Adaptive Droop Control of Multi-Terminal HVDC Network For Frequency Regulation and Power Sharing
Adaptive Droop Control of Multi-Terminal HVDC Network For Frequency Regulation and Power Sharing
1, JANUARY 2021
Abstract—This paper proposes a novel adaptive droop control HVDC power transmission grid. The multi-terminal HVDC
strategy for multi-terminal HVDC (MTHVDC) network to provide (MTHVDC) system provides an effective solution as it as-
frequency support and power sharing via a distributed consensus sociates the connection of several AC grids into the HVDC
algorithm. The droop coefficient is adaptively tuned by an effective
integration of optimization algorithms and frequency consensus transmission grid [4]–[7]. In addition, the voltage source con-
method. The optimization control aims to minimize generation cost, verter (VSC) control can be operated in different modes in the
frequency deviation cost, and converter losses. Furthermore, the MTHVDC network to achieve an optimal solution in the entire
coordination between the consensus-based droop controller and AC/DC hybrid system [8], [9].
the optimization technique is established by a feedback loop of
coupling gain solution method, which is based on coupling gain
matrix. The proposed control is successfully applied to a four- A. Related Work and Motivations
terminal MTHVDC system and modified New England 39-bus
system on PSCAD. Power sharing is further improved by adding In a VSC-MTHVDC system, the classical control strategy
an adjustment factor based on frequency deviation with coupling consists of regulating either fixed reference dc link voltage
gain matrix, which redirects power flow in ac and dc sub-grids. In control or dc link voltage droop control. Moreover, the frequency
addition, the loss reduction of voltage source converters is presented droop is facilitated with this classical dc link control, where the
to demonstrate the proposed control performance.
frequency controlling is superimposed on dc-link voltage [10].
Index Terms—Adaptive droop, coupling gain index, Constant power control mode is also one of the approaches that
multiterminal HVDC, power sharing, voltage source converter several research papers have presented [11]–[13]. Nevertheless,
(VSC).
the controlling of dc voltage and frequency with appropriate
power sharing is a major concern and poses a technical chal-
I. INTRODUCTION
lenge. In this regard, the design of the droop coefficient is
ODERN power systems have significantly evolved with
M the demand to interconnect AC and DC subsystems.
With the necessity to balance generation and consumption
extremely important in any of the power sharing methods. As the
MTHVDC network involves several AC grid areas connecting
into an HVDC link, incorporating fixed droop coefficient into
mismatches, the high-voltage direct current (HVDC) power the control might bring fluctuations in the network as it cannot
transmission grids have emerged due to the integration of large coordinate with the diverse operating conditions. A variety
distributed generation systems into the grid [1]–[3]. Thus, the of solutions have been presented in recent research work on
regulation of voltage and frequency in a large hybrid sys- adaptive power sharing [14]–[17]. An adaptive voltage droop
tem along with improved power sharing is a challenge in has been proposed in [14] on multiterminal VSC-HVDC sys-
tems for dc voltage deviation and power sharing. The droop
Manuscript received December 19, 2019; revised March 30, 2020; accepted coefficients are based on power sharing and frequency deviation
May 26, 2020. Date of publication June 2, 2020; date of current version January factors; however, factors from other VSCs for coordination could
6, 2021. This work was supported in part by the ARC Discovery under Grant
DP170103427 and Grant DP180103217, and in part by the ARC Research Hub improve the power flow. Furthermore, a grid reconnection event
for Integrated Energy Storage Solutions under Grant IH180100020. Paper no. on contingency analysis has not been considered. Frequency
TPWRS-01900-2019. (Corresponding author: Mir Nahidul Ambia.) regulation techniques on MTHVDC network have been explored
Mir Nahidul Ambia and Weidong Xiao are with the School of Electrical and
Information Engineering, University of Sydney, Sydney, NSW 2006, Australia in [15]–[18]. Frequency deviation sharing among various ac
(e-mail: [email protected]; [email protected]). systems in MTHVDC system has been facilitated in [16], while
Ke Meng and Zhao Yang Dong are with the School of Electrical Engineering selective routing is proposed in [17].
