VIRTUAL INERTIA CONTROL OF DFIG-BASED WIND

TURBINES FOR DYNAMIC GRID FREQUENCY SUPPORT

Xiaorong Zhu1, Yi Wang1, Lie Xu2, Xiangyu Zhang1, Heming Li1
1
School of Electrical and Electronic Engineering, North China Electric Power University, Baoding 071003, China
2
School of Electronics, Electrical Engineering and Computer Science, Queen’s University of Belfast, Belfast BT9 5AH, UK
E-mail:͑[email protected]

Keywords: Virtual inertia control (VIC), doubly-fed DFIG control system decouples the mechanical and electrical
induction generator (DFIG), dynamic frequency support, systems and thus it prevents the generator from responding to
maximum power point tracking (MPPT). system frequency changes.

Abstract Similar to conventional generators, wind turbines have a

significant amount of kinetic energy stored in the rotating
This paper investigates virtual inertia control of doubly fed mass of their blades. In case of variable-speed wind turbines,
induction generator (DFIG)-based wind turbines to provide this energy will not naturally contribute to the inertia of the
dynamic frequency support in the event of abrupt power grid. Therefore, additional control is needed to make the
change. The model and control scheme of the DFIG is “hidden inertia” available to the grid. In [1] the factors
analysed. The relationships among the virtual inertia, the rotor affecting inertial response of a variable-speed wind energy
speed and the network frequency variation are then conversion system (WECS) are studied. In [2-3] some
investigated. The “hidden” kinetic energy that can be released techniques to emulate additional inertia using WECS are
to contribute to the grid inertia by means of shifting the presented. The frequency support is a combination of inertial
operating point from the maximum power tracking curve to a control, using the kinetic energy stored in the rotating masses,
virtual inertia control curve is investigated. The virtual inertia and a proportional control of frequency deviation. Once the
control strategy based on shifting power tracking curves of transient contribution of the wind generator finishes, its rotor
the DFIG is proposed and the calculation method for speed differs from the optimal value. In [4] an additional
determining these virtual inertia control curves is presented. control action is added to recover the rotor speed to the
A three-machine system with 20 percent of wind penetration optimal value. However, an oscillation of the injected power
is used to validate the proposed control strategy. Simulation is clearly observed. An algorithm to extract the maximum
results show that by the proposed control strategy, DFIG kinetic energy without stall the turbine is proposed in [5], in
based wind farms have the capability of providing dynamic which the electrical torque is increased step-wise by 20% at
frequency support to frequency deviation, and thus improving the time of the disturbances and is ramped down considering
the dynamic frequency performance of the grid with high the governor time constant. However, controlling power or
wind power penetration. torque requires different strategies. In [6], an inertia control
scheme of DFIG to extract energy from the turbine in a stable
1 Introduction way is proposed, however the strategy is not investigated in
detail and the rotor speed can not recover to its optimal value
Wind power generation is developing fast globally. However, after the dynamic frequency support.
the increasing penetration of wind generation in power system
has brought up new technical challenges, such as frequency In this paper, a novel inertial control of DFIG-based wind
control, system protection, balance control, transient stability turbines is investigated. In the event of dynamic frequency
and voltage stability. Among these technical issues, the changes, by shifting the operating point from the maximum
system inertia plays an extremely important role as it power tracking curve to the virtual inertial control curve, a
determines the sensitivity of system frequency to supply step electrical power reference of the rotor side converter
demand imbalances. During a frequency event, the dynamic (RSC) is generated and the kinetic energy is released /
grid frequency response is initially supported by the system increased to contribute to grid inertia. The rotor speed
inertia. recovery strategy is proposed to recover it to the optimal
value when the system frequency returns to a safe margin.
DFIG-based wind turbines have become the preferred option Simulation studies based on a power network with 20%
for large scale wind farms due to their ability to provide the installed wind generation capacity are performed to
benefits of variable-speed operation cost-effectively. demonstrate the benefits of the proposed schemes.
However, operating a large number of DFIG-based wind
turbines in place of conventional plants will reduce the 2 Modelling and control of DFIG
effective inertia of the system. This is due to the fact that the
The proposed virtual inertia control (VIC) of DFIGs is ­koptZr3 (Z0  Zr  Z1 )
developed from the maximum power point tracking (MPPT) °
3
control. Therefore, the dynamic model and MPPT control of * °° (Pmax  koptZ1 ) (3)
Popt ® (Zr  Zmax )  Pmax (Z1  Zr  Zmax )
DFIGs are briefly summarized at first.
° (Zmax  Z1 )
°P (Zr ! Zmax )
In the synchronous dq reference frame, the flux vector, °̄ max
voltage vector, power and electrical torque can be expressed where kopt is defined as the MPPT curve coefficient..
as
ȥs Ls I s  Lm I r
ȥr Lr I r  Lm I s
dȥs
Vs Rs I s   jZeȥs
dt
dȥ
Vr Rr I r  r  j Ze  Zr ȥ r (1) Tr Zr T s Ze
dt
(T r , T s )
3 ˆ 3
Ps  jQs  Vs I s  Vs ȥˆ s  Lm Iˆr (Zr , Zs )
2 2 Ls Vr_dq
*

