Control of a Grid-Connected Variable Speed WECS Based on an

Induction Generator Fed by a Matrix Converter

Roberto Cardenas I, Ruben Penal, Pat Wheeler, Jon Clare2, Ramon Blasco-Gimenez(3)
(I) Electrical Engineering Department, University of Magallanes, P.O. Box I I3-D,
Punta Arenas Chile, email [email protected].
(2) School of Electrical and Electronics Engineering, University of Nottingham, email
[email protected]
(3) Technical University of Valencia, email [email protected]

Keywords: Induction Generator, Matrix Converter, Wind

Energy. Wind
Cage induction Pi
Abstract Machine ----+ Grid

In this paper the control of a grid-connected Wind Energy

Conversion System (WECS), based on a sensorless vector
~:::~~::~
i ~
controlled cage induction machine fed by a matrix converter is -----------------_.'
Variable Speed
/
presented. The matrix converter is controlled using a space Wind Turbine
vector modulation algorithm. Stability issues related to the
operation of the WECS connected to the grid using a matrix
converter are discussed in this work. The influence of a Output
synchronous rotating filter in the dynamic of the proposed current
WECS is analysed. The implementation of a Model Reference Fig. 1. WECS proposed in this work.
Adaptive System (MRAS) observer for sensorless control of the
proposed WECS is also presented in this work. 2 Control of the Variable Speed WECS
1 Introduction The control system proposed in this work is shown in
Fig. 2. The induction generator is driven by a variable speed
Matrix converters have many advantages, which are well wind turbine. The matrix converter is controlled using the
documented in the literature [I]. The matrix converter space vector modulation algorithm discussed in [3]. Zero
provides bi-directional power flow, sinusoidal input/output displacement factor at the matrix converter input is used in
currents and controllable input power factor [2-3]. The matrix this work. However variable displacement factor control is
converter allows a reliable and compact design due to the lack also feasible [2-3].
of dc-link capacitors for energy storage. Because of its The induction machine is vector controlled using a direct
relatively small size, in some applications the matrix vector control system orientated along the rotor flux vector
converter can be embedded in the machine itself [4]. [7]. Because the WECS is not required to operate at high
Considering its robustness, reduced size and reliability, rotational speed, the induction generator is operated at
matrix converters are good candidates for wind energy nominal flux. Therefore the current demand ids * is constant for
applications. the whole operating range.
The advantages of cage induction machines, for wind It can be demonstrated that in steady state, when the wind
energy applications, are also well known [5]. These machines turbine is operating at the maximum aerodynamic efficiency,
are relatively inexpensive, robust and require low the captured power Pm and the induction machine rotational
maintenance. When induction machines are operated using speed are related by [5],[8]:
vector control techniques, fast dynamic response and accurate
torque control are obtained. In this paper a variable speed Pm = koptm; (I)
WECS based on a grid-connected induction generator fed by where k opt depends on the blade aerodynamics, gear box ratio
a matrix converter is proposed. A Space Vector Modulation and wind turbine parameters [5]. From (I) and neglecting the
(SVM) algorithm [3] is used to control the matrix converter, losses, the WECS is operating in the optimal aerodynamic
regulating the torque and the magnetising current in the efficiency point when the induction generator torque is
generator. The proposed WECS is shown in Fig. 1. The controlled to:
induction generator is vector controlled using a conventional
Te = koptm; (2)
control system orientated along the rotor flux vector. To avoid
the use of a position encoder in the control system, in this using the state equations of a cage induction machine
work an MRAS observer [6] is implemented. orientated along the rotor flux [7], it can be shown that for
optimal power capture, the torque current demand i qs • is
obtained from (2) as:

55
 relationship is obtained as [3]:

Vo = %[vjmjC + v{md] (4)

