Study on the simulation of a

Vertical Axis Wind Turbine

(VAWT)
Document:
Report

Author:
Jordi Serrat Guevara

Director:
Pau Nualart Nieto

Degree:
Bachelor’s in Aerospace Technology Engineering

Examination session:
Spring, 2022.
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

Acknowledgements
I would like to express my gratitude to the people
who helped me during the course of this project, as
well as the entire degree.

To Pau Nualart Nieto, for his help, commitment

and dedication from the beginning of the project.
To my family, for their support and patience during
these four years. To my friends for making this
period of my life memorable.
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

Abstract

This project represents the final thesis of the Bachelor’s Degree in Aerospace Technologies
Engineering. In this thesis the development of a simulation code for H-Type Vertical Axis Wind
Turbines based on the Double Multiple Streamtube Model is presented.
The aim of the project, aside from the coding of the program itself, is to apply knowledge
gained during the degree to an area of expertise which was not explored in class, as well as
learning more about it. Another objective is to develop the programming abilities in Python, a
language which is not taught in any course and is important in engineering applications.
Regarding the study itself, first a general overview on wind energy is developed: a few
concepts related to this field and important to the development of the project are introduced,
and then the state of the art in the wind turbine field is presented.
Then the code is developed: both formulation and its implementation in the code are detailed.
Once the raw model has been implemented, a convergence study is carried out as the preliminary
validation, and then several corrections are added in order to account for specific phenomena
which are not initially taken into consideration. Next, conclusions about their effect on the
results are extracted.
When the code is closed, it is applied to the calculation of instantaneous torque generated by
a rotor, and the effect of the number of blades in its variability is quantified. Finally the results
of an experimental campaign are compared with those obtained with the simulation code.
Some discrepancy is found between the program’s results and experimental measurements,
and possible reasons for it are given at the end of the project. However, the code has proven
useful to perform analyses during the early stages of a turbine’s design, mainly due to its low
computational requirements and the ability to capture the effects of design features on the rotor’s
performance.
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

Resum

Aquest projecte representa el Treball de Final d’Estudis del Grau en Enginyeria en Tecnolo-
gies Aeroespacials. En aquesta tesi es presenta el desenvolupament d’un codi de simulació per
aerogeneradors d’eix vertical de tipus H, basat en el model de múltiple tub de corrent doble
(DMST).
L’objectiu del projecte, a més de programar el codi com a tal, és aplicar coneixements ad-
quirits durant el grau a una àrea que no ha sigut explorada a classe, aixı́ com aprendre’n més.
Un altre objectiu és desenvolupar les habilitats de programació en Python, un llenguatge que no
s’ensenya a cap assignatura i és important en aplicacions d’enginyeria.
Pel que fa l’estudi com a tal, primer es dóna una visió general de l’energia eòlica: s’intro-
dueixen alguns conceptes relacionats amb aquest camp i importants per al desenvolupament del
projecte, i després es presenta l’estat de l’art en el camp dels aerogeneradors.
Després es desenvolupa el codi: es detallen tant la formulació com la seva implementació en
el codi. Un cop el model base s’ha implementat, es duu a terme un estudi de convergència com
a validació preliminar, i llavors s’afegeixen diverses correccions per tenir en compte fenòmens
especı́fics que inicialment no es consideren. Llavors s’extreuen conclusions sobre com afecten els
resultats.
Quan el codi està tancat, s’aplica al càlcul del parell instantani generat per un rotor, i es
quantifica l’efecte del nombre de pales en la seva variabilitat. Finalment els resultats d’una
campanya experimental es comparen amb aquells obtinguts amb el codi de simulació.
S’ha trobat alguna discrepància entre els resultats del programa i les mesures experimentals,
i possibles raons per aquest fet es donen al final del projecte. Tot i això, el codi ha demostrat
ser útil per realitzar anàlisis durant les etapes inicials d’un disseny de turbina, principalment
degut als seus reuduı̈ts requeriments computacionals i a la capacitat de capturar els efectes dels
diversos paràmetres de disseny en el rendiment del rotor.
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

Contents
1 Introduction 9
1.1 Aim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.2 Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3 Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.4 Justification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.5 Schedule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

2 Important concepts of wind energy 17

2.1 Power coefficient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2 Betz limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3 Tip speed ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.4 Cp - λ curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

3 State of the art 19

3.1 Classification of wind turbines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.1.1 Horizontal Axis Wind Turbines (HAWT) . . . . . . . . . . . . . . . . . . . . 19
3.1.2 Vertical Axis Wind Turbines (VAWT) . . . . . . . . . . . . . . . . . . . . . . 21
3.2 Current VAWT technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.3 Simulation methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.3.1 Double Multiple Stream Tube . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.3.2 Vortex lattice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.3.3 Cascade model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.3.4 Computational Fluid Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . 31

4 DMST method algorithm 33

4.1 Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.1.1 Coordinate system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.1.2 Domain discretization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.1.3 Momentum Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.1.4 Blade Element theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.2 Algorithm flowchart . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

5 Validation and preliminary results 48

5.1 Convergence analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
5.2 Preliminary results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

1
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

6 Tip loss correction 52

7 Strut correction 55

8 Dynamic stall correction 58

8.1 Considered models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
8.1.1 Leishman-Beddoes model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
8.1.2 Gormont model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
8.1.3 Massé and Berg correction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
8.1.4 Strickland correction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
8.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

9 Tower shadow correction 67

10 Additional studies 71
10.1 Torque value variability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
10.2 Comparison with experimental data . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

11 Conclusions 77
11.1 General conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
11.2 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

2
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

List of Figures
1 Cumulative power capacity of the installed wind energy worldwide since 2001. Sources:
[1], [2], [3]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2 Representative Cp − λ curve. Source: [8]. . . . . . . . . . . . . . . . . . . . . . . . . 18
3 Resistance HAWT, commonly known as American windmill. Source: [9] . . . . . . . 20
4 Vestas V162-6.8MW wind turbine. Source: [10] . . . . . . . . . . . . . . . . . . . . . 20
5 Three different commercial 600W VAWTs. Source: Amazon . . . . . . . . . . . . . . 21
6 Bus stop in Iceland. Source: [11]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
7 Representation of different VAWT concepts. Source: [12] . . . . . . . . . . . . . . . . 23
8 Diagram of the forces generated by the airfoil of a 4-blade VAWT. . . . . . . . . . . 24
9 Comparison between typical values of power coefficient for different types of wind
turbines. Source: [15] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
10 Illustration of the DMST model principle. Source: [16]. . . . . . . . . . . . . . . . . 28
11 Representation of the bound and shed vortices of a blade element. Source: [17]. . . . 29
12 Representation of a series of bound ring vortices. Source: [19]. . . . . . . . . . . . . 29
13 Cascade representation of an axial pump or compressor. Source: [23] . . . . . . . . . 30
14 Cascade representation of a vertical axis wind turbine. Source: [22] . . . . . . . . . . 30
15 Streamline representation of the flow around a VAWT. Source: [26] . . . . . . . . . . 31
16 Coordinate system: top view. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
17 Coordinate system: front view. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
18 Angular discretization of the domain. Source: [30]. . . . . . . . . . . . . . . . . . . . 34
19 Front view of the rotor’s discretization. Source: [31], modified. . . . . . . . . . . . . 35
20 Diagram of the streamtube and relevant parameters defined by an actuator disk.
Source: [4] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
21 Thrust coefficient against induction factor, theoretical results and empirical approx-
imations. Source: [32]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
22 Representation of base suction. a) a = 1, b) 0 < a < 1. Source: [33]. . . . . . . . . . 39
23 Diagram of the relevant velocities, angles and forces acting on the airfoil. Source:
[30], modified. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
24 Flowchart of the main function. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
25 Flowchart of the solver function. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
26 Lambda-averaged relative error as a function of number of divisions. . . . . . . . . . 49
27 Relative error as a function of number of divisions for one tip speed ratio. . . . . . . 49
28 Power coefficient curve for the raw DMST code. . . . . . . . . . . . . . . . . . . . . . 51
29 Power coefficient comparison: raw code and tip loss corrected. . . . . . . . . . . . . . 54

3
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

30 Power coefficient comparison: raw code, tip loss corrected and strut corrected. . . . . 57
31 Example of the dynamic stall effect on an airfoil’s polar curves. Source: [38] . . . . . 58
32 Lift coefficient against azimuthal position for various TSRs using Strickland adaptation. 62
33 Lift coefficient against azimuthal position for various TSRs using Strickland and Berg
adaptations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
34 Power coefficient comparison: raw code, tip loss, strut and dynamic stall corrected. . 64
35 Lift coefficient vs angle of attack with Strickland model + Massé and Berg correction,
λ = 2.53. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
36 Drag coefficient vs angle of attack with Strickland model + Massé and Berg correc-
tion, λ = 2.53. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
37 Graphic representation of the tower wake model. Source: [47] . . . . . . . . . . . . . 67
38 Power coefficient comparison: raw code, tip loss, strut and tower corrected. . . . . . 68
39 Upwind speed as a function of azimuthal position for different tower diameters and
λ = 3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
40 Generated torque at λ = 3 during one revolution for different numbers of blades. a)
Nb = 2 b) Nb = 3 c) Nb = 4 d) Nb = 5 . . . . . . . . . . . . . . . . . . . . . . . . . . 72
41 Drawing of the rotor of study. Source: [50] . . . . . . . . . . . . . . . . . . . . . . . . 73
42 Simulated and experimental results comparison. . . . . . . . . . . . . . . . . . . . . . 74
43 Blade Reynolds number against azimuth angle. . . . . . . . . . . . . . . . . . . . . . 75
44 Simulated and experimental results comparison with low Reynolds polars. . . . . . . 76
45 Depiction of the flow curvature effect. Source: [52]. . . . . . . . . . . . . . . . . . . . 79
46 Depiction of the streamtube expansion’s effect on the discretization of the rotor.
Source: [40]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

4
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

List of Tables
1 Rotor parameters for the convergence study. . . . . . . . . . . . . . . . . . . . . . . . 48
2 Computational time required for different discretizations. . . . . . . . . . . . . . . . 50
3 Rotor parameters for the results analysis. . . . . . . . . . . . . . . . . . . . . . . . . 51
4 Maximum power coefficient comparison. . . . . . . . . . . . . . . . . . . . . . . . . . 54
5 Maximum power coefficient comparison. . . . . . . . . . . . . . . . . . . . . . . . . . 57
6 Maximum power coefficient comparison. . . . . . . . . . . . . . . . . . . . . . . . . . 64
7 Maximum power coefficient comparison. . . . . . . . . . . . . . . . . . . . . . . . . . 69
8 Power coefficient at λ = 3 for different tower diameters. . . . . . . . . . . . . . . . . 70
9 Maximum and average torque relationship for different numbers of blades. . . . . . . 72
10 Rotor parameters for the experimental results. . . . . . . . . . . . . . . . . . . . . . 74

5
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

Nomenclature
α̇ Angle of attack temporal variation rate

α Angle of attack

Ar Rotor frontal area

αref Dynamic stall reference angle

Ast Streamtube cross-sectional area

x, y, z Axis nomenclature

BEM T Blade Element Momentum Theory

BET Blade Element Theory

c Blade chord

Cd Drag coefficient

Cl Lift coefficient

Cn Normal force coefficient

Cp Power coefficient

Ct Tangential force coefficient

CT Thrust coefficient

∆h Blade element height

∆θ Streamtube azimuth increment

F Tip loss factor

H Rotor height

6
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

HAW T Horizontal Axis Wind Turbine

a Induction factor

L Langth

λ, T SR Tip speed ratio

MT Momentum Theory

Nb Number of blades

Nel Number of elements per blade (vertical discretization)

Nst Number of total streamtubes, or number of actuator disks per half-cycle.

Nstrut Number of struts

ω Rotor angular speed

p−
d Pressure at the downwind surface of the actuator disk

p+
d Pressure at the upwind surface of the actuator disk

p∞ Upwind pressure

R Rotor radius

r Radial position

ρ Air density

t Thickness

τ Torque

t/c Thickness to chord ratio

θst Azimuthal angle of an actuator disk integration point

7
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

θ Azimuthal angle

T Thrust

Ud Incident wind speed at the actuator disk or airfoil

Uin Actuator disk upwind speed

U∞ Upwind speed

Un Normal relative speed

Ur Relative speed

Ut Tangential relative speed

Uw Actuator disk downwind speed

V AW T Vertical Axis Wind Turbine

8
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

1 Introduction

1.1 Aim

The aim of this project is to create a Darrieus type Vertical Axis Wind Turbine (VAWT) simulation
code and assess its performance, both in numerical convergence and comparing to data found in
other academic papers.

The code will start from a simplified standpoint and increase in complexity and capabilities as the
project advances. The target is to observe the sensitivity that the different implemented features
on the code have on the results.

After that the program will be applied to a case study in order to find out the effect of a design
parameter on certain performance metrics.

9
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

1.2 Scope

The study will include:

• An explanation of the most significant concepts in the wind energy field.

• An introduction and description of the current VAWT technology and simulation methods.

• The derivation of the main mathematical expressions used in the simulation code, based on
the Double Multiple Stream Tube model.

• A flow-chart of the code.

• The code of the program itself, as well as the routine developed to post-process the results.

• A comparison between the versions of the code.

• Conclusions regarding the validation of the code.

