Professional Documents
Culture Documents
net/publication/3353056
CITATIONS READS
324 169
3 authors, including:
Some of the authors of this publication are also working on these related projects:
Control Systems and Hardware Topologies for Modular Multilevel Converter Applications View project
All content following this page was uploaded by R. Pena on 29 May 2017.
R. Pena
J.C. Clare
G.M.A s her
Indexing terms: Doubly f e d induction motor, P WA4 converters, Vector control, Wind energy
= slip angle
Abstract: A doubly fed induction generator S = slip
(DFIG) designed to be driven at variable speed Te = electromagnetic torque
from a wind turbine and supplying an isolated Topt = optimum torque
load is presented. Two back-to-back PWM
voltage-fed inverters connected between the stator Pop, = optimum power
and the rotor allow sub- and super-synchronous 6 = chopper duty cycle
operation with low distortion currents. The load V = wind velocity
voltage is maintained at constant frequency and B = friction coefficient
its magnitude is regulated through control of the
stator flux of the generator. An auxiliary load is SufJces
connected in parallel with the main load, and the d, q = d-q (synchronous) axes
auxiliary power is controlled to allow the DFIG a, p = a, p (stationary) axes
to track the optimal wind turbine speed for S, Y = stator, rotor
maximum energy capture from the wind. An
indirect stator-orientated vector control scheme is Superscripts
used to control the DFIG and this results in *
= demanded (reference) value
constant load voltage and frequency for
variations in both load and wind speed. The 1 Introduction
techniques have been implemented and validated
on a 7.5kW experimental rig. The doubly fed induction generator (DFIG) can supply
power at constant voltage and constant frequency while
the rotor speed varies. This makes it suitable for varia-
ble speed wind energy applications. Additionally, when
List of symbols a bidirectional AC-AC converter is used in the rotor
circuit, the speed range can be extended above synchro-
Vds, vqa, Vdr, vqr = 2-axis machine voltages nous speed and power can be generated both from the
id,s,i,,, idr,iqr = 2-axis machine currents stator and the rotor. An advantage of this type of
h = flux linkage DFIG drive is that the rotor converter need only be
rated for a fraction of the total output power, the frac-
L,. L,, Lo, L, = machine inductances per phase tion depending on the allowable sub- and super-syn-
R,. Rr = machine resistances per phase chronous speed range.
0, = leakage factor A good introduction to the operational characteris-
= magnetising current referred to stator tics of the DFIG connected to the grid can be found in
h S
[l, 21 in which Scherbius schemes using either cyclocon-
T,S = stator time constant
verters or six pulse naturally commutated DC-link con-
P = pole pairs verters are used. Problems affect one or both types of
a,?Lor = supply, rotor angular frequency converter: restricted speed range, extra forced commu-
0, = rotor position tation at synchronous speed and poor quality converter
currents. These may be overcome by using back-to-
0, = stator flux angle
back PWM converters in the rotor circuit. Such con-
0, = stator voltage angle verters are characterised by low distortion currents
0IEE. 1996 both in the machine rotor and the supply grid. They
IEE Proceedings online no. 19960454 also provide control of the overall system power factor.
Paper first received 2nd November 1995 and in revised form 1I th March A grid-connected DFIG system using such a scheme
1996 has been simulated in [3] and shown to be experimen-
The authors are with the Department of Electrical and Electronic tally feasible [4]. A full experimental investigation of
Engineering, The University of Nottingham, University Park, the grid-connected scheme has been done by the
Nottingham NG7 2RD, UK present authors [5] who have demonstrated the excel-
380 IEE Proc.-Electr. Power Appl.. Vol. I43. No. 5, September 1996
lent dynamic performance of the vector control torque control so as to emulate a wind turbine.
schemes regulating the system power factor and the Dynamic results are presented illustrating the regula-
converter link voltage while providing for maximum tion of output voltage and frequency under load
energy generation under variable wind speed condi- impacts and variable wind speed while maintaining
tions. maximum energy capture from the wind.
The possibility of a DFIG supplying an isolated load
has been indicated in [6, 71 in which some mention is 2 Experimental system
made of the steady state control problem. In [8] results
from a simulation are presented in which the stator A schematic of the ovierall system feeding an isolated
voltage is controlledl directly rather than via a field ori- load is shown in Fig. 1. The system shown is in fact a
entated stator flux controller such as that -used in the modified version of the grid-connected system
present study. Direct control of the stator voltage leads described by the authors in [5] and the reader is
to considerable practical difficulties (which are over- referred to that work for a complete description and a
come with flux control) since the control characteristics discussion of the selection of the rated operating
are load dependent. In [9] a system is p-resented in points.
which the rotor is supplied from a battery via a PWM The DFIG used was a 7.5kW, 415V, 50Hz 6-pole
converter with experimental results from a 250 W pro- machine whose parameters are given in the Appendix.
