IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 34, NO.

1, JANUARY 2019 163

Enhanced Frequency Regulation Using Multilevel

Energy Storage in Remote Area Power
Supply Systems
Yingjie Tan , Member, IEEE, Kashem M. Muttaqi , Senior Member, IEEE,
Phil Ciufo , Senior Member, IEEE, Lasantha Meegahapola , Member, IEEE, Xiaobin Guo,
Bo Chen, and Haomin Chen

Abstract—Frequency support from renewable power genera- based remote area power supply (RAPS) systems [1]. One ma-
tors is critical requirement to ensure the frequency stability of jor problem with renewable power generation is that it is not
remote area power supply (RAPS) systems with high penetration dispatchable as in the case of conventional energy resources
of renewable power generation. However, traditional control
strategies and the stochastic nature of wind resource constrain such as thermal and hydro. Therefore, it becomes onerous for
wind energy conversion system (WECS) such as permanent conventional generators to meet load side requirements in RAPS
magnet synchronous generator (PMSG) from participating in systems due to the increasing penetration level of renewable en-
frequency regulation. This work proposes to integrate hybrid ergy resources.
energy storage including ultracapacitors (UCs) and lead-acid The low inertia characteristic of small scale RAPS system re-
batteries (LABs) into a PMSG to provide frequency support. The
UCs deal with fast changing frequency by emulating conventional sults in system’s high sensitivity to the mismatch between gener-
inertial response, whereas the LABs mimic automatic governor ation and load demand. Large frequency excursions are common
response (i.e., primary frequency response). The mechanical power in such systems [2]. The adoption of power electronics based
reserved in wind turbine using suboptimal maximum power point renewable power generators such as wind energy conversion
tracking strategy is utilized to restore system frequency (i.e., systems (WECSs) aggravates the frequency stability problem
secondary frequency response). Moreover, supplementary control
strategies are proposed to enable the UCs and LABs to assist since these generators are not naturally involved in frequency
primary frequency response and secondary frequency response, regulation. The capability of frequency regulation using doubly-
respectively. Simulation study and experimental test are carried fed induction generator (DFIG) based WECS has been widely
out to validate the effectiveness of frequency response provided investigated [3]–[5] due to the large market share of DFIG in
by the multilevel energy storage. The multilevel energy storage renewable rich power networks. The WECS is being developed
solution can effectively regulate RAPS system frequency while
avoiding abrupt and frequent charging/discharging of the LABs to very high capacity and achieving popularity in remote lo-
and significant mechanical/electromagnetic stress on the WECS. cations like offshore where wind resource is abundant but the
harsh environment demands a greater degree of reliability on
Index Terms—Frequency regulation, multilevel energy storage,
permanent magnet synchronous generator (PMSG), remote area
WECS [6]. Permanent magnet synchronous generator (PMSG)
power supply system (RAPS), ultracapacitor, wind power. based WECS with full-scale converter can be a good option to
meet this requirement. PMSG is suitable for wind power gener-
I. INTRODUCTION ation due to its high torque-to-volume ratio, elimination of the
URAL electrification and reduction in dependency on requirement for excitation windings and its capability of direct
R fossil-fuel rely on renewable energy (e.g., wind, solar) drive variable speed operation without a gearbox. PMSG is gain-
ing attention on its contribution to frequency regulation. In [7],
Manuscript received November 5, 2017; revised March 9, 2018 and June a frequency controller based on virtual inertia and droop loop is
8, 2018; accepted August 19, 2018. Date of publication August 27, 2018; introduced to PMSG. The kinetic energy stored in rotating me-
date of current version December 19, 2018. Paper no. TPWRS-01666-2017.
(Corresponding author: Yingjie Tan.) chanical components of the WECS is extracted and injected into
Y. Tan, X. Guo, B. Chen, and H. Chen are with the Electric Power grid according to the variation in system frequency. The inferior
Research Institute, China Southern Power Grid, Guangzhou 510663, China aspect of the controller is a requirement for precise termination
(e-mail:,[email protected]; [email protected]; [email protected]; [email protected]).
K. M. Muttaqi and P. Ciufo are with the Australian Power Quality and of the frequency support to avoid a second frequency event (as
Reliability Centre, School of Electrical, Computer, and Telecommunications the WECS accelerates and restores the kinetic energy at the post-
Engineering, University of Wollongong, Wollongong, N.S.W. 2522, Australia disturbance stage). Similar frequency regulator is also applied
(e-mail:,[email protected]; [email protected]).
L. Meegahapola is with the School of Electrical and Computer Engineering, in [8] and the frequency support is enhanced with power reserve
Royal Melbourne Institute of Technology, Melbourne, Vic. 3000, Australia in WECS apart from kinetic energy. Nevertheless, the utilization
(e-mail:,[email protected]). of mechanical power in frequency regulation is a slow process.
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. Additionally, due to the frequent frequency fluctuation and thus
Digital Object Identifier 10.1109/TPWRS.2018.2867190 abrupt torque variation, the mechanical/electromagnetic stress

