QBlade v0.

01
Guidelines
David Marten
July 7, 2010
Contact: qblade@web.de
Contents
1 Introduction 4
1.1 The BEM in the wind turbine industry . . . . . . . . . . . . . . . . 4
1.2 The software project . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2 Software implementation 7
2.1 Code limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Code structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3 The QBlade module in XFOIL . . . . . . . . . . . . . . . . . . . . . 9
2.4 The blade design and optimization sub-module . . . . . . . . . . 10
2.5 360

polar extrapolation sub-module . . . . . . . . . . . . . . . . . 13

2.6 The rotor simulation sub-module . . . . . . . . . . . . . . . . . . . 14
2.7 The turbine denition and simulation sub-module . . . . . . . . . 14
3 Simulation parameters and corrections 18
3.1 Number of Elements . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.2 Epsilon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.3 Maximum number of iterations . . . . . . . . . . . . . . . . . . . . 20
3.4 Relaxation factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.5 Density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.6 Corrections to the simulation . . . . . . . . . . . . . . . . . . . . . 22
3.7 Foil interpolation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4 Simulation results 25
4.1 Rotor simulation variables . . . . . . . . . . . . . . . . . . . . . . . 26
4.2 Turbine simulation variables . . . . . . . . . . . . . . . . . . . . . . 27
4.3 General validity checks on simulation results . . . . . . . . . . . . 28
4.4 Validation of the results against a reference BEM code . . . . . . . 28
4.5 Global variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.6 Local variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.7 Sensitivity analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5 How to create a blade simulation 40
Bibliography 47
3
1 Introduction
Solving the worlds energy problem is one of the great topics of our time. Re-
cently, an increase in global awareness has led to a boom in the market for re-
newable energy technology. To reduce the dependence on fossil fuels, efcient
power generation methods in the elds of solar, wind, wave, biomass and ther-
mal energy are in demand. In Germany wind energy plays a central role to reach
the goal of an almost carbon dioxide free energy production till 2050 [1]. Because
there is a limit to the number of adequate sites for turbines, that do not interfere
with natural protection laws or residents, a part of the strategy to keep the de-
velopment of wind energy on a high level is to replace older turbines at good
wind sited with newer, more efcient ones.
A condition for an efcient conversion of the wind energy into mechanical en-
ergy with wind turbines is the optimal design of the rotor blades. Methods for
rapid development, reliable and robust predictions of the aerodynamic charac-
teristics and simulation of the ow conditions around a rotor blade are essential
for this design task.
1.1 The BEM in the wind turbine industry
The blade design methods for the wind turbines originate from the aircraft de-
sign industry and apply the same techniques. But because the ow conditions
that a turbine blade experiences are quite different to those affecting a plane
a lot of the assumptions made in ight aerodynamics cantt be applied to the
complex ow eld around a wind turbine. This complex ow eld is unsteady,
incompressible, three dimensional, turbulent and often separated. The aerody-
namics of a Wind Turbine are inuenced fromthe far eld conditions, far up and
downstream of the rotor and at the same time depend on small scale turbulent
ow conditions around the blades. This implies the need for a large simulated
domain as well as a ne spatial resolution. Conducting a full CFD analysis that
fullls these requirements and accounts for all these effects, is very time consum-
ing and expensive. An alternative to CFD simulations are the vortex methods,
with the limitation that they can not model viscous behavior since they are based
on potential ow theory. That is why only [8] design and evaluation tools that
4
1 Introduction
are based on the Blade Element Momentum Method are used in the industry. The
other (CFD, RANS and vortex) methods are almost only used in research envi-
ronments. The main advantage of the BEM compared to CFD is that it is very
cost efcient and the computational time is signicantly less. The prediction of a
wind turbines performance, operating in a uctuating wind eld complicates the
application of the BEM, that assumes a steady state wind eld. The BEM, which
is in fact a two dimensional method extrapolated into the third dimension ap-
plies semi-empirical corrections, derived from correlations with measurements
or full CFD computations, to account for three dimensional effects. TANGLER
states that in general the BEM under predicts the overall performance of a Tur-
bine and over predicts the peak power [12]. But nevertheless the blade element
momentum method is widely applied in the wind turbine industry because the
use of analysis techniques of lower order-accuracy greatly simplies the turbine
design. With the BEM its possible to rapidly develop and test different rotor de-
signs against one another, commit small changes and test again, and in this way
evolve a preliminary design that can be studied in greater detail with other tech-
niques, like CFD, later. This, and the verication of BEM simulations with wind
tunnel and eld measurements, justify the use of BEMcomputational methods to
analyze the blades from a two dimensional point of view. The BEMs ability for
robust analysis and low computational cost make up for the shortcomings and
inaccuracies. Virtually all the modern rotors for Horizontal Axis Wind Turbines
that exist today were designed using the Blade Element Momentum Method.
1.2 The software project
This software project is realized being a part of the wind energy group at the
Berlin Technical University Department of Experimental Fluid Mechanics, led
by Prof. Dr. Christian Oliver Paschereit. The aim of this work is to provide
an open source turbine calculation software that is seamlessly integrated into
XFOIL, an airfoil design and analysis tool. The motivation for this is to create
a one solution software for the design and aerodynamical computation of wind
turbine blades. The integration in XFOIL allows for the user to rapidly design
customairfoils and compute their polars, extrapolate the polar data to a range of
360

