Sei sulla pagina 1di 7

1

Doubly-fed Induction Generator Modeling and Control in DigSilent PowerFactory


C. Hamon, K. Elkington, M. Ghandhari Royal Institute of Technology, Stockholm, Sweden

AbstractSeveral computer programs exist to carry out dynamic simulations and this study will focus on one of them, namely DigDilent PowerFactory. It offers two built-in models of doubly-fed induction generator. A new model has also been developed, based upon a controllable voltage source. These three models are compared, in terms of dynamic behavior and simulation time. One of them is then used to study the impact of an input control signal based on the single machine equivalent method. This signal provides power oscillation damping. Index TermsDFIG, PowerFactory

inuence of this new signal on the damping of the system is studied. II. M ACHINE A. Working Principles In the case of a DFIG, both stator and rotor have three sinusoidally distributed windings, corresponding to three phases, displaced by 120 . The rotor is connected to the grid through converters and a DC-link. A schematic of such a system is presented in gure 1. The size of the converters limits the amount of power owing between the rotor and the grid.
THEORY

I. I NTRODUCTION

NTERNATIONAL agreements have set high demands on the share of renewable energy in the total energy mix. Among renewable sources, signicant investments have been made in wind power. In Europe, for instance, the total installed capacity of wind power was 76 GW at the end of 2009, out of which almost 10.5 GW was newly installed during 2009. This represents nearly 40% of new installations of all energy sources combined [1]. The number of large wind parks is due to rise, which will signicantly impact existing power systems. These effects must be studied, using exhaustive models of wind turbines. However, due to the intricacy of electric grids, models which are too detailed take a long time to simulate. Therefore a trade-off between simplicity and accuracy must be made, and comprehensive models should be simplied. The aim of this work is to use a specic simulation software package, DigSilent PowerFactory version 14 [2], to study the behavior of doubly-fed induction generators (DFIG). Comprehensive studies of DFIG [3][6] have been made. Models of different orders have been presented and different control strategies have been studied. PowerFactory offers two built-in DFIG models, one comes as a built-in example in version 13 and the second is a builtin component. PowerFactory has already been used in some studies [7], [8] but these have only used one of the models. There is a need to compare these models to study whether they differ from each other, and if so, in which aspects. Also, PowerFactory offers possibilities to create new models. A third new DFIG model has been created using a voltage source. The behaviors of the two built-in models and the new model are compared. With increasing wind power penetration, control of DFIG to secure operation of power systems will become all the more important. A specic control strategy that uses the single equivalent machine (SIME) signal is examined and the

DFIG

Grid

DC link DC C AC Rotor-side converter DC G rid-side converter AC

Fig. 1: DFIG with its converters

B. Electrical equations The electrical equations for DFIG are conveniently expressed in the dq -reference frame, rotating at synchronous speed: vds vqs vdr vqr ds qs dr qr = Rs = Rr = = ids iqs idr iqr + 1 d s dt 1 d s dt , , ds qs dr qr + qs ds qr dr , (1) + +s , (2) Xs ids + Xm idr Xs iqs + Xm iqr Xr idr + Xm ids Xr iqr + Xm iqs (3) (4)

where d and q denote the d and q components. The symbols v , i, R, X and stand for voltages, currents, resistances, reactances and ux linkages per second, respectively. The

subscripts s and r denote the stator and rotor quantities, m is used for the mutual reactance. The symbols s and s are the synchronous speed and the slip. Complex substitution is used in the following. To each vector quantity fdq of two elements can be associated one complex quantity f dq : fdq = fd fq fd + jfq = f dq . (5)

III. T HIRD - ORDER MODEL A very common assumption in power system simulations is that stator transients are neglected. The latter varies at high frequency and is not of interest when studying electromechanical transients [10]. In the dq -reference frame rotating s at synchronous speed, it means neglecting the d dt term in the stator electrical equation [11]. It eliminates the differential terms in equation (6) of the detailed model which then becomes a so-called third-order model. The stator resistance is often assumed to be very small so that it is usually neglected to further simplify the model. The stator voltage equation can then be rewritten as dqs . v dqs = j Dening (16)

Also, the usual per-unit system is used to normalize all electrical quantities. Equations (1), (2), (3) and (4) can then be rewritten as: v dqs = Rs dqs + v dqr dqs dqr dqs 1 d dqs , + j s dt dqr 1 d dqr , = Rr dqr + + js s dt = Xs dqs Xm dqr , = Xr dqr Xm dqs , (6) (7) (8) (9)

dqr , = j Xm E Xr X2 X = Xs m , Xr

(17) (18)

with all quantities in per-unit.