Ii = -3 L
T -
~omi + 10 m
TC-d] (5)
2
The demanded values of md and fFfi are calculated as:
Input * -c*
Grid
_* V _* V
filter
md = o- m· =o- (6)
3-c 1 3v- c
vif if
where the vectors vo,io are the output voltage and output
current respectively; the vectors Vi' Ii are the input voltage
and input current respectively. In (6) "c" stands for complex
conjugated and V o * is the reference output voltage used by
Variable Speed the modulation algorithm. In Fig. 2, V o * is the output of the
Wind Turbine current controllers (i.e. V o *=Vds *+jvqs * ).
Fig. 2. Control system proposed.
The small signal models of (4-5) are obtained as:

Avo = %[AVjMiOC + ~oAm{ +Avf M dO + ~bAmd] (7)

.* 2 3 koptLr
I qs = ktUJr kt = 2
2PLo ids
* (3) Mj = ~ [MiOMo + loAmj + M doAio + [8 Amd ] (8)
2
where L n L o are the rotor and magnetising inductances Using (6) small signal functions for m; and mi* are obtained
respectively, P is the pole number and ids * is the direct current
as:
demand. Considering that sensorless operation is used in the
-* + 3MiO~vifC = ~vo
3Vifo~mi
-c*
(9)
proposed WECS shown in Fig. 2, the rotational speed
estimated by the MRAS observer is used to regulate the -*d + 3MdO~vifc-*
torque current demand of (3). The MRAS observer used in 3Vifo Mn = Avo (10)
this work has been discussed before in [5-6]. Further The stability of the matrix converter is improved when the
information about vector control of induction generators for input voltage used in the modulation algorithm is filtered
wind energy applications is outside the scope of this paper using a first order filter. In order to avoid a phase shift
and the interested reader is referred elsewhere [5],[8]. between the MC input voltage and the voltage used by the
SVM algorithm, the filtering is carried-out using a
2.1 Modelling of the Matrix Converter synchronous d-q rotating axis. The d and q components of the
input voltage are filtered using a first order low pass filter
The matrix converter topology used in this work is shown with a cut-off frequency of 1/Lf. In d-q coordinates the
in Fig. 3. Nine bidirectional switches, implemented using filtered voltage is obtained as (see Fig. 2):
IGBTs, are required. A second order input filter is used to vd·,_ vqi
reduce the harmonic content in the current supplied to the vd· =__ v· = - - - (11)
if Si1+ 1 qif Si1+ 1
grid. A resistor connected in parallel with the input
inductance may be required to improve the input filter A small signal model of the synchronous rotating filter is
damping coefficient. Assuming that the switching frequency obtained as:
is much higher than the input/output fundamental frequencies, ~V·
~vif =__ ,_ (12)
then the input/output voltages can be represented by their sil +1
average values over a cycle period. The MC input/output A simplified model in d-q coordinates can be obtained
Bidirectional
Grid input Input FiIter Switch assuming that the cage machine stator is a high inductance
--fi--------------1 load and Llio~O. If the output current does not change, then
:A the output of the current controller Llvo * ~O.
In the quiescent point it is assumed that ViO= V~fV= ViOc= VifDc
and Vo=qViO, where q is the voltage transfer ratio. Using these
assumptions the small signals models of (8) can be modified
to:
3 r- - I C _ ]
~li
T

= -llo~mi + OAmd (13)

2
a b c
the small signal models of (9-10) can be simplified to:
To cage induction * -q *-q
generator Ami = --~vif Amd = --~vif (14)
3ViO 3V;o
Fig. 3 Matrix converter topology used in this work.