, and directly integrate theminto a wind turbine rotor-simulation. This skips

the step of exporting and importing foil and geometry data between different
codes, and the troubles associated with that. At the same time the integration of
the BEM code into XFOILs sophisticated GUI will make this software accessible
to many more interested people than the usual command line interface software
tools. The software is especially adequate for teaching, as it provides a hands
on feeling for HAWT rotor design and shows all the fundamental relationships
and concepts between twist, chord, foils, turbine control and type and the power
5
1.2 The software project
curve in an easy and intuitive way. The GUI serves as well as a post processor to
conducted rotor simulations and gives deep insight into all relevant blade and
rotor variables for verication, to compare different rotor congurations, or even
to study the numerical algorithmand the dependencys among the aerodynamic
variables themselves. In addition to that the resulting software is a very exible
and user-friendly platform for wind turbine blade design that can also act as
a modular system for future implementations that can exploit the possibilities
that a combination of manual and parametric airfoil design and analysis coupled
with a blade design and simulation tool offers.
The functionality of the BEM software should include the following features:
Extrapolation of XFOIL generated or imported polar data to 360

AoA
Blade design and optimization, including 3D visualization, using XFOIL
generated or imported proles
Turbine denition (rotor blade, turbine control, generator type, losses...)
Computation of rotor performance over range
Computation of turbine performance over wind speed range
Annual yield computation with WEIBULL distribution
Manual selection of BEM correction algorithms
Manual selection of all simulation parameters
Data browsing and visualization as post processing
Export functionality for all created simulation data
Blade geometry export functionality
Storing of projects, rotors, turbines and simulations in a runtime database
6
2 Software implementation
2.1 Code limitations
Like the original XFOIL written by MARK DRELA and XFLR, written by ANDRE
DEPERROIS, QBlade has been developed and released according to the princi-
ples of the General Public License. One important point about the GPL is that this
program is distributed without any warranty. Whether the warranty of mer-
chantability, nor the warranty of tness for a particular purpose. The resulting
software is not intended as a professional product and does not offer any guar-
antee of robustness or accuracy. It is distributed as a personal use application
only. This software may not be default-free and there will certainly be more
bugs discovered after the distribution. However, a validation against other BEM
software permits some trust in the results this software provides.
2.2 Code structure
An overview of the data objects, in which the blades, turbines, polars and simu-
lations are stored, and their relation to XFOIL can be found in Fig. 2.1
360

polar Object
A 360

polar object is created in the 360

polar extrapolation sub-module. It is

dened by a name and a parent polar (the polar that was extrapolated). The
lift and drag coefcients over the whole 360

range of the AoA are stored as

data. If the parent polar is deleted the extrapolated polar will be deleted as well
to ensure consistency. If a 360

polar object is deleted all blades, turbines and

simulations that include this object are deleted.
7
2.2 Code structure
Figure 2.1: Data objects and data ow in the BEM module
Blade object
A blade object is created in the blade design and optimization submodule. It
stores the geometric blade data (chord, radial position, twist and offset) and the
associated foils and 360

polars. It is dened by a name. If a blade object is

deleted all associated simulations and turbines are deleted as well.
Turbine object
A turbine object is created in the turbine denition sub-module. It stores all
turbine parameters (pitch, stall, single-, variable-, 2-step transmission,...) and a
rotor-blade that needs to be associated with the turbine. If a turbine object is
deleted all associated simulations deleted as well.
8
2 Software implementation
Simulation object
A simulation object is created when a simulation is dened. It stores the simula-
tion parameters (max. iterations, corrections, elements, epsilon) and, if a simula-
tion was executed, the global results. Global results are results that characterize
the whole rotor (, C
P
, C
T
, ...). If a simulation object is deleted all associated blade
data objects are deleted as well.
Blade data object
Blade data objects are created automatically during a simulation. They store all
the data that is computed for every element along the blade. One blade data
object is created for every increment of the simulation. The blade data objects
require the most memory space.
2.3 The QBlade module in XFOIL
QBlade consists of four sub-modules, that will be discussed in greater detail on
the following pages. These Modules are:
Blade design and optimization
360

polar extrapolation
Turbine denition and simulation
Rotor simulation
9
2.4 The blade design and optimization sub-module
2.4 The blade design and optimization sub-module
Figure 2.2: Blade design window
In the Blade design and optimization sub-module the user can create a rotor-
blade. A blade consists of an arbitrary number of sections, see Fig. 2.3. Every
section is dened by chord, twist, offset (offset is not relevant for the BEM algo-
rithm, it is just included for the geometry export function), prole and the associ-
ated 360

polar. A new blade can only be created when there is at least one 360

polar object present in the runtime database, it can only be saved when a foil and
a polar is selected at every station. Each rotor-blade has a number of blades
associated with it that denes how many blades the rotor has.
Figure 2.3: Sections along a Blade
10
2 Software implementation
Optimization of blade geometry
While a blade is edited, or created, and every section is fully dened the user can
optimize the blade geometry with the blade optimization dialog, see Fig. 2.4.
Figure 2.4: Optimization dialog
The user has to choose a tip speed ratio
0
to optimize for, and the sections
(positions) that are to be optimized. From this tip speed ratio an assumed inow
angle is computed for every section:

loc
= tan
1
_
1

0,loc
2
3
_
. (2.1)
Optimize for lift/drag sets the twist, at the specied
0,loc
at which the blade
section operates, to an AoA that yields the highest glide ratio. The option to de-
crease or increase this angle exists for the case that the AoA yielding the highest
glide-ratio is close to the stall point. If the user optimizes for stall, the twist is
set in such a way that all the stations, at the same time, experience stall at the
selected
0
value. The third option allows to set a linear twist.
The chord distribution can be optimized according to BETZ [4, p.202]:
c(r) =
16
9
R
BC
L

0
1
_
_

0
r
R
_
2
+
4
9
, (2.2)
or SCHMITZ [4, p.202]:
c(r) =
16r
BC
L
sin
2
_
1
3
tan
1
_
R