0885-8950 © 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.
164 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 34, NO. 1, JANUARY 2019

on the WECS will deteriorate the reliability of the WECS and

increase lifetime cost.
Energy storage can be a promising technique to relieve
WECS from the stress caused by frequency regulation. Par-
ticularly, ultracapacitor (UC) is a decent short-term energy stor-
age device due to its high round-trip efficiency, high number
of charging/discharging cycles, and fast charging/discharging
rates. These advantages enable UC to support instantaneous load
spikes [9]. UC is applied in [2], [10] for dynamic frequency sup-
port under frequency disturbance in an isolated power system.
However, due to the limited energy density of UC, the capability
of dynamic frequency support is limited to primary frequency
response that lasts less than 30 s [4]. Batteries have high en- Fig. 1. Control diagram of a PMSG with multilevel energy storage.
ergy density and suitable for long-term frequency regulation [9].
Particularly, lead-acid batteries (LABs) are still widespread in A diode bridge converts the variable magnitude variable fre-
RAPS systems due to their low initial costs, in spite of the avail- quency power generation to DC. A second stage involves a
ability of advanced battery technology such as lithium based boost converter to control the power output from PMSG. Hence,
batteries [11]. Nevertheless, the drawback of faster degradation typical maximum power point tracking (MPPT) or the SOPPT
rate demands sophisticated approaches to reduce lifetime cost. strategy can be implemented here. The PMSG is interfaced
An existing approach is to combine LAB with UC, which alle- with the RAPS network with the grid side converter (GSC) that
viates the drawbacks associated with the single energy storage. maintains a constant DC link voltage. The two types of energy
Hence, the hybrid energy storage may radically improve fre- storage device are connected to the DC link through two sepa-
quency regulation in a self-reliant RAPS system. In [12], [13], rate bidirectional DC/DC converters. The structure grants more
sizing methods are proposed to determine the optimal capacity flexibility in controlling the energy storage. The structure also
for battery and UC to cope with frequency deviation. A central- avoids an additional inverter by sharing the GSC to couple the
ized hybrid storage and converter are implemented at the point energy storage to the network, which is feasible due to: 1) The
of common coupling (PCC) in [14] to smooth the generation PMSG is not always operating at its full rating under variable
from wind farm. Contrarily, a grid connected PMSG with inte- wind speed; 2) The implementation of SOPPT strategy vacates
grated hybrid energy storage is studied in [15]. Majority of the capacity for frequency regulation service. Hence, this configu-
studies are focused on numerical study of the balance between ration can be more cost-effective than centralized configuration.
generation and load demand, which provide good guidelines Dynamic models for ultracapacitors and lead-acid batteries are
for the optimization of system component sizing. The contri- presented in [17].
bution of hybrid energy storage for improving RAPS system
dynamic performance remains to be explored. Authors in [16]
implemented centralized hybrid energy storage to emulate a III. MULTILEVEL ENERGY STORAGE BASED FREQUENCY
synchronous generator, but the coordination with generation re- REGULATION STRATEGY
sources is not considered. Conventionally, upon the occurring of a frequency distur-
Based on hybrid energy storage, a multilevel energy storage bance in power system, synchronous generators release their
solution is proposed in this work: 1) the hybrid storage, consist- stored kinetic energy inherently to provide inertial response and
ing of buffer storage (i.e., LAB) and cache storage (i.e., UC), limit the rate of change of frequency (ROCOF). Primary fre-
is integrated on the DC link of a PMSG. The UC is mainly quency response reserve is activated in a few seconds when
utilized to provide virtual inertia whereas LAB mainly supplies the frequency excursions exceed the noncritical frequency de-
primary frequency response. 2) a pre-defined mechanical power viation band. This process is generally achieved by increas-
reserve is established in WECS using suboptimal maximum ing/decreasing mechanical power output from the prime mover
power point tracking (SOPPT) strategy [3] to coordinate with using a droop-based governor. Secondary frequency response
hybrid storage for assisting restoration of system frequency. En- restores the active power balance between generation and load
hanced frequency regulation strategies are proposed to improve demand thus recovers the system frequency to the nominal value.
frequency regulation service provided by PMSG. Components Secondary frequency response lasts up to 15 mins until tertiary
in simulated RAPS system apply detailed model to study the frequency control is activated to free up the secondary frequency
dynamic performance of the PMSG. Experiments carried out response reserve [18].
on laboratory RAPS system validate the effectiveness of the In this paper, the hybrid energy storage and power reserve
proposed frequency regulation solution. available in the de-loaded PMSG provide frequency response in
different time scale. The control mechanism is sketched in Fig. 2.
Three levels of energy storage (i.e., UC, LAB, and mechanical
II. PMSG WITH MULTILEVEL ENERGY STORAGE
power reserve) and three types of frequency response (i.e., in-
The modular hybrid energy storage (i.e., UC and LAB), ertial response, primary frequency response, and secondary fre-
integrated into a PMSG based WECS, is shown in Fig. 1. quency response) are presented. The tertiary frequency control
TAN et al.: ENHANCED FREQUENCY REGULATION USING MULTILEVEL ENERGY STORAGE IN REMOTE AREA POWER SUPPLY SYSTEMS 165