• A case study using the developed tool.

• The necessary budget for the development of the study.

The study will not include:

• A thorough explanation of the theories used outside the DMST model, such as those possibly
dynamic stall, strut and pole interference, etc.

• The study of the coupling between the wind turbine and electric generator. The simulation
will consider just the aerodynamic performance and output of the rotor, which would then
have to be coupled to a generator in order to harness the electrical power. This also means
that mechanical and electrical losses, as well the control system will not be studied.

• The scenario of initiation of operation, where for example a small Savonius turbine can be
used to start the turbine’s movement, or any transient scenario, for that matter.

• Study of the physical construction of the turbines, including materials, construction methods
and load analyses on any of the components.

10
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

1.3 Requirements

The requirements for the study are the following:

• The program must be written in Python, with exception of the post-processing routine, which
can be done in Matlab if it proves to be too time-consuming otherwise (finding and learning
how to use the necessary libraries).

• The code must be structured in a clear manner and the program must be relatively easy to
understand and use. This includes comments within the code, easy to modify input data and
easy to obtain output data.

• The theory on which the code will be based is the Blade Element Momentum Theory, applied
to VAWTs with the Double Multiple Stream Tube model.

• The entirety of the project must be completed before the established deadline (June 22nd )
and have a total workload of 300 hours.

11
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

1.4 Justification

During the last decades, wind energy generation has experienced a massive growth across the
globe, and the need to further transition to sustainable sources of energy will continue to push the
development of those renewable, including wind energy.

Figure 1: Cumulative power capacity of the installed wind energy worldwide since 2001. Sources:
[1], [2], [3].

As seen in figure 1, there is clearly an upward trend, and according to experts the installation of this
type of energy will have to continue to increase in order to accomplish the environmental objectives
set for the future [3].

Currently, virtually the only choice for large scale wind energy harnessing is the use of conventional
Horizontal Axis Wind Turbines (HAWT), since in the past they were deemed to offer the most
advantages for this type of operation, and subsequently have been the main focus of development
since then.

However, saturation is starting to become an issue for onshore installations, particularly in Europe
[4], and so recently focus is being given to offshore facilities, as well as decentralised generation in
buildings and houses [5].

In this regard VAWTs could be competitive with HAWTs, in the first case due to their advantageous
upscaling behaviour, and in the second case because of the reduced noise values, as well as a better
performance in irregular wind conditions and smaller space requirements. These concepts will be
explained further in section 3.

12
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

Though research of large scale offshore VAWTs is still at an initial stage [6], small-scale onshore
devices are already commercially available. For this reason, there is currently a niche which can be
filled by VAWTs in urbanized areas or household applications, where the aforementioned factors
are a big issue.

The presence of this need for the development of vertical axis wind turbines implies a need for
simulation tools for their design.

Models based on Blade Element Momentum Theory have proved to be useful in the design of
HAWTs, mainly in the early stages, due to their relative simplicity and reduced computation time
compared to full Computational Fluid Dynamics simulations, as well as the accuracy of their results.
For the same reason, the preliminary design of a VAWT could also benefit from such a tool.

BEMT for VAWTs is not a new method, since the first algorithms of this kind were already developed
as early as the 1980’s (for example the DARTER code developed in the Sandia laboratories [7]).
However, some of the theories that it includes have evolved over the years, and the creation of such
a tool represents a valuable exercise, on the one hand because of the concepts it applies, and on the
other hand because of the code development and the post-processing and analysis of the results.

13
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

1.5 Schedule

DESCRIPTION OF THE TASKS

1. Planning: Definition of the aim of the project and the preliminary tasks and schedule.

2. Project charter: Elaboration of the project charter document, exact definition of the scope of
the project and its requirements, detailed definition of the tasks and schedule.

3. Introduction: Redaction of the introduction of the project, which contains the aim, scope,
requirements and justification.

4. State of the art: Research and writing of the state of the art on wind turbines and specifically
on vertical axis turbines. This includes a classification of wind turbines, an overview of the
current VAWT technology and an exposition of different simulation methods.

5. Important concepts of wind energy: General concepts relevant to wind energy, such as power
coefficient, tip speed ratio, etc. The aim is to introduce the concepts which will be used later
on in the project, both to simulate and evaluate the performance of the desired VAWTs.

6. Description and formulation of the used method: Include in the report the detailed explanation
and development of the formulation of the DMST method, which is used in the simulations.
This includes:

• Assumptions and limitations

• Momentum Theory thrust coefficient.

• Discretization of the VAWT surface.

• Blade Element Theory formulation.

• BET and MT combination.

• Integration of the results.

7. Polar curves: Obtain polar curves for one or more airfoils for a complete angle of attack range,
from existing databases.

8. Algorithm flow chart: Develop the flow chart for the implementation of the method in a
program.

9. Basic simulation code: Code the basic simulation program and obtain the benchmark results.

14
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

10. Increase code complexity and capabilities: Implement corrections to increase the code’s capa-
bilities, such as tip loss, dynamic stall, tower shadow, etc.

11. Post processing routine: Develop the routine to plot results and obtain the relevant perfor-
mance values from the simulations. The routine must at least obtain the Cp − λ curve and
the torque variation during a revolution of the rotor, aside from other interesting studies such
as the effect of blade solidity on the results.

12. Validation of the code: Perform a convergence study and provide preliminary results to be
used as a base for the further modifications of the program. in order to find whether results
are valid or not. Make this comparison with the different versions of the code.

13. Case study: Apply the code to a study of interest and extract conclusions on it. Search
experimental data on a VAWT and compare it with results obtained with the code, with the
objective of assessing its correlation with real results.

14. Budget: Elaborate the budget for the development of the project.

15. Conclusions and future work: Extract conclusions on the project and the developed code.
Explain further developments which could be made to the project in the future.

16. Formalities and revision of the final documents: Final writing of necessary elements in the final
documentation, such as abstract, table of contents, list of tables, list of figures, nomenclature,
references, etc. These will be developed during the course of the project but a time period at
the end of it has been allocated for their revision and modification, if needed.

17. Presentation: Elaboration and preparation of the presentation.

GANTT CHART

The described tasks have been organised in a Gantt chart, found in the following page. Highlighted
in red are the deadlines for the documentation hand-in (June 22nd ) and the presentation dates
(starting on July 11th ).

15
 February March April
Week 1 Week 2 Week 3 Week 4 Week 4 Week 4 Week 5 Week 6 Week 7 Week 8 Week 9 Week 10 Week 11
Task id. Task
01/02/2022 07/02/2022 14/02/2022 21/02/2022 28/02/2022 07/03/2022 14/03/2022 21/03/2022 28/03/2022 04/04/2022 11/04/2022 18/04/2022 25/04/2022
1 Planning
2 Project charter
3 Introduction
4 State of the art
5 Wind energy important concepts
6 DMST method formulation
7 Polar curves
8 Algorithm flow chart
9 Simulation code (basic)
10 Post processing routine
11 Increase code complexity and capabilitites
12 Code validation
13 Case study
14 Budget
15 Conclusions
16 Formalities and revision of the final documents
17 Presentation

May June July

Week 12 Week 13 Week 14 Week 15 Week 16 Week 17 Week 18 Week 19 Week 20 Week 21 Week 22 Week 23
Task id. Task
02/05/2022 09/05/2022 16/05/2022 23/05/2022 30/05/2022 06/06/2022 13/06/2022 20/06/2022 27/06/2022 04/07/2022 11/07/2022 18/07/2022
1 Planning
2 Project charter
3 Introduction
4 State of the art
5 Wind energy important concepts
6 DMST method formulation
7 Polar curves
8 Algorithm flow chart
9 Simulation code (basic)
11 Post processing routine
10 Increase code complexity and capabilitites
12 Code validation
13 Case study
14 Budget
15 Conclusions
16 Formalities and revision of the final documents
17 Presentation
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

2 Important concepts of wind energy

Before getting deep into the project’s contents, a few important concepts which will be used through-
out the project are introduced.

2.1 Power coefficient

The power coefficient or Cp is the main parameter used in wind turbines to characterize performance.
It is defined as:
P
Cp = 1 3
(2.1)
2 ρU∞ Ar

Although it does not give direct power figures, it provides a measure of how much kinetic energy is
extracted from the wind, out of the total available wind power which goes through the rotor area.

2.2 Betz limit

Using the actuator disk theory, which will be developed further in section 4.1.3, an important
conclusion can be extracted [4]. It starts from the formulated value of the thrust coefficient:

T
Ct = 1 2
= 4a(1 − a) (2.2)
2 ρAd U∞

Now using the definition of power coefficient and knowing that P = T · Ud and Ud = U∞ (1 − a):

T · U∞ (1 − a)
Cp = 1 3
= CT (1 − a) = 4a(1 − a)2 (2.3)
2 ρA U
d ∞

This expression can be used to find a theoretical maximum power coefficient:

dCp 1
= 0 −→ 4(1 − a)2 − 8a(1 − a) = 0 −→ a = (2.4)
da 3
Cp,max ≈ 0.6 (2.5)

This value is known as the Betz limit, and no wind turbine, no matter how efficient, can exceed it.
Due to the idealistic nature of the hypotheses made in the theory used to derive the aforementioned
expressions (see section 4.1.3), the power coefficient values tend to be relatively far from this number,
as it will be displayed in figure 9.

17
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

2.3 Tip speed ratio

The tip speed ratio is a dimensionless measure of the speed of the rotor compared to the wind speed.
It is defined as:

ωR
λ= (2.6)
U∞

2.4 Cp - λ curves

Using the aforementioned concepts of power coefficient and tip speed ratio, a curve can be obtained
of the relationship between both values. This curve depends on each turbine’s design parameters
and it is used to determine the rotor power for any combination of wind and rotor speed [8], which
is helpful since it concentrates this information in a single plot instead of having to use multiple
lines to plot power against angular speed, for example, where one line for each wind speed is needed.

Moreover, it is helpful to visualize in a clear manner the effect on performance of any design
modifications. It typically looks like figure 2.

Figure 2: Representative Cp − λ curve. Source: [8].

18
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

3 State of the art

A wind turbine is a power generating device that converts kinetic energy from wind into usable
energy, either to directly use the generated momentum -such as grain grinders and water pumps
used in the past, in which case they are called windmills-, or to produce electricity. In this last case,
the turbine’s shaft is connected to a generator which uses electromagnetic induction to generate the
voltage.

Various types of turbines exist, each with its own particularities. The aim of this section is to
present them and to expand on Darrieus type Vertical Axis Wind Turbines, which are the focus of
the project.

3.1 Classification of wind turbines

The most common criterion for the classification of wind turbines is their axis orientation, through
which two types of wind turbines can be distinguished.

3.1.1 Horizontal Axis Wind Turbines (HAWT)

As the name suggests, in this type of turbine the axis is located on the same plane as the wind
speed, with its blades perpendicular to it. They are the most commonly used nowadays.

They normally require a yaw control system, since they need to be aligned with the wind direction,
This is widely implemented via sensors and servo actuators, though in some turbines the pointing
can be done by a simple wind vane or even without the need for any mechanism, which is the case
of downwind turbines. In this last case, the momentum caused by the misalignment between wind
and rotor corrects it. However, they suffer from the aerodynamic interference caused by the nacelle
and are not predominantly implemented at large scales.

Within the HAWT group, two main different concepts can be distinguished, which differ in the
medium with which the torque is generated: lift or drag.

• Resistance HAWT:

These turbines use the drag generated by the air coming in contact with the blades to rotate.
They were the first type to appear and are characterised by their low efficiency and low tip
speed ratios. Nowadays they are out of use, and in their time they were used to grind grain
or pump water.

19
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

Figure 3: Resistance HAWT, commonly known as American windmill. Source: [9]

• Lifting HAWT:

These turbines use the lift generated by the blade, whose cross-section has the shape of an
airfoil. They work at a high tip speed ratio and they are are the most common type of turbines
due to their high efficiency, low torque ripple, overall reliability and the potential to produce
the largest amount of power out of all the current technologies.

However, they do still have some disadvantages, such as the fact that large turbines have to
be installed high above the ground due to the rotor radius and thus the tower must endure
demanding loads. Moreover, because of the size of the components the transportation to the
installation site is challenging, and the assembly process once there is complex and requires
the use of large cranes.

Figure 4: Vestas V162-6.8MW wind turbine. Source: [10]

20
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

Aside from this, HAWTs can be classified according to other characteristics, such as the number
of blades, type of power control (stall or pitch), safety system (aerodynamic or mechanic brakes),
rotor position with respect to the tower (upwind or downwind), etc. [4].

3.1.2 Vertical Axis Wind Turbines (VAWT)

In this case, the axis of the wind turbine is oriented perpendicularly to the direction of the wind,
with the blades running along the axis tough not necessarily parallel to it.

This has some major implications on the operation of the turbine: unlike HAWTs, where the
air speed vector relative to the blade remains constant throughout the movement of the rotor, in
VAWTs this vector changes direction depending on the rotor’s azimuthal position. Primarily, this
means that VAWTs do not need to be pointed towards the wind in order to operate, although other
implications appear which will be expanded further in section 3.2.

Opposite to HAWTs, which have converged into one main conceptual design, there is a large dispar-
ity in VAWT solutions, being there many different commercially available solutions for a particular
power output. Such an example can be found in figure 5.