0
r
__
. (2.3)
It is important to note that C
L
is computed from the expected angle (
0
) and
the twist angle for every station individually. It may happen, especially for stall
11
2.4 The blade design and optimization sub-module
Figure 2.5: Comparison of dimensionless chord distribution after BETZ and
Schmitz [4]
blades, that a low C
L
value at
0
causes a very high value for the chord. This
is not a failure of the optimization equations. One has to be careful with large
chord values. The value for solidity, =
cB
2r
(the section of an annulus that is
covered by blades), must be between zero and unity. For a section whose solidity
is greater than unity the BEM algorithm does not converge. Attention has to
be paid especially in the root region because both the BETZ and the SCHMITZ
equation lead to large chord values here.
Scaling of blade geometry
Figure 2.6: Scale blade dialog
The scale option allows to scale the blades twist, chord and/or position. This can
be done by simply giving a new value for the variable that is scaled. The scaling
ratio is then computed automatically.
12
2 Software implementation
2.5 360

polar extrapolation sub-module

Figure 2.7: The 360

polar extrapolation sub-module

Polars, that are a result of an XFOIL analysis or imported, can be extrapolated
to the full 360

AoA range in the 360

polar extrapolation sub-module. A po-

lar can be selected from the drop-down menu in the tool bar. When creating
an extrapolation the user can set the CD
90,2D
value for the extrapolation man-
ually. The two points C
L1
and C
L2
, from which the interpolation function f for
the positive extrapolation is constructed can be manipulated using the sliders A+
(for CL1) and B+ (for C
L2
). The sliders A- and B- manipulate the corresponding
point for the negative extrapolation. The extrapolation is carried out exactly as
described in Methods for Root Effects, Tip Effects and Extending the Angle of Attack
Range to +-100

, with Application to Aerodynamics for Blades on Wind Turbines and

Propellers [9]. Whenever a slider is moved, or the value for C
D90,2d
is changed,
the whole extrapolation is computed again. With the option edit current po-
lar the currently selected 360

polar can be edited manually after it has been

extrapolated.
13
2.7 The turbine denition and simulation sub-module
Figure 2.8: The Rotor simulation sub-module
2.6 The rotor simulation sub-module
In the rotor simulation sub-module the user can commit rotor-blade simulations
over a range of tip speed ratios. A rotor simulation can only be dened when at
least one rotor blade is present in the runtime database. When dening a rotor
simulation the user has to select the desired corrections to the BEM algorithm
and the simulation parameters. Once a simulation is dened the user can select
a range of lambda values (tip speed ratios), and the incremental lambda step for
the simulation. A rotor simulation is always carried out dimensionless. The free
stream velocity is assumed to be unity and the rotor radius is normalized for the
computation. This implies that no power curve or load curves, like the bending
moment, can be computed during a rotor simulation.
2.7 The turbine denition and simulation sub-module
In the turbine denition and simulation sub-module the user can dene a wind
turbine. To dene a wind turbine a rotor-blade must be present in the runtime
database. To create a turbine the turbine type and the turbine parameters have
to be specied. The turbine type is dened by:
14
2 Software implementation
Figure 2.9: The turbine denition and simulation sub-module
Regulation: pitch or stall
Transmission: single, 2-step or variable
If a pitch regulated turbine is dened the user has to specify a nominal power
output. When the wind speed , that yields the nominal power output is reached
the blades are pitched to reduce the power for higher wind speeds to the nomi-
nal output. A stall regulated turbine has no pitch control and the power output
is limited solely when stall occurs at the rotor. Designing a stall turbine that
limits it power to the desired output and at the desired wind speed requires an
iterative approach.
For a single speed transmission the user has to select only one rotational speed,
in which the turbine operated over the whole range of wind speeds. For 2-step
transmission two rotational speeds and a wind speed at which the transmission
changes between these to rotational speeds have to be selected. Avariable trans-
mission turbine has a minimum and a maximum value for he rotational speed.
Additionally the user selects a desired tip speed ratio,
0
, from this ratio a rota-
tional speed is computed for every given wind speed during the simulation. If
the computed rotational speeds are lower or larger than the bounding minimum
or maximum values, these values give the rotational speed.
The turbine parameters that dene a turbine are:
15
2.7 The turbine denition and simulation sub-module
Rotor-blade
Cut in wind speed
Cut out wind speed
Fixed losses
Variable losses
Every turbine needs to have a rotor blade dened. This can be any blade that is
stored in the runtime database. At the cut-in wind speed the turbine starts and at
the cut-out wind speed the turbine stops operation. To account for power losses,
that are not of aerodynamical nature, caused by the efciency of the generator
and the gearbox a value for xed losses and a value for variable losses can be
selected. The equation in which these losses are implemented is:
P
out
= (1 k
v
)P
0
P
f ixed
. (2.4)
k
v
is the variable loss factor and P
f ixed
the xed loss factor.
Figure 2.10: Different variable loss factors - effect on power output
If a turbine simulation has been conducted the user may calculate the annual
yield of the turbine by specifying an annual wind speed distribution via the two
parameters k and A of the WEIBULL distribution. The probability for a wind
speed to occur is:
h
W
(V
0
) =
k
A
_
V
0
A
_
k1
exp
_

_
V
0
A
_
k
_
. (2.5)
16
2 Software implementation
The probability f (V
i
< V
0
< V
i+1
) that a wind speed lies between V
i
and V
i+1
is:
f (V
i
< V
0
< V
i+1
) = exp
_