Fig. 2. Multilevel energy storage based frequency regulation strategy.

is usually implemented manually and not considered in this pa-

per. Details on the multilevel frequency regulation scheme are Fig. 3. Virtual inertia controller for the UC.
presented in following subsections.

to absorb energy when the frequency increases. Consequently,

A. Virtual Inertial Response
primary frequency response is attenuated and recovery of sys-
UC is featured as having high power density and fast charg- tem frequency will be slowed down. In practice, stabilization
ing/discharging capability, which enables UC to handle fast and recovery period of a frequency event should be as short as
power variations under frequency disturbances. UC is also capa- possible, and this period is limited to 5 minutes in islanded sys-
ble of coping with frequent power absorption/injection due to its tems as stated in Australian National Electricity Market (NEM)
high number of charging cycles. Therefore, it is advantageous mainland frequency operating standards [20]. Hence, the nega-
to apply UC for providing power surge and emulating the iner- tive effect on the primary frequency response caused by UC in
tial response of conventional synchronous generators. ROCOF the frequency recovery section is better to be avoided. It is im-
is expected be reduced to a level where other generators in the portant to note that the power absorption is natural behavior for
system stay online without tripping. a synchronous generator during the frequency recovery period
The inertial response of a conventional generator can be rep- whereas the power consumed by UC is controllable. Therefore,
resented by (1) [19]. it is not compulsory for the UC to emulate conventional inertial
dωm response during the frequency recovery period. Contrarily, the
Pm = 2H · ωm · (1) inertial term in (2) can be inversely assigned to the UC power
dt
reference to enhance the primary frequency response. Hence,
where Pm is the mechanical power which is represented as the rather than absorbing power from the network, the UC main-
variation in kinetic energy, H is the inertia constant, and ωm is tains its power injection after the frequency deviation section.
the rotor rotating speed. Similarly, the virtual inertia term has 3) Quasi Steady Section: When frequency reaches lower
the form given by (2). limit of the noncritical operating frequency band, the virtual
df0 inertial term is assigned back to the UC power reference di-
PH = kc · (2) rectly, then the UC is ready for next frequency disturbance.
dt
As shown in Fig. 3, a low pass filter with a time constant
where f0 is the system frequency and kc is the inertia coefficient.
of Tf is applied in front of the derivative term to alleviate the
Since ωm is close to rated value (i.e., 1 p.u.), kc can be set as
impact of high-frequency noise on the controller. Pcc is the
(2H) by comparing (1) and (2).
ancillary UC charging set point. The error between the inductor
In order to enhance the primary frequency response to be
current reference (i∗L c ) and the actual current (iL c ) is fed to a
explained in Section III-B, the virtual inertial response term
current regulator to generate the duty ratio of gate signals for
is proposed to be directly or inversely assigned to the power
the bidirectional DC/DC converter as shown in Fig. 1.
reference (Pc∗ ) for the UC depending on ROCOF during a fre-
quency event. An example is shown in the control diagram of
UC in Fig. 3. The frequency event is divided into three sections: B. Primary Frequency Response
frequency deviation section, frequency recovery section, and Primary frequency response aims to raise the frequency nadir
quasi-steady section. above particular level (ruled by local frequency operating stan-
1) Frequency Deviation Section: When frequency de- dards) to avoid load shedding. For example, in Australian energy
creases, the virtual inertial response term is assigned to the market (mainland), automatic load shedding will be triggered
UC power reference to increase its power output. Hence, virtual when frequency reduces below 49 Hz as specified in [20]. Con-
inertial response is supplied to mitigate the ROCOF. sidering the small size of the RAPS system studied in this paper,
2) Frequency Recovery Section: Due to the activation of pri- a lower frequency (i.e., 48.5 Hz [2]) is chosen.
mary frequency response, the system frequency reaches the fre- Primary frequency response is generally obtained using P-f
quency nadir and starts to recover. A fraction of the electric droop mechanism. LAB is applied to provide long-term fre-
power will be absorbed by synchronous generators to increase quency response due to its higher energy density. Since the
their rotational speed and restore their kinetic energy. If the tra- droop control is relatively slow [21], the surging current would
ditional virtual inertia control strategy is used, the UC also starts not stress the LAB. Furthermore, a normal operating frequency
166 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 34, NO. 1, JANUARY 2019