Figure 5: Three different commercial 600W VAWTs. Source: Amazon

As with HAWTs, in VAWTs the same main classification can be made according to the way that
wind is used to generate power.

• Resistance VAWT:

Also known as Savonius rotor turbines, they use drag to generate the torque. In this case,
there is a difference in aerodynamic drag between the blades which are facing towards and
away from the wind due to their scoop-like shape.

21
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

These turbines are less efficient than the lifting type and have less power output potential,
but are functional at low wind speeds. Nowadays they are primarily used in low output
applications, such as household or more recently in urban infrastructures: for example, since
2015 they have been used in Iceland’s bus stops, which are also shelters from the cold and
rain, to power their lighting and heating [11] (see figure 6). Outside the wind energy field
these are the type of devices found in instruments such as cup anemometers.

Their shape and size varies widely from one design to another, with the number of blades
ranging from two up to tens, some twisting them to reduce torque ripple, others with semi-
spherical shape or a more complex one such as the red rotor seen in figure 5.

Figure 6: Bus stop in Iceland. Source: [11].

• Lifting VAWT:

Also known as Darrieus turbines, the blades use an airfoil to generate the lift that turns the
rotor. They work at a slightly lower tip speed ratio than lifting HAWTs and have a low
starting torque. This means that the turbine needs a relatively fast wind speed in order to be
able to start by itself.

For this reason, lifting VAWTs normally require a starting system, which can consist on the
generator working as a motor, an independent motor or a Savonius type rotor installed on the
shaft.

22
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

As it has been mentioned, conceptually there are many different alternatives, the most common
of which are:

– Troposkien type, with curved blades attached to the tower by the top and bottom part
resembling an egg beater.

– H type, with simple straight blades.

– Helix type, a variation of the H-type which rolls the blades forming a helix in order to
reduce torque ripple

These examples can be seen represented in figure 7, together with a Savonius rotor.

Figure 7: Representation of different VAWT concepts. Source: [12]

More radical designs have appeared over the years, some built or just conceived to solve
different design challenges or to optimize the turbine to a particular application. However,
these designs have not seen much success due to the added complexity they introduce against
the benefits they provide.

3.2 Current VAWT technology

Although they currently represent a fraction of the installed wind power, VAWTs had had an
important presence in history, being in fact believed to be the first type of wind turbine to be built
around 200 B.C., which at the time was used for grinding grain [13]. More recently and in the field
of interest, VAWTs were heavily researched during the 1970s through the 1980s and 90s, with a
hiatus since the late 1990s and regaining attention just recently [6].

23
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

As it has been said, the main advantage is the capability to generate power regardless of the wind
direction, which would in this regard imply the need for fewer components and reduced complexity
compared to HAWTs, and therefore lower cost.

Moreover, due to their shape they do not need to be built on top of high towers with the gearbox
located at the highest point, which makes HAWTs more dangerous, difficult and expensive to build
and service. On the contrary, VAWTs’ gearboxes are located at the base of the shaft, near the
ground, reducing risk and cost [14].

The operation of VAWTs, although not as straightforward as that of HAWTs, is based on the same
principle: when the blades move through the air current they generate aerodynamic lift and drag
forces, the resultant of which has a component which is tangential to the turning movement of the
rotor an thus generates a positive torque. This is not as simple to visualize as in HAWTs, where the
blades see the same angle of attack regardless of their azimuthal position, but instead it is variable
during the movement. This is depicted in figure 8, where the force decomposition of each of the
blades at a given azimuthal position is depicted.

Figure 8: Diagram of the forces generated by the airfoil of a 4-blade VAWT.

If they present so many advantages, why do they currently represent such a small part of the wind
energy market? It all comes down to a few reasons:

24
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

Firstly, VAWTs have been thought to be inherently unreliable until recent years, which is the main
reason why in the 70s ans 80s they fell short to HAWTs, which were deemed to be more competitive.
This is due to structural failures in early VAWT prototypes and models during those years, caused
by the cyclical stresses that the components of the turbines endure. The available materials and
production methods back then, like the fact that the blades were predominantly made of extruded
aluminum panels, as well as a certain lack of knowledge in terms of material science, particularly
fatigue loading and its effects, made VAWTs specially susceptible to crack under this type of stresses
[6].

Secondly, there is the cost of energy: while HAWTs produce power during the entire revolution and
they do so in a constant manner, except for some interference when they go over the tower, VAWTs
only do it during part of it and irregularly. This means that the amount of energy which can be
harnessed with HAWTs is higher and it can be done in a more efficient manner, since their blades
can be designed so that the airfoil works the majority of the time at the most efficient angle of attack
[13]. Figure 9 shows a comparison of the typical values of power coefficient expected from various
types of turbines as a function of tip speed ratio. The turbine which offers the best performance is
the three-bladed HAWT, while the Darrieus rotor has about 20% lower Cp .

Figure 9: Comparison between typical values of power coefficient for different types
of wind turbines. Source: [15]

25
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

Technological developments in recent years, such as the introduction of composite materials and
new production methods, as well as a better understanding of the challenges of VAWT design,
have allowed to mitigate the initial problems that ultimately hindered the implementation and
investigation of VAWTs in favor of HAWTs in the past.

This has been the main driver of the interest gain in VAWT technology, rather than big design or
concept changes.

Currently, the commercial applications of VAWTs are limited to small-scale operations in houses
and buildings, with relatively simple and robust designs.

Scientific research, however, is focused on the scaling of the turbines to multi-megawatt applications
for offshore use, which arises from the fact that VAWTs could be theoretically scaled to higher
power outputs than HAWTs due to the presence of cyclical loads of aerodynamic nature instead of
gravitational nature, which should be easier to manage when upscaling.

For this reason, new possibilities such as blade pitch control are being studied in order to determine
whether the increase in performance they imply is worth the added complexity. Research in this
direction is still in an early stage, with some small prototypes being tested onshore and a few
offshore [6]. Time will tell if they will end up being competitive with HAWTs or not.

3.3 Simulation methods

In order to simulate a VAWT, one could take different approaches. This section offers a brief
overview of some of the methods which can be used with this purpose.

3.3.1 Double Multiple Stream Tube

The Double Multiple Stream Tube method, or DMST, is a model based on the Blade Element
Momentum Theory (BEMT), which is broadly used in preliminary performance calculations in
HAWTs.

The first part of the method is based in the so-called actuator disks, which represent the area where
kinetic energy is extracted from the wind into the wind turbine’s rotor.

This area is traditionally portrayed as circular, although in reality its shape is not relevant. The
airflow that passes through the actuator disk moves within a streamtube, during which flow velocity
and pressure experience changes: On the one hand, the velocity profile is continuous, since a sudden

26
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

change would imply infinite acceleration (not realistic); on the other hand, pressure does experience
an abrupt change between both sides of the actuator disk. See figure 20 for graphic reference.

At the inlet of the streamtube, the airflow travels at the upstream velocity and is at ambient
pressure. At the outlet (downstream), the pressure matches that from its surroundings (ambient)
but the velocity is not recovered, hence the extracted kinetic energy.

In this streamtube the conservation equations can be applied and, with some simplifications, the
pressure increment experienced through the actuator disk can be obtained as a function of an
induction factor a, which characterizes the velocity at the actuator disk as following:

Ud = U∞ · (1 − a) (3.1)

This is known as Momentum Theory, and it represents half of the BEMT method.

The other half of the method is the Blade Element Theory (BET), which is based on the forces
generated by the blades’ airfoil, extracted from its polar curves. The angle of attack is a function
of the incident speed on the airfoil, and thus depends on the induction factor. The experienced
pressure increment can be obtained through the coefficients found in the polar curves, and it can
then be equated to the one obtained by Momentum Theory. Since BET does not provide an
analytical expression, the induction factor is found iteratively. Once the result is obtained, the
desired parameters are calculated.

In HAWTs its implementation is based on the discretization of the domain in radially-distributed

actuator disks, in which the blade element has the same angle of attack during the entire revolution.

In VAWTs, however, this is not possible since the angle of attack is a function of the azimuth angle.
For this reason, the discretization is grid-like, with actuator disks located perpendicularly to the
airflow.

However, this poses the question of the downstream actuator disks, and in which way they are
affected by the presence of the upstream ones. The adopted solution is for each streamtube to
contain two actuator disks: one for the upstream part of the rotor and one for the downstream
one. Then, the downstream velocity of the air that passes through the first streamtube will be the
upstream velocity for the second streamtube. This is shown in figure 10.

Although the used equations are different, the process to be followed is similar as for HAWTs, and
so the induction factor is calculated for each streamtube, as well as other relevant parameters which
are finally integrated along the entire rotor revolution.

27
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

Figure 10: Illustration of the DMST model principle. Source: [16].

3.3.2 Vortex lattice

The vortex lattice model, also known as vortex line model, is based on the modelling of the blades
and their wake by a number of discrete vortex lattices or lines, each of them with their determined
strength (circulation) [17].

In this model the supposition is made that the flow is both incompressible and inviscid in all the
regions except the thin layer around lifting bodies and their wake.

As seen in figure 11, the blade is discretized in elements, each of which contain one or a series
of bound vortices, which can be followed by a trail of shed vortices. They are what is known
as horseshoe vortices, the succession of which one after the other creates elements known as ring
vortices.

The lifting line is located at the quarter-chord point of each element so as to make it coincide
with the aerodynamic centre. According to the thin airfoil theory, in order to enforce the stream
boundary condition (flow is tangent to the surface of the airfoil when in contact with it), a control
point is located at the 3/4 chord point, where the zero tangential velocity condition will be applied
[18].

28
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

Figure 11: Representation of the bound and shed Figure 12: Representation of a series of bound
vortices of a blade element. Source: [17]. ring vortices. Source: [19].

By virtue of the expressions of induced velocity by a vortex line at a certain distance, the contri-
butions of all the lifting lines can be added for each control point (plus the wind speed, of course)
and the normal component equated to zero, obtaining a linear system of equations. When solved,
the system provides the strength of each vortex line [20].

Following the Kutta-Joukowsky theorem, the lift generated per unit spam by each bound lifting
line is:
l = ρVrel Γb (3.2)

where Γb is the strength or circulation of the vortex line. In reality, this lift component is not
perfectly perpendicular to the airspeed, since each specific bound lifting line receives induced down-
wash components from all the other vortex lines. This allows for, once the circulation is known, the
calculation of an effective angle of attack and an induced drag [21]:

d = ρVrel Γb sin(αi ) (3.3)

where αi is the induced angle of attack. Finally, the total lift and induced drag are found by adding
their value multiplied by the span for all bound vortex lines.

The advantage of this method is the speed, since nowadays solving a large linear system of equations
is computationally relatively easy, although one clear disadvantage is the inability to obtain viscous
drag values, which have to be calculated in some other way in order to obtain accurate results.
Moreover, the model relies on conclusions derived from the thin airfoil theory, which are not always
true in the case of VAWTs since relatively thick airfoils tend to be used.

29
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

3.3.3 Cascade model

The cascade model was first introduced to VAWT analysis by Hirsch and Mandal [22]. The model is
based on the cascade theory that is used extensively in the performance analysis of turbomachinery
(figure 13). In this model the VAWT is considered to be a turbomachine without a casing and the
turbine’s blades are assumed to be positioned in a plane or a cascade, a representation of which can
be seen in figure 14.

Figure 13: Cascade representation of an axial pump or compressor. Source: [23]

Figure 14: Cascade representation of a vertical axis wind turbine. Source: [22]

30
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

The advantage of this model mainly lies on the possibility to calculate blade interactions at a
relatively low computational cost. However, it has some weak points, mainly the absence of per-
turbations from the wake of the upstream half of the rotor, affecting the downstream part of the
rotor, since the blades are assumed to be in the same plane perpendicular to the wind [24].

3.3.4 Computational Fluid Dynamics

Arguably the most precise method of them all, CFD solvers have one thing in common which
separates them from the aforementioned models: instead of the mathematical modelling of the
aerodynamic phenomena taking place in the turbine rotor, they solve, in one way or the other, the
Navier-Stokes equations or other fluid governing equations [25].

This is done by creating a mesh around the rotor, and after solving the flow equations integrating the
results on the blades’ surfaces, and thus obtaining the relevant performance and load parameters.

Aside from this, this type of programs allow for the visualization of the flow passing around the rotor
and its components, allowing a detailed characterization of aspects such as the wake, aerodynamic
interferences and turbulence, with the aid of specific turbulence models.

Figure 15: Streamline representation of the flow around a VAWT. Source: [26]

The difference from method to method lies on the discretization scheme, which in turn determines
how the Navier-Stokes equations will be solved. Some of the most well known schemes are:

31
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

• Finite Volume Method (FVM): The main characteristic of this discretization scheme is that
it solves the integral form of the conservation laws, making it less susceptible to inaccuracies
and instabilities caused by sharp gradients and generally simplifying the calculations [27].

• Finite Element Method (FEM): Commonly used in structural calculations, it can also be
applied to fluid dynamics cases. The main characteristic is the resolution of the equations
in their differential form, although this is done by assuming the shape of the solution via
what is known as shape functions. This means that the solution within the discrete elements
follows a distinctive shape, in which certain parameters can vary but not the basic shape.
The advantage is that simple shape functions can be used, but through a fine discretization
an arbitrary distribution can be represented [28].