_
V
i
A
_
k
_
exp
_

_
V
i+1
A
_
k
_
. (2.6)
Thus the annual energy production is calculated as:
AEP =
N1

i=1
1
2
(P(V
i+1
) + P(V
i
)) f (V
i
< V
0
< V
i+1
) 8760 . (2.7)
A turbine simulation is carried out over a range of wind speeds, with the chosen
incremental step size. Depending on the specied rotational speed of the turbine
a tip-speed ratio is computed for every wind speed. Then a BEMsimulation over
he computed tip speeds, that is equivalent to a rotor simulation, is carried out.
17
3 Simulation parameters and
corrections
Figure 3.1: Simulation denition dialog
When dening a simulation the following parameters have to be set:
Number of elements along blade
Maximum epsilon for convergence
Maximum number of iterations
Relaxation factor
Density
Viscosity
18
3 Simulation parameters and corrections
3.1 Number of Elements
The number of elements species into how many elements the blade is divided.
This number is independent from the number of sections that a blade has. The
BEM algorithm is executed once for every element. The input values, like chord
and twist, are interpolated between the blade stations, where they are dened,
and computed for the centers of the elements. The elements are distributed us-
ing sinusoidal spacing. This allows for more elements to be placed in the tip and
root region where largest gradients usually are to be expected. By using sinu-
soidal spacing, the overall number of elements that are required is less and the
computational time is reduced.
Figure 3.2: Sinusoidal spaced elements along the blade
All the variables, resulting from the BEM computation, are computed for the
center of an element and are treated as the averaged values over an element. This
solves a problem that arises if the values would be computed for the boundaries
of an element and then interpolated to the center fromthere. When the PRANDTL
tip loss factor F goes to zero at the tip or at the hub numerical instabilities arise
and result in a non converging iteration. This problem is skipped because the
center of an element can be arbitrarily close to the tip or hub of the blade but
never be on the same position, it is always

i
2
away (
i
is the width of the i
th
element). The forces per length, P
N
and P
T
, that were computed for the element
center are integrated over the whole element, to yield the elements contribution
to the total torque and thrust.
3.2 Epsilon
The epsilon value, , denes when an iteration is converged. The maximum
of the difference of axial and radial induction factor between the last and the
19
3.4 Relaxation factor
current iteration has to be below for convergence.
max(|a a
old
|, |a

old
|) < . (3.1)
A recommendation for epsilon is 10
5
. See section Sensitivity analysis in the Ap-
pendix for details.
3.3 Maximum number of iterations
The maximum number of iterations prevents that the algorithm may get stuck
in an innite loop.
3.4 Relaxation factor
A common problem, during the iteration loop of a BEM computation, is the
uctuating behavior of the axial induction factor. The reason for this uctuation
is the periodical switching of the turbines loading state between light and heavy
loading [7] (see section The turbulent wake state). This may lead to a stop of the
iteration, after the maximum number of iterations is reached and impacts both
the codes performance and its accuracy. In [7] MAHERI proposes to introduce a
relaxation factor,
relax
, to overcome these uctuations.
Figure 3.3: Fluctuation of the axial induction factor around the light and heavy
loading state [7]
The relaxation factor is introduced in the iteration after a newvalue, a
k+1
, for the
axial induction factor has been calculated:
a
k+1
=
relax
a
k+1
+ (1
relax
)a
k
; 0 <
relax
< 1 . (3.2)
20
3 Simulation parameters and corrections
Figure 3.4: Damped uctuation of the axial induction factor for different relax-
ation factors [7]
The convergence rate of the BEM code strongly depends on the initial guess
value for the axial induction factor. If the initial guess (a = 0 in this implemen-
tation) is in the neighborhood of the nal result, convergence is achieved signi-
cantly faster. To further accelerate the convergence rate of the BEMcode MAHERI
proposes that for the rst few iterations a relaxation factor
relax
= 1 should be
applied, to let the rst few oscillations happen. These oscillations then mark the
boundary of the neighborhood of the nal result. With a three-point-equation
the axial induction factor is then placed inside this neighborhood.
a
k+1
=
1
4
a
k+1
+
1
2
a
k
+
1
4
a
k1
. (3.3)
From there the iteration proceeds as normal with the desired relaxation factor
and eq. 3.2 applied.
Figure 3.5: Accelerated convergence by placing the induction factor inside the
neighborhood of the nal result [7]
21
3.6 Corrections to the simulation
3.5 Density
The density for the uid around the wind turbine is needed to calculate the
power output:
P =
1
2
AV
3
0
C
P
. (3.4)
The density is only used to compute the power output for a turbine simulation.
During a rotor simulation all variables are dimensionless and only depend on
the tip speed ratio.
Dynamic viscosity
The dynamic viscosity is needed to compute the local REYNOLDS number along
the blade:
Re(r) =
V
rel
(r)c(r)