56
using (14) in (13) assuming Mni* = ~mi ' ~m; = ~d; and Therefore for a disturbance Ltv; a current Lti is produced. If
!j is zero and assuming that the d-q components of the
considering VilFVO/g yields:
voltages and currents in the input filter are decoupled, then
-.- q2 [_ c
~li=-2V 10 +1 0 ~vif
-c] (15)
the equivalent resistance in the d-q axis can be calculated as:
o R eq d =- - - -1- - -
Defining Yo(mo}=1/Vo as the equivalent admittance connected - q2 Re(Yo(mo»
at the MC output, in d-q components (15) can be written as: (20)
Req q =- - -1- - -
Mid =_q2 Re[Yo({Oo)]~ - q2 Re(Yo (mo»
ST f +1
(16)
q Re Yo (mo ) -~vi
2 [ ] Therefore for the matrix converter operating at unity power
A •
uliq = --
factor there is always a negative resistance in either the d or q
ST f +1
axis (see (16». Therefore the proposed WECS can become
In (15) some digital effects have been neglected. For instance
unstable when the matrix converter is feeding a relatively
the dynamic effects of the zero order hold and the time delay
large load at the output.
produced because the duty cycles calculated in a given
To increase the stability, the synchronous rotating filter is
sampling time are used in the next processing cycle. These
used. For instance, if !j ~ 00, then the resistance calculated
effects can be important when the switching frequency is low.
The digital effects can be easily considered if the values of from (16,20) is a large value (i.e. Req~CX) and the WECS is
stable in the whole operating range. However, to use a very
md* and mi* are calculated as:
high value of 'J is not appropriate because in this case the

~;T ) Mnd = Mn;e -STd (1~;ST ) (17)

SVM cannot compensate disturbanceslharmonics in the
1
Mni = Mnte-STd[ MC input voltage and the output current can have a relatively
high harmonic distortion. Therefore a trade-off between
where Td is the time delay, T is the sampling time and system stability and output current waveform quality is
(l-e-ST)/(s1) is the representation of the zero order hold [3]. required in order to regulate the cut-off frequency of the
When the digital effects are important, (17) can be used to synchronous rotating filter.
obtain the small signal model of (16).
If the MC is operated at unity power factor, then the matrix 3 Experimental Results
converter can be represented by an equivalent resistance R.eq•
Therefore the input filter and the MC can be represented by The control system of Figs. 1 and 2 have been implemented
the block diagram of Fig. 4. In this graphic it is assumed that in the experimental rig shown in Fig. 5. The variable speed
the MC is connected to a strong grid (i.e. the grid impedance~ wind turbine is emulated using a speed controlled ac machine.
0). It can be shown that the damping coefficient of the system To implement the emulation, a wind speed profile is sent
shown in Fig. 4 can be obtained as: from the host PC to a second order model of the WECS
implemented in the DSP [5,9]. The power coefficient curve,
~Lf/Cf Cp(TSR, fJ), of [5] has been discretised and stored in a look-up
t; = 2R + t;filter (18)
eq table. Linear interpolation is used to obtain the power
Where is the no-load damping coefficient (i.e. when
e;,lter coefficient from the look-up table.
Req~oo). If I/R eq is negative, the damping coefficient may DSP-based
become negative producing unstable operation. Control and
When a small signal model is derived, then the equivalent Emulation system
resistor can be obtained as:
R = 8v i ~ ~vi (19) Modulation FPGA Based
eq a·'i A·
u1i and Control Switching

,_~ D
Grid
-----R.;----------
Model
~""H~~t"pc"~ L L__
-- __
- __
- __...--_---J
-_

3<1>
Input

Cage
Cage
induction
Motor
Generator
Fig. 4. Single phase models of the proposed WECS considering a strong grid.
Fig. 5. The experimental system.