• Finite Difference Method (FDM): This method differs from the rest in that the differential
equations are simplified locally as Taylor series expansions, so that they cease to be derivateves
per se, and increments are evaluated at each element of the grid. Its relevance relies on the
fact that it is the easiest to program due to its relative simplicity in the formulation [29].

Although these methods offer by far the best post-processing capabilities and do not rely on models
of different phenomena, but simulate the fluid itself, they are computationally very expensive and
almost impossible to program by oneself. This large simulation time makes it difficult to make
extensive batch analyses to compare the effect of certain parameters quickly, and so its use is more
focused in detailed design and optimisation.

In most cases, moreover, simulating and post processing tools are not open-source and require
paying for a license.

32
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

4 DMST method algorithm

As previously mentioned, the simulation method developed in this project is the Double Multiple
Stream Tube Model (DMST).

4.1 Formulation

The first part of the method relies on the conclusions extracted from the momentum theory, so the
first part of this section will be devoted to their obtainment.

4.1.1 Coordinate system

First of all, for coherence reasons it is important to set a coordinate system, which will be used in
the entire project.

Figure 16: Coordinate system: top view.

Figure 17: Coordinate system: front view.

In the front view the observer is located in the same direction as the wind.

4.1.2 Domain discretization

As mentioned in section 3.3, the domain is discretized in a series of streamtubes, each of which has
two actuator disks. First of all, let us talk about the angular discretization, the coordinates for
which will be those of figure 18.

33
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

Figure 18: Angular discretization of the domain. Source: [30].

With regards to the width, each streamtube takes the same ∆θ so that a finer discretization is
obtained near the edges of the rotor, where the greatest change in angle of attack is expected due to
the circular trajectory of the blades. The definition of the azimuthal angle increment is correlated
to the desired number of streamtubes as:

π
∆θ = (4.1)
Nst

where Nst represents the number of streamtubes in a rotor’s half-circle. The integration point for
each streamtube is located at the medium point of the arc that they describe, and it will be the θst
angle used to calculate the necessary velocities, angle of attack, direction of the forces, etc.

For the vertical discretization, which can be seen in figure 19, an uniform distribution has been
implemented, with a characteristic parameter ∆h calculated as:

H
∆h = (4.2)
Nel

where H is the rotor’s height and Nel is the number of streamtubes in the vertical direction, which
coincides with the number of elements in which the blade is discretized.

34
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

Figure 19: Front view of the rotor’s discretization. Source: [31], modified.

With this parameters, the front area of each streamtube can be obtained:
 
∆θ ∆θ
Ast = ∆hR cos(θst − ) − cos(θst + ) (4.3)
2 2

With this step the discretization of the domain is completed and the formulation can be developed.

4.1.3 Momentum Theory

Classical Momentum Theory

Also known as the Rankine-Froude theory, it introduces the following assumptions:

• Perfect flow

• Steady movement

• Uniform velocity in sections parallel to the actuator disks

• Incompressible flow

• No air rotation effects

35
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

These assumptions will be used to apply various simplifications to the fluid mechanics’ conservation
equations in their integral form. To start with, figure 20 shows a diagram which will be followed to
develop the formulation (see section 3.3 for a brief explanation on the behaviour of the variables).

Figure 20: Diagram of the streamtube and relevant parameters defined by an actuator disk. Source:
[4]

The first examined equation is the conservation of mass:

Z Z
d
ρdV + ρ(⃗v · ⃗n)dS = 0 (4.4)
dt V S

Since the flow is considered steady, the derivative can be eliminated. Then, the surface integral can
be evaluated in three places: Upstream, the actuator disk and downstream, yielding the following:

ρA∞ U∞ = ρAd Ud = ρAw Uw (4.5)

where, as seen in section 3.3, Ud is defined as:

Ud = U∞ (1 − a) (4.6)

Next the momentum conservation equation is applied as following:

Z Z Z Z Z
d
ρ⃗v dV + ρ⃗v (⃗v · ⃗n)dS = − p⃗ndS + ⃗nτ dS + ρFm dV (4.7)
dt V S S S V

36
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

Eliminating the time derivative, the viscous term and the body forces, the equation can be simplified
as:
−
ρUw2 Aw − ρU∞
2
A∞ = −(p+
d − pd )Ad = −T (4.8)

where the T denotes that the right hand side of the equation is the equivalent to a thrust force,
which will be important later on.

Next, the energy conservation equation can be developed:

Z   Z   Z Z Z Z
d 1 2 1 2 ⃗
ρ e + v dV + ρ e + v ⃗v ·⃗ndS = ρfm ·⃗v dV + (⃗n·τ )·⃗v dS+ (−⃗q)·⃗ndS+ QdV
V dt 2 S 2 V S S V
(4.9)
From this equation, the time derivative, viscous term, body forces term and heat transfer terms
can be eliminated due to the initial hypotheses, and what is left can be expressed as Bernoulli’s
equation:
1 2
ρU + p + ρgy = cnst (4.10)
2

This can be applied to either side of the actuator disk, so as to obtain the following system:

1 2 1
ρU + p∞ = ρUd2 + p+ (4.11)
2 ∞ 2 d
1 2 1
ρUw + p∞ = ρUd2 + p−
d (4.12)
2 2

which yields:
− 1
p+ 2 2
d − pd = 2 ρ(U∞ − Uw ) (4.13)

Rearranging equations (4.5), (4.6), (4.8) and (4.13) the value of the downstream velocity can be ob-
tained as a function of the induction factor, as well as the area of the streamtube in the downstream
section:

Uw = U∞ (1 − 2a) (4.14)
1−a
Aw = Ad (4.15)
1 − 2a

With a similar process, and taking advantage of the thrust found in equation (4.8), the following
expression can be found:
2
T = 2ρAd U∞ a(1 − a) (4.16)

37
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

which can be written as a dimensionless parameter only as a function of the induction factor:

T
CT = 1 2
= 4a(1 − a) (4.17)
2 ρAd U∞

This is the ultimate conclusion from this theory, and it is used in the BEMT method to obtain the
value for the induction factor.

Glauert correction

As ideal as the developed theory may seem, upon closer inspection it can be seen that for a > 0.5
the theory gives negative values of downstream velocity and downstream streamtube area, which
are not physically possible.

Moreover, at values near a = 0.5 results start deviating from experimental measurements. For this
reason, Glauert fitted a parabola to experimental data taken from rotors operating in the turbulent
windmill state [32], which can be seen in figure 21.

Figure 21: Thrust coefficient against induction factor, theoretical results and empirical approxima-
tions. Source: [32].

The correction meets the classical momentum theory curve at a = 0.4, and from then on the thrust
coefficient’s expression is:
0.0203 − (a − 0.143)2
CT = 0.889 − (4.18)
0.6427

38
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

Advanced Momentum Theory

It has been mentioned that for induction factor values near a = 0.5, results cease to be accurate.
One of the reasons why this happens, is because the model assumes that no mixing occurs between
the wake and the flow surrounding it. This means, that the boundary of the actuator disk matches
the boundary of the streamtube.

This is a fairly acceptable hypothesis at low induction factors, where a minimal amount of mixing
would occur. However, at larger induction factors, the abrupt decrease in pressure at the downwind
face of the actuator disk would cause a significant flow of air to spill into the wake. Another way
of seeing it, would be that not all the airflow that reaches the actuator disk goes through it, but
some of it circumvents the disk and afterwards rejoins the flow which has gone through the disk
(see figure 22).

This is known as base suction, and its effects can be taken into account according to a relatively
recent study (see reference [33]).

Figure 22: Representation of base suction. a) a = 1, b) 0 < a < 1. Source: [33].

The study expands the theory developed by Taylor ([34]), which yields the same result as the
classical Momentum Theory but characterizes the actuator disk as a distribution of sources of equal
strength. The new model develops this idea and allows for a region where flow mixing occurs and
base suction appears. The most important conclusions regarding this theory are the following:

39
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

4 3−a
CT = ·a· (4.19)
3 1+a
1−a
Uw = U∞ (4.20)
1+a

Although more suitable than the classical Momentum Theory, at high induction factors (a > 0.7)
Glauert’s empirical correction fits experimental data in a better way, so its use is favorable in this
induction factor range.

As an additional conclusion the Betz limit can also be proved with this expression, following the
same process as depicted in section 2.

4 (3 − a)(1 − a)
Cp = CT (1 − a) = ·a· (4.21)
3 1+a
dCp
= 0 −→ a = 0.3705 (4.22)
da
Cp,max = 0.5966 ≈ 0.6 (4.23)

The obtained power coefficient maximum value is also around 0.6.

4.1.4 Blade Element theory

This is the part which concerns the aerodynamic forces generated by the blades.

In order tho characterize them, it is important to define the relevant angles and vectors that will be
calculated in order to be able to establish relationships between them. With this aim, the diagram
shown in figure 23 will be used.

40
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

Figure 23: Diagram of the relevant velocities, angles and forces acting on the airfoil. Source: [30],
modified.

The ultimate aim is to obtain the value of the thrust represented in figure 23 as Th , from now on
denominated T . To do so, the following process is followed:

First, the incident wind speed at the airfoil is calculated:

Ud = Uin (1 − a) (4.24)

Note that the freestream velocity is not necessarily U∞ , since it will only take this value on the
front half of the movement. On the other hand, at the rear half this value will be the downstream
velocity obtained for the front half.

Then the wind speed and the velocity caused by the rotation of the blades can be decomposed in
normal and tangential components, which will be helpful to find the angle of attack and later the
total air relative velocity. The components are:

Un = Uin (1 − a)sin(θ) (4.25)

Ut = ωR + Uin (1 − a)cos(θ) (4.26)

41
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

And so the angle of attack is simply:

 
Un
α = atan (4.27)
Ut

Once the angle of attack is found, the lift and drag coefficients are obtained from the airfoil’s polar
curves and next, these coefficients are projected onto the normal and tangential axis as following:

Cn = Cd sin(α) + Cl cos(α) (4.28)

Ct = Cl sin(α) − Cd cos(α) (4.29)

Once this step is completed, the thrust coefficient can be found as a function of the azimuth angle
(θ).
CT = Cn sin(θ) − Ct cos(θ) (4.30)

The thrust generated by a blade at a given moment can be calculated with the following expression:

1
T = ρAb Ur2 CT (4.31)
2

where Ab is the blade element area:

Ab = ∆h · c (4.32)

and Ur is the total relative velocity, which is found by adding the normal and tangential components
calculated in equations 4.25 and 4.26, respectively:
q p
Ur = Un2 + Ut2 = (Uin (1 − a)sin(θst ))2 + (ωR + Uin (1 − a)cos(θst ))2 (4.33)

In order to be able to compare this thrust to the one found via the Momentum Theory, two things
must be done:

1. The resultant thrust acting in a particular streamtube is the sum of each blade’s cycle-average
value. This happens beacuse the blade only passes through the streamtube for a θ increment
out of the entire revolution [30] and so, the equivalent value of thrust is:
Z θf
Nb
Teq = T dθ (4.34)
2π θ0

42
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

2. The thrust coefficient must be nondimensionalized with the same magnitudes as in the MT
case, which are the freestream velocity and the streamtube area:

T
CT = 1 2
(4.35)
2 ρAst Uin

Combining equations 4.34 and 4.35 the following is obtained:

R θf
Nb θ0 T dθ
CT = (4.36)
2π 12 ρUin
2A
st

When the rotor’s discretization is fine, the ∆θ that each streamtube represents can be considered
infinitesimal. In this case, the integral in equation 4.36 can be in practice substituted by the
instantaneous value of T multplied by ∆θ:

Nb T ∆θ
CT = (4.37)
2π 12 ρUin
2A
st

This is the value which can be compared to the value found with the Momentum Theory, and all
together is what is known as the Blade Element Momentum Theory.

As it can be seen, an analytical dependence between this value and the induction factor cannot
be established, partly due to the relative complexity of the expressions but mainly because of the
presence of the polar curves, which do not follow an analytical function. For this reason, the only
way to solve this problem is to guess an induction factor value in the beginning and perform this
process iteratively in order to find the solution.

Once this itreative process is completed, a similar process to the thrust coefficient calculation can
be followed to find the torque and, finally, the power coefficient.

Firstly, the instantaneous torque produced by one blade element is:

1
τ = ρUr2 Ab RCt (4.38)
2

This can be used to calculate the power generated along a θ span as:

Z θf
Nb
P = τ ωdθ (4.39)
2π θ0

43
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

Which can then be applied to the contribution of all the streamtubes to calculate the total power
generated. Furthermore, the integral can be once again simplified as in equation 4.37.

Nel 2N
!
st
Nb X X
Ptot = τij ω∆θ (4.40)
2π
j=1 i=1

Finally, the power coefficient can be calculated.

Ptot
CP = 1 3
(4.41)
2 ρU∞ Ar

Where Ar is the rotor’s frontal area.

4.2 Algorithm flowchart

In order to perform the calculations which have been explained until this point, a Python code has
been made from scratch with the following structure:

In the main file, the first section is devoted to the inputs: here the user defines parameters such
as rotor radius, height, blade number, chord, airfoil polar files, etc. After that the preliminary
calculations are performed, mainly the actuator disks’ integration points azimuth and y coordinate,
streamtube areas, etc. As well as reading the polar files and saving them in array form.