'Vrq
3 3Lm
Te pD Im ª¬ȥs Iˆs º¼  pD Im ª¬ȥs Iˆr º¼

+
'Vrd

+
2 2 Ls

+
+
Vrd' Vrq'
I rq*

+
where Ȧe is the grid synchronous angular speed and Ȧr is the
rotor angular speed. Vs and Vr are the stator and rotor voltage *
I rd* Popt
vectors respectively. Is and Ir are the stator and rotor current

+
vectors respectively. ȥs and ȥr are the stator and rotor flux
respectively. Rs and Rr are the stator and rotor resistances, and Zr
Ls and Lr are the stator and rotor inductances respectively. Lm
is the mutual inductances and Te is the electrical torque. Ps Fig.1 Overall vector control scheme of the RSC
and Qs are the stator output active and reactive power
respectively, and pD is the number of DFIG’s pole pairs.

Using stator voltage oriented control, ignoring stator

resistance and assuming constant stator flux , the electrical
P /pu

torque, total output active power and stator reactive power

can be given by [7]

3 pLm
Te Vs ird
2Ze Ls
Z0 Z1 Zmax
3Zr Lm Zr /pu
Pe Vs ird (2)
2Ze Ls Fig.2 Maximum wind power point tracking curve
3 §V ·
Qs  Vs ¨ s  Lm irq ¸
2 Ls © Ze ¹ Under the MPPT control, the electromagnetic power can only
be regulated according to the rotor speed, which is decoupled
from the grid frequency. Therefore, the DFIG cannot
It can be seen from (2) that the decoupled control of the
contribute to system inertia in case of system power
electromagnetic power/torque and the stator reactive power
imbalance. In order to emulate the inertia response as
can be achieved by regulating the d- and q-axis rotor currents
synchronous generators using DFIGs, an improved control
respectively.
scheme by introducing the grid frequency deviation needs to
be investigated.
Fig. 1 shows the overall vector control scheme of the RSC.
The outer-loop control regulates the electromagnetic power
and reactive power independently and generates the reference 3 Virtual inertia control of DFIGs
signals Ird* and Irq* for the inner-loop current regulation. The System inertia determines frequency change rate during a
*
reference signal of the electromagnetic power Popt is sudden power change. Synchronous generators and fixed
determined by the MPPT curve, which is shown in Fig.2. speed wind turbines can automatically provide inertia support
According to Fig.2, the reference value of the electromagnetic for an abrupt frequency change, while the variable speed wind
power can be expressed as turbines can not due to the decoupled operation between the
rotor speed and grid frequency. However, a variable speed (2Zr0  'Zr )'Zr
wind turbine can emulate inertia even greater than its natural J vir J DFIG
(2Ze  'Ze )'Ze
inertia by fast active power control. (9)
'Zr Zr0 Z
| ˜ J DFIG O r0 J DFIG
3.1 Definition of DFIG virtual inertia 'Ze Ze Ze
The kinetic energy stored in the rotating mass of a where O='Zr/'Ze, is defined as the virtual inertia coefficient.
synchronous generator can be expressed as
1 Due to DFIG’s asynchronous operation, the allowed DFIG
Ek ³ ( Pm  Pe澡dt ³ J Zm d Zm J Zm 2 (4)
2 speed variation is far wider than that of a typical synchronous
where Zm is the rotor mechanical angular speed, J is the generator. According to (5), the inertia constant of the DFIG
moment of inertia in the shaft. can be defined as
J virZe 2
Generally, the inertia constant of a synchronous generator is H vir (10)
2 pD 2 PN_DFIG
defined as the ratio of the total stored energy at the nominal
speed ZmN to the rated power PN, i.e.
where PN_DFIG is the rated capacity of the DFIG. Thus the total
J ZmN 2 inertia time constant of the grid can be defined as
H (5)
2 PN
§ 1 m § ·
2·
n