Fig. 4. Droop based primary frequency response controller for the LAB. Fig. 5. Control diagram of the boost converter.

band is usually set around the nominal frequency to avoid un-

necessary activation of the primary frequency response. Hence,
frequent charging/discharging of the LAB can be avoided and
lifetime of LAB can be prolonged. In mainland Australia, the
band is set to be 49.5 to 50.5 Hz for islanded system [20]. Power
reference for the LAB is given by (3).

Pb∗ = Pbc + Pbs − kb · (f0 − f n om ) (3)

where fnom is the nominal frequency (e.g., 50 Hz in this study),

kb is the droop coefficient, and Pbc and Pbs are LAB charging Fig. 6. Architecture of the test RAPS system.
and secondary frequency response set point respectively. The
control diagram of the LAB is presented in Fig. 4. A deadband
is introduced for the frequency error (Δf) to set the normal and the AGC set point.
operating frequency band. Pbs =
⎧
⎨0, Pts = Ps − P0 (v)
⎪ Ps < Ptm ax (v)
C. Secondary Frequency Response Ps − Ptm ax (v), Pts = Pr es Ptm ax (v) < Ps < Ptm ax (v)+Pbm ax
⎪
⎩ m ax
Secondary frequency response usually implements central- Pb , Pts = Pr es Ps > Ptm ax (v) + Pbm ax
ized automatic generation control (AGC) to modify the active (4)
power set points of interconnected generators and restore fre-
quency to nominal value. The integration of LAB enables PMSG where Pts is the secondary frequency response set point of the
to participate in secondary frequency response. Apart from the wind turbine, P0 is the actual wind turbine power output, Pres
energy in LAB, mechanical power reserve is established in the is the mechanical power reserve in the wind turbine, and Pbm ax
wind turbine by operating the turbine away from its typical is LAB rated power. The control diagram of the boost converter
maximum power point intentionally. The strategy to deload the is shown in Fig. 5. Ramp rate limiter is applied to the turbine
PMSG is referred as SOPPT by the authors [3]. The mechanical power set point to reduce the mechanical/electromagnetic stress
power reserve can also be utilized as charging energy source on the WECS.
for the UC and LAB. A constant amount of mechanical power
(e.g., 0.01 p.u. in this work) is reserved by manipulating the IV. PERFORMANCE OF THE MULTILEVEL ENERGY STORAGE
wind turbine to follow the SOPPT curve.
The proposed multilevel energy storage solution is validated
As explored in [22], WECS is capable of participating in
on SimPowerSystem platform. A detailed RAPS system sim-
AGC using set point control. With two secondary frequency
ulation model (see Fig. 6) is established based on a practical
response reservoirs (i.e., LAB and mechanical power reserve),
laboratory scale RAPS system. Parameters of the system com-
a coordination strategy is required to accomplish power shar-
ponents are extracted from the components’ datasheets.
ing between the wind turbine and LAB. Equation (4) presents
the strategy which prioritizes the mechanical power reserve for
secondary frequency response. When the power set point (Ps ) A. Frequency Regulation Using Multilevel Energy Storage
assigned by AGC for the WECS is less than the available turbine Transient dynamics of the RAPS system under a frequency
power (Ptm ax (v)), secondary frequency response set points for disturbance generated by a sudden load increase (10%) are
the LAB and the turbine are set to be zero and (Ps − P0 ) re- shown in Fig. 7. Four scenarios are investigated: 1) WECS
spectively. If Ps is larger than the total power available in LAB without frequency regulation; 2) WECS equipped with UC
and wind turbine, the LAB power set point and turbine power providing virtual inertial response; 3) WECS with LAB
set point are assigned to their maximum. For other scenario, the participating in primary frequency response; 4) WECS with
turbine power set point is set at its maximum value whereas both UC and LAB being activated. Upon the increase in load
LAB fills the power gap between the maximum turbine power demand, frequency drops in all scenarios as shown in Fig. 7-(a).
TAN et al.: ENHANCED FREQUENCY REGULATION USING MULTILEVEL ENERGY STORAGE IN REMOTE AREA POWER SUPPLY SYSTEMS 167

Fig. 9. Effect of the virtual inertial coefficient. (a) Frequency. (b) WECS
power output.