totype. A control s,ystem based on regulating the rms The PWM converters used were standard 7.5kW com-
voltage is used which results in large voltage: deviations mercial bipolar transistor PWM inverters with a maxi-
and very slow recovery following load changes. Again, mum switching frequency of IkHz. For all of the
the control design is load dependent. In [lo] Joeng and experimental investigations the DC-link voltage was set
Park present a working scheme for a DFIC; supplying at 550V and the DFIG was operated with a reduced
an isolated load. This system is unable to operate voltage of 220V because of the voltage limitation of the
super-synchronously and the harmonic currents drawn converters [5]. The incllusion of line inductance in the
by the diode rectifier are undesirable. Only steady state rotor circuit is explained in [5]. The turbine and gear-
results are shown. box is emulated in the experimental system by a thyris-
This paper presents the analysis, control and experi- tor converter fed D C machine which simulates a
mental validation of a vector controlled variable speed 7.5kW turbine with cut-in and rated wind speeds of
DFIG supplying an isolated load. Two back-to-back 4ms-' and lorns-', respectively, corresponding with
PWM voltage-fed inverters connected between the sta- generator speeds of 5001 and 1500rpm.
tor and the rotor are used to allow bidirectional power The DC motor was a 15kW, 1500rpm machine oper-
flow. A novel indirect stator flux orientation is pro- ated under either speed or torque control. Under
posed for vector control of the generator in which the torque control, wind speed data is fed from a PC to a
stator flux angle is not derived from voltag,e measure- microprocessor which calculates the instantaneous tur-
ment and is thus shielded from noise and possible har- bine torque from a given turbine blade characteristic
monic contamination on the stator voltage. This angle (see the Appendix). This torque forms the torque
is also used to derive the voltage vector orientation for demand to the DC drive after compensation for drive
vector control of the supply-side PWM converter. The losses. The speed of the turbine-generator set is deter-
experimental validation of the scheme is presented first mined by an optimal speed tracking algorithm that
with the generator driven by a speed controlled D C effects maximum energy capture from the wind. When
motor at both sub- and super-synchronous speeds. the generated power exceeds the load power under
Under steady speed conditions, the output regulation these conditions, the surplus power is dumped in an
performance is investigated for impact changes in both auxiliary load. The exact nature of an auxiliary load
active and reactive load. For investigating the perform- depends on the stand-alone application; resistive heat-
ance under variable speed, an auxiliary load control ing and water pumping for irrigation are two common
system is presented and the DC motor operated under examples. In this study, the auxiliary load is a diode
DC d r i v e
- - -- -- - - -
I encoder(
current
control
PC host
user
Interface
&- 1- I
382 IEE Proc.-Electr. Power Appl.. Vol 143, No. 5, September 1996
rotor demand voltages uses the slip angle derived from: stator or the stator converter. If supplied by the stator,
= 8, .- 8, =
I w,*dt - 8,
where 0, is measured using a 720-line encoder. Eqn. 6
(6)
the rotor excitation curi~ent,id,., will naturally compen-
sate under the action of the i, loop described in the
previous section. This is investigated experimentally in
Section 5. Sourcing the reactive power from :he stator
forces the stator flux to rotate at the demand frequency converter can be achieved by forming the i demand
during both steady state and dynamic conditions for from the $,i PI controller; this can be done after the iq;
any shaft speed. demand reaches a givein limit (and then held at that
limit); or alternatively the ,i PI controller output may
3. I Stator flux and voltage regulation supply both demands through a sharing algorithm. The
Stator flux is regulated through control of .$,i This is optimum source of reactive power will depend on rela-
obtained from: tive loss considerations in the machine and the two
converters and is beyond the scope of this paper. In
(7)
this study iq*is normally set to zero.
Ads = ,A:, cos 8, + Xp, sin 8, A complication arising from the stand-alone applica-
tion concerns the derivation of the voltage vector posi-
,A, = /(vas - R,z,,)dt tion angle, 0,, for the stator converter. In a grid-
(8) connected application, the grid voltages may be
assumed to be free of harmonics and 0, is derived
Xps = / ( u p s - Rsips)dt through voltage measurements as described in [5]. In
the stand-alone case. there is no low impedance voltage
where a, B denotes the stationary reference frame. The source and stator voltage harmonics will predominantly
angle 8, is used to demodulate both the stator flux and arise from the PWM voltage harmonics of the stator
the stator currents although id, is not used. The derived converter and from the current harmonics drawn by a
,i is compared with the demand value andl the error converter controlled auxiliary load. It has been deter-
forms the command idv* via a PI controller a's shown in mined through experimental investigation that, even
Fig. 2. The i,, loop, shown in Fig. 3, was designed with with the LC line filters of Section 2 included, these
a closed-loop natural frequency of 6.7rads-'. This is voltage harmonics are amplified to an unacceptable
limited by the available converter voltage. degree if the voltage angle 8, is calculated from voltage
1 +as measurement. Fortunately, the angle 0, may be derived
-us from:
RS
8, = 13, 7r/2+ (9)
I
Fig.3 Magnetising current control scheme
*T- _1
where 0, is derived from the frequency reference which
is a clean software variable. The error in 0, due to the
DFIG stator impedance voltage drop under load is
negligible in practice. The resulting viability of the
stator side converter operation is a significant
advantage of the indirect orientation scheme described
In the present implementation, the magnetising cur- in Section 3.