After this the solver begins, in which three loops are performed:

1. Tip speed ratio (λ): The user defines a tip speed ratio array, so the program performs a batch
analysis of as many cases as desired.

2. Vertical (y) position: the vertical position is also swept to calculate for all the rotor’s ”layers”.

3. Azimuth (θst ): Calculates each actuator disk at a given heigt.

With this process, results are obtained first for an entire revolution at the lowest vertical position,
and this process is performed for all heights. Once this has finished, a new tip speed ratio value is
selected and the process is performed again for the remaining TSRs.

Within the azimuthal angle’s loop is where the iterative process to find the induced factor is made.
It is performed with the fsolve function, which is implemented in the scipy library and uses Powell’s
method in order to find the roots of a system of equations by minimising the sum of squares [35].

44
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

In this case only one equation must be solved. The function internally gives values to a and
ultimately tries to minimise the result until a certain tolerance is achieved of a variable commonly
denominated F , which is calculated with equations 4.19 (or the Glauert correction) and 4.37 as:

F = CT,M T − CT,BET (4.42)

Upon solving the problem, the function returns the induction factor’s value, with which the func-
tion is called again, although this time to calculate and retrieve other relevant variables such as
downstream velocity, blade element relative velocity, angle of attack, torque, etc. Most of these are
saved for post-processing purposes, and torque will be used to calculate the power coefficients.

These values are saved first in a one-dimensional array where every position represents an azimuthal
position. When the azimuth loop is done, these vectors are appended into another array, where
every cell represents a vertical position and contains the vector of azimuthal values. Finally, when
the vertical position loop is also done, these two-dimensional arrays are appended into a final array,
every position of which contains one tip speed ratio case.

With this method, results are accessible, for example for the angle of attack as:

α[λ][y][θst ] (4.43)

where it is understood that λ, y and θst are their respective indexes.

After this, the power coefficient values are calculated using the expressions depicted in 4.40 and
4.41 and lastly the post-processing routine is called, where the desired plots are made.

The flowchart followed by the algorithm is now presented. It has been divided between the main
function, which shows the external processes that the program makes (initial calculations, sweep of
the domain, post-process, etc.) and the equation’s solver function.

45
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

START

Input data
Polar curves

Previous calculations

Assign λ=λ[0]
y=y[0], θ=θst[0]

Assign Uin

Solve equation for a

Save a, α, τ, Uw ,etc.

No
θ=θst[-1]? Next θ

Yes

θ=θst[0]

No
y=y[-1]? Next y

Yes

y=y[0]

No
λ=λ[-1]? Next λ

Calculate Cp

Postprocess

END

Figure 24: Flowchart of the main function.

46
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

SOLVER FUNCTION
CALLED

Guess a value

Yes CT, MT with Glauert

a > 0.624?
correction

No

CT, MT with theoretical

expression

Calculate Ua, Un, Ut, Ur

Calculate α

Extract Cl and
Cd from polars

Calculate Cn and Ct

Calculate cycle average CT, BET (thrust

coefficient)

F = CT, MT - CT, BET

No
|F| < tolerance ?

Yes

RETURN a VALUE

Figure 25: Flowchart of the solver function.

47
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

5 Validation and preliminary results

With the theory developed in the previous section, and following the flow charts which have been
presented, the simulation code has been built. In order to assess its validity, the first test which
must be performed is a convergence analysis.

5.1 Convergence analysis

This test will provide information regarding the variation of the results as a function of the number of
azimuthal divisions in which the rotor has been separated. Due to the lack of an analytical solution,
the study will be performed taking as a reference a simulation performed with a large number of
elements, about ten times more than the amount which is expected to be finally accepted.

The parameters of the rotor upon which the study will be performed are presented in table 1.

Table 1: Rotor parameters for the convergence study.

Parameter Value

Number of blades 3
Airfoil NACA 0021
Chord [m] 0.086
Rotor height [m] 1
Rotor radius [m] 0.515
Upwind velocity [m/s] 3.1

Two methods are presented: The first one calculates the mean relative error produced in each tip
speed ratio by a determined number of elements, which could be expressed as following:

PNλ
i=1 |(Cp,i − Cp,ref )/Cp,ref |
ϵ(Nst ) = · 100 (5.1)
Nλ

where Nλ is the length of the tip speed ratio vector.

The second one calculates the evolution of the relative error with the number of elements, but in
this case for a single tip speed ratio value. The λ corresponding to a power coefficient close to the
maximum has been selected in order to ensure its relevance. In this case the error is simply:

Cp − Cp,ref
ϵ(Nst ) = · 100 (5.2)
Cp,ref

48
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

The obtained results are:

Figure 26: Lambda-averaged relative error as a function of number of divisions.

Figure 27: Relative error as a function of number of divisions for one tip speed ratio.

49
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

In both cases the reference results have been obtained for a number of streamtubes equal to Nst =
400. It can be seen that the error decreases rapidly upon increasing the number of streamtubes,
and there is a large portion of the cases which have an error that is invisible to the naked eye.

This indicates both that the reference value is large enough and that the solution converges. Taking
the case with the largest error (figure 26), a suitable number of streamtubes to perform further
analysis is deemed to be:
Nst = 30

which implies an average error well under 1% and will help keeping computational time lower than
if a larger number of streamtubes were to be used.

The program’s execution time is an important factor in the determination of the discretization
number. With the used laptop computer 1 , the following average times per tip speed ratio case
were obtained:

Table 2: Computational time required for different discretizations.

Nst Time per TSR case [s]

30 0.75
50 1.10
100 2.16

Seeing how low these times are, increasing the number of streamtubes per semi-circle to 50 or 100
for a simple study with 10 or 20 tip speed ratios is not much of a sacrifice, with the condition that
there is no vertical discretization (with no tip loss effects it makes no sense to make it). When
the vertical discretization is introduced, times in table 2 are multiplied by the number of vertical
elements, so the Nst parameter becomes more important.

5.2 Preliminary results

Once the converge of the code has been assessed, some preliminary results are presented for this
first version. The idea is to then apply a set of corrections for different phenomena and analyze the
changes they induce.

1
find model in the budget document

50
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

For this analysis, the following rotor has been used.

Table 3: Rotor parameters for the results analysis.

Parameter Value

Number of blades 3
Airfoil NACA 0021
Chord [m] 0.5
Rotor height [m] 10
Rotor radius [m] 5
Rotor speed [rpm] 25

With these inputs, results are shown in figure 28.

Figure 28: Power coefficient curve for the raw DMST code.

It can be seen that the curve, although never exceeding the Betz limit, is quite high compared to
what is expected from VAWTs (see figure 9). This can be attributed to the highly ideal nature
of the model, which does not consider effects such as tip loss, dynamic stall and pole and struts
interference, among others. The purpose of the following sections is to increase the code capabilities
to take these effects into consideration.

As a demonstration, some additional results that the code extracts from every study are shown in
annex G. These include the induction factor values as a function of azimuth for every tip speed
ratio case, as well as lift coefficient, etc.

51
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

6 Tip loss correction

The first and probably most obvious correction which can be made is the tip losses. This phe-
nomenon has two implications:

1. Downwash component caused by the pressure difference on the upper and lower parts of the
blade. It is the one taken into consideration in Prandtl’s well known finite wing theory.

2. The spanwise component of the velocity, which is not so obvious but relevant to this project.
Its cause is found in the aforementioned downwash: Since part of the airflow goes around
the blade near its tips, air must travel towards the blade tips, especially in the regions close
to them. This effect is known as vertical streamtube expansion, and it effectively causes a
reduction of the momentum transferred to the airfoil, in benefit to the increase of momentum
given to the airflow in order to travel towards the blade tip.

The methodology applied to implement tip losses is taken from reference [31] which, for the spanwise
velocity, proposes the use of a correction factor thought of especially for VAWTs following Willmer’s
modified version of the Prandtl method, which was developed for propellers.

The idea behind it is to replace the wake of the turbine by a series of flat disks, around which a
potential flow can be described. This allows for the definition of a factor F , which accounts for a
momentum transfer reduction in a given streamtube. In practice, the factor models the fraction
of the velocity normal to the blade which remains after some of the airflow gets deflected in the
spanwise direction. It is defined as:

acos(e−πa/s )
F = (6.1)
acos(e−πH/(2·s) )

Which is defined taking H as the entire height of the rotor, and a and s are calculated as:

πUw
s= (6.2)
Nb ω
H
a= − |y| (6.3)
2

where y is the vertical coordinate of the streamtube, taking the central point of the rotor as the
origin. Upon analysing these expressions, it is observed that the F value is equal to 1 in the equator
of the rotor, while it reaches 0 at the tips of the blades.

52
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

The implications of this phenomenon lie on the calculation of the normal speed presented in equation
4.25:

Un = F Uin (1 − a)sin(θst ) (6.4)

The tangential velocity is calculated equally as previously defined, as in equation 4.26.

The angle of attack and relative velocity however, do suffer changes since they are calculated using
the normal velocity:

   
Un F Uin (1 − a)sin(θst )
α = atan = atan (6.5)
Ut ωR + Uin (1 − a)cos(θst )
q p
Ur = Un2 + Ut2 = (F Uin (1 − a)sin(θst ))2 + (ωR + Uin (1 − a)cos(θst ))2 (6.6)

As for the downwash induced loss, the paper proposes the use of Prandtl’s finite wing theory, which
assumes an elliptical lift distribution along the span of the blade. The model corrects the lift
coefficient given by the entire blade with a function of the aspect ratio.

This makes it difficult to implement in an algorithm such as the present one, since the Cl values
are calculated point by point to find the thrust coefficient and then the induction factor. Prandtl’s
finite wing correction would modify the wing’s CL as a whole, and so it could not be taken into
consideration in the Blade Element Theory.

For this reason, in the present study the tip loss correction will only be made for the spanwise
component of the flow, with the aforementioned F factor.

Once the correction is applied, a comparison between the corrected and the raw code can be es-
tablished. Using the same parameters as in section 5.2 (see table 1), the following results are
obtained:

53
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

Figure 29: Power coefficient comparison: raw code and tip loss corrected.

Table 4: Maximum power coefficient comparison.

Case Cp,max ∆Cp,max [%] λ(Cp,max )

Raw code 0.561 - 3.44

Tip loss 0.536 −4.46 3.56

Although not drastically, the power coefficient values go down as expected. The general shape of the
curve remains constant, with the maximum Cp values around the same tip speed ratio. However, it
seems that the loss in performance is larger at the TSRs with the highest power coefficient values.

As a final comment, good convergence was observed for a vertical discretization of Nel = 20, which
increased the computational time required according to what was mentioned in section 5.1.

54
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

7 Strut correction
The blades’ supporting elements play a crucial role in the structural integrity of the wind turbine.
However, due to their rotation along with the blades they have an unavoidable undesired effect in
the form of a resistive torque.

At low tip speed ratios, this effect is less predominant: due to the high wind speed in comparison
with the struts, in one half of the rotation the strut does travel against the wind and therefore a
resistive torque appears, while in the other half the strut travels in the same direction as the wind
and thus is ”pushed”. Although the struts which travel upwind receive a higher relative velocity
than those which travel downwind, due to the low tip speed ratio this difference is comparatively
small and therefore the resistive torque effects are small.

At larger tip speed ratios, on the other hand, the speed of the struts moving in the same direction
as the wind starts to become larger than the wind speed, and so they start to generate resistance
instead of positive torque. At the same time the struts which travel against the wind receive an even
bigger relative speed. This effect is magnified as tip ratios become even larger, hence the increased
importance of this phenomenon at large tip speed ratios.

Another phenomenon which arises from the presence of struts is the disturbances in the airflow that
they generate, affecting the blades when they go thorough the downwind portion of the rotation.
This disturbance would depend both in the thickness and shape of the struts, as well as their number
and overall dimensions in relation with the diameter and height of the rotor.

Implementing a correction for this effect would incorporate a lot of complexity to the code, and
its characterisation would be characterized much better with CFD. For this reason it is considered
out of the scope of the project and therefore the hypotheses that the strut’s dimensions are small
compared to the rotor will be made.

To calculate the aforementioned resistive torque, the struts are modeled as cylinders, and they are
discretized for every azimuthal position (i), in as many elements as streamtubes they cross. Then
the normal relative velocity is calculated for the center point of each of the elements (j) in the
following manner:

Un,ij = −(Uij · cos(θi ) + ω · rij ) (7.1)

55
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

where r is the radial distance of the element’s centre and U is the wind speed going through
the streamtube, which in the case of the upwind part of the rotation is interpolated between the
velocity at the disk Ud and the downwind velocity Uw . In the downwind part of the movement this
interpolation is done between the input velocity Ui n and the velocity at the disk.

Then the elemental torque can be easily obtained by virtue of the section’s drag coefficient as:
 
1 2
τij = ρU Lij tij CD,ij (7.2)
2 n,ij

Note that the drag coefficient is referenced to the surface seen in the direction of the wind (in this
case the length of the element multiplied by its thickness). This coefficient was initially set as an
input of the program. However, this number would be expected to suffer large variations from
one relative airspeed to another since it can be dependant on the Reynolds number, and so an
interpolation has been applied following the data acquired in [36] (see annex C).

Now the elemental torque can be integrated along the entire strut as following:

Nel,j
X
τi = τij (7.3)
j=1

Once this process is completed for the entire cycle, a torque value will be obtained for each azimuthal
position, which is the same as what was obtained by the BEMT analysis. So, both torque values
are then added and cycle averaged as shown in equation 4.40.