57
 From the wind turbine model the rotational speed of the Fig. 7 shows the real rotational speed, the estimated rotational
WECS generator, m,.*, is calculated in each sampling time. speed and the speed error. For this test a standard MRAS
The driving ac motor forces the induction machine rotational observer similar to that presented in [5-6] is implemented.
speed to this value. With this emulation technique the cage The sensorless performance of the proposed WECS, based on
machine rotates at the same speed as that of a generator a matrix converter is quite good even when the wind turbine
driven by a real wind turbine. A complete discussion of the inertia has been reduced to 1kgm2, which is less than a half
emulation technique can be found in [9]. the inertia of a wind turbine of this size [5]. As shown in Fig.
The matrix converter is controlled using the SVM 7b, for this test the speed error is less than 2% for the whole
algorithm presented in [3]. A switching frequency of 12.5kHz wind profile.
is used to control the bidirectional switches. The matrix For the experimental test shown in Figs. 6-7, the cut-off
converter is controlled from a DSP and FPGA-based external frequency of the synchronous rotating filter has been adjusted
hardware. The commutation is controlled using the four-step to ~50Hz to achieve a good performance. In order to test the
method implemented in the FPGA [1]. A speed encoder of system stability, the cage machine is operated at 1150rpm,
10,000 pulses per revolution is used for comparison with a steady state torque current of 6A.The generated power
purposes and for the control of the ac machine. The MC is about 1.5kW. The cut-off frequency of the synchronous
input filter is implemented using a inductor of 0.625mH, a rotating filter is varied from 20Hz to 3000Hz in 6s. Fig. 8
delta connected capacitor of 2J.lF and a parallel resistor of shows the experimental results obtained in this test.
lOOn. The experimental prototype is rated at about 2.5kW. In Fig. 8a the variation of the cut-off frequency is shown.
Fig. 6 shows the performance of the proposed system when Fig. 8b shown the matrix converter input current. When the
the wind profile of Fig. 6a is used in the wind turbine cut-off frequency is increased, the damping coefficient of (18)
emulator. Fig. 6b, shows the power generated by the wind is reduced, and the oscillations in the input current increases.
turbine when the torque current of the cage machine is When the cut-off frequency of the rotating filter is ~1000Hz,
regulated according to (3). The performance of the matrix the system is unstable and the overcurrent protection trips.
converter based WECS is good, similar to that obtained from Fig. 8c shows the voltage transfer ratio q. Again, when the
a WECS implemented using back to back converters [5]. cut-off frequency increases, the oscillations in the voltage
Fig. 6c shows the d-q axis machine currents (ids, iqs ) and the transfer ratio q increases until the system becomes unstable.
power component of the matrix input current (idi ). As stated in Fig. 9 shows the instantaneous matrix converter input
Section 2, the direct current is regulated to a constant value current, in steady state operation, considering two cut-off
and the matrix converter input current is proportional to the frequencies. Fig. 9a shows the current for Ie =20Hz. In this
cube of the rotational speed for driving the WECS to the point case the damping coefficient of (18) is high and the input
of maximum aerodynamic efficiency (see (1)). current waveform is good without much distortion. Fig. 9b
shows the MC input current waveform, for the system close
to unstable operation. In this case the matrix converter input
--g
14

00 12 current has harmonic distortion with oscillations of

frequencies close to the input filter resonant frequency.
] 10
~
] 8 4 Conclusions
~ 6
In this paper a control system suitable for the operation of a
60

~ -1000

~ ]
o
1000
£ -2000
'E
~ 800

10 20 30 40 50 60 o 10 20 30 40 50 60

$
5
~
5
~------------..- --....- ......
........... ~ 1.5
Time(s)
.----,..----,..----~--~--~-----.

u
C'"
0
g
Q)
0.5
-6 -5 ~
Q)
0
& -0.5
r./)
-1
10 20 30 40 50 60 -1.5 '----_ _.L...-_ _.L...-_ _.L...-_ _.L...-_ _. l - _ - - - '
Time (s) o 10 20 30 40 50 60
Fig. 6. Experimental results using a wind turbine emulator. a) Wind profile Time(s)
used in the test. b) Output power of the system. c) d-q stator currents and Fig. 7. a) Rotational speeds corresponding to the test of Fig 6. b) Speed
matrix converter input current idi . error.

