ U=0. noSlip boundary conditions in close relation to the wall shear stress.
➢ Turbulent viscosity (t ).
➢ Logarithmic wall function in harmony with the wall roughness
➢ Numerical approach adjusted empirically for boundary layer without pressure gradients
➢ It takes into account the roughness of the terrain and so, affecting to the TI produced by the terrain,
wind shear and shear stress.

• Inlet:
• “Exact” solution of the k-epsilon model for boundary layer flow without pressure gradients ( flat
terrain in neutral atmospheric conditions): Richards-Hoxey profiles 1993
 CFD full RAN
Other turbulent models. Canopy Model▪ decreases in velocity
▪ increases in
[Continuity equation] turbulence
 ui ▪increases in
=0
xi dissipation
[Average equation]
  uj 
 ui  ui u j   p 2     iu
+ = + 
k +  +  −F
xi   3  x j  t  x j xi 
i
t x j
  
[k transport equation]

k  u j k   t k
 a
C f a ui
2
+ = + Pk −  + Fk Fi u
t x j x j   x j  j

[ transport equation] Fk ui Fi
  u j      t      C p  Fk
+ = + (C P − C2 )+ F
 k 1 k
F k
t x j x j   x j 
u  u j  
 iu
Pk =  t  i +
x j xi  x j : fraction of the area covered with trees
 
－Fi: extra term added to the momentum equation a : leaf surface area density
+ Fk: extra term added to the transport equation of k Cf: drag coefficient for canopy
+ F: extra term added to the transport equation of  Cp1: model coefficient for F
CFD full RANS
Other models. Actuator Disk ui  ~p   D   
+ (uu )= −2  u − −
p0(x, y)
− ( )
 ij − g 3 z  k+
1
f iT
t x j xi xi 0 x j xi  0  0
ji i3k 3k

• Three approximation levels:

• Constant thrust coefficient Ct

• BEM Actuator Disk Model

➢ Rotor model -> Disk discretized azimuthally and radially .
➢ 2.5D geometry needed.
➢ Pure Airfoil Polar curves needed.
➢ Polar curve shave preserving interpolation.
➢ 3D corrections included.
➢ Steady simulation.
➢ CFD/actuator disk coupling through Navier-Stokes
momentum equation.
Thank you!