The code allows for the implementation of any number of hoizontal struts, the necessary inputs
being a vector of the vertical positions where the struts are located and their thickness. Then the
program finds the index of the height division where each strut is located and performs the depicted
process for each of them.

The most complex part of this correction lies in the discretization, since for each azimuthal point
the number of elements and their locations change, and there are a lot of particular situations which
have to be accounted for, such as the number of streamtubes per half-cycle being even or odd, and
whether the studied azimuthal position is located at a top or bottom quadrant.

Using the same parameters as until now, and considering the presence of two struts per blade at
heights of −4m and 4m with a circular section and thickness of t = 0.1m, the results shown in
figure 30 are obtained.

56
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

Figure 30: Power coefficient comparison: raw code, tip loss corrected and strut corrected.

Table 5: Maximum power coefficient comparison.

Case Cp,max ∆Cp,max [%] λ(Cp,max )

Raw code 0.561 - 3.44

Tip loss 0.536 −4.46 3.56
Tip loss + strut 0.409 −23.41 2.94

Note that every relative increment is calculated with respect to its previous case.

A quite drastic change can be observed, where both the maximum power coefficient value and the
tip speed ratio at which it occurs have decreased, as it can be seen in table 5.

As expected, the power loss becomes more and more pronounced as the tip speed ratio increases,
to the point where at λ ≈ 5 the power coefficient is already zero, while without the correction this
value was reached for λ ≈ 7.7.

57
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

8 Dynamic stall correction

Normally, an airfoil or wing reaches the stall point upon reaching a given angle of attack, which
implies as a drop in lift coefficient and an increase in drag coefficient due to the flow detachment
on the upper surface of the wing.

Although this is generally taken as true, the measurements or calculations to characterize this
phenomenon are made in steady-state conditions. In some situations, however, changes in angle of
attack can be sudden or oscillating, and that is where dynamic stall appears.

Dynamic stall can be seen as a hysteresis phenomenon, where upon suddenly increasing the angle of
attack beyond the static stall angle (αss ) produces a delay in the flow separation, which momentarily
maintains high lift until the flow finally separates and steady-state is reached. The same effect
happens when the flow is already detached and the angle of attack is lowered: the flow reattachment
suffers a delay and momentarily lift is different than expected, as well as the drag [37].

Figure 31: Example of the dynamic stall effect on an airfoil’s polar curves. Source: [38]

As seen in figure 31, static measurements can suffer large modifications under dynamic stall con-
ditions, and so for some applications its effects are not negligible, which is the case of the current
implementation.

The aerodynamic sections found in VAWTs’ blades suffer large changes in their angle of attack
during their rotation cycle, and in some cases these angles can be close or over the stall point,
particularly in relatively small tip speed ratios, where additionally the time variation of the angle
of attack is the highest.

58
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

8.1 Considered models

Due to the difficulty to predict the dynamic stall effect, methods outside of computational fluid dy-
namics must rely on empirical parameters to determine the deviation from steady-state coefficients.
In this regard a few different formulations have arised over the years.

8.1.1 Leishman-Beddoes model

One of the most widely used of such methods is the Beddoes-Leishman model [39], which is pre-
dominant in HAWT BEMT methods. It was first formulated in the late 1980s for use in helicopter
blades and over the years several modifications have been performed to it. However, the main
principles of the method remain the same.

The model has three parts:

1. Unsteady attached flow

2. Stall onset

3. Separated flow, also known as deep stall.

In each of these phases the coefficients are calculated by virtue of both potential flow expressions
and empirical coefficients, which vary with time according to circulatory and impulsive functions
[40]. After an iterative process the final results are obtained.

The large amount of empirical coefficients needed for the calculations, as well as the fact that an
iterative process is needed, make its implementation complicated and cause it to increase the overall
computational cost of the code, since an iteration has to be internally performed for each iteration
of the BEMT code.

8.1.2 Gormont model

A simpler method formulated by Gormont first appeared in 1973 [41], also originally developed for
helicopter blades. The model is based on the calculation of the coefficients from those corresponding
to a reference angle of attack, αref :

αref = α − K1 ∆α · sign(α̇) (8.1)

59
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

where the parameters ∆α and K1 are found through a series of expressions which can be seen in
annex A, which have been taken from reference [42]. It is important to mention that the angle
of attack variation rate α̇ refers to the change in absolute value (positive if the absolute value
increases and negative if it decreases). When applying the method this value is calculated as the
difference between the current and the previous angle of attack’s absolute values divided by the
time increment.

|α| − |αold |
α̇ = (8.2)
∆t
∆θ
∆t = (8.3)
ω

Once the parameters are calculated, the final lift coefficient is calculated as:

Cldyn = Cl (α0 ) + m(α − α0 ) (8.4)

where α0 is any convenient angle of attack, normally the zero-lift angle, α is the angle of attack
obtained through the BEM analysis, and m is a parameter which is defined as following:
 
Cl (αref ) − Cl (α0 ) Cl (αss ) − Cl (α0 )
m = min , (8.5)
αref − α0 αss − α0

Then the drag coefficient is simply:

Cddyn = Cd (αref ) (8.6)

8.1.3 Massé and Berg correction

Due to the fact that Gormont’s method was developed for helicopter blades, the angle of attack
they reach is substantially different compared to the values reached in a VAWT at certain regimes,
which can cause an overestimation of the effect of dynamic stall.

For this reason first the Massé modification to the Gormont method was proposed [43], which
corrects the dynamic coefficients with an interpolation between them and the static coefficients
according to the following expressions [42]:

( dyn
AM αss −α
Cl + AM αss −αss (Cl − Cl ) when α ≤ AM αss ,
Clmod = (8.7)
Cl when α > AM αss ,

60
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

( dyn
AM αss −α
Cd + AM αss −αss (Cd − Cd ) when α ≤ AM αss ,
Cdmod = (8.8)
Cd when α > AM αss ,

where AM is an empirical constant. Originally Massé proposed its value to be set at 1.8, although
later Berg [44], who whoud ultimately give the method its name, found better correlation with the
Sandia experiments for a value of AM = 6.

8.1.4 Strickland correction

Strickland [45] also proposed a modification to Gormont’s model, which considered two main hy-
potheses:

• In VAWT applications the used airfoils typically were thicker than those used in helicopter
blades.

• The blades do not reach high Mach numbers, so the flow can be considered as incompressible.

Due to the first consideration, the parameter Sc which appears in the original Gormont method (see
annex A), is set to zero. Due to the second consideration the Mach-dependant term also disappears,
and so the model is simplified to the following expressions [46]:

αref = α − K1 ∆α · sign(α̇) (8.9)

s
cα̇
∆α = γ (8.10)
2Ur

(
1.4 − 6(0.06 − t/c) for lift,
γ= (8.11)
1 − 2.5(0.06 − t/c) for drag,

K1 = (0.75 + 0.25 · sign(α̇)) · sign(α̇) (8.12)

(
α−α0
Cl (α0 ) + (Cl (αref ) − Cl (α0 )) when α ≥ αss .
Cldyn = αref −α0
(8.13)
Cl (α) when α < αss .

61
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

(
Cd (αref ) when α ≥ αss .
Cddyn = (8.14)
Cd (α) when α < αss .

Note that in this case, the correction is only applied when α ≥ αss .

8.2 Results

The chosen method has been the Gormont method with the Strickland adaptation, both for
simplicity reasons and the fact that it was derived to be applied in VAWTs, with blades which use
thicker airfoils. The value of stall angle of attack is set to αss = 10◦ for the NACA 0021, which is
not the maximum lift angle but it is where the lift slope ceases to be linear.

Upon implementing the method, some unreasonable results are observed. As seen in figure 32, the
method can grossly overpredict the dynamic stall effect at the lowest tip speed ratios when it is
coupled with the tip loss correction. This happens because the larger the angle of attack variation
rate, the further the method extends the linear part of the polar curve, which can lead to results as
the ones obtained in low tip speed ratios, where the α̇ values are the highest.

Figure 32: Lift coefficient against azimuthal position for various TSRs using Strickland adaptation.

This figure shows the lift coefficient evolution at an element located near the tip of the blade.

62
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

To mitigate this phenomenon, the Massé and Berg correction has been applied with a parameter
AM = 6. Once this has been performed the following results are obtained:

Figure 33: Lift coefficient against azimuthal position for various TSRs using Strickland and Berg
adaptations.

As it can be seen, results make much more sense since peak lift coefficients are not so high. This
happens because the correction’s interpolation between the dynamic and static coefficients dampens
the effect of the initial dynamic stall model, particularly in the most extreme situations.

There is still some oscillations in the results at λ = 1.08 between 180◦ and 200◦ , but that could be
due to the low relative velocity, which can cause large sudden variations in the angle of attack.

Using this model a new power coefficient curve can be calculated for the rotor:

63
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

Figure 34: Power coefficient comparison: raw code, tip loss, strut and dynamic stall corrected.

Table 6: Maximum power coefficient comparison.

Case Cp,max ∆Cp,max [%] λ(Cp,max )

Raw code 0.561 - 3.44

Tip loss 0.536 −4.46 3.56
Tip loss + strut 0.409 −23.41 2.94
Tip loss + strut + dyn. stall 0.418 +2.2 3.03

The effect of dynamic stall in the power coefficient curve can be seen in the lower tip speed ratios,
where stall occurs. In this range this phenomenon increases power, which can be explained by the
delay in flow separation when reaching angles of attack above the stall angle, which causes the blade
to produce higher lift forces and thus more torque. This results in a slight increase in maximum
power coefficient, given that at that point there is still a bit of stall. The change in maximum Cp
TSR is negligible.

64
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

Taking advantage of the results given by the program, the hysteresis-like cycle of the relation-
ship between lift and drag coefficients and angle of attack can be observed. Figure 35 shows this
phenomenon for the rotor of study and a tip speed ratio of λ = 2.53.

Figure 35: Lift coefficient vs angle of attack with Strickland model + Massé and Berg correction,
λ = 2.53.

Figure 36: Drag coefficient vs angle of attack with Strickland model + Massé and Berg correction,
λ = 2.53.

65
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

In both cases the cycle can be easily observed, and as expected the static curve lies in the middle
of the loops formed at each end of the linear part of the curve.

It can also be seen that for the lift coefficient, the points where the line is prolonged are not fully
tangential to the static curve, which they should be. The most likely reason for which this happens
is the fact that the lift slope is already not linear at the selected stall angle, which is difficult to
determine in thick airfoils such as the NACA 0021, where the true linear section is small. With a
thinner airfoil, where linearity is maintained until close to the sudden lift drop, the model would
adapt better to the static curve.

66
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

9 Tower shadow correction

Depending on the construction of the turbine, there can be a shaft at the center of the rotor, whose
presence interferes with the incident flow. This can be particularly detrimental to performance if
the shaft-to-rotor diameter ratio is relatively large, since wind speed is reduced at a large portion
of the rear part of the rotor.

In early investigation (1970s and 1980s) this effect was not paid too much attention, mainly because
research was focused in large scale turbines and the mentioned diameter ratio was large enough so
that performance reduction was small.

However, in the case of small-scale turbines the shaft can play a big role in the airflow received by
the blades and therefore, in performance. In this study the effects of tower wake will be considered
with the basis of a static cylinder.

It is first worth mentioning that quantifying the velocity profile after a cylinder is no easy task, since
its wake is rarely uniform due to phenomena like the von Karman vortices, for example. However,
for simplicity purposes in this study the wake will be considered as fully developed, and therefore
its effect will be quantified as a velocity deficit which results in a mean velocity profile.

Figure 37: Graphic representation of the tower wake model. Source: [47]

The used formulation follows the proposal made in reference [48], which, according to the axis which
have been used during this study, considers the following expression:

" 1/2 #
13y 2
 
CD,twr Dtwr
U (x, z) = U0 · 1 − 1.2 · exp (9.1)
−x CD,twr Dtwr x

67
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

which is applied to the downstream disks within a boundary of:

|zb | = 0.59 · (CD,twr Dtwr (−x))1/2 (9.2)

In these equations, the parameters x and z are calculated as:

x = R · sin(θst ) (9.3)
z = R · cos(θst ) (9.4)

and the calculation is performed for all the azimuthal positions of the rear actuator disks, if the
corresponding z value falls within the boundary set with equation 9.2.

In equation 9.1, U0 is the incident velocity of the given actuator disk, calculated as the downwind
velocity of the streamtube’s first actuator disk, Dtwr is the tower’s diameter and CD,twr is its drag
coefficient. The final upwind velocity that the affected actuator disk receives is U (x, z).

According to reference [48], these expressions are expected to be valid for x values bigger than a
few tower diameters.

Applying this correction, and setting a tower diameter of Dtwr = 1m to the turbine which has been
tested until now, the power coefficient curve is represented in figure 38.

Figure 38: Power coefficient comparison: raw code, tip loss, strut and tower corrected.

68
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

Table 7: Maximum power coefficient comparison.

Case Cp,max ∆Cp,max [%] λ(Cp,max )

Raw code 0.561 - 3.44

Tip loss 0.536 −4.46 3.56
Tip loss + strut 0.409 −23.41 2.94
Tip loss + strut + dyn. stall 0.418 +2.2 3.03
Tip loss + strut + dyn. stall + tower 0.396 −5.26 3.03

As expected, the Cp values diminish, although not too drastically. One interesting remark is that
the performance decrease seems to be the most important around the maximum power coefficient
values, being almost unnoticeable at the lowest and highest tip speed ratios. The optimum tip
speed ratio, however, does not suffer any changes.