58
 3000 profiles of relatively high variabili~. For all the tests the
,-.. performance of the proposed WECS has been excellent.
::t 2000
'-"
~u 1000 Acknowledgements
(a)
0
,-..
0 2 3 4 5 6 This work has been funded by Fondecyt Chile, contract
5 ~~~~~~~~~~~~~~~~~~~ N~ 1060498.Thes~portofilieUn~~~~ofMagallan~~
<
'-"
I also acknowledged.
=t::
(!)

~
0
(b)
U
~
-5
'-" References
~
=
~
-10
0 2 3 4 5 6
1
[1] Wheeler, P.W., Rodriguez, J., Clare, J.C., Empringham,
L., Weinstein, A., "Matrix converters: a technology review",
0.8 IEEE Transactions on Industrial Electronics, Vol. 49, Nr. 2,
0.6
(c) pp. 276 - 288, April 2002.
cr
0.4 [2] A. Alesina and M. G. B. Venturini, "Analysis and design
of optimum amplitude nine-switch direct AC-AC
0.2
0 2 3 4 5 6 converters," IEEE Trans. Power Electron., vol. 4, pp. 101-
Time (s) 112, Jan. 1989.
[3] Casadei, D.; Serra, G.; Tani, A.; Zarri, L., "Matrix
Fig. 8. Experimental results showing the variation of the cut-off frequency of
converter modulation strategies: a new general approach
the synchronous rotating filter. a) Cut-off frequency fc in Hz. b) MC input based on space-vector representation of the switch state",
current. c) Voltage transfer ratio q. IEEE Trans. On Industrial Electronics, Vol. 49, Nr. 2, pp.
370-381, April 2002.
[4] P. Wheeler, J. Clare, D. Lampard, S. Pickering, K.
Bradley, L. Empringham, "An Integrated 30kW Matrix
Converter based Induction Motor Drive", Proceedings of
$ IEEE Power Electronics Specialist Conference, PESC05,
=
(1)

~ -1
Recife Brazil, June 2005.
U [5] R. Cardenas, R. Pefia, "Sensorless Vector Control of
Induction Machines for Variable Speed Wind Energy
-3
Applications", IEEE Transactions on Energy Conversion,
-5 0 10 20 30 40 50 60 70
Vol. 19, Nr. 1, pp. 196-205, March 2004.
Time (ms) [6] C. Schauder, 'IoAdaptive Speed Identification for Vector
6 Control of Induction Motors without Rotational Transducers",
IEEE Transactions on Industry Applications, Vol. 28, Nr. 5,
pp. 1054-1061, October 1992.
$2 [7] W. Leonhard, "Control of electrical drives", Springer-
=(1)
t::
::s
Verlage, Berlin 1985.
U [8] Mirecki A., Roboam X., Richardeau F.," Architecture
-2 Complexity and Energy Efficiency of Small Wind Turbines",
IEEE Transactions on Industrial Electronics, Vol. 54, Nr. 1,
-6 pp.660-670, Feb. 2007.
o 10 20 30 40 50 60 70 80
[9] R. Cardenas, R. Pefia, G. Asher, and J. Clare, "Emulation
Time (ms)
of wind turbines and flywheels for experimental purposes," in
Fig. 9. Experimental results showing the instantaneous input current. a) MC
input current for a cut-off frequency of 20Hz. b) MC input current for a cut- Proc. Eur. Power Electron. Conf., Graz, Austria, Aug. 2001.
off frequency close to 1000Hz.

WECS based on a cage induction generator fed by a matrix

converter has been presented. The control system proposed in
this work used a standard MRAS observer for sensorless
vector controlled operation of the induction machine.
The stability of the proposed WECS is improved by using
a synchronous rotating filter in the input voltage Vi. Filtering
the voltage used in the SVM algorithm enhances the damping
coefficient of the input filter.
Experimental results have been presented in this work.
The performance of the WECS have been tested using wind

59