Hence, the inertia constant of the power system with DFIG- 1
base wind farms can be expressed as ¦ 2 p 2 i e ¦ ¨¨ 2 p 2 J vir, jZe2 ¸¸
¨ J Z ¸ 
i 1© i ¹ j 1 © D, j ¹
H tot (11)
S N_tot
n § 1 · m
¦ ¨ 2 p 2 J iZe2 ¸  ¦ Ek_DFIG, j
i 1© i ¹ j 1 It can be seen from (9) that the virtual inertia of the DFIG is
H tot (6)
S N_tot determined not only by its nature inertia, but also by the pre-
disturbance rotor speed Ȧr0 and the virtual inertia coefficient Ȝ.
where pi and Ji are the number of pole pairs and moment of Different from synchronous generators, whose rotor speeds
inertia for synchronous generator i respectively. SN_tot is the are coupled directly to the system frequency, the speed
nominal capacity of the grid, and Ek_DFIG,j is the kinetic energy variation of the DFIG may be much greater than the system
stored in the rotating mass of DFIG-based wind turbine j. As frequency variation due to the asynchronous operation.
Ek_DFIG,j can not be seen if MPPT control is used, thus there is Therefore, the virtual inertia of the DFIG can be several times
of its natural inertia.
m
¦ Ek_DFIG, j | 0 (7) 3.2 Virtual inertia control of DFIG
j 1
Based on the MPPT control, the VIC scheme proposed in this
Therefore, operating a large number of DFIG-based wind paper is designed to regulate the DFIG output active power
turbines in place of conventional plants can significantly reference according to the system frequency deviation to
reduce the effective inertia of the system. supply inertial contribution to the grid. The regulation of the
active power reference is achieved by changing the MPPT
When the rotor speed of a DFIG changes from Zr0 to Zr1, the curve coefficient kopt to the VIC curve coefficient kVIC, which
kinetic energy available from the DFIG can be expressed by is a function of system frequency deviation. From (3), it can
considering an equivalent synchronous generator whose rotor be seen that different kVIC will generate a series of power
speed changes proportionally from Ze to Ze1 as tracking curves, defined as VIC curves. The upper and lower
limit of the VIC curves are defined as PVIC_max and PVIC_min
1 (shown in Fig.3), respectively, to ensure that a steady
'Ek_DFIG J DFIG ª¬(Zr0  'Zr ) 2  Zr0 2 º¼ / pD 2 operating point can be reached under any wind velocities.
2 (8)
1
J vir ª¬(Ze  'Ze ) 2  Ze 2 ) º¼ / pD 2 3.2.1 Calculation of kVIC
2
As shown in Fig.3, while the wind velocity is 9m/s, the DFIG
where 'Ze=Ze1-Ze, and 'Zr=Zr1-Zr0. JDFIG is natural inertia of operates at point A under the MPPT control initially. During
the wind turbine and Jvir is DFIG’s virtual inertia. an increase of system load, the grid frequency decreases. The
DFIG is switched from the MPPT control to the VIC and the
According to (8), the virtual inertia of the DFIG can be power tracking curve is shifted from the MPPT curve Popt to
obtained as the VIC curve (i.e., PVIC_max) immediately. The operating
point is thus moved from A to O and consequently the rotor
decelerates and the kinetic energy stored in the rotating mass
is released to support the grid frequency. After the initial immediately. The DFIG operating point is shifted from A to
dynamic frequency response period, the DFIG operates at O and its output power changes from PA to PO. Since the
point B under the VIC. electrical power is greater than the mechanical power, the
rotor decelerates and the operating point moves along the
Pvic_max curve to B.

After the initial dynamic frequency response period, the

frequency deviation will gradually reduce to zero with the
power system primary and secondary frequency control.
P /pu