Fig. 7. Comparison of various frequency regulation strategies.

Fig. 10. Effect of droop coefficient. (a) Frequency. (b) LAB power output.

B. Effects of Virtual Inertial and Droop Coefficients

Effects of virtual inertial coefficient in (2) and droop coef-
ficient in (3) on the performance of the frequency regulation
Fig. 8. Charging/discharging of the LAB and UC under frequency regulation.
are investigated by varying kc and kb , and results are shown in
Fig. 9 and Fig. 10 respectively. By increasing the virtual inertial
Without the presence of frequency regulation, scenario (1) has coefficient of the UC, the reduction in ROCOF and increase in
the lowest frequency nadir which is below 49 Hz, which may frequency nadir can be observed in Fig. 9-(a). The zoomed fig-
activate load shedding (load shedding is not implemented in the ure in Fig. 9-(b) demonstrates that the increase in virtual inertial
test). The high ROCOF may also trigger the disconnection of coefficient augments the initial power output from WECS when
interconnected generators. With the presence of UC, ROCOF the frequency disturbance occurs. Nevertheless, the improve-
is reduced, and the frequency nadir is raised slightly. The ment in frequency nadir is limited and further increase in kc
inclusion of LAB significantly reduces the frequency deviation will cause large swings in WECS power output and deteriorate
and raises the frequency nadir. This is achieved by injecting the stability of the RAPS network.
a significant amount of LAB power into the RAPS network In regard to the primary frequency response provided by LAB,
as seen in Fig. 7-(b). In scenario (4), both the frequency nadir the effect on the ROCOF is negligible whereas frequency nadir
and ROCOF are improved, and optimal frequency regulation is significantly improved with larger droop coefficient as shown
is obtained. Due to the application of droop-based frequency in Fig. 10-(a). Particularly, the frequency nadir is raised above
response, the WECS power output increases in comparison with 48.5 Hz when kb is set at 0.04. Consequently, the load shedding
the pre-disturbance value. This attributes to the permanent fre- will not be activated during the frequency disturbance. Since in-
quency deviation caused by load increase (secondary frequency terconnected generators share loads in proportion to their droop
regulation is not activated to restore the system frequency at coefficients, permanent change in WECS power output at post-
this stage). Meanwhile, the power output from parallel diesel disturbance stage increases with larger kb (the diesel generator
generator also increases as shown in Fig. 7-(c). It is important is also governed by droop control strategy).
to note that the wind turbine power output does not change
in all scenarios. Hence, mechanical/electromagnetic stress on
C. Effect of Enhanced Primary Frequency Response
the WECS is eliminated compared to the case when frequency
regulation is provided through turbine control as studied As discussed in Section III-A, the proposed supplementary
in [7], [8]. control enhances the primary frequency response by involving
Fig. 8 presents the power output from UC and LAB in sce- virtual inertia of the UC in aiding the primary frequency re-
nario (4). It can be seen that UC covers all the fast changing sponse provided by the LAB. The performance of the enhanced
power profile whereas LAB supplies slow changing power. The primary frequency response is compared with original strat-
elimination of frequent charging/discharging cycle on the LAB egy, and the results are shown in Fig. 11. Higher climbing rate
is beneficial for extending lifetime of the LAB. of the frequency during the frequency recovery section can be
168 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 34, NO. 1, JANUARY 2019

Fig. 11. Enhanced primary frequency response.

Fig. 12. Architecture of the laboratory RAPS system.

observed in Fig. 11-(a) in the case of the enhanced controller.

Overshoot of frequency in quasi-steady state section is flattened.
The faster frequency restoration is due to the increased power
output from WECS as shown in Fig. 11-(b) during the fre-
quency recovery period. As expected, the enhanced controller
has no impact on frequency nadir.