and Telecommunications, University of New South Wales, Sydney, NSW 2052,
Australia (e-mail: [email protected]; [email protected]). Power sharing through voltage deviation and frequency analy-
Ahmed Al-Durra is with Advanced Power & Energy Center, EECS Depart- sis from other VSCs might improve the dynamic response of the
ment, Khalifa University, Abu Dhabi, UAE (e-mail: [email protected]). control. A virtual synchronous generator control is proposed in
Color versions of one or more of the figures in this article are available online
at https://ieeexplore.ieee.org. [19]. A hierarchical security-constrained optimal power model
Digital Object Identifier 10.1109/TPWRS.2020.2999443 is presented in [20] to minimize the total generation and security
0885-8950 © 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.
Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY DURGAPUR. Downloaded on June 23,2023 at 13:51:40 UTC from IEEE Xplore. Restrictions apply.
AMBIA et al.: ADAPTIVE DROOP CONTROL OF MULTI-TERMINAL HVDC NETWORK 567
Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY DURGAPUR. Downloaded on June 23,2023 at 13:51:40 UTC from IEEE Xplore. Restrictions apply.
568 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 36, NO. 1, JANUARY 2021
algorithm receives information from j, providing the coefficient 4) Analysis of Convergence: The Lebesgue theorem is ap-
aij , which reflects the link between the entities. Based on the plied to identify the convergence. According to the prin-
concept of consensus algorithm, the formulation for frequency ciple, if f (t)is a continuous function, f (s) is bounded in
∞
is presented by: the half-plan(s) ≥ σ, where f (s) = 0 e−(s)t |f (t)|dt ≤
∞ −σt
t 0 e |g(t)dt < ∞|. Hence, the theorem provides f (s) =
f¯i = fi + ψn,i aij fj − f¯i dt (3) sf (s) − f (0), (s) ≥ σ. In this analysis, the assumption is that
∞
0 j∈N
i
f (tk ) is continuous and 0 e−σt |g(t)dt < ∞|. Based on the
bounded condition, the integration by parts of Lebesgue leads
A scenario-based frequency consensus is formulated, where, to (6) based on the formulation from (4).
tk reflects any entity taken from the system. However, an entity
taken from the consensus output is considered at tk+1 . Based on T
the assumption, the formulation in (3) is derived into (4), where f (s) = lim e−st
fj (tk ) is the estimated frequency from jth VSC and fi (tk ) is T →∞ 0
the frequency from ith VSC. The representation of f¯i (tk+1 ) is ⎧ ⎫
⎨ t ⎬
the estimated frequency, which is utilized to keep the system × fi (tk )+ψn,i aij fj (tk ) − f¯i (tk+1 ) dt dt (6)
in the close-loop form in (4). Based on the dynamic consensus ⎩ 0 j∈N ⎭
i
observer [34], [36], the node i receives the estimates of data
from vj ∈ E and compares with its own estimation f¯i (tk+1 )
and local measurement fi . The proposed consensus algorithm The equation in (6) can be further simplified to integrate by
on finding an appropriate adaptive droop coefficient is applicable parts of Lebesgue, which is based on that the coupling gain and
for both neighbor side information and estimation; however, the consensus coefficients are equal to 1. The integration part of (6)
estimation is used to generate a frequency correction through can be derived further to (7).
integration controller to adjust an optimal set point. The de-
signed adaptive Kd droop coefficient is formulated in (5). The T
∞ −st
frequency consensus method on the global system is highlighted e−st 1
f (s) = lim fi (tk ) + e fi (tk )
in Fig. 3. T →∞ −s 0 s 0
t T
∞ −st
f¯i (tk+1 ) = fi (tk ) + ψn,i aij [fj (tk ) − f¯i (tk+1 )]dt e−st t
1 t
+ lim f¯j (tk ) dt + e ¯
fj (tk ) dt
0 j∈N
i
T →∞ −s 0 0 s 0 0
(4) T
∞ −st
e−st t
1 t
t + lim f¯j (tk+1 )dt + e f¯i (tk+1 )dt
Kd = ψn,i aij fj (tk ) − f¯i (tk+1 ) dt (5) T →∞ −s 0 i 0 s 0 0
0 j∈N
i (7)
Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY DURGAPUR. Downloaded on June 23,2023 at 13:51:40 UTC from IEEE Xplore. Restrictions apply.
