Abstract
—
The paper presents the development of a wind turbine simulator which consists of an induction motor driven by a torquecontrol inverter. The wind turbine simulation system includes: wind speed simulation, mathematical model of wind turbines, modeling of rotor blade characteristics, modeling of tower effect and emulation of rotor inertia. Wind speed can be easily programmed according torecorded wind speed data or Van Der Hoven model or manual set up. The developed algorithms were implemented by a lowcost, high performance DSC controller with C language and the system was tested in the laboratory with 1 kW dc generator. The power responses,torque responses and tip speed ratio responses confirms that the system can operate very well under step change of power reference and load disturbances. The advantages of the simulators are that various wind profiles and wind turbines can be incorporated as desired in thecontrol software and it includes the data acquisition to verify the control algorithms and display the parameters. The experimental resultsconfirmed the wind turbine simulator can perform satisfactory under steady state wind profile, turbulence and tower effect.
Keywords
—
Wind turbine simulator, Wind speed generation, Power spectrum, Torque control
NOMENCLATURE
1
c
P
= power coefficient of turbine [pu]c
T
= torque coefficient of turbine [pu]f = frequency [Hz]G = gear ratio [pu]J
g
= inertia of generator [kg.m
2
]J
m
= inertia of motor [kg.m
2
]J
t
= inertia of turbine [kg.m
2
]N = sampling operationP = output power of turbine [W]P
turb
= power produced by turbine [W]P
wind
= power in wind [W]R = radius of blade of blade [m]S
vv
= power spectral debsity [ m
2
/s]T
f1
= friction torque of wind turbine system [N.m]T
f2
= friction torque of MG set [N.m]
Τ
comp
= compensation torque [N.m]
Τ
f1
= friction torque [N.m]
Τ
g
= generator torque [N.m]
Τ
m
= motor torque [N.m]
Τ
tower
= tower effect turbine ripple [N.m]
Τ
turb
= turbine torque [N.m]T’
turb
= turbine torque without tower effect [N.m]v
t
= wind speed [m/s]v = mean wind speed [m/s]
α
= angular acceleration of turbine [rad/s
2
]
β
= pitch angle [rad]
ϕ
i
= phase angle with uniformly distributed randomnumber in a domain of [rad].
λ
= tip speed ratio [pu]
ω
0
= starting radian frequency [rad/s]
ω
g
= angular velocity of generator [rad/s]
ω
t
= angular speed of turbine [rad/s]
B. Neumanee and S. Sirisumrannukul are with Department of ElectricalEngineering, Faculty of Engineering, King Mongkut’s Institute of Technology North Bangkok, Thailand, Pibulsongkram Rd., Bangsue,Bangkok 10800, Thailand, Email: bln@kmitnb.ac.th, spss@kmitnb.ac.thS. Chatratana is a Deputy Director of the Technology ManagementCenter, National Science and Technology Development Agency(NSTDA), 111 Thailand Science Park, Paholyothin Rd., Klong 1, KlongLuang, Phathumthani 12120, Thailand, Email : somchaich@nstda.or.th
ρ
=
density of air [kg/m
3
]
1. INTRODUCTION
Wind power has become one of the most attractive energyresources as it is almost pollutionfree (if noise is not consideredas pollution) when used for electricity production. As a result, agreat deal of research has been focused on the development of new turbine design to hoe to reduce the costs of wind power andhow to make wind turbines more economical and efficient. Theinvestigation of wind power system involves high performancewind turbine simulator, especially for the development of optimal control solutions. At present, such simulator has becomea necessary tool for research laboratories to enhance the qualityof the wind energy conversion system.The basic requirement for a wind simulator is that its staticand dynamic characteristics must be as close as possible to thoseof real wind turbine. For the last few decades, the most commonstructure of a wind simulator was based on a DC motor withcurrent control (i.e., torque control on the shaft of the DC motor).However, the simulator requires a relatively largesized DCmotor. This constraint makes DC motor system unattractive, dueto its unavailability and maintenance requirement. In addition, itis rather expensive. Later the DC motor system was replaced byan induction motor system which eliminated the abovementioned disadvantages.In this paper, a squirrel cage induction motor (IM) isproposed for a wind simulator as a torquegenerating source. Thewind simulator consists of two main parts as shown in Fig 1. Thefirst part (left hand side of the figure) is used to create requiredwind speed. The powerspeed pattern of the wind can begenerated from the data based on the Van Der Hoven powerspectrum model or from the actual recorded wind speed data orfrom manual set up. This part includes the mathematical modelof the wind turbine to calculate reference signals which will beused as torque reference for the inverter in the second part. Thesecond part is an electromechanical tracking system (ETS) whichgenerates shaft torque whose characteristics are governed by thepattern determined by the first part.Bunlung Neammanee, Somporn Sirisumrannukul and Somchai Chatratana
Development of a Wind Turbine Simulatorfor Wind Generator Testing
Fig.1 Wind speed simulation system with DSC board andelectromechanical systems.