V. EXPERIMENTAL TESTS
Experimental tests are carried out on a laboratory RAPS sys-
tem to further validate the multilevel energy storage solution.
The architecture of the laboratory RAPS system (see Fig. 12) has
the same configuration as the simulation model shown in Fig. 6
(Synchronous generator is not mounted adjacent to the PMSG
test rig and not shown in the figure.). The PMSG is coupled
to an induction motor which is controlled by a variable speed
Fig. 13. Performance of the transient frequency regulator. (a) Without fre-
drive (VSD) to emulate torque-speed characteristics of a wind quency support from the WECS. (b) With virtual inertia and primary frequency
turbine. The diesel generator is emulated by a synchronous gen- support from the WECS. (c) With virtual inertia and enhanced primary fre-
erator coupled to another induction motor. The induction motor quency support from the WECS.
is also controlled by a VSD. A LAB bank and a UC bank are
connected to the WECS using two independent bidirectional
any frequency support, as shown in Fig. 13-(a), a frequency
DC/DC converters.
dip of 0.71 Hz occurs. Due to the frequency regulation of the
synchronous generator, the frequency returns to the noncrit-
A. Performance of the Transient Frequency Regulator ical operating frequency range, and the recovery takes 1.06s
In this scenario, the synchronous generator operates in while the overshoot in frequency exceeds the noncritical range.
speed control mode to maintain the system frequency around Fig. 13-(b) presents the result for WECS providing virtual
50 Hz (for this test, a noncritical operating frequency range of inertial response and primary frequency response (kb = 0.1,
[49.95 Hz, 50.05 Hz] is chosen). A sudden load change (i.e., kc = 0.01). The frequency dip is reduced to 0.47 Hz and the re-
600 W) occurs (which is carried out intentionally) to gener- covery time is reduced to 0.74 s. As discussed in Section III-A,
ate a frequency disturbance. If the WECS does not provide the traditional virtual inertia control slows down the frequency
TAN et al.: ENHANCED FREQUENCY REGULATION USING MULTILEVEL ENERGY STORAGE IN REMOTE AREA POWER SUPPLY SYSTEMS 169

Fig. 14-(a) when the synchronous generator is supplying a frac-

tion of the load. Therefore, an extended noncritical operating
critical frequency of [49.5 Hz, 50.5 Hz] is set to avoid the activa-
tion of primary frequency response from the WECS at the initial
condition (zero LAB current can be observed in Fig. 14-(b)).
The sudden load increase causes a sharp decrease in frequency.
Excessive frequency deviation triggers the virtual inertial
response and the enhanced primary frequency response. At the
end of the transient period, the frequency stabilizes at a value
(with a deviation of 0.65 Hz) outside the noncritical operating
range. The secondary frequency response provided by mechan-
ical power reserve in the wind turbine is activated first. The
wind turbine shifts from SOPPT state to typical MPPT state and
an increase (around 100 W) in turbine power thus WECS power
output can be seen in Fig. 14-(a). Meanwhile, the LAB power
output decreases on the slight increase in frequency, but the
frequency is still lower than the lower limit of the noncritical op-
erating frequency range. Subsequently, the secondary frequency
response from LAB is also activated to restore the frequency.
The LAB power output increases at a ramp rate of 100 W/S. At
the moment when frequency reaches the noncritical operating
range, primary frequency response is deactivated and a minor
turning point can be observed on the WECS power curve. There
is also another nonzero section on the UC current curve in
Fig. 14-(b), which is due to the fact that enhanced primary
frequency response from the UC is active during the frequency
recovery period. It is also important to note that current of high
frequency variation is handled by the UC whereas the LAB
only manages slow changing current, which protects the LAB
from inrush currents and unnecessary charging/discharging
cycles. The release of the secondary frequency response reserve
can be achieved by tertiary frequency control [18].

Fig. 14. Performance of the secondary frequency regulator. (a) Frequency VI. CONCLUSION
deviation and WECS status. (b) UC and LAB currents.
A multilevel energy storage system comprising LAB, UC
and mechanical power reserve in a wind turbine was developed
recovery due to the UC’s power absorption at the frequency re-
and investigated to assess their contribution to RAPS system
covery section, which is highlighted by the negative part of the
frequency regulation. The matching nature between the energy
UC power curve in Fig. 13-(b). The performance of the proposed
density of the three energy storage techniques and the response
enhanced primary frequency response control strategy is pre-
timeframe associated with three types of frequency response
sented in Fig. 13-(c). By comparing Fig. 13-(b) and Fig. 13-(c),
schemes enables the multilevel energy storage as a favorable
it can be seen that the main difference is the frequency recovery
option for frequency regulation. Control strategies were devel-
section. When the frequency reaches the frequency nadir, the
oped to accomplish the frequency regulation capability for each
UC injects power (see the zoomed figure in Fig. 13-(c)) instead
energy storage technique. Moreover, an improved virtual in-
of absorbing power as in the case of traditional virtual iner-
ertial control strategy was proposed by involving UCs to pro-
tia control. Hence, supplementary primary frequency response
vide supplementary primary frequency response. Consequently,
is provided by the UC to assist the frequency recovery. Conse-
frequency regulation performance provided by the WECS can
quently, the frequency recovery time is further reduced to 0.61 s.
be enhanced and the frequency recovery period can be short-
ened. Simulation studies and laboratory tests were carried out
B. Performance of the Secondary Frequency Regulator to validate the effectiveness of the proposed control strategies.
The secondary frequency response strategy discussed in The multilevel energy storage solution can effectively regu-
Section III-C is tested in this scenario. The synchronous gener- late RAPS system frequency. Moreover, the solution eliminates
ator operates in droop control mode, which means the speed set frequent charging/discharging cycles on LAB by utilizing com-
point of the VSD is proportionally reduced with the increase plementary characteristics of LAB and UC. The solution also
in load torque. The scaling factor is chosen to be 0.1. Hence, relieves WECS from abrupt mechanical/electromagnetic stress
an initial frequency deviation of 0.35 Hz can be observed in when the WECS is engaged in frequency regulation. Hence,
170 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 34, NO. 1, JANUARY 2019

these advantages can be beneficial to the extension of service [19] J. Ekanayake and N. Jenkins, “Comparison of the response of doubly fed
life and reduction in lifetime cost for WECS and LAB. Further- and fixed-speed induction generator wind turbines to changes in network
frequency,” IEEE Trans. Energy Convers., vol. 19, no. 4, pp. 800–802,
more, the benefit of the multilevel energy storage solution is not Dec. 2004.
restricted to LAB. The contribution is also applicable to general [20] AEMC Reliability Panel, “Application of frequency operating standards
battery technique for which the number of charging/discharging during periods of supply scarcity, final determination,” Aust. Energy Mar-
ket Commission, Sydney, NSW, Australia, Final determination, 15 Apr.
cycles presents as a problem. 2009.
[21] T. Goya et al., “Frequency control in isolated island by using parallel
operated battery systems applying h control theory based on droop char-
REFERENCES acteristics,” IET Renewable Power Gener., vol. 5, no. 2, pp. 160–166,
2011.
[1] Y. Tan, L. Meegahapola, and K. M. Muttaqi, “A review of technical chal- [22] L.-R. Chang-Chien, C.-C. Sun, and Y.-J. Yeh, “Modeling of wind farm
lenges in planning and operation of remote area power supply systems,” participation in AGC,” IEEE Trans. Power Syst., vol. 29, no. 3, pp. 1204–
Renewable Sustain. Energy Rev., vol. 38, pp. 876–889, 2014. 1211, May 2014.
[2] G. Delille, B. Francois, and G. Malarange, “Dynamic frequency control
support by energy storage to reduce the impact of wind and solar gener-
ation on isolated power system’s inertia,” IEEE Trans. Sustain. Energy,
vol. 3, no. 4, pp. 931–939, Oct. 2012.
[3] Y. Tan, L. Meegahapola, and K. M. Muttaqi, “A suboptimal power-point-
tracking-based primary frequency response strategy for DFIGs in hybrid
remote area power supply systems,” IEEE Trans. Energy Convers., vol. 31, Yingjie Tan (S’12–M’17) received the B.Sc. (Eng.) (Hons.) degree in electrical
no. 1, pp. 93–105, Mar. 2016. engineering and automation from China University of Petroleum (Huadong),
[4] G. Ramtharan, J. B. Ekanayake, and N. Jenkins, “Frequency support from Qingdao, China, in 2012, and the Ph.D. degree in electrical engineering from
doubly fed induction generator wind turbines,” IET Renewable Power the University of Wollongong, Wollongong, NSW, Australia, in 2017. He is
Gener., vol. 1, no. 1, pp. 3–9, 2007. currently a Research Fellow with Electric Power Research Institute, Guangzhou,
[5] K. V. Vidyanandan and N. Senroy, “Primary frequency regulation by China Southern Power Grid Co., Ltd. His research interests include power
deloaded wind turbines using variable droop,” IEEE Trans. Power Syst., system automation, renewable distributed generation, and microgrids.
vol. 28, no. 2, pp. 837–846, May 2013.
[6] D. Zhou, F. Blaabjerg, M. Lau, and M. Tonnes, “Thermal behavior op-
timization in multi-MW wind power converter by reactive power circu-
lation,” IEEE Trans. Ind. Appl., vol. 50, no. 1, pp. 433–440, Jan./Feb.
2014.
[7] J. F. Conroy and R. Watson, “Frequency response capability of full con- Kashem M. Muttaqi (M’01–SM’05) received the Ph.D. degree in electrical
verter wind turbine generators in comparison to conventional generation,” engineering from Multimedia University, Selangor, Malaysia, in 2001. He is
IEEE Trans. Power Syst., vol. 23, no. 2, pp. 649–656, May 2008. currently a Professor with the School of Electrical, Computer, and Telecommu-
[8] I. D. Margaris, S. A. Papathanassiou, N. D. Hatziargyriou, A. D. Hansen, nications Engineering, University of Wollongong, Wollongong, NSW, Australia.
and P. Sorensen, “Frequency control in autonomous power systems with
high wind power penetration,” IEEE Trans. Sustain. Energy, vol. 3, no. 2,
pp. 189–199, Apr. 2012.
[9] X. Luo, J. Wang, M. Dooner, and J. Clarke, “Overview of current develop- Phil Ciufo (SM’07) received the Ph.D. degree in electrical engineering from
ment in electrical energy storage technologies and the application potential the University of Wollongong, Wollongong, NSW, Australia, in 2002. He is
in power system operation,” Appl. Energy, vol. 137, pp. 511–536, 2015. currently an Associate Professor with the University of Wollongong.
[10] Y. Tan, K. M. Muttaqi, P. Ciufo, and L. Meegahapola, “Enhanced fre-
quency response strategy for a PMSG-based wind energy conversion sys-
tem using ultracapacitor in remote area power supply systems,” IEEE
Trans. Ind. Appl., vol. 53, no. 1, pp. 549–558, Jan./Feb. 2017.
[11] E. M. Krieger, J. Cannarella, and C. B. Arnold, “A comparison of
lead-acid and lithium-based battery behavior and capacity fade in off- Lasantha Meegahapola (S’06–M’11) received the Ph.D. degree from Queen’s
grid renewable charging applications,” Energy, vol. 60, pp. 492–500, University of Belfast, Belfast, U.K., in 2010. He is currently a Senior Lecturer
2013. with Royal Melbourne Institute of Technology, Melbourne, VIC, Australia.
[12] Y. Liu, W. Du, L. Xiao, H. Wang, S. Bu, and J. Cao, “Sizing a hybrid energy
storage system for maintaining power balance of an isolated system with
high penetration of wind generation,” IEEE Trans. Power Syst., vol. 31,
no. 4, pp. 3267–3275, Jul. 2016.
[13] J. Cao, W. Du, H. Wang, and M. McCulloch, “Optimal sizing and control
strategies for hybrid storage system as limited by grid frequency devi- Xiaobin Guo is currently the Director of Smart Grid Department, Electric Power
ations,” IEEE Trans. Power Syst., vol. 33, no. 5, pp. 5486–5495, Sep. Research Institute, Southern Power Grid Co., Ltd. (CSG), Guangzhou, China.
2018.
[14] C. Abbey, K. Strunz, and G. Joos, “A knowledge-based approach for
control of two-level energy storage for wind energy systems,” IEEE Trans.
Energy Convers., vol. 24, no. 2, pp. 539–547, Jun. 2009.
[15] L. Wei, G. Joos, and J. Belanger, “Real-time simulation of a wind tur-
bine generator coupled with a battery supercapacitor energy storage sys- Bo Chen is currently a Senior Research Fellow with Electric Power Research
tem,” IEEE Trans. Ind. Electron., vol. 57, no. 4, pp. 1137–1145, Apr. Institute, Southern Power Grid Co., Ltd. (CSG), Guangzhou, China.
2010.
[16] J. Fang, Y. Tang, H. Li, and X. Li, “A battery/ultracapacitor hybrid en-
ergy storage system for implementing the power management of virtual
synchronous generators,” IEEE Trans. Power Electron., vol. 33, no. 4,
pp. 2820–2824, Apr. 2018.
[17] Y. Tan and K. M. Muttaqi, “Multilevel energy storage based frequency Haomin Chen is currently a Research Fellow with Electric Power Research
regulation in remote area power supply systems,” in Proc. IEEE Int. Conf. Institute, Southern Power Grid Co., Ltd. (CSG), Guangzhou, China.
Power Syst. Technol., 2016, pp. 1–6.
[18] OH-P1-Team, “Operation handbook policy 1: Load-frequency control and
performance,” Union Co-Ordination Transmiss. Elect., Brussels, Belgium,
2009.