and using equations (8) and (9), the stator voltage equation (16) can also be written as [12] jX v dqs = E dqs , (19)

C. Mechanical equation The dynamics of the generator shaft relate the rotor speed and the electromagnetic torque: J dm = Tm Te , dt (10) E

behind a and it can then be represented as a voltage E reactance X as depicted in gure 2. jX dqs

where J is the inertia of the machine, Tm is the mechanical torque and Te is the electromagnetic torque. The mechanical torque Tm is related to the mechanical power Pm extracted from the wind and available on the turbines shaft: Pm p Tm = = Pm , (11) m r where r is the rotor electrical speed and p is the number of pole pairs. Introducing the electrical active power dened as Pe = qr idr dr iqr = ds iqs qs ids , the electromagnetic torque can be written as [9] p Pe . Te = s Dening the acceleration time constant as Tag = Jbase , Tbase (14) (13) (12)

v dqs

Fig. 2: Stator equivalent circuit Introducing the transient open-circuit time constant T0 : T0 = Xr , s Rr (20)

and using equations (9) and (16), the rotor voltage equation (7) can be expressed as dE 1 = dt T0 Xs E jss T0 E X .

Xm Xs X v dqr v dqs + jT0 s + X Xr

(21)

Considering equation (16), the electrical active power Pe dened in (12) can be rewritten Pe = vsd ids + vsq iqs , (22)

where base and Tbase are the base speed and the base torque, the mechanical equation can be written as: 1 dm (Tm Te ) , = dt Tag with all quantities in per-unit. (15)

which means that it is equal to the stator active power Ps , which is also dened by the right-hand side of equation (22). The electrical torque given by equation (13) can then be written p Te = Ps . (23) s

IV. C ONTROL

STRATEGY

The machine is controlled by the rotor-side converter. Let: VDC be the voltage of the DC-link md and mq be the modulation depths set by the rotor-side converter for d and q components, respectively. The rotor voltage components are therefore set as: 3VDC vrd = md , (24) 2 2 3VDC (25) vrq = mq . 2 2 The controlled quantities are either total active and reactive powers or total active power and stator voltage. Only a small part of the total active power is exchanged with the grid via the rotor. This means that stator active power can be controlled to control total active power. Using equations (8) and (16), expression (22) for stator active power can be rewritten as Xm Ps = (vsd idr + vsq iqr ). (26) Xs The grid-side converter is controlled in this study to prevent any ow of reactive power between the rotor and the grid. Controlling stator reactive power means therefore controlling total reactive power. Stator reactive power is expressed as: Qs = Im( vdqs dqs ) = vsq ids vsd iqs . Qs = 1 Xm (vsd iqr vsq idr ) |v s |2 . Xs Xs (27)

the critical machines and one set S2 with the non-critical machines, as dened in [13]. The following quantities are dened: Hi i sime =
i S1

Hi i
i S2

, Hi

(31)

Hi
i S1

i S2

Hi i sime =
i S1

Hi i
i S2

, Hi

(32)

Hi
i S1

i S2

where the symbols i , i and Hi denote the rotor angle, the angular speed and the inertia of synchronous generator i. From these the SIME signal is dened: ysime = sin (sime ) sime . (33)

This signal can be amplied and is used as shown in gure 3. The amplication gains KSp and KSq (or KSv if voltage is controlled instead of reactive power) can be different. In this gure, the subscript meas denotes measured values and the subscript set setpoint values.

ysime P meas KSp

Using equations (8) and (16), this expression can be rewritten: (28) P set

This expression shows the coupling between reactive power and stator voltage. Expressions (26) and (28) can be simplied when all quantities are expressed in the stator ux reference frame. The latter has it xaxis aligned with the stator ux. Given relation (16) between stator voltage and stator ux, this means that the y axis is aligned with stator voltage. Expressions for stator active and reactive powers become: Xm vs iyr , (29) Xs Xm 1 2 Qs = vs ixr v . (30) Xs Xs s Therefore, in this new reference frame, active power is controlled by acting upon the y -component of rotor current and reactive power (or voltage) by acting upon the x-component of rotor current. The two components of rotor current are controlled by acting upon rotor voltage, set by the rotor-side converter as described by equations (24) and (25). Two PI controllers are used to control either active and reactive powers or active power and stator voltage. Their outputs are modulation depths mx and my , dened in the stator ux reference frame, which are then rotated back to the dq reference frame to get md and mq . Another control strategy based upon the introduction of a new signal will also be studied. In this new strategy, a system with two sets of generators is considered, one set S1 with Ps =