The developed wind turbine simulator employs a 4kWinduction motor as prime mover. A digital signal controller(DSC) board is used to interface the wind speed generator and atorque control inverter which drives induction motor. A controlprogram is developed to obtain output torque from wind profiles.The program also computes the theoretical shaft torque of thewind turbine from turbine characteristics and rotation speed of the induction motor.
2. WIND SPEED SIMULATION
The modeling of wind speed modeling is very importantbecause it dictates the performances of wind generators anddetermines the features offered by a simulator for prediction of the energy output and analysis of the energy conversion andsystem dynamics. The nature of wind speed is generally assumedto be composed of two components: steady state mean flow andturbulence. Turbulence is characterized by random fluctuation of speeds. Simulation of these two components is usually performedseparately. The mean wind speed is the steady part of temporalaverage over some period and increases with the elevation. Theturbulent of wind speed is random with time and space and iscommonly assumed to be a stationary Gaussian process [1].As wind speed data are time series, the identification of thedata can be captured in forms of time and frequency domains. Inthis paper, frequency power spectrums based on the Van derHoven model shown in Fig. 2, is chosen [2], [3]. This model isregarded as one of the best known reference of wind speedmodels. In frequency domain, the equivalent description of windfluctuation component can be obtained through power spectrumdensity and coherence functions.The power spectrum in Fig. 2 shows the contribution of harmonic components in a range from 0.0007 to 900 cycles/h(i.e., more than six decades). This spectrum gives a completedescription of the energy content of turbulence. The frequencyrange contains the spectral domain that describes the mediumand longterm variations, as well as the spectral range of theturbulent component.From Van der Hoven model, a numerical wind speedsimulation procedure has been developed based on the samplingof the spectrum. For every value of discrete angular frequency
ω
i
, i = 1, 2,…, N+1, Van der Hoven spectral model gives acorresponding value of power spectral density, S
vv
(
ω
i
). Theamplitude A
i
of the wind harmonic at frequency
ω
i
is given by))]((S)(S[
21A
i1i1ivvivvi
ω−ωω+ω=
++
(1)The wind speed as a function of time v(t) is simulated by)tcos(AA)t(v
iN1ii0
ϕ+ω+=
∑
=
(2)with A
0
=v,
ω
0
=0 and
ϕ
0
=0. Note that the first and second termsof Equation (2) represent the average and turbulence of windspeed respectively. The parameter vis calculated on a timehorizon greater than the largest period in the Van der Hovencharacteristic.
Fig. 2 Van der Hoven spectral model.
3. MATHEMATICAL MODEL OF WINDTURBINE
A typical horizontal wind turbine with three bladescoupled with a gear box to capture and transfer energy to thegenerator, is shown in Fig. 3. Many wind turbines are directlycoupled with low speed generator without a gear to reduce thegear’s losses, weight, and maintenances. A vertical wind turbinehas the same characteristics but it can be operated at a low tipspeed ratio. The characteristics such as power, torque, and speedfor both types of the wind turbines can be modeled by a motorand generator set with a motor torque controller. The controlleruses a torque reference which depends on wind speed, angularspeed and the aerodynamics of the rotor blade to generate adesired torque.
Fig. 3 a) Horizontal wind turbine coupling with gear box.b) Vertical twistedH Rotor with direct coupling togenerator.3.1 Modeling of Rotor Blade Characteristics
Wind speed generally varies according to elevation of theblades (i.e., every single spot on the turbines would not have thesame wind speed). Modeling of wind speed must take intoaccount all different positions on the blades and therefore, it isvery difficult. For the sake of simplicity, single value of windspeed is applied to the whole wind turbines. Modeling of rotorblade characteristic requires tip speed ratio and the relationshipof torque and power coefficient versus tip speed ratio. The tip
speed ratio (TSR),
λ
is obtained from
tt
vR
ω=λ
(3)The power captured by the blades, P
turb
, can be calculatedusing),(cvR
2P
P3t2turb
βλπρ=
(4)The aerodynamic torque acting on the blades,
Τ
turb
, isobtained by),(cvR
2T
T2t3turb
βλπρ=
(5)If c
P
is known, the aerodynamic torque can also becalculated from
tP3t2turb
/ ),(cvR
2T
ωβλπ
ρ=
(6)It can be seen from the above two equations that c
T
and c
P
are a function of
λ
and
β
. But in this paper,
β
is kept constant;namely, pitch angle is fixed and this is generally true for a smallwind turbine. Therefore, c
T
and c
P
depend only on
λ
.Figure 4 shows a relationship of torque coefficient versusTSR of a real 3 kW, three blade horizontal axis wind turbine witha rotor diameter of 4.5 m [4]. This curve represents an importantcharacteristic which determines the starting torque of the windturbine. In general, this curve is available from the manufactureor can be obtained from a field test. With this curve, c
P
, whichindicates the efficiency of power conversion of the rotor blades,can be calculated by multiplying c
T
with
λ
[5]. Fig. 4 also showsthe c
P
TSR profile corresponding to the c
P
curve. It is importantto note that the power and torque coefficient of a wind turbinedepends on aerodynamic design of the blades. With Equations(3)  (5) and Fig. 4, a block diagram for the system can be builtas shown in Fig. 5.