It is important to say that in order to fully capture the tower’s effect, it has been deemed necessary
to increase the number of azimuthal divisions per semi-cycle from Nst = 30 to Nst = 60, therefore
increasing the required computational time.

Also worth mentioning, the tower’s drag coefficient was initially defined as an input to the program.
However, finally the same interpolation as in section 7 has been applied.

Once this correction is applied an interesting study can be carried out in order to visualize how the
incident speed is affected by the presence of a tower depending on its size. Figure 39 shows the
current rotor’s actuator disk upwind speed as a function of azimuthal angle, for different tower sizes
and a tip speed ratio of λ = 3.

In this plot, it can be seen that during the first half of the rotation the upwind speed is constant,
since the rotor receives undisturbed air. In the second part, however, air is affected by the presence
of the first actuator disks and therefore the velocity profile ceases to be constant.

As for the effect of the tower, the air speed loss that it causes can clearly be seen as a distribution
around θ = 270◦ . A spike is created, the shape of which is wider and longer as the tower diameter
increases. This means that a wider azimuth range is affected and overall velocity is lower, which
was expected.

69
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

Figure 39: Upwind speed as a function of azimuthal position for different tower diameters and
λ = 3.

Seeing figure 39, the question appears as to how much this phenomenon affects the power output of
the turbine at different tower diameter values. Power coefficient for the previously shown diameters
can be seen in table 8, as well as the the loss percentage compared to the tower-free case.

Table 8: Power coefficient at λ = 3 for different tower diameters.

Dtwr [m] D/Dtwr Cp,max ∆Cp,max [%]

0 (no tower) ∞ 0.418 -

0.5 20 0.404 −3.35
1 10 0.396 −5.26
2 5 0.380 −9.09

As expected, the larger the tower, the larger the power loss. The fact that for a relatively small
diameter tower the power loss is at ≈ 3% shows that although the effects of tower wake can be
small, they are non-negligible, let alone for larger towers.

70
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

10 Additional studies
Once all the presented corrections are applied, the code can be used for additional studies. This
section presents an analysis on the aerodynamic torque variability of the rotor which has been
simulated in the rest of sections, as well as a comparison with some experimental data.

The code is provided in annexes E and F, and the modifications done in the algorithm’s flowchart
can be found in annex D.

10.1 Torque value variability

One of the issues that VAWTs have is the variability of the torque generated by each blade during
a revolution, or torque ripple, which is important mainly because it can have a detrimental effect
on the turbine’s components, subject to cyclical loads. Moreover, the electric signal created by the
generator will also be affected by these variations, and therefore it will have to be treated with
components that can withstand these possible voltage swings.

One of the most determining factors in this phenomenon is the number of blades of the rotor. Using
the developed tool a comparison can be made between different rotors, with the aim to observe how
the shape of the generated torque changes.

In this case torque is calculated instantaneously, rather than averaged as when calculating the
thrust and power coefficients. The expression used is then equation 4.38, with which the following
summation is performed for each azimuthal position:

Nel
X
τi = τij (10.1)
j=1

Figure 40 shows the comparison between the torque generated along the rotation cycle for rotors
with 2, 3, 4 and 5 blades. Note that neither the struts or the tower have been considered, and the
tip speed ratio has been set to λ = 3.

71
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

Figure 40: Generated torque at λ = 3 during one revolution for different numbers of blades. a)
Nb = 2 b) Nb = 3 c) Nb = 4 d) Nb = 5

The figure shows clearly how as the number of blades increases, the variability in torque output
decreases, as well as the peak torque value generated by each individual blade. In order to help
quantify this, table 9 is presented.

Table 9: Maximum and average torque relationship for different numbers of blades.

Nb τavg [N m] τmax [N m] ∆τmax /τavg [%]

2 1530.90 786.76 94.58

3 1382.86 1020.98 35.44
4 1290.17 1094.69 17.86
5 1179.85 1092.85 7.96

72
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

Observing these values, it can be seen that the biggest change is given from two to three blades,
and improvement is still observed from four to five blades. This does not mean, however, that more
blades is always better, since average torque decreases and also a larger amount of blades will cause
variations in the power curve (it increases solidity, a parameter which is known for lowering the
optimal TSR value and narrowing its useful range for a specific rotor [49]). Moreover, a reduction
in torque ripple can also be accomplished with other strategies, such as twisting the blades (helix
type rotors).

10.2 Comparison with experimental data

The last study that will be performed is a comparison of the code with some experimental data.
The airfoil which has been used until now was chosen in the beginning due to the available data on
their polar curves and its use in simulations and experiments.

The program will be compared to the wind tunnel experiment performed in reference [50], which
has been chosen due to the extensive documentation on it, as well as on the turbine’s dimensions.
Figure 41 shows a drawing of the rotor, as well as the designation of the provided measurements,
the most relevant of which have been collected in table 10.

Figure 41: Drawing of the rotor of study. Source: [50]

73
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

Table 10: Rotor parameters for the experimental results.

Parameter Value

Number of blades 3
Airfoil NACA 0021
Chord [m] 0.085
Struts per blade 2
Strut thickness [m] 0.0063
Strut distance to origin [m] 0.364
Rotor height [m] 1.46
Rotor radius [m] 0.515
Rotor speed [rpm] 400

Introducing the data to the code the following results are obtained:

Figure 42: Simulated and experimental results comparison.

Although the curve follows the same shape, with the maximum power coefficients being obtained
around the same tip speed ratio for both numerical and experimental results, the figures provided
by the code are noticeably larger.

74
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

Upon a closer look, one possible reason for this discrepancy is the used polar curves, which for all
the simulations until now have been those corresponding to a Reynolds number of Re = 2e6. This
number makes sense for large rotors, but in this case the blades reach nowhere near this figure. To
visualize the number around which the blade Reynolds are, figure 43 is plotted.

Figure 43: Blade Reynolds number against azimuth angle.

It can be observed that the values vary between a bit less than Re = 250000 and a bit more than
Re = 0, with an average between Re = 100000 and Re = 150000. The polars for the closest
Reynolds value found in reference [51] are for Re = 160000, so this value is taken (see polars in
annex B). With this data, the power coefficient curve shown in figure 44 is obtained.

In this case the correlation improves significantly with respect to the previous simulation. The
model still overpredicts the turbine’s performance, specially at high tip speed ratios, which could
be due to the model and its hypotheses themselves. For example, one aspect which would hinder
performance and has not been taken into consideration is the downwash effect of the finite wing.
More improvements to the program which were outside the scope of this project, but could be
implemented in the future, are given in section 11.2.

75
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

Figure 44: Simulated and experimental results comparison with low Reynolds polars.

Another possible source of discrepancy is the nature of the experimental results, which are most
likely affected by mechanical and/or electrical losses in the bearings of the rotor and power mea-
surement equipment. The code is ideal in the sense that it measures only aerodynamic power.

Despite these differences the model still gives reasonable values, and it proves to be useful in being
able to predict the effect on performance of certain design changes, such as the number of blades,
among others.

76
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

11 Conclusions

11.1 General conclusions

In this project a DMST code for H-Type VAWT simulation has been developed, both in its most
essential form and adding corrections to it to increase its capabilities.

First, general aspects of the project have been presented: its aim, scope, requirements, justifica-
tion and scheduling. Then, a few important concepts of wind energy which are relevant to the
understanding of the rest of the project have been explained.

After that, an overview on the state of the art of wind turbines, particularly VAWTs, and their
simulation methods has been provided, which serves as the setting in which this project lies.

Then the development of the code has begun: the theory applied in the deduction of the expressions
of the raw model has been thoroughly explained, as well as implementation details such as discretiza-
tion of the domain and the code’s flowchart. A preliminary validation has also been performed in
the form of a convergence study.

As for the corrections, the theory in which they are based has been introduced and their implemen-
tation detailed. Conclusions on the implications of each correction have been drawn.

Then one possible study in which the code can be applied has been showcased: the instantaneous
torque generated by a rotor, and the effect of different numbers of blades in its variability. Conclu-
sions have been drawn from this study.

Finally the parameters of a real rotor have been introduced to the program, and the obtained results
have been compared with those resulting from the experimental measurements.

Opon completion of the project, conclusions on the accomplishment of its aims and requirements
can be extracted.

• Regarding the the original aim, a DMST code for VAWT simulation has been made, both raw
and with corrections, later comparing it to experimental data and applying it to a practical
case. In this aspect the initially set objectives have been accomplished.

77
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

• The introduced corrections have mainly hindered the turbine performance predicted by the
raw model, the strut correction having the most impact in this aspect. This was expected
due the corrections being aimed at reducing the ideal nature of the initial formulation. An
exception is the dynamic stall effect, which increases the power coefficient at low tip speed
ratios because of the delay in the lift coefficient drop experienced by the blades.

• As for the correlation between the results extracted from the code and experimental studies,
some discrepancy was observed. Possible reasons for it are discussed in the report: some of
them are related to possible mechanical and electrical losses in the experiments, which are not
considered in the simulation, and others refer to additional corrections and changes which can
be made to the code but fell out of the scope of the project. These are presented in section
11.2.

• In terms of the requirements, the main one was to program the code in Python. This proved to
be quite challenging particularly in the beginning, since there was no previous experience with
this language before the development this project. However, overtime it became easier to work
with it and even the post-processing routine was made in Python, therefore accomplishing the
goal of becoming familiar with this language.

• The set time frame has also been met, finishing the project before the June 22nd deadline. At
some points a delay was suffered with respect to the original Gantt chart due to unforeseen
complications in the code, some of which required time to overcome. However, during the
planning stage of the project extra time was allocated in order to compensate for these possible
disruptions, and so they have not had an impact in the overall completion of the project in
its due time.

As a final comment, this project has dived into an area which will inevitably have to be researched
in the coming years, due to the rising environmental concerns pushing the need to find better
renewable energy solutions. By any means does it prove that Vertical Axis Wind Turbines are
the best of these solutions, but further studies should be carried out particularly to evaluate their
suitability and optimise their design for urban environments, where they could most likely make
the biggest impact.

In this regard the development of computational simulation codes such as this one, which have
a very low environmental impact (see the Budget document for the total energy needed for its
development), is necessary in order to reduce the amount of time, cost and negative effects on the
environment that the construction, testing and dismantling of experimental turbine models entail.

78
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

11.2 Future work

Due to the time limitations imposed on this project, with a total workload of 300 hours, some extra
additions which could have been done to the program are not yet implemented. This section aims
to explain them in order to set the tracks for future work arising from this project.

1. Finite wing downwash

The first modification which could be added is the downwash component of the finite wing,
which has not been implemented in this project for reasons mentioned in section 6. However,
with more time to dedicate to this matter a model could be found to take into consideration
this effect, which would benefit the model since it would reduce its over-predicting tendency.

2. Flow curvature

Another effect which has not had the time to be deeply studied is flow curvature, which is
an effect that takes place because of the curvilinear trajectory of the blade across the airflow.
This causes the airfoil to ”see” different angles of attack across its chord. As depicted in figure
45, this effect’s consequence can be seen as a variation in the airfoil’s camber.

Figure 45: Depiction of the flow curvature effect. Source: [52].

There are some analytical models to take into consideration this effect, the simplest and most
well known of which is the Goude formulation (see reference [40]), which proposes to maintain
the original airfoil polars and modify the angle of attack.

79
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

3. Streamtube expansion

Another phenomenon which would be interesting to implement is the streamtube expansion,

since in the original DMST formulation the streamtubes are not allowed to expand from the
front actuator disk to the rear one. However, it impacts the structure of the code massively
since it affects the discretization of the domain, which ceases to be symmetrical at the front
and rear halves of the rotor, and therefore it was not possible to implement in the available
timeframe.

Figure 46: Depiction of the streamtube expansion’s effect on the discretization

of the rotor. Source: [40].

The main effect that this change in the discretization would introduce is the fact that the
downstream area would become larger, which at a first glance would reduce the predicted
performance because a larger fraction of the rotor would receive disturbed air.

4. Tangential induction factor

As a final suggestion for future projects, there are a few models which introduce, aside from
an axial induction factor, a tangential one which affects this component wind speed that the
blade element receives. This is standard practice in HAWTs, but is not common in VAWT
simulation codes. Its expected effect is similar to the one observed in HAWTs: to increase
the tangential velocity received by the blades, therefore decreasing their angle of attack and
hurting performance, most notably at high tip speed ratios and high rotor solidities [4].

80
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

5. Extra capabilities

Aside from these modifications of the model’s theory itself, the program could benefit from a
few extra implementations, which would broaden its capabilities.

• Reynolds polar interpolation: Seeing the impact that the polar Reynolds number can
have on the results, it would be useful to take the time to collect polar curves at different
Reynolds numbers, enough to be able to perform an interpolation for the local number
calculated at each blade element, aside from the angle of attack.

• Implement a variable radius rotor: Troposkein rotors have been historically widely stud-
ied, and therefore the code would benefit from the possibility to implement such a rotor
shape.

• Implement a helix-shape rotor: These rotors are of interest due to their torque ripple
reduction, and it would be interesting to be able to observe this effect.