. (3.5)
The dynamic viscosity is only used to compute the REYNOLDS number during
a turbine simulation. During a rotor simulation all variables are dimensionless
and only depend on the tip speed ratio.
3.6 Corrections to the simulation
The following correction algorithms can be selected to be included in the BEM
simulation:
PRANTL tip loss, found in [5]
PRANTL root loss, found in [5]
New tip loss model after SHEN et al. [11]
New root loss model after SHEN et al. [11]
22
3 Simulation parameters and corrections
3D correction after Snel, found in [6]
Foil interpolation
Any combination of these corrections can be added to a simulation with one ex-
ception. The tip or root loss model after SHEN can never be used in combination
with the PRANDTL root or tip loss model. That is because the PRANDTL tip loss
factor F is included in the models by SHEN a priori.
3.7 Foil interpolation
The foil interpolation, in effect, is not a correction to the BEM algorithm. It
merely is the most simple solution to a problem that arises during the discretiza-
tion of the blade. As stated previously, a blade is dened in sections. Every
section may have a different airfoil, that denes the sections geometry. The ge-
ometry in between two sections, of a real blade, is a liner interpolation between
the two airfoils. The problem now is that only polar data for the airfoils at every
section, but no data for the interpolated airfoils in between are present in the
database. If the option Foil interpolation is not selected, the BEM treats the blade
as in Fig. 3.6
Figure 3.6: Foil distribution along the blade without interpolation
All elements, whose centers lie between section 1 and section 2 are linked to the
polar data from foil 1. The last airfoil, at position Z, is not included at all in the
simulation. From the element that lies just before section 2 , to the rst element
that lies after section 2 there is a discontinuity, as the foils rapidly change from
23
3.7 Foil interpolation
foil 1 to foil 2. This is expressed in the simulation results, if interpolation is not
included.
When the foil interpolation is switched on, the polar data, that is used for the
BEM computation of an element, is a linear interpolation between the polar data
of the bounding airfoils. The equation used is:
Polar
int
= Polar
2

r
center
r
sec
1
r
sec
2
r
sec
1
+ Polar
1
(1
r r
sec
1
rsec
2
r
sec
1
) . (3.6)
This interpolation more accurately represents a true blade geometry. Strictly
speaking, the linear interpolation between two polars never represents the true
polar of the intermediate airfoil. This interpolation is just the most simple ap-
proximation to the real polar, that is not present in the database. However, the
accuracy can be arbitrarily improved by importing the geometric data for these
intermediate airfoils and create new sections where the intermediate airfoils are
placed. Another possibility is to use XFOILs dynamic coordinate mixing function,
where intermediate airfoil geometries can be created and simulated in XFOIL.
24
4 Simulation results
There are two different types of simulation results for a BEMcomputation. Global
variables, or rotor variables, are values that characterize the rotor, or turbine, as
a whole. The C
P
value is such a global variable. Every increment of tip speed ra-
tio or wind speed, that was simulated yields one C
P
value. The C
P
over curve
gives only information about the overall rotor performance, but not about the
local events taking place at the blades. These global variables are computed out
of the local variables. Every point in the C
P
curve represents a BEMcomputation
for one tip-speed ratio or wind speed.
The thrust T is calculated by adding up the normal forces acting on the elements.
The local variables, or blade variables, like the AoA , give insight in the local
conditions at the blade. Every curve of a local variable represents a BEM com-
putation for one tip-speed ratio or wind speed.
Figure 4.1: Rotor graph to the left and blade graph to the right, context menu
In the context menu of a graph, the user can set the graph as rotor graph, to
display global variables, or as blade graph to display local variables.
25
4.1 Rotor simulation variables
4.1 Rotor simulation variables
As stated before all variables calculated in a rotor simulation are dimensionless.
The global variables, computed during a rotor simulation are:
Power coefcient C
P
Thrust Coefcient C
T
Tip-speed ratio
Power coefcient based on tip-speed K
P
=
C
P

3
Inverse tip-speed ratio
1

The computed local variables of a rotor simulation are:

Axial induction factor a
Radial induction factor a

Local tip-speed ratio

loc
Radial position r
Dimensionless normal force component C
n
Dimensionless tangential force component C
t
Inow angle
Relative angle
Twist angle
Chord c
Lift coefcient C
L
Drag coefcient C
D
Lift to Drag Ratio
C
L
C
D
26
4 Simulation results
PRANDTL tip loss factor F
Number of iterations n
Annulus averaged axial induction factor
Annulus averaged radial induction factor
4.2 Turbine simulation variables
In a turbine simulation these global variables are computed additionally to all
the global variables from a rotor simulation:
Power P [W]
Thrust T [N]
Wind speed V
0
_
m
s

Angular frequency [rpm]

Pitch angle
WEIBULL probability h
w
WEIBULL probability wind speed
3
h
w
V
3
0
Root bending moment M[Nm]
Power coefcient including losses C
P,loss
These additional blade variables are computed during a turbine simulation:
Local REYNOLDS number Re
loc
=
V
rel
c

Re deviation from polar simulation Re

loc
Re
polar
Critical roughness k
critical
100

V
rel
[mm] from [10]
Resultant velocity V
rel
_
m
s

27
4.4 Validation of the results against a reference BEM code
Tangential force per length P
T
_
N
m

Normal force per length P

N
_
N
m

Mach number Ma
Circulation =
1
2
C
L
V
rel
c
_
m
2
s
_
4.3 General validity checks on simulation results
Because of the assumptions made in the derivation of the BEM method, two
checks need to be performed by the user to ensure the validity of the computa-
tion. It is assumed that the blade elements possess radial independence and that
there is no cross ow in radial direction from one element to another. To ensure
consistency with these assumptions the annulus averaged axial induction factor
F a and the circulation have to be relatively constant along the blade. A large
gradient in the averaged induction factor involves cross ow and a large gradi-
ent in circulation causes a radial dependent down wash. These checks should
especially performed at wind speeds where the turbine operates at rated power
and most notably in the outer part of the blade [6]. To check for the validity of
the used polars, the user may check deviation of the REYNOLDS number at which
the polar was computed to the local REYNOLDS number at the blade. When the
difference is higher that 10
6
, the polar data should be adapted to the ow condi-
tions at the turbine.
4.4 Validation of the results against a reference BEM
code
In this section the results, computed with the QBlade will be compared to the
results produced by a reference inhouse BEM code, based on an EXCEL spread-
sheet. Utilizing a classical BEM algorithm with optional PRANDTL tip and hub
loss corrections the reference software is suitable to validate the BEM algorithm
that has been implemented in XFOIL. The reference software utilizes a database
of precalculated polars. In order to compare the results of the two BEM al-
gorithms, the blade parameters and the airfoil polars have to be exactly the
same. For this comparison the polar data has been exported from the refer-
ence code and imported into QBlade. Both simulations were conducted with
28
4 Simulation results
only the PRANDTL tip and hub loss corrections turned on. Foil interpolation
was switched on in the new BEM code. In the reference software an element is
always bounded by two sections, thus the number of elements is xed.
Compared Turbine:
Regulation: Stall;
Transmission: Single;
Rotational speed: 15.47 rpm;
Cut in wind speed: 3.5
m
s
, Cut out wind speed: 25.00
m
s
;
Variable losses: 0.06, Fixed losses: 69000 W.
The turbine has a 3 bladed rotor:
Radius [m] Chord [m] Twist [