P Ip

my

Qset or V set

P Iq

+ +

mx

Qmeas or V meas

KSq or KSv ysime

Fig. 3: Control strategy with additional signal The signal may add damping in the system if it is properly tuned. V. DFIG
MODELS

PowerFactory offers two built-in DFIG models. The rst is depicted in gure 4 and will be referred to as Model A in the following. The second is depicted in gure 5 and will be referred to as Model B in the following. In Model A, the rotor sides power factor is controlled to be one. The controller is however not instantaneous and small amounts of reactive power are either produced or consumed. In Model B, the DC-link is hidden and the user does not have

20,00 kV

BB 1
2-Winding Transformer
900,00 MVA 5,36 %

900,00 MVA 5,36 %

3,30 kV

900000. kVA

2-Winding Transformer(1)

900,00 MVA 0,43 %

consider gure 5 once again, the amount of power owing from the transformer to the grid can be dened. This model is therefore much more interesting than the other one. Finally, PowerFactory allows the user to dene a control strategy. The built-in components are controllable via input signals. Model A is controlled by the modulation depths of the rotor-side controller in a reference frame rotating with the rotor angle. With Model B, the user denes the two components of the rotor voltage directly, in the rotor-angle oriented reference frame. In addition to these two built-in models, a new DFIG model has been built upon a voltage source as represented in gure 6.

900,00 MVA 0,43 %

U11
400,00 MVA 1,15 kV

eq X eq E v g = v s : grid Fig. 6: Equivalent circuit One of the greatest advantages with this approach is that most power system simulation software provides controllable voltage source models. This means that even software packages which do not allow the user to create new components can be used. This can be done in PowerFactory. The user can create control schemes for existing components but the creation of completely new components is not possible. The built-in voltage source will therefore be controlled to emulate the DFIGs behavior, according to the strategy depicted in gure 7. In this gure, Eeq and represent the voltage magnitude and angle of the controlled source, and Ps , Pr and Qs are stator active power, rotor active power and stator reactive power, respectively.
1

400,00 MVA

400,00 MVA

PWM U1
900000. kVA
900000. kVA 1939. V

G ~
G1d 1 x 90000..

DC

Fig. 4: Built-in model with DC link in PowerFactory

Terminal

110,00..

Terminal(1)

110,00..

2-Winding..

1 138491,88..

Control scheme mx , my
G ~
Asynchron..

Computation of the power theoretically delivered by the DFIG Ps , Qs , Pr

Fig. 5: Built-in model without DC link in PowerFactory Computation of the controlled sources voltage magnitude and angle Eeq , Voltage source

any control over the grid-side converter. Other interesting differences exist, specically concerning the load ow calculations. With Model A, it is only possible to dene the active and reactive powers produced by the stator. There is no possibility to dene it as a PV (to set active power and voltage) or PQ (to set active and reactive powers) bus. This puts some strong limitations on the load ow behavior of this model, since we are generally interested in total power and not only stator power. With Model B however, the user can dene the generator bus as a PV or PQ bus. It is even possible for the user to dene the power ow at another bus. Thus, if we

P meas , Qmeas

Fig. 7: Strategy to compute Eeq and From the modulation depth components, the DFIG equa-

tions are used to calculate the theoretical active and reactive power transmitted to the grid. The total powers transmitted to the grid are: Pg = Ps + Pr , Qg = Qs . (34) (35)

same dymamics. For disturbance 2 however, Model B differs from the other two. The difference in behavior with the other two models is however very small and it can be seen in the generated active power and speed, Model B deviates from Model A and Model C mostly during the rst oscillation.

Once these powers are obtained, the voltage which must be set by the equivalent controlled voltage source to feed the grid is calculated. The actual powers transmitted to the grid are then measured, compared with the setpoints and used by the control scheme to compute the control signals mx and my . This model will be referred to as Model C in the following. The new model is based upon a voltage source but it might be noted that any built-in component which can be controlled to produce the right amount of power could also be used. VI. S TUDY
CASES

0.9 active power (pu) speed (pu) 5 6 7 time (s) 8 0.85 0.8 0.75 0.7

1.11 1.1 1.1 1.09 1.09 5 6 7 time (s) 8

(a) Active powers generated by the DFIGs 0.3 reactive power (pu) voltage (pu) 5 6 7 time (s) 8 0.28 0.26 0.24 0.22 0.2

(b) Speeds of the generators

In this paper, two sets of simulations are carried out. The rst is used to compare the three models presented in section V. The second studies the effect of the SIME signal introduced in section IV. Both sets of simulations use the system depicted in gure 8. G1 1 5 6 7 8 9 10 11 3 G3

1.04

1.02

1 5 6 7 time (s) 8

(c) Reactive powers generated by the DFIGs

(d) Voltages at bus 1

2 G2 L7 L9

4 G4

Fig. 9: Disturbance 1, magnied parts, for Model A [dotted], Model B [dashdotted] and Model C [dashed], setpoints [solid]

Fig. 8: Two-area system The system is composed of two areas, each containing two generators. Generators G2, G3 and G4 are synchronous generators. Generator G1 is a DFIG generator, modeled by either Model A, Model B or Model C. Except for the values used for the DFIG, all values can be found in [10]. As for the DFIG, the values can be found in the appendix in table II. A. Model comparison In this section, the synchronous generators are each equipped with an excitation system with power system stabilizer (PSS) whose values can be found in [10]. The active and reactive powers produced by the DFIG are controlled by the scheme described by gure 3 without additional signal. The following disturbances have been simulated: disturbance 1: a three-phase short-circuit at bus 9 cleared by tripping a line between buses 8 and 9 after 100 milliseconds disturbance 2: a reactive power setpoint increase of 0.3 p.u. (dened as Qset in gure 3) Magnied parts depicting the transient responses of the three models can be found in gures 9 and 10. For disturbance 1, all models behave in the same way. They have the
0.78 active power (pu) speed (pu) 6 8 time (s) 10 1.1 1.1 1.1 1.1 1.09 5 6 7 time (s) 8

0.78

0.78

(a) Active powers generated by the DFIGs 0.31 reactive power (pu) voltage (pu) 5 6 7 time (s) 8 0.31 0.3 0.3 0.29

(b) Speeds of the generators

1.07

1.06

6 7 time (s)

(c) Reactive powers generated by the DFIGs

(d) Voltages at bus 1

Fig. 10: Disturbance 1, magnied parts, for Model A [dotted], Model B [dashdotted] and Model C [dashed], setpoints [solid]

TABLE I: Simulation time comparison - elapsed time in seconds


Model A Simulation 1 Simulation 2 55 10 Model B 3 3 Model C 5 3

0.6 0.4 Active power (pu) 0.2 0 0.2 0.4 0 5 10 15 time (s) Fig. 11: Active power along a line between buses 8 and 9 after disturbance 1, with [dashed] and without [solid] the additional signal 20 25 30

Computation time is also an important criterion. A speed comparison has therefore been performed. The elapsed time for each simulation and each model is gathered in table I. The simulation times of Model B and the Model C compared to Model A can be nearly 20 times faster, depending on simulations. This is a considerable difference and makes Model A only interested to use for specic simulations such as studies of the DC-link behavior. B. SIME signal In this part, the DFIG is Model B (depicted in gure 5). The control strategy will be altered to include the additional signal as presented in section IV and compared to the scheme without this extra signal. The system depicted in gure 8 is known for interarea oscillations [10]. Generators G3 and G4 are the critical machines and generator G2 is the non-critical machine. Voltage is controlled instead of reactive power. All values for the control scheme can be found in the appendix in table IV. The excitation system for the synchronous generators is IEEE type AC4, whose parameter values can be found in [10]. The PSSs have been taken away from the synchronous generators. The aim is to study whether the SIME signal can add damping to power oscillations. A three-phase short-circuit at bus 8 cleared by tripping a line between buses 8 and 9 after 100 milliseconds is simulated. The active power along the remaining line between buses 8 and 9 reects the interaction between the two areas of the system and is plotted in gure 11. Without the additional signal, the system becomes unstable after a few oscillations. In gure 12, the output of the voltage regulator of generator 2 is plotted. This output is limited and the lower limit is hit only one oscillation after the fault. After two oscillations, the output begins oscillating rapidly between the lower and the upper limits. The same phenomenon occurs in the voltage regulators of generators 3 and 4. With the altered control scheme, the power oscillations are stabilized and damped and the equilibrium point of the system is altered. Since one line was tripped between buses 8 and 9, the power ow along the remaining line increases as observed in gure 11. Also, although the voltage regulator output of generator 2 hits the limit once, it is set back to its setpoint value as seen in gure 12. VII. C ONCLUSION The theory lying behind DFIG has been used to create a new model in PowerFactory based upon a voltage source. This model and two built-in models were compared in this study. Model B offers many more possibilities for load ow, can be dened either as a PQ or a PV bus and can be controlled to set the load ow values at a remote bus. Also, the difference

4 AVR output (pu) 2 0 2 4 0 5 10 15 time (s) Fig. 12: Disturbance 1, outputs of the voltage regulator of generator 2, with [dashed] and without [solid] the additional signal 20 25 30

in speed with Model A is very large. Considerable time can be gained by using this model in simulations. Finally, it is simpler to use and comes as a all-in-one component whereas the other model comes with a DC link, another converter and a transformer for the rotor side. This makes Model A suitable only when modeling the converters is of interest. The new model that was developed has been tested just in a few cases in this study. It is more complicated to use than built-in components but it is more exible because it relies on code and can be made to do exactly what the user wishes. The SIME signal has shown great potential to add damping to the system when using DFIG. R EFERENCES
[1] European Wind Energy Association. Statistics & targets, April 2010. [2] DigSilent GmbH. Manuals DigSilent PowerFactory, 14.0 edition, 2007. [3] JB Ekanayake, L. Holdsworth, and N. Jenkins. Comparison of 5th order and 3rd order machine models for doubly fed induction generator (dg) wind turbines. Electric Power Systems Research, 67(3):207215, 2003.

[4] Feng Wu, Xiao-Ping Zhang, K. Godfrey, and Ping Ju. Modeling and control of wind turbine with doubly fed induction generator. In Power Systems Conference and Exposition, 2006. PSCE 06. 2006 IEEE PES, pages 14041409, Atlanta, GA, October/November 2006. [5] R. Pena, J. C. Clare, and G. M. Asher. Doubly fed induction generator using back-to-back PWM converters and its application to variablespeed wind-energy generation. In Electric Power Applications, IEE Proceedings -, volume 143, pages 231241, May 1996. [6] P. Ledesma and J. Usaola. Doubly fed induction generator model for transient stability analysis. IEEE Transaction on Energy Conversion, 20(2):388397, June 2005. [7] M.A. Poller. Doubly-fed induction machine models for stability assessment of wind farms. In Power Tech Conference Proceedings, 2003 IEEE Bologna, volume 3, page 6 pp. Vol.3, 23-26 2003. [8] A.D. Hansen, F. Iov, P.E. Srensen, N.A. Cutululis, C. Jauch, and F. Blaabjerg. Dynamic wind turbine models in power system simulation tool DIgSILENT. Ris National Laboratory for Sustainable Energy, 2007. [9] P.C. Krause, O. Wasynczuk, and S.D. Sudhoff. Analysis of electric machinery. McGraw-Hill New York, 1986. [10] P. Kundur, N.J. Balu, and M.G. Lauby. Power system stability and control. McGraw-Hill Professional, 1994. [11] P. C. Krause, F. Nozari, T. L. Skvarenina, and D. W. Olive. The theory of neglecting stator transients. IEEE Transactions on Power Apparatus and Systems, 98(1):141148, January 1979. [12] K. Elkington, V. Knazkins, and M. Ghandhari. Modal analysis of power systems with doubly fed induction generators. In Bulk Power System Dynamics and Control - VII. Revitalizing Operational Reliability, 2007 iREP Symposium, pages 18, Charleston, SC, August 2007. [13] M. Ghandhari, G. Andersson, M. Pavella, and D. Ernst. A control strategy for controllable series capacitor in electric power systems. Automatica, 37(10):1575 1583, 2001.

TABLE III: Values used in the control scheme for the model comparison
Parameter Kp Tp Kq Tq KSp KSv Value 0.2 (p.u.) 2 (s) 0.2 (p.u.) 2 (s) 0 (p.u.) 0 (p.u.)

A PPENDIX A VALUES USED IN THE SIMULATIONS The values used for DFIG are gathered in table II. TABLE II: DFIG parameters
Name Rated Power Rated Voltage Number of poles Nominal frequency Stator resistance Rs Stator reactance Xs Rotor resistance Rr Rotor reactance Xr Mutual reactance Xm Acceleration time constant Tag DC Voltage Value 900 (MW) 3.3 (kV) 2 50 (Hz) 0 (p.u.) 3.1 (p.u.) 0.01 (p.u.) 3.1 (p.u.) 0.1 (p.u.) 1.145767 (s) 1.3225 (p.u.)

TABLE IV: Values used in the control scheme with SIME signal
Parameter Kp Tp Kv Tv KSp KSv Value 0.1 (p.u.) 2 (s) 0.1 (p.u.) 2 (s) -10 (p.u.) 5 (p.u.)

The PI controllers used in this paper to control active power, reactive power or voltage are dened as: PIp = Kp 1 + PIq = Kq PIv = Kv 1 sTd 1 1+ sTq 1 1+ sTv , , . (36) (37) (38)

The values in tables III and IV have been used for the model comparison in section VI-A and the study of the SIME signal in section VI-B, respectively.

Potrebbero piacerti anche