The amount of work done, together with the fact that there are still many corrections and
modifications which can be applied to the program, highlights the complexity of the different
phenomena, as well as their modelling, that take place in a Vertical Axis Wind Turbine.

81
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

References
[1] W. W. E. Association, Worldwide wind capacity reaches 744 gigawatts – an unprecedented 93
gigawatts added in 2020, https://wwindea.org/worldwide-wind-capacity-reaches-744-
gigawatts/, (Accessed on 05/12/2022), Mar. 2021.
[2] M. Jaganmohan, Global installed wind energy capacity 2021, https : / / www . statista .
com / statistics / 268363 / installed - wind - power - capacity - worldwide / # : ~ : text =
Installed % 20wind % 20power % 20capacity % 20worldwide % 202001 % 2D2021 & text = The %
20cumulative % 20capacity % 20of % 20installed , about % 20780 % 20gigawatts % 20that %
20year., (Accessed on 05/12/2022), Apr. 2022.
[3] V. Petrova, World adds 93.6 gw of wind in 2021, gwec says even more is needed, https :
//renewablesnow.com/news/world-adds-936-gw-of-wind-in-2021-gwec-says-even-
more-is-needed-779796/, (Accessed on 05/12/2022), Apr. 2022.
[4] F. J. S. Cano, “Class notes”, Wind turbine design (220038), 2022.
[5] J. Damota, M. Lamas, A. Couce-Casanova, and J. Rodriguez-Garcia, “Vertical axis wind
turbines: Current technologies and future trends”, in International Conference on Renewable
Energies and Power Quality (ICREPQ’15), vol. 1, 2015, pp. 530–535. doi: https://doi.
org/10.24084/repqj13.389.
[6] B. Hand and A. Cashman, “A review on the historical development of the lift-type vertical axis
wind turbine: From onshore to offshore floating application”, Sustainable Energy Technologies
and Assessments, vol. 38, p. 100 646, 2020. doi: https://doi.org/10.1016/j.seta.2020.
100646.
[7] S. F. Johnston Jr., “Proceedings of the vertical axis wind turbine (vawt) design technology
seminar for industry”, Sandia Labs., Albuquerque, NM (USA), Tech. Rep., Aug. 1980.
[8] TU-Delft, The cp-lambda curve, http://mstudioblackboard.tudelft.nl/duwind/Wind%
20energy%20online%20reader/Static_pages/Cp_lamda_curve.htm#:~:text=The%20CP%
2D%CE%BB%20curves,from%20CP%2D%CE%BB%20curves., (Accessed on 06/07/2022).
[9] R. Schnurr, The iconic windmills that made the american west - atlas obscura, https://www.
atlasobscura.com/articles/windmills-water-pumping-museum-indiana, (Accessed on
03/09/2022), Jan. 2018.
[10] W. S. Magazine, Vestas introduces v162-6.8 mw wind turbine, https://www.windsystemsmag.
com/vestas-introduces-v162-6-8-mw-wind-turbine-2/, (Accessed on 03/09/2022), Feb.
2022.
[11] I. Magazine, Icelandic wind-powered bus stops could soon be coming to your city, https :
/ / icelandmag . is / article / icelandic - wind - powered - bus - stops - could - soon - be -
coming-your-city, (Accessed on 03/29/2022), Apr. 2016.

82
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

[12] F. Castellani, D. Astolfi, M. Peppoloni, F. Natili, D. Buttà, and A. Hirschl, “Experimental

vibration analysis of a small scale vertical wind energy system for residential use”, Machines,
vol. 7, p. 35, May 2019. doi: 10.3390/machines7020035.
[13] T. M. Letcher, Wind energy engineering: A handbook for onshore and offshore wind turbines.
Academic Press, 2017, isbn: 978-0-12-809451-8.
[14] D. Nigam, What are vertical axis wind turbines (vawts)?, https://blog.arborwind.com/
what-are-vertical-axis-wind-turbines, (Accessed on 03/19/2022), Sep. 2019.
[15] E. Hau and H. von Renouard, Wind turbines: fundamentals, technologies, application, eco-
nomics. Springer New York, 2003, isbn: 978-3-642-06348-0.
[16] P. Tchakoua, R. Wamkeue, M. Ouhrouche, E. Benini, and G. Ekemb, “Electric circuit model
for the aerodynamic performance analysis of a three-blade darrieus-type vertical axis wind
turbine: The tchakoua model”, Energies, vol. 9, p. 820, Oct. 2016. doi: 10.3390/en9100820.
[17] F. Meng, H. Schwarze, F. Vorpahl, and M. Strobel, “A free wake vortex lattice model for
vertical axis wind turbines: Modeling, verification and validation”, in Journal of physics:
conference series, IOP Publishing, vol. 555, 2014, p. 012 072. doi: 10.1088/1742-6596/555/
1/012072.
[18] E. Ortega, “Class notes - module 2.4: Discrete vortex method”, Aerodynamics (220024), 2020.
[19] C. Xie, Y. Liiu, and C. Yang, “Static aeroelastic analysis of very flexible wing based on
the 3-d lifting-line theory”, in 53rd AIAA/ASME/ASCE/AHS/ASC Structures, Structural
Dynamics and Materials Conference 20th AIAA/ASME/AHS Adaptive Structures Conference
14th AIAA, 2012, p. 1459. doi: 10.2514/6.2012-1459.
[20] E. Ortega, “Class notes - module 3.2: The finite wing”, Aerodynamics (220024), 2020.
[21] A. for Students, Subsonic aerofoil and wing theory, http://www.aerodynamics4students.
com/subsonic-aerofoil-and-wing-theory/3d-vortex-lattice-method.php, (Accessed
on 06/04/2022).
[22] I. H. Hirsch and A. Mandal, “A cascade theory for the aerodynamic performance of darrieus
wind turbines”, Wind Engineering, vol. 11, pp. 164–175, 1987.
[23] C. E. Brennen, Chapter 1 - hydrodynamics of pumps, https://authors.library.caltech.
edu/25019/1/chap3.htm, (Accessed on 06/05/2022), Dec. 2000.
[24] B. Hand, A. Cashman, and G. Kelly, “An aerodynamic modelling methodology for an offshore
floating vertical axis wind turbine”, in 2015 International Conference on Renewable Energy
Research and Applications (ICRERA), 2015, pp. 273–277. doi: 10 . 1109 / ICRERA . 2015 .
7418708.
[25] Wikipedia, Computational fluid dynamics, https://en.wikipedia.org/wiki/Computational_
fluid_dynamics#Discretization_methods, (Accessed on 06/05/2022), Jun. 2022.

83
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

[26] R. Neal, Finite element analysis – computational fluid dynamics, https : / / rossrneal .
wordpress.com/portfolio/finite-element-analysis-cfd/, (Accessed on 06/05/2022).
[27] S. P. Neill and M. R. Hashemi, “Chapter 8.2.3 - finite volume method”, in Fundamentals of
Ocean Renewable Energy, Academic Press, 2018, pp. 193–235, isbn: 978-0-12-810448-4.
[28] E. Dick, “Introduction to finite element methods in computational fluid dynamics”, in Com-
putational Fluid Dynamics. Springer Berlin Heidelberg, 2009, pp. 235–274, isbn: 978-3-540-
85056-4.
[29] A. J. Wain and E. J. Dickinson, “Chapter 1 - nanoscale characteristics of electrochemical
systems”, in Nanoscale Electrochemistry, vol. 18, Elsevier, 2021, pp. 1–48, isbn: 978-0-12-
820055-1.
[30] A. A. Ayati, K. Steiros, M. A. Miller, S. Duvvuri, and M. Hultmark, “A double-multiple
streamtube model for vertical axis wind turbines of arbitrary rotor loading”, Wind Energy
Science, vol. 4, no. 4, pp. 653–662, 2019. doi: https://doi.org/10.5194/wes-4-653-2019.
[31] I. Paraschivoiu, P. Desy, and C. Masson, “Blade tip, finite aspect ratio, and dynamic stall
effects on the darrieus rotor”, Journal of Propulsion and Power, vol. 4, Jan. 1988. doi: 10.
2514/3.23034.
[32] M. L. Buhl Jr, “New empirical relationship between thrust coefficient and induction factor for
the turbulent windmill state”, National Renewable Energy Lab.(NREL), Golden, CO (United
States), Tech. Rep., 2005.
[33] K. Steiros and M. Hultmark, “Drag on flat plates of arbitrary porosity”, Journal of Fluid
Mechanics, vol. 853, Oct. 2018. doi: 10.1017/jfm.2018.621.
[34] G. Taylor, “Air resistance of a flat plate of very porous material”, Aeronautical Research
Council, Reports and Memoranda, vol. 2236, pp. 159–162, 1944.
[35] Wikipedia, Powell’s dog leg method, https://en.wikipedia.org/wiki/Powell%27s_dog_
leg_method, (Accessed on 06/06/2022), Apr. 2022.
[36] N. K. Delany and N. E. Sorensen, “Low-speed drag of cylinders of various shapes”, National
Advisory Committee for Aeronautics, Tech. Rep., 1953.
[37] A. Choudhry, R. Leknys, M. Arjomandi, and R. Kelso, “An insight into the dynamic stall lift
characteristics”, Experimental Thermal and Fluid Science, vol. 58, pp. 188–208, 2014. doi:
https://doi.org/10.1016/j.expthermflusci.2014.07.006.
[38] A. Filippone, Dynamic stall, unsteady aerodynamics, https://aerodyn.org/dstall/, (Ac-
cessed on 05/31/2022), 2003.
[39] J. G. Leishman and T. Beddoes, “A semi-empirical model for dynamic stall”, Journal of the
American Helicopter society, vol. 34, no. 3, pp. 3–17, 1989. doi: https://doi.org/10.4050/
JAHS.34.3.3.

84
 Study on the simulation of a Vertical
Axis Wind Turbine (VAWT)

[40] E. Dyachuk and A. Goude, “Simulating dynamic stall effects for vertical axis wind turbines
applying a double multiple streamtube model”, Energies, vol. 8, pp. 1353–1372, Feb. 2015.
doi: 10.3390/en8021353.
[41] R. E. Gormont, “A mathematical model of unsteady aerodynamics and radial flow for appli-
cation to helicopter rotors”, Boeing Vertol Co Philadelphia Pa, Tech. Rep., 1973.
[42] C. Masson, C. Leclerc, and I. Paraschivoiu, “Appropriate dynamic-stall models for perfor-
mance predictions of vawts with nlf blades”, International Journal of Rotating Machinery,
vol. 4, Jan. 1998. doi: 10.1155/S1023621X98000116.
[43] B. Massé, Description de deux programmes d’ordinateur pour le calcul des performances et
des charges aerodynamiques pour les eoliennes a axe vertical. Institut de recherche de l’Hydro-
Québec, 1981.
[44] D. E. Berg, “Improved double-multiple streamtube model for the darrieus-type vertical-axis
wind turbine”, Sandia National Labs., Albuquerque, NM (USA), Tech. Rep., 1983.
[45] J. Strickland, B. Webster, and T. Nguyen, “Vortex model of the darrieus turbine: An analytical
and experimental study”, NASA STI/Recon Technical Report N, vol. 80, p. 25 887, 1980. doi:
https://doi.org/10.1115/1.3449018.
[46] A. G. Sanvito, V. Dossena, and G. Persico, “Formulation, validation, and application of a
novel 3d bem tool for vertical axis wind turbines of general shape and size”, Applied Sciences,
vol. 11, no. 13, p. 5874, 2021. doi: https://doi.org/10.3390/app11135874.
[47] A. Bianchini, L. Ferrari, and S. Magnani, “Start-up behavior of a three-bladed h-darrieus
vawt: Experimental and numerical analysis”, in Turbo Expo: Power for Land, Sea, and Air,
vol. 54617, 2011, pp. 811–820. doi: https://doi.org/10.1115/GT2011-45882.
[48] R. D. Blevins, “Forces on and stability of a cylinder in a wake”, J. Offshore Mech. Arct. Eng.,
vol. 127, no. 1, pp. 39–45, 2005. doi: https://doi.org/10.1115/1.1854697.
[49] H. Zhu, W. Hao, C. Li, and Q. Ding, “Numerical study of effect of solidity on vertical axis
wind turbine with gurney flap”, Journal of Wind Engineering and Industrial Aerodynamics,
vol. 186, pp. 17–31, 2019. doi: https://doi.org/10.1016/j.jweia.2018.12.016.
[50] L. Battisti, G. Persico, V. Dossena, et al., “Experimental benchmark data for h-shaped and
troposkien vawt architectures”, Renewable Energy, vol. 125, pp. 425–444, 2018. doi: https:
//doi.org/10.1016/j.renene.2018.02.098.
[51] R. E. Sheldahl and P. C. Klimas, “Aerodynamic characteristics of seven symmetrical airfoil
sections through 180-degree angle of attack for use in aerodynamic analysis of vertical axis
wind turbines”, Mar. 1981. doi: 10.2172/6548367.
[52] A. Bianchini, F. Balduzzi, G. Ferrara, and L. Ferrari, “Virtual incidence effect on rotating
airfoils in darrieus wind turbines”, Energy Conversion and Management, vol. 111, pp. 329–
338, 2016. doi: https://doi.org/10.1016/j.enconman.2015.12.056.

85