] Foil
1.2 2.19 12 Cylinder
2.7 2.19 12 Cylinder
4.7 2.49 12 TRANSIT
7.7 3.09 12 TRANSIT
9.2 3.2 11.009 DU 00-W-401
12.2 3.13 8.874 DU 00-W-350
15.2 2.97 7.253 DU 97-W-300
18.2 2.775 5.924 DU 91-W2-250
21.2 2.533 4.724 NACA 63(4)-421
24.2 2.285 3.636 NACA 63(4)-421
27.2 2.073 2.71 NACA 63(4)-421
30.2 1.863 1.977 NACA 63(3)-418
33.2 1.636 1.415 NACA 63(3)-418
36.2 1.414 0.911 NACA 63(3)-418
39.2 1.217 0.466 NACA 63(3)-418
42.2 0.994 0.037 NACA 63(3)-418
43.2 0.84 0 NACA 63(3)-418
44 0.52 0 NACA 63(3)-418
44.25 0.3 0 NACA 63(3)-418
44.45 0.05 0 NACA 63(3)-418
29
4.5 Global variables
Figure 4.2: The compared rotor blade
4.5 Global variables
For the WEIBULL distribution with k = 2 and A = 9.591
m
s
the predicted annual
yield of both programms is:
reference code: 7671 MWh
QBlade: 7495 MWh
This important result shows good agreement and has a relative difference of only
2.3%.
Figure 4.3: Power and thrust over wind speed
In general the power curves of the reference software and QBlade match well
(Fig. 4.3). The reference software predicts the maximumpower at 14.5
m
s
, QBlade
at 15
m
s
. A larger discrepancy can be found in the curves for thrust. Here the ref-
erence software predicts a thrust thats approximately 5% higher than predicted
by QBlade, however the overall tendency of the curves is the same.
The reference software predicts a lower C
P
for low wind speeds (or high tip-
speed ratios) as QBlade (Fig. 4.4). Even though this difference is quite large, its
effect on the power curve is little, since the low wind speeds do not contribute
much to the power output of the turbine. In contrast smaller differences between
30
4 Simulation results
Figure 4.4: C
P
and C
T
over windspeed
the two C
P
curves, in the region of high wind speeds are amplied in the power
curve. In general there is a good matching between the computed C
P
and C
T
curves.
Figure 4.5: Bending moment over wind speed
The characteristics of the bending moment curves match (Fig. 4.5). QBlade pre-
dicts a moment that is about 3% higher than the moment computed by the refer-
ence software.
4.6 Local variables
The comparison of the distribution for local variables, at a wind speed V
0
=
7.5
m
s
, is shown on the following pages.
31
4.6 Local variables
Figure 4.6: Axial and radial induction factor over radial position
The curves for the axial induction factor match everywhere except for the tip and
hub region (Fig. 4.6). This might be due to a different implementation of the tip
loss correction. The curves for the radial induction show a large difference in the
hub region. The reference software might limit the radial induction factor to a
value that cant be smaller than zero. In QBlade no such limit is implemented.
Figure 4.7: and over radial position
The curves for circulation and AoA both hardly show any differences (g.
4.9).
Because the AoA curves were matching and the polar data in both codes are
the same the C
L
and C
D
curves show good agreement (Fig. 4.7). The peak in
C
D
curve computed by QBlade is due to the higher radial resolution. In the
reference software the number of elements is dened by the number of sections
minus one. In QBlade the user can dene an arbitrary number of elements for
any blade.
32
4 Simulation results
Figure 4.8: C
L
and C
D
over radial position
Figure 4.9: V
rel
and Re over radial position
The V
rel
and Re curves show very good agreement (Fig. 4.8). Generally all the
compared curves have the same characteristics. There are slight differences for
some variables but the reasons behind it have not been investigated. In any
way the BEM method is a lower order accuracy analysis technique and draws
its usefulness from predicting overall performance characteristics and only load
and power approximations. The validation of QBlade can therefore be recorded
as successful.
4.7 Sensitivity analysis
In this section the inuence of the different simulation parameters and correc-
tions on a BEM simulation are pointed out. All graphs presented in this section
33
4.7 Sensitivity analysis
are screen shots from QBlade. The rotor with which this analysis was conducted
has three blades:
Radius [m] Chord [m] Twist [