02468101200.050.10.150.20.250.30.350.40.45
tip speed ratio
p o w e r a n d t o r q u e c o e f f i c i e n t [ c P , c T ]
cPcT
[λ]
Fig. 4 c
T

λ
and c
P

λ
characteristic of a real wind turbine.Fig. 5 Block diagram of rotor blade characteristics.3.2 Periodic Torque Ripple caused by Tower Effect
There is a phenomenon that the output torque of a windturbine contains periodically rippled torque [6]. The phenomenoncan be graphically explained by Fig. 6 (a). As the wind passesthrough the tower, the wind speed in front of the tower decreasesdue to the resistance of the tower structure. At this time, if thetower and one of the blades are in alignment, the wind speedstriking the blades will reduce and cause the ripple effect asshown in Fig 6 (b), where the shape of torque ripple is modeledas a ramp function [7]. The frequency of the rippled torque isequal to the main rotational frequency multiplies by the numberof blades. The shape and amount of torque ripple depend on thecharacteristic of physical structures. Net output torque whichtakes the tower effect into account, can be calculated by
towerturbturb
TTT
−′=
(7)
1f ggtturb
TGT)GJJ(T
++α+=
(8)(a)(b)
Fig. 6 (a) Rotor position at 0,
π
/3, 2
π
/3 rad (black bladecreates the tower effect). (b) Periodic torque ripple modelingwith tower effect of 3 blades wind turbine.3.3 Emulation of Rotor Inertia
Figure 7 shows the physical structure of a wind turbineconsisting of blades (left part), a gear box (middle part), and agenerator (right part). A torque equation of the wind turbine isgiven in Equation (8) with an assumption that all of the rotatingparts are considered as a lumped mass. Figure 8 shows the block diagram of the wind turbine, which is developed from Equations(3) and (8).
Fig. 7 System moment of inertia: gear box with turbine andgenerator torque.Fig. 8 Block diagram of the wind turbines with finite bladesand friction losses.
If an induction motor is employed in the wind turbinesimulator, the torque behavior generated by motor should besimilar to that generated by a real wind turbine. This can beachieved by a dynamic equation
2f ggmm
TGT)GJJ(T
++α+=
(9)Subtracting equation (8) from equation (9) with theassumption that the friction and gear losses are negligible (i.e.,T
f1
and T
f2
=0) and rearranging, results in
α−−=
)JJ(TT
mtturbm
(10)
α−=
)JJ(T
mtcomp
(11)Figure 9 shows an extension of Fig. 8 to includecompensation moment of inertia, (J
t
J
m
), and derivative of rotational speed used to produce the compensation torque. Thecompensation torque varies according to the acceleration ordeceleration of the wind turbine. A low pass filter (LPF) isincluded for noise elimination.
Fig. 9 The block diagram of wind turbine with compensatethe moment of inertia.
4.
IMPLEMENTATION OF WIND SIMULATOR SYSTEM
4.1 Wind simulator hardware
The hardware components in Fig. 10 compose of 1) threephase 4kW induction motor coupled with 1 kW dc generator, 2)1000 pulses rotary encoder, 3) torque control inverter (FRENIC5000 G11), 4) Digital signal controller (DSC) board, and 5) apersonal computer. The controller board uses a high performance16 bits dsPIC30f4011 which combines the advantage of highperformance 16bit microcontroller (MCU) and highcomputation speed digital signal processors (DSP). Speed signalfrom the encoder is sent to the torque controller inverter and theDSC board via a data acquisition interface circuit inside the DSCboard. The DSC board is linked with the personal computer viatwo RS232 ports: one port for transferring wind speed data to theDSC board and another port for sending the values of differentparameters (i.e., P,
ω
t
,
α
,
λ
, c
T
) to the computer.
Fig. 10 Wind simulator hardware structure.4.2 Wind simulator software
Figure 11 shows the diagram for signal flow of thesoftware implemented on the DSC. The command sets areprogrammed in C language. The algorithm starts by receivingrotational speed signal from the encoder and wind speed that canbe selected from three sources as shown by a block at the lowerend on the left hand corner of Fig. 11. The three sources are 1)manual input from a potentiometer, 2) a data file from a windspeed recorder and 3) a wind speed simulator software which isable to generate the timeseries wind speed based on a specifiedmodel such as the Van der Hoven power spectrum model.
Fig. 11 Signal Flow of Wind simulator software.
With wind speed data, the DSC is able to calculateparameters such as
ω
g
,
λ
, and
α
. The parameter c
T
, which isneeded for the aerodynamic torque calculation, can be directlyobtained from a look up table. The DSC computes the torquecompensation and the periodic torque ripple due to the towereffect represented by a block diagram at the top of Fig.11. Once
Τ
m
is obtained, it will be sent to D/A to convert to motor torquereference command in the form of voltage ranging from 0 to10V. This command will be passed to the torque control inverterto drive the motor.
5. EXPERIMENTAL RESULTS
Three experiments were conducted on the wind turbinesimulator: 1) Determination of the wind turbine characteristic, 2)