rent demand ims*is set directly from the user PC and
the stator voltage is not compensated against load vari-
ations (the uncompensated steady state voltage regula-
tion is about 3% for maximum load which is
considered acceptablle for the purposes of the present
investigation). For all of the experimental investiga-
tions, imJ*was set to 6A corresponding to a stator volt-
age of 220V. More effective regulation could be
obtained deriving i,\* from the vqs equation in eqn. 1
10-
taking into account the stator resistanc;e. Under
dynamic loads, the voltage regulation will have the 8-
same bandwidth as the ,i loop; this is of c'ourse una-
voidable.
a
-
6-
4-
-.. --YW
2-
4 Vector control of the stator side converter 0 J
-
verter uses a reference frame oriented along the stator
voltage vector position. The analysis and iniplementa-
tion is fully described in [5]. Under stator voltage orien-
tation, the direct component of the AC side converter
8-
current, id, is proportional to the rotor power and is Irns
used as the actuating variable in a DC-link voltage ? 6-
control loop, the DC-link voltage reference being set by - 4t
the user from the P C The quadrature current compo-
nent, iq, represents the reactive power and directly 0 20 40 60 80 100 120
1
140
determines the displacement factor of the converter t,ms
current. In a stand-alone application, any reactive cur- Fig.5 Experimental response to a step change in resistive load: magnet-
rent drawn by the load must be supplied either by the ising current
-3001 I
0 20 40 60 80 100 120 140
t.ms Ol I I ,
0 100 200 300 400 500 600
Fi . 7 Experimental response to a step change in resistive load: stator
,,fag, t,ms
Fig. 10 Experimental response to a step change in reactive load powei
magnetising current
Figs. 4-7 show the response to a step change in the
active load power from 550W to 1650W (approx 15%
6 Optimal wind turbine operation using
to 40% of rated load with the reduced line voltage) auxiliary load
with the speed at 700rpm (s = 0.3). Fig. 4 shows the
d-q axis motor currents. After the impact there is a When the generator is driven by a wind turbine it will
momentary loss of field orientation due to the finite be necessary to match the power extracted from the
response time of the iqr loop; the loss of orientation wind with the isolated load power and the electrical
reduces the iqs component at the expense of a negative and mechanical power loss. Fig. 11 shows the turbine
ids component that serves to reduce the magnetising power-shaft speed characteristics (for various wind
current, as shown in Fig. 5 , and hence the voltage. The speeds) of the 7.5kW wind turbine emulated with the
orientation recovers in about 40ms (this does not relate
DC drive. The maximum energy capture curve (Pop,in
to the response speed of the i, loop because of the
Fig. 11) for below rated wind speed operation (BRWS)
reduction in i, and hence iqr*). Fig. 6 shows that the
is defined by
increase in i, draws an increase in active power
through the rotor circuit resulting in an increase in the
front-end converter current which is of course expected
at sub-synchronous speed. Fig. 7 shows the corre- where Kept is a function of blade parameters. To ensure
sponding voltage dip lasting barely one cycle. Both the the system operates on the maximum energy capture
magnitude and duration of the dip can be reduced curve, an auxiliary load is used to make up the differ-
through increasing the speed of the current loops. ence between the optimum power and the main load
power. For wind speeds higher than rated, the turbine
200 10 energy capture must be limited by using pitch control
100 6 or stall control [11].
.' ?
>O
-1 00
O -2 -
-6
-200 -10
0 20 40 60 80 100 120
i.ms
Fig.8 Experimental response to a step change in reactive load power:
voltage, stator current and magnetising current
z
k
3
a
6
4
- 3$zrnLsk’
-T , ( S ) - netising current, d-q axis rotor currents
6(s) 7-1s +1 (13)
1200
250
3
I
/
g
L
900
=-.
0 J
3’ 600 =d
0 5 10 I5 20 25 30 200
t.5 300
Fig. 17 Experimental response to a step change in loud power while
operating at constant (optimum) speed: d y axis rotor currents
0 30 60 90 120 150 180
t*s
Fig.21 Experimental results showing voltage control and speed tracking
for operation with a realistic wind profile
Q 60 S
I
‘ 4 40
1600 r
t,s
Fig. 18 Experimental response to a step change in loud power while
o erating ut constant (optimum) speed: magnetising current; auxiliary loud
k t y cycle
100
60
lo 403
20
01 1
I Y
5 6 7 8
wind speed, m / s
Fig.22 Experimental optimum speed tracking accuracy
9 10
0 5 10 15 20 25 30 35 40 45
+ optimum speed, 0 generator speed
t,s
Fig. 19 Experimental response to a step change in load during speed
tracking: auxiliary load duty cycle 8 Conclusions