] Foil
3 2 0 Cylinder
5 2 0 Cylinder
7 4 14.62 NACA 4450
9 4 14.62 NACA 4450
11 3 6.89 NACA 4425
15 1.83 4 NACA 4420
19 1.4 1.73 NACA 4418
23 1.24 1.27 NACA 4418
27 1.15 1.17 NACA 4414
31 1.09 0.37 NACA 4414
33.5 0.8 0.37 NACA 4414
34.5 0.7 0.37 NACA 4414
34.8 0.5 0.37 NACA 4414
35 0.1 0 NACA 4414
Foil Interpolation
Figure 4.10: Effect of Foil Interpolation
Figure 4.10 shows the effect of Foil interpolation on a blade that was simulated
with 50 elements. The left graph shows the effect on the C
P
curve. The C
P
value
34
4 Simulation results
is signicantly higher with the interpolation turned on. This is because the whole
root region from r = 3m to r = 7m is treated as being fully cylindric, with zero
lift without the interpolation. When the interpolation is turned on, the root re-
gion contributes to the total torque of the rotor because from the interpolation
between a cylindric polar with a NACA 4450 polar results a positive lift for el-
ements between 5-7m. This is more realistic since the real geometry is not fully
cylindric between 5-7m and should have a positive lift.
The distribution of the axial induction factor along the blade can be found with
and without interpolation in the right graph. In the interpolated curve the jumps
of the axial induction factor are smoothed to a more continuous behavior. For all
the following investigations the foil interpolation is turned on.
Epsilon
Figure 4.11: The effect of different epsilon values on simulation results
The left graph in Fig. 4.11 shows a section of the C
P
curve for different epsilon
values. The right graph shows the distribution of the axial induction factor. From
these graphs an epsilon value of 10
5
should provide enough accuracy to any
simulation.
Number of elements
In Fig. 4.12 different C
P
curves for a discretization with 10, 50 and 100 elements
are shown. For high tip-speed ratios a large difference between 10 and 50 ele-
ments is found. There is hardly a difference between 50 and 100 elements for the
whole range. The right graph shows the effect of different element numbers for
the distribution of the axial induction factor along the blade. Its obvious here
that 10 elements are a relatively coarse distribution that lter a lot of the peaks
35
4.7 Sensitivity analysis
Figure 4.12: Effect of the number of elements
in the curve. From this a minimum element number of 50 is recommended for
all simulations.
Tip loss corrections
Figure 4.13: Different tip loss corrections
In Fig. 4.13 the effect of the different tip loss corrections on the C
P
curve is il-
lustrated. The right graph shows that the tip loss corrections only affect the
distribution of the thrust coefcient in the tip region. C
T
curves show that the
New tip loss correction model after SHEN was motivated by the need to reduce
the thrust coefcient to zero at the tip.
The left graph in Fig. 4.14 shows the effect of the different corrections on the
axial induction factor, the right graph shows PRANTLs tip loss factor.
36
4 Simulation results
Figure 4.14: Different tip loss corrections
Figure 4.15: Different root loss corrections
Root loss corrections
The effect of the different root loss corrections on the C
P
curve is little. It becomes
more obvious in the distribution of the axial induction factor in the root region
(Right graph, Fig. 4.15).
3D correction
The 3D correction accounts for the HIMMELSKAMP effect. An impact on the C
P
curve for low tip speed ratios is visible in the left graph, Fig. 4.16. The right
graph shows that the airfoils close to the root experience an increased lift coef-
cient.
37
4.7 Sensitivity analysis
Figure 4.16: The 3D correction
Figure 4.17: CD90 = 1.0 (green), CD90 = 3.0 (blue)
CD 90 Value / Extrapolation
To investigate the impact of the polar extrapolation on the accuracy of the simu-
lation results the following blade was simulated:
38
4 Simulation results
Radius [m] Chord [m] Twist [

] Foil
4 3.96 27 NACA 4414
6.375 3.18 16.43 NACA 4414
9.75 2.33 9.38 NACA 4414
16.5 1.47 3.62 NACA 4414
18.1875 1.34 2.82 NACA 4414
19.875 1.23 2.17 NACA 4414
23.25 1.06 1.14 NACA 4414
30 0.83 9 NACA 4414
For the rst simulation the polar data from the NACA 4414 prole was extrap-
olated with a CD90 value of 1.0. The second simulation was carried out with
polar data that was extrapolated with a CD90 value of 3.0. The different polars
can be found in Fig. 4.17.
Figure 4.18: The effect of the extrapolation
The left graph in Fig. 4.18 shows that the large difference in the polar data only
has an effect on the C
P
curve for low tip-speed ratios, when the AoAs are high
enough to be in the extrapolated region of the polars. When the turbine is op-
erating at the highest efciency the AoAs are relatively small, they are mostly
on the linear part of the lift curve. The right graph shows the dimensionless
power curve K
P
over a range of tip-speed ratios. Here the extrapolated polars
make a large difference in the region of high wind speeds (V
0

1

), a xed speed
turbines operates at low tip speed ratios and the AoAs increase. The quality of
the extrapolated polar data especially matters for the simulation of single speed
stall turbines, where the point of maximumpower output and the corresponding
wind speed is of importance.
39
5 How to create a blade simulation
This section will give a short introduction on how to design and simulate a ro-
tor or turbine with QBlade. Only the very basic functionalities are mentioned
here and this is merely an overview in which succession the modules and sub-
modules are to be run through to create a blade and simulate it. All the functions
not mentioned in this part are more or less self-explanatory and the user may ex-
periment freely with them.
Figure 5.1: QBlade startup screen
Figure 5.1 shows the start screen of QBlade. From the File menu the user can
select the different applications (the menu point Rotor and Turbine Design distin-
guishes QBlade from XFLR. During this tutorial the modules Direct Foil Design,
XFoil Direct analysis and Rotor and Turbine Design are used. More information
about XFOIL and XFLR5 in general can be found in [3] and [2].
40
5 How to create a blade simulation
Figure 5.2: Direct Foil Design module
The introduction starts in the Direct Foil Design module. The rst thing that needs
to be done when creating a rotor is to create its airfoils. Airfoils can be created
using splines, a NACA airfoil generator or via an import function in XFLR5. In
this case only one airfoil, the NACA 5518 was created.
41
Figure 5.3: XFOIL Direct Analysis module
In the XFOIL Direct Analysis module the ow around the airfoils is simulated to
create a polar. An analysis can be dened under the menu point Polars, then the
simulation can be started. The analysis will only converge for a limited range
of AoA values, typically from about -5