570 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 36, NO. 1, JANUARY 2021
Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY DURGAPUR. Downloaded on June 23,2023 at 13:51:40 UTC from IEEE Xplore. Restrictions apply.
572 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 36, NO. 1, JANUARY 2021
Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY DURGAPUR. Downloaded on June 23,2023 at 13:51:40 UTC from IEEE Xplore. Restrictions apply.
AMBIA et al.: ADAPTIVE DROOP CONTROL OF MULTI-TERMINAL HVDC NETWORK 573
TABLE I
COMPARISON ANALYSIS BETWEEN PROPOSED CONTROL AND METHOD IN REF [14]
Fig. 9. Pole zero map of control transfer function for SCR 1 to SCR 5 A. Power Sharing Performance of the Adaptive Control
(a) kd = 0.03 (b) kd = 0.07.
Strategy During N-1 Contingency
The case study is designed to show the feasibility of the control
the control system. A similar analysis is demonstrated in Fig. 8(c, strategy during a contingency in the network. Four-terminal
d, e, f) showing the stable nature of the system as none of the MTHVDC network is assumed. PDG1 , PDG2 , PDG3 , and PDG4
Nyquist plots encircle (−1,0) point. However, Fig. 8(e, f) depicts are considered as generations into the network, where PDG4 is
that even though the control system at kd = 0.08 is stable, it is the output power of the distributed generations. Three ac grids
close to the (−1,0) point. For a system like MTHVDC, a safer output power are depicted as PDG1 , PDG2 , and PDG3 . In this
margin is be considered for designing the proposed controller case study, a contingency occurs at 10.12s on DG-3 and the ac
in case any unintentional disturbances occur in the system. grid is disconnected from the MTHVDC network. The frequency
Therefore, for the final selection maximum, kd is chosen as regulation performance and power sharing has been analyzed
0.07. in this section. According to the proposed control strategy, the
output power of the DGs are increased to initiate the power
sharing within the neighboring ac areas.
C. Stability Margin for Selected Droop Design Furthermore, the frequency droop coefficient is changed
The stability margin of the maximum and minimum kd is adaptively to improve the performance of the control. Fig. 10
investigated in this section. Fig. 9 represents the pole-zero map represents the output generation of DGs. The contingency of
of the entire control transfer function (22) with SCR varying PDG3 is demonstrated in Fig. 10c. At 10.12s, PDG4 increases
from 1 to 5. In Fig. 9a and 9b, the pole-zero location movements its generation as per the proposed control. The responses of
are observed for kd at 0.03 and 0.07, respectively. For both the proposed method and control strategy reported in [14] are
values of kd , the poles move from left plane to right plane as shown with red solid lines and blue dashed lines respectively.
grid impedance increases. As the grid impedance increases, the Furthermore, solution with consensus but without optimization
impact on the voltage-current controller increases in terms of has been represented with black dashed lines. According to
Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY DURGAPUR. Downloaded on June 23,2023 at 13:51:40 UTC from IEEE Xplore. Restrictions apply.
574 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 36, NO. 1, JANUARY 2021
Fig. 12. (a) Frequency regulation (b) voltage response by adaptive and tech-
nique in Ref [14].
Fig. 10. Distributed generations of the four AC areas in the MTHVDC network
(a) PDG1 (b) PDG2 (c) PDG3 (d) PDG4 .
Fig. 13. DC-link voltage of the VSCs using adaptive and technique in
Ref [14].
Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY DURGAPUR. Downloaded on June 23,2023 at 13:51:40 UTC from IEEE Xplore. Restrictions apply.
AMBIA et al.: ADAPTIVE DROOP CONTROL OF MULTI-TERMINAL HVDC NETWORK 575
Fig. 14. Distributed generations of the four AC areas in the MTHVDC network
(a) PDG1 (b) PDG2 (c) PDG3 (d) PDG4 .
Fig. 17. Single line diagram of modified New England 39-bus test system with
MTHVDC network.
Fig. 15. Active power through the VSCs in the MTHVDC network. The proposed control strategy is implemented on a modified
New England 39 Bus system with four-terminal MTHVDC
network. The VSCs are connected to the New England 39 bus
system through bus numbers 37 and 32. PDG2 and PDG3 are
the output power by DG2 and DG3 accordingly, which are
AC grids. DG-1 ac grid is connected to DG-3 through the
ac sub-grids. The distributed generation, PDG4 extracts output
power and proposed control is applied on DG-4 side VSC. The
Fig. 16. (a) Frequency regulation & (b) Terminal voltage by adaptive and one-line diagram of the test system is represented in Fig. 17.
conventional control contingency.
Two scenarios are considered in this case study. Scenario-2 takes
place at 6th sec, when DG-3 is disconnected and contingency
occurs. Furthermore, at 10th sec, DG-3 is reconnected considered
grid and distributed generation output power and Fig. 15 depicts
as scenario-3.
active power through VSCs of the four-terminal MTHVDC.
The output power of DGs are shown in Fig. 18. In scenario-2 at
Frequency regulation and the terminal voltage of the system
6th sec during a contingency, the adaptive droop ensures proper
are demonstrated in Fig. 16. The proposed method is indicated
frequency regulation in the system. The coupling gain index
with red solid lines, and the method in [14] is indicated with
ψn,i provides necessary information to the adaptive droop con-
blue dashed lines. It is depicted in Fig. 14 and 15 that the opt f f
proposed controller can generate an optimal solution on extract- troller to adjust the active power. Furthermore, ψn,i Cn,i ΔPn,i
ing the droop coefficient. Furthermore, the distributed genera- and φloss loss
n,i Pn,i cost minimization strategy ensures adaptabil-
tion and grid power is adjusted according to the optimization ity. Based on the updating of current injection reference idref
Si
control, reflecting that the optimal operating point is achieved as per proposed control strategy, improved power sharing
during N-2 occurs.
To test the effectiveness of the proposed controller, different The solid lines represent result by proposed control, and
combinations of N-2 contingencies have been tested. The testing dashed lines represent established technique in Ref [14] in
results on the stable behavior of the control have been summa- Fig. 18, 19, and 20. At 6th sec, oscillation is observed by the
rized in Table II of Appendix. Both four-terminal MTHVDC method in [14]; however, a smooth transition is achieved by
Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY DURGAPUR. Downloaded on June 23,2023 at 13:51:40 UTC from IEEE Xplore. Restrictions apply.
576 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 36, NO. 1, JANUARY 2021
VI. CONCLUSION
A novel adaptive droop control strategy is presented to provide
power sharing and frequency regulation in MTHVDC systems.
The droop coefficient is designed by the coordinated opti-
mization strategy and frequency consensus. The power sharing
Fig. 20. DC-link voltage of the VSCs using adaptive and traditional droop strategy by coupling gain index solution provides adjustment to
control on modified test system.
ensure power flow through AC and DC grids. The frequency
consensus method uses frequency mismatch from AC grid side
VSCs and provides DG side VSC to operate at an optimal point.
adaptive droop as Fig. 18. The power sharing through MTHVDC The optimization minimizes generation, frequency deviation,
network and New England-39 bus system is shown in Fig. 19. In and converter loss cost. The performance of the proposed control
scenario-2, power sharing through MTHVDC to New England strategy has been tested on the modified New England 39 bus
39 bus system through VSC-3 is shown. Moreover, scenario-3 system with 4-terminal MTHVDC system. The results from case
at 10th sec shows that power is injected by DG2 to MTHVDC study on various scenarios show that proposed controller can
network. Interconnection between area 1 and 3 on AC side is reduce oscillation during transitions like contingency and load-
zero in scenario-2 due to necessary power reference changes. ing variations. Furthermore, the proