Renewable Energy 50 (2013) 780e785

Contents lists available at SciVerse ScienceDirect

Renewable Energy
journal homepage: www.elsevier.com/locate/renene

The impact of wind farms with doubly fed induction generators on power system electromechanical oscillations
M. Jafarian*, A.M. Ranjbar
Department of Electrical Engineering, Sharif University of Technology, Tehran, Iran

a r t i c l e i n f o
Article history: Received 18 January 2012 Accepted 12 August 2012 Available online Keywords: Small signal stability Wind farm Doubly fed induction generator Dynamic interaction Electromechanical oscillation

a b s t r a c t
Introduction of large amounts of new wind generation can affect the small signal stability of power systems with three mechanisms: displacing synchronous generators (SGs); reducing SGs power generation; and the dynamics of wind farms (WFs) interacting with the electromechanical mode of SGs. In this paper a novel approach is developed to investigate the impact of the latter mechanism on existing power systems oscillations. In this approach, the dynamic behavior of grid connected WFs is studied independent of the dynamic behavior of system SGs. This approach helps to identify the conditions in which the dynamics of WFs may interact with the electromechanical mode of SGs. Also it helps to foresee the impact of these probable interactions on the frequency and damping of system oscillations. By using this approach in a test system, it was shown that under some circumstances these dynamic interactions considerably decrease the damping of system oscillations but they barely change the frequency of system oscillations. The frequency of system oscillation and the operating point of WF are the two major parameters determine the severity of the decrease in oscillation damping. Comparison of the SG electromechanical eigenvalues calculated before and after the introduction of the WF in the test system, conrmed the prospects of the proposed approach. 2012 Elsevier Ltd. All rights reserved.

1. Introduction During the last decade, wind power has been the worlds fastest growing energy source [1] and many large WFs have been installed and integrated into power systems, caused the share of wind power to reach a considerable level [2]. As long as wind power penetration is insignicant, SGs determine the overall dynamic behavior of power systems, but introduction of large amounts of new wind generation can affect the stability of power systems [3]. These effects in the elds of frequency stability [4,5], transient stability [6], voltage stability [7] and small signal stability have been treated and addressed in recent research efforts. Among them, the small signal stability problem of power systems with high penetration levels of wind power is one of the major challenging elds. Small signal stability is the ability of the power system to maintain synchronism when subjected to small disturbances. In todays power systems, the small signal stability problem is usually the lack of sufcient damping torque for system oscillations [8]. Among the several wind generation technologies developed until now, variable speed wind turbines utilizing DFIGs are the

* Corresponding author. Tel.: 989133011810; fax: 982166164037. E-mail addresses: mojedt@yahoo.com, mojedt@gmail.com (M. Jafarian), ranjbar@ sharif.edu (A.M. Ranjbar). 0960-1481/$ e see front matter 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.renene.2012.08.033

most popular scheme in power system industry [9]. DFIGs do not introduce new electromechanical oscillations in power system but they can affect existing system oscillations by three mechanisms: displacing SGs; reducing SGs power generation; and the dynamics of WFs interacting with the electromechanical mode of SGs. The impact of DFIG based WFs on the small signal stability of power systems has been the subject of many recent contexts. The impact of large scale DFIG based wind power generation on power system oscillations is investigated in [10]. The dynamics of WFs are not considered and it is supposed that WFs displace existing network generators. In [11] an approach based on the sensitivity of system electromechanical modes with respect to the changes of system inertia is developed with the assumption that WFs replace network generators and thereby decrease system effective inertia. In this study WF dynamics are not considered too. Modal analysis technique is used in [9] [12], and [13] to analyze the impact of DFIG based WFs on power system oscillations. Up to now in most of the studies devoted to this subject, either WF dynamics have been neglected or their impact on power system oscillations have been investigated with the help of modal analysis technique. Modal analysis technique can afford the identication of dynamic interactions only in the simulated conguration and does not suggest a general pattern. This work develops a novel approach to investigate the interactions between the dynamics of WFs and the electromechanical

M. Jafarian, A.M. Ranjbar / Renewable Energy 50 (2013) 780e785

781

mode of SGs. The basis of this approach is studying the dynamic behavior of grid connected WFs and system SGs independently. In this regard, rst the element of network Jacobian matrix with respect to which the SG electromechanical eigenvalue has the most sensitivity is identied; then by evaluating the inuence of the WF dynamics on that element, conditions in which the dynamics of WF may interact with the electromechanical mode of SGs are detected. Also the impact of these probable interactions on the frequency and damping of system oscillations is foreseen. 2. WF model For power system dynamic simulations, it is common to model WFs as a single equivalent wind turbine [11]. A wind turbine is consisted of the turbine, generator, drive train and converter. 2.1. Turbine In general, for stability studies, the dynamics related to the turbine, yaw system, and tower can be ignored [14,15]. Therefore wind turbines input mechanical power is assumed constant in this paper. 2.2. Generator Equation (1) describes the dynamic behavior of a DFIG. These equations are obtained by transforming the machine three-phase voltage equations into a synchronously rotating frame, referred to as the dq frame.
Fig. 1. DFIG controls.

mechanical and electrical torques respectively; and k and c are the shaft stiffness and damping coefcients respectively. 2.4. Converter In case of DFIGs, a three-phase voltage is injected into the rotor through a back-to-back converter system. This converter is composed of three parts: grid side converter; DC link; and rotor side converter. If the switching frequency is high enough and the switching losses are ignored, for power system stability studies it is possible to neglect the dynamics related to the grid side converter and DC link [16]. To model the rotor side converter, the decoupling control strategy for the active and reactive power generation of DFIGs, that was proposed in [17], is used in this paper. Fig. 1 shows the block diagram of this control strategy. Similar control structures have been used in [16,18], and [19]. In this paper only the voltage control mode is considered for WFs. In this mode the reactive power reference (Q* in Fig. 1) comes from the simplied equivalent of the supervisory controller of the WF which is depicted in Fig. 2. In this mode, the terminal voltage reference (V* in Fig. 2) is set by the WF operator. DFIG parameters and WF controller parameters are given in the Appendix. 3. The impact of WFs on system electromechanical oscillations To investigate the impact of WFs on system electromechanical oscillations, Test system A is developed and used. Fig. 3 depicts the oneeline diagram of this test system, where impedances are given in p.u. on a 1000 MVA base. In this paper for the sake of simplicity, only the oscillation of one SG against a strong system (represented as an innite bus) is considered. The fourth-order model with exponential modeling of magnetic saturation is considered for the SG. Constant mechanical torque input is assumed. The IEEE-type AC4A excitation system model is used. Parameters of the SG are given in the Appendix. The SG supplies a 150 MW load. One half of the load is represented as constant impedance and the other half is modeled as constant power. In Test system B, as depicted in Fig. 4, a 150 MW DFIG based WF is introduced in Test system A. It is assumed that the WF is operating at its rated operating point (rated wind speed).

  1 d j vsqd Rs isqd us jsdq ueb dt sqd   1 d j vrqd Rr irqd us ur jrdq ueb dt rqd

(1)

jsqd Xss isqd Xm irqd jrqd Xrr irqd Xm isqd

In these equations v, i and j represent the voltage, current and ux respectively; us and ur are the stator and rotor rotating speeds respectively; ueb is the rotating speed of the synchronous reference frame and equals 2p60 rad/s; R is the resistance; Xss, Xrr and Xm are the stator, rotor and mutual reactances respectively; and subscripts s, r, d and q stand for the stator, rotor, d-axis and q-axis variables respectively. 2.3. Drive train In general, to model the drive train of a wind turbine, it is enough to consider a two-mass model (one for the turbine, the other for the generator). The following equations are used to model the drive train of the wind turbine,

2Ht

dut dqtw Tm kqtw c dt dt (2)

1 dqtw ut ur ueb dt dur dqtw kqtw c Te 2Hr dt dt Te Xm iqr ids idr iqs

where ut and ur are the turbine and generator rotating speeds respectively; qtw is the shaft twist angle; Ht and Hr are the turbine and generator inertia constants respectively; Tm and Te are the

Fig. 2. Voltage control mode.

782

M. Jafarian, A.M. Ranjbar / Renewable Energy 50 (2013) 780e785 Table 1 SG electromechanical eigenvalues. SG capacity (MVA) 1000 1000 850 SG power generation (MW) 800 650 800 Electromechanical eigenvalue 0.119 8.970j 0.114 8.663j 0.113 10.259j

Fig. 3. Test system A.

and two outputs: Dp2 and Dq2. The dynamic behavior of System 1 can be described using the state equations of the SG as follows:

Introduction of new wind generation can affect system oscillations by three mechanisms: displacing SGs (and thereby reducing installed SGs capacity and system effective inertia); reducing SGs power generation; and the dynamics of WFs interacting with the electromechanical mode of SGs. The impact of each of these mechanisms is studied separately in this section. 3.1. Reduction in the capacity and active power generation of SGs To analyze the impact of reduction in the capacity and active power generation of SGs on system oscillations, rst we reduced the capacity of the SG in Test system A from 1000 to 850 MVA; then its active power generation from 800 to 650 MW. Table 1 shows the SG electromechanical eigenvalues for the base case and these two developed scenarios. Considering the results in Table 1, it can be concluded that reduction in the capacity of SGs increases the frequency of system oscillations but does not affect the damping of oscillations. Also reduction in the active power generation of SGs barely changes the frequency and damping of system oscillations. 3.2. Interactions between the dynamics of WFs and the electromechanical mode of SGs In this paper a novel approach is developed to investigate the interactions between the dynamics of WFs and the electromechanical mode of SGs. In this approach the dynamic behavior of WFs and SGs is studied independently. In this regard, Test system B is modeled as two separate systems: System 1 that contains the SG; and System 2 that contains the WF and the transmission network. The small signal model of System 1 has two inputs: Dp2 and Dq2 (incremental changes in the SG active and reactive power generation respectively) and two outputs: Dq2 and Dv2 (incremental changes in the bus 2 voltage angle and magnitude respectively). The small signal model of System 2 has two inputs: Dq2 and Dv2,

d xg Ag xg Bg Dp2 Dq2 T dt Dq2 Dv2 Cg xg Dg Dp2 Dq2 T

T

(3)

where xg is the vector containing the state variables of the SG. The dynamic behavior of System 2 can be described using its transfer function matrix as follows:

Dp2 Dq2 T GDq2 Dv2 ! GPq GPV G GQ q GQV

(4)

where G is the two-by-two matrix represents the transfer function of System 2. In Test system A, which contains no WF, G turns into a scalar matrix (network Jacobian matrix). In this case, substituting Equation (4) into Equation (3) gives Equation (5) which describe the state equations of Test system A. In these equations, Atot is the state matrix of the test system.

d xg Atot xg dt  1  Atot Ag Bg G1 Dg Cg

(5)

Introduction of the WF in Test system B changes the magnitude and phase of the elements of G (mainly because of the dynamics of the WF). In the proposed approach, interactions between the dynamics of WFs and the electromechanical mode of SGs are investigated in two steps: rst the element of G with respect to which the SG electromechanical eigenvalue has the most sensitivity is identied; then by evaluating the inuence of WF dynamics on that element, the impact of WF dynamics on the electromechanical mode of SGs is foreseen.

3.3. Sensitivity of the electromechanical eigenvalue of SGs with respect to the elements of G The sensitivity of the electromechanical eigenvalue of the SG in Test system A with respect to parameter a can be obtained using Equation (6). In these equations lEM is the SG electromechanical eigenvalue and JEM and VEM are its associated left and right eigenvectors respectively.

Table 2 Sensitivity values of the SG electromechanical eigenvalue. Sensitivity parameter jGPqj jGQqj jGPVj jGQVj Sensitivity value 0.296:91  0.008:183  0.027:297  0.001:38 Sensitivity parameter :GPq :GQq :GPV :GQV Sensitivity value 1.511:182  0.006:274  0.017:36  0.004:198


Fig. 4. Test system B.

M. Jafarian, A.M. Ranjbar / Renewable Energy 50 (2013) 780e785

783

Fig. 5. Bode plots of GPq in Test system A and Test system B.

Fig. 6. The electromechanical eigenvalues of the SG.

vlEM vAtot F JEM va va EM  1 2  2 vG vAtot G1 Bg G1 Dg Cg va va

state equations of the WF. In these equations, xd is the vector containing the state variables of the WF.

(6)

To obtain the sensitivity values with respect to the phases and magnitudes of the elements of G, vG/va in Equation (6) should be replaced with appropriate matrixes. For example, in the cases of magnitude and phase of GPq, vG/va should be replaced with the following matrixes respectively:

Dp2 Dq2 Dp3 Dq3 T JDq2 Dv2 Dq3 Dv3 ! Jg Jgd J Jdg Jd dxd Ad xd Bd Dp3 Dq3 T dt Dq3 Dv3 Cd xd Dd Dp3 Dq3
T T

(9)

(10)

vG vG va vjGPq j vG vG va v:GP q

1 0

0 0

! (7) 0 0 ! (8)

Substituting Equation (9) into Equation (10) gives Equation (11) which describes the state equations of System 2 (in Test system B). In these equations I is the two-by-two identity matrix.

jGPq j*1:90+ 0

dxd T Asys2 xd Bsys2 Dq2 Dv2 dt Dp2 Dq2 T Csys2 xd Dsys2 Dq2 Dv2   Asys2 Ad Bd I Jd Dd 1 Jd Cd   Bsys2 Bd I Jd Dd 1 Jdg   Csys2 Jgd Cd Jgd Dd I Jd Dd 1 Jd Cd   Dsys2 Jgd Dd I Jd Dd 1 Jdg Jg
T

The sensitivity values in Test system A are calculated and listed in Table 2. In this table the sensitivity values with respect to the phases of the elements of G are reported in degrees. By considering the results in Table 2, it can be concluded that the SG electromechanical eigenvalue has the most sensitivity with respect to the changes of GPq, so that an incremental change in the magnitude and phase of this element will result in an increase in the frequency, and damping of this mode respectively. Variations in the other elements of G inconsiderably affect this mode. 3.4. The impact of WF dynamics on GPq To form the transfer function of System 2 (G) in Test system B, each of the elements of System 2 (the WF and the transmission network) is modeled individually and then their models are combined together. The transmission network is modeled using its Jacobian matrix as in Equation (9). The inputs of this model are: Dq2; Dv2 (incremental changes in the voltage angle and magnitude of bus 2 respectively); Dq3; and Dv3 (incremental changes in the voltage angle and magnitude of bus 3 respectively), and its outputs are: Dp2; Dq2 (incremental changes in the active and reactive power injected in bus 2); Dp3; and Dq3 (incremental changes in the active and reactive power injected in bus 3 respectively). In these equations J is the network Jacobian matrix. The WF is modeled as a system with two inputs: Dp3 and Dq3 and two outputs: Dq3 and Dv3. To form the small signal model of the WF, the WF model described in Section 2 is linearized. Equation (10) describes the

(11)

The transfer function of System 2 can be formed using its state equations as follows:

Dp2 Dq2 T GDq2 Dv2 ! 1 GP q GPV Csys2 sI Asys2 G Bsys2 Dsys2 GQ q GQV

(12)

As it was mentioned, the electromechanical eigenvalue of SGs has the most sensitivity with respect to the changes of GPq. The Bode plots of GPq in Test system A and Test system B are depicted in Fig. 5. The frequency range of 0.6e30 rad/s (about 0.1e5 Hz), which
Table 3 Operating points of a DFIG wind turbine. Operation Sub-synchronous speed point Synchronous Super-synchronous speed speed 1.1 1 1.2 1 1.25 1

ur (p.u.)
PWF (p.u.)

0.67 0.74 0.8 0.89 0.93 1 0.3 0.4 0.5 0.7 0.8 1

784

M. Jafarian, A.M. Ranjbar / Renewable Energy 50 (2013) 780e785 Table 4 SG electromechanical eigenvalues in different WF operating points.

ur (p.u.)
No WF 0.74 1 1.25

PWF(p.u.) 0 0.4 1 1

lEM
0.116 0.128 0.081 0.179 14.656j 14.672j 14.665j 14.807j

Fig. 7. The Bode plot of GPq e sub-synchronous speed.

is the general frequency range of power system electromechanical oscillations, is chosen for this plot. As shown in Fig. 5, the dynamics of the WF decrease the phase of GPq in the frequency range of 13e18 rad/s (about 2e2.9 Hz). The most decrement happens in frequencies near 15 rad/s (about 2.4 Hz). Therefore, this approach foresees that introduction of the WF in Test system B decrease the damping of system oscillations with frequency in the range of 2e2.9 Hz and the most damping decrement occurs for system oscillations with frequency near 2.4 Hz. To verify the results, we calculate the electromechanical eigenvalues of the SG in Test system A and Test system B with considering different values for the SG inertia constant (from 1 to 4 s). Results (which are depicted in Fig. 6) conrm the prospects of the proposed approach. In this gure the direction of the arrow indicates increasing SG inertia. 4. Impact of WFs dynamics on system oscillations in different operating points of WFs The dynamic behavior of DFIG based WFs is totally dependent on their operating point because they operate in a large slip range. Table 3 describes a typical relationship between the steady state rotor speed and the active power generation of a DFIG [14]. Until now, we have assumed that the WF is operating at its rated operating point. Fig. 7 and Fig. 8 depict the Bode plots of GPq in cases

the WF is operating in sub-synchronous speed (WFs steady state rotor speed equals 0.74 p.u.) and super-synchronous speed (WFs steady state rotor speed equals 1.25 p.u.) respectively. By comparing Figs. 6e8, it can be said that the phase of GPq peaks at the same frequency of about 15 rad/s (2.4 Hz) in all the considered operating points but peak values are different (positive in the cases of sub-synchronous and super-synchronous speed operating points and negative in the case of synchronous speed operating point). By using the proposed approach, it can be foreseen that the introduction of the DFIG based WF in Test system B increases in the former cases and decreases in the latter case the damping of SG electromechanical oscillations with frequency near 2.4 Hz. For the considered operating points, in case the SG inertia constant equals 1.5 s, we calculate the electromechanical eigenvalues of the SG in Test system A and Test system B. Results (shown in Table 4) support the prospects of the proposed approach. 5. Conclusion In this paper a novel approach is developed to investigate the interactions between the dynamics of DFIG based WFs and the electromechanical mode of SGs and the impact of these interactions on power system oscillations. The basis of this approach is to study the dynamic behavior of grid connected WFs and SGs independently. In this regard, rst the element of network Jacobian matrix with respect to which the electromechanical eigenvalue of SGs has the most sensitivity is identied; then by evaluating the inuence of the WFs dynamics on that element, conditions in which the dynamics of WFs may interact with the electromechanical mode of SGs are detected. Also the impact of these probable interactions on the frequency and damping of system oscillations are foreseen. By using this approach in a simple test system, it was shown that these interactions occur in some frequencies. These interactions have no signicant impact on the frequency of system oscillations but they can change the damping of oscillations. The frequency of the oscillation and the operating point of the WF are the two parameters which determine the severity and direction of these changes. In this regard, it was shown in a test system that at rated operating point, the dynamics of the WF decrease the damping of system oscillations with frequency in the range of 2e2.9 Hz and at sub-synchronous and super-synchronous speed operating points, they will increase the damping of SG electromechanical oscillations with frequency near 2.4 Hz. Comparison of the electromechanical eigenvalues of the SG calculated before and after the introduction of the WF in the test system conrmed the prospects of the proposed approach. In this paper, the developed approach was used in a small test system and the impact of DFIG based WFs was investigated only on one type of system oscillations (oscillation of one SG against a strong system). This approach can be extended to be used in more complicated systems with different types of system oscillations. Appendix Parameters of SG: Rating 1000 MVA, Pg 800 MW, Vbase 26 kV, fbase 60 Hz, H 4 s,

Fig. 8. The Bode plot of GPq e super-synchronous speed.

M. Jafarian, A.M. Ranjbar / Renewable Energy 50 (2013) 780e785

785

Ra 0.003 p.u., Xl 0.16 p.u., Xd 1.81 p.u., Xq 1.76 p.u. 0 0 Xd 0.3 p.u., Tdo 8 s. Parameters of SG excitation system: KA 0.04, TA 0.04, TB 12, TC 1, TR 0.05. Parameters of wind turbine: Sbase 2 MVA, Vbase 690 V, us 2p60 rad/sec. Ht 4.33 s, Hr 0.62 s, k 1.11 p.u./el. rad, c 1.5 p.u.*s/el. rad. Rs 0.00488 p.u., Rr 0.00549 p.u. Xm 3.9530 p.u., Xss 4.0453 p.u., Xrr 4.0525 p.u. Parameters of DFIG controller: Kp1 0.1, Ki1 4, Kp2 0.1, Ki2 4, Kp3 0.01, Ki3 2, Kp4 0.1, Ki4 2. Kiv 5, Kpv 18, TC 0.15 s, TR 0.05 s, TV 0.05 s. References
[1] Fernandez RD, Jurado F, Saenz JR. Aggregated dynamic model for wind farms with doubly fed induction generator wind turbines. Renew Energ 2008;33:129e40. [2] Yingcheng X, Nengling T. Review of contribution to frequency control through variable speed wind turbine. Renew Energ 2011;36:1671e7. [3] El-Moursi MS, Bak-Jensen B, Abdel-Rahman MH. Novel STATCOM controller for migrating SSR and damping power system oscillations in a series compensated wind park. IEEE T Power Electr 2010;25:429e41. [4] Doherty R, Mullane A, Nolan G, Burke DJ, Bryson A, OMalley M. An assessment of the impact of wind generation on system frequency control. IEEE T Power Syst 2010;25:452e60. [5] Fernandez RD, Mantz RJ, Battaiotto PE. Impact of wind farms on a power system: an eigenvalue analysis approach. Renew Energ 2007;32:1676e88.

[6] Faried SO, Billinton R, Aboreshaid S. Probabilistic evaluation of transient stability of a wind farm. IEEE T Energy Conver 2009;24:733e9. [7] Vittal E, OMalley M, Keane A. A steady-state voltage stability analysis of power systems with high penetrations of wind. IEEE T Power Syst 2010;25: 433e42. [8] Kundur P. Power system stability and control. New York: McGraw-Hill; 1994. [9] Tsourakis G, Nomikos BM, Vournas CD. Contribution of doubly fed wind generators to oscillation damping. IEEE T Energy Conver 2009;24:783e91. [10] Slootweg JG, Kling WL. The impact of large scale wind power generation on power system oscillations. Electr Pow Syst Res 2003;67:9e20. [11] Gautam D, Vittal V, Harbour T. Impact of increased penetration of DFIG-based wind turbine generators on transient and small signal stability of power systems. IEEE T Power Syst 2009;24:1426e34. [12] Tsourakis G, Nomikos BM, Vournas CD. Effect of wind parks with doubly fed asynchronous generators on small-signal stability. Electr Pow Syst Res 2009; 79:190e200. [13] Vowles DJ, Samarasinghe C, Gibbard MJ, Ancell G. Effect of wind generation on small-signal stability; A New Zealand example, in power and energy Society general Meeting, 2008, 1e8. [14] Mei F, Pal B. Modal analysis of grid-connected doubly fed induction generators. IEEE T Energy Conver 2007;22:728e36. [15] Petru T, Thiringer T. Modeling of wind turbines for power system studies. IEEE T Power Syst 2002;17:1132e9. [16] Miao Z, Fan L. The art of modeling and simulation of induction generator in wind generation applications using high-order model. Simulat Pract Theory 2008;16:1239e53. [17] Yamamoto M, Motoyoshi O. Active and reactive power control for doubly-fed wound rotor induction generator. IEEE T Power Electr 1991;6:624e9. [18] Fan L, Kavasseri R, Miao ZL, Zhu C. Modeling of DFIG-based wind farms for SSR analysis. IEEE T Power Deliver 2010;25:2073e82. [19] Miao Z, Fan L, Osborn D, Yuvarajan S. Control of DFIG-based wind generation to improve interarea oscillation damping. IEEE T Energy Conver 2009;24: 415e22.