AoA to +25

AoA. It is also possible to

import polar data in this module, whenever importing polar data, an arbitrary
airfoil with exactly the same name as the imported polar data needs to be created
to connect the polar data with this airfoil.
42
5 How to create a blade simulation
Figure 5.4: Rotor and Turbine Design module / 360

polar interpolation
To simulate a wind turbine the AoA range of the polar needs to be extrapolated
to 360

, this is done in the submodule for polar extrapolation. Only extrapolated

polar data can be used to simulate a turbine or rotor. The sliders on the left side
of Fig. 5.4 can be used to better match the extrapolation with the computed polar
data. Below the sliders a value for C
D,90
can be specied. When the extrapolated
polar is saved it is stored in the runtime database. In the 360 Polar menu, it is
possible to automatically load a cylindrical foil. This foil has zero lift and a drag
value that can be specied by the user. Cylindrical foils are used in the root
region of a blade for structural reasons.
43
Figure 5.5: Rotor and Turbine Design module / Blade design
If one or more 360

polars have been created a blade can be designed. The Op-

timize and Scale buttons open the corresponding dialog windows. Stations can
be added or removed, one airfoil and one polar has to be specied at every sec-
tion. Additionally every blade has a Number of Blades dened, that species
how of how many blades the rotor exists. When a blade is fully dened it can be
saved. Instead of creating a new blade it is possible to edit stored blades. When
this is done a copy of the stored blade is edited. When a blade from the database
is deleted or overwritten all associated simulations and turbines are deleted as
well.
44
5 How to create a blade simulation
Figure 5.6: Rotor and Turbine Design module / Rotor simulation
All blades in the database can be simulated in the Rotor Simulation sub-module.
Whenever a simulation is dened all simulation parameters need to be specied.
The simulation then is conducted over the desired range of tip-speed ratios. The
three graphs showthe simulation results. By double-clicking on a graph the user
may change the variables on the x- and y-axis. By right-clicking on a graph the
user may change the graph type to display local or global variables. The local
variables are always displayed for every computed lambda value. The curve
belonging to the currently selected lambda value (in the right drop-down menu)
is highlighted. In the graphs context menu the user may isolate the highlighted
curve and then compare it to the local variable curve of other simulations. If
multiple simulations are stored in the database, the curves of global variables
are always shown simultaneously.
45
Figure 5.7: Rotor and Turbine Design module / Turbine denition and
simulation
In the Turbine denition and simulation sub-module a turbine can be dened only
if a blade is stored in the database. The user species all turbine parameters and
saves the turbine. Then a turbine simulation, over a range of wind speeds, can be
conducted. When the user sets the k and A values for the WEIBULL distribution,
the annual yield is computed automatically.
In the le menu the user can save the whole conducted work, including foils,
polars, blades and simulations as a .wpa le.
46
Bibliography
[1] BMU Kurzinfo Windenergie, 2010 [online], Available from:
http://www.erneuerbare-energien.de/inhalt/4642/ [Accessed 26 May 2010]
[2] DEPERROIS, A.: XFLR5 Analysis of foils and wings operating at low reynolds
numbers, 2009 [online], Availiable from: http://xr5.sourceforge.net/xr5.htm
[Accessed 19 February 2010]
[3] DRELA, M.; YOUNGREN, H.: XFOIL 6.94 User Guide, MIT Aero & Astro,
2001
[4] GASCH, R.; TWELE, J.: Windkraftanlagen Grundlagen, Entwurf, Planung und
Betrieb, Teubner, Wiesbaden, 2007
[5] HANSEN Martin O. L.: Aerodynamics of Wind Turbines. Earthscan, London,
2nd Edition, 2008.
[6] VAN LANGEN, P.J.: Blade Optimization Tool User Manual, ECN-C-06-006, 2006
[7] MAHERI, A.; NOROOZI, S.; TOOMER, C.; VONNEY, J.: Damping the uctu-
ating behavior and improving the convergence rate of the axial induction factor
in the BEMT-based aerodynamic codes, University of West England BS16 1QJ,
Bristol, 2006
[8] MIKKELSEN, R: Actuator Disk Methods Applied to Wind Turbines, Dissertation
MEK-FM-PHD 2003-02, Technical University of Denmark, 2003
[9] MONTGOMERIE, B: Methods for Root Effects, Tip Effects and Extending the An-
gle of Attack Range to +-100

, with Application to Aerodynamics for Blades on

Wind Turbines and Propellers, FOI Swedish Defence Research Agency, Scien-
tic Report FOI-R-1035-SE, 2004
[10] PECHLIVANOGLOU, G.: The Effect of Distributed Roughness on the Power Per-
formance of Wind Turbines, GT2010-23512, Berlin University of Technology,
Berlin, 2010
47
Bibliography
[11] SHEN, W.Z.; MIKKELSEN, R.; SORENSEN, J.N.; BAK, C.: Tip Loss Corrections
for Wind Turbine Computations. Wind Energy 2005. Wiley, 2005
[12] TANGLER, J.: The Nebulous Art of using Wind-Tunnel Airfoil Data for Predicting
Rotor Performance, NREL/CP-500-31243, National Renewable Energy Labo-
ratory, Golden, Colorado, 2002
48