According to (14), the power reference curve will recover to

MPPT curve, and the operating point is moved back from B
to A. Thus the rotor speed of the DFIG recovers to the
A optimal value after the dynamic frequency support.
'Zr
Zr1 Zr0 4 Simulation studies
Zr /pu
Fig.3 The scheme of the virtual inertia based power point tracking curve Simulations have been carried out to illustrate the ability of
the DFIG to emulate system inertia and to validate the
Assuming that the wind velocity and the mechanical power of proposed control scheme using Matlab/Simulink. A three-
the DFIG remain constant, the relationship between the active machine power grid shown in Fig.5 is used, which consists of
power reference at point A and that at point B can be two conventional power plants (G2, G3), three aggregated
expressed as loads (L1, L2 and L3) and a DFIG-based wind farm rated at
1000 MW(500*2 MW). The G2 and G3 are rated at 3500
kVICZr13 | koptZr03 (12)
MW and 700 MW, respectively. The three loads L1, L2 and
L3 are 1200 MW, 1500 MW and 600 MW respectively. The
where Ȧr1 can be expressed by frequency deviation
inertia constant of G2 and G3 are 5.2 s and 3.84 s respectively,
Zr1 Zr 0  'Zr Zr 0  O'Ze Zr 0  2ʌO'f (13) and the natural inertia constant of the DFIG is 3 s. The G2
and G3 regulate frequency by their governors with 4% droop
Substituting (13) into (12), the VIC curve coefficient kVIC can setting. Other parameters of the DFIG, G2 and G3 are given
be calculated as in appendix A. Comparative studies of network frequency
Zr 0 3 responses with and without the proposed VIC following load
kVIC 3
kopt (14) sudden changes were carried out.
Zr 0  2ʌO'f
Therefore, the VIC curve coefficient kVIC is the function of
the frequency deviation. The inertia control curves can be
achieved by replacing kopt with kVIC in Eq. (3).

3.2.2 Principle of the VIC

The principle diagram of the VIC scheme is shown in Fig.4.
A washed out function is used to eliminate the steady-state
DC component of the frequency error.
Fig.5 Structure of the simulation system
Zr
4.1 Sudden load increase
sT2 kVIC *
Popt_vic
1  sT2 The wind velocity is 9 m/s and the DFIG is initially under the
O MPPT control. Load L1 is increased from 600 MW to 900
MW at 3s. Under the VIC scheme , the DFIG is switched
Fig.4 Principle diagram of the VIC from the MPPT control to the VIC when the frequency
deviation is greater than 0.1Hz. Fig. 6 compares the frequency
The dynamic response of the VIC can be divided into two response with and without the VIC scheme. The virtual inertia
stages: fast dynamic frequency support stage (A-O-B) and of the DFIG is about 8 times of its natural inertia in this case
slow rotor speed recovery stage (B-A). (i.e. O=8). It can be seen from Fig. 6 that the virtual inertia
control improves the system performance significantly by
In the event of load increasing, the frequency will drop at the increasing the minimum frequency point from 49.61 to
first stage. Thus kVIC increases from the original value kopt and 49.76Hz. It also considerably reduces the rate of the change
reaches its upper limit rapidly. The corresponding power of frequency. It can also be seen that the DFIG recovers to the
reference curve will then be shifted from Popt to Pvic_max MPPT control after 20s.
 power and the rotor speed remain constant, whereas with the
VIC the rotor speed increases by 0.05 pu and the output active
power decreases by 0.08 pu. At the beginning of the
frequency increase, the VIC curve coefficient kVIC decreases
f / Hz

rapidly. When the frequency deviation reduces to zero, kVIC

again goes back to kopt.

Comparing the frequency responses shown in Fig.6 with O=8

and that in Fig.8 with O=6, it can be seen that larger O can
emulate more virtual inertia, and DFIGs with larger virtual
Fig.6 Comparison of the network frequency response during load L1 sudden inertia can provide better dynamic frequency support.
increasing 300MW

f / Hz
Zr / pu
Zr / pu

Fig. 8 Comparison of the network frequency response during load L1 sudden

decreasing 300MW

(a) MPPT control (b) VIC

Fig. 7 The dynamic response of the DFIG during load L1 sudden increasing
300MW
Zr / pu

Zr / pu
Figs. 7 (a) and (b) compare the dynamic responses of the
active power, rotor speed and kVIC with and without the VIC.
It can be seen that for the conventional MPPT control the
DFIG’s active power and rotor speed remain constant, which
indicates that the DFIG has no contribution to system
frequency variation. In contrast, when VIC is implemented,
the rotor speed decreases by 0.08 pu and the output active (a) MPPT control (b) VIC
power increases by 0.18 pu. At the beginning of the frequency Fig. 9 The dynamic response of the DFIG during load L1 sudden decreasing
change, the VIC curve coefficient kVIC increases rapidly and 300MW
reaches its maximum value. When the frequency deviation
reduces to zero, kVIC goes back to kopt gradually. 4.3 Comparison of the frequency responses of DFIG and
SG
4.2 Sudden load decrease
In order to illustrate the effect of VIC on system frequency
Under similar conditions as previous case, simulation results response, three cases are considered, i.e., G1 being a
of reduction of load L1 from 600 MW to 300 MW at 3s are synchronous generator with same capacity and inertia as the
shown in Figs. 8 and 9. Again, when the frequency deviation DFIG, G1 being the DFIG without inertia control, and G1
is greater than 0.1Hz, the DFIG is switched from the MPPT being the DFIG with the proposed inertia control. When load
control to the VIC. Compared the frequency response with L1 is increased from 600 MW to 900 MW at 3 s, the system
and without the VIC shown in Fig. 8, it can be seen that the frequency responses for synchronous generator (line fa),
VIC significantly improves system performance by DFIG without inertia control (line fb) and DFIG with the
decreasing the maximum frequency point from 50.39Hz to proposed inertia control (line fc) are shown in Fig.10.
50.28Hz. After 25 s the DFIG recovers to the MPPT control.
In this case, O=6. As shown in Fig.10, DFIG-based wind farm without inertia
control results in the largest frequency drop due to the
Fig.9 (a) and (b) compare the dynamic responses of the active reduced effective total system inertia.. However, with the
power, rotor speed and kVIC with and without the VIC. It can proposed VIC scheme, the DFIG can emulate inertia which is
be seen that with the conventional MPPT control the active higher than that of the equivalent synchronous generator and
consequently, it results in a significant improvement of frequency control,” IEEE Trans.Power Syst., vol. 21, no. 1,
system performance. pp. 433–434, Feb. 2006.
[3] G. Lalor, A. Mullane, and M. O’Malley, “Frequency
control and wind turbine technologies,” IEEE Trans.
Power Syst., vol. 20, no. 4, pp.1905–1913, Nov. 2005.
[4]. Juan Manuel Mauricio, Alejandro Marano, Antonio
f / Hz

Gómez-Expósito, and José Luis Martínez Ramos.

fa “Frequency regulation contribution through variable-
fb speed wind energy conversion systems”, IEEE Trans.
fc Power Syst., vol. 24, no. 1, pp. 173-180, Feb. 2009.
[5] Mustsfa Kayikci, Jovica V. Milanovic. “Dynamic
contribution of DFIG-based wind plants to system
Fig. 10 Dynamic frequency response to a load increase frequency disturbances”, IEEE Trans. Power syst., vol. 24,
no. 2, pp. 859-867, May 2009.
5 Conclusions [6] Yi Wang, Xiaorong Zhu, Lie Xu, Heming Li.
“Contribution of VSC-HVDC connected wind farms to
In the event of active power variation, system inertia grid frequency regulation and power damping”, in Proc.
determines the rate of frequency change. This paper 2011 Indus. Electro. Conf., Phoenix, USA, 2010.
investigates ways to make the “hidden inertia” of DFIGs [7] Yi Wang, Lie Xu. “Coordinated control of DFIG and
available to the grid. The model and independent active and FSIG-based wind farms under unbalanced grid
reactive power control strategy of DFIGs have been conditions”, IEEE Trans. Power Del., vol. 25, no. 1, pp.
illustrated. The virtual inertia of the DFIG is defined and the 367-377, 2010.
virtual inertia control strategy is proposed. Fast dynamic
frequency support is obtained by changing the active power
reference of the RSC from the MPPT curve to the virtual Appendix A
inertia control curves. The rotor speed recovery to the optimal
value is achieved. The proposed strategy is validated by Tab.1 Parameters of the 2MW wind generator pu
simulation studies of a three-machine network. The results Rs Ls Rr Lr Lm H p
show that DFIG-based wind farms are able to emulate inertia
by the proposed control and the virtual inertia can be larger 0.0108 0.102 0.01 0.11 3.362 3s 2
than that of a equivalent synchronous generator due to
DFIG’s asynchronous operation. Tab.2 Parameters of the synchronous generator G2 pu
Xd Xdc Xds Xq Xqs Rs
Acknowledgements 2 0.35 0.252 2.19 0.243 0.0045
c s s
This project is supported by National Natural Science Xl Td0 Td0 Tq0 H p
Foundation of China (50807015, 50977028) and Fundamental 0.117 8 0.0681 0.9 5.2s 1
Research Funds for the Central Universities.
Tab.3 Parameters of the synchronous generator G3 pu
References
Xd Xdc Xds Xq Xqc Xqs Rs
[1]A. Mullane and M. O’Malley, “The inertial response of 2.13 0.308 0.234 2.07 0.906 0.234 0.005
induction-machine-based wind turbines,” IEEE Trans.
Power Syst., vol. 20, no. 3, pp. 1496–1503, Aug. 2005. Xl Td0c Td0s Tq0c Tq0s H p
[2] J. Morren, S. de Haan, W. Kling, and J. Ferreira, “Wind 0.117 6.09 0.033 1.653 0.029 3.84s 1
turbines emulating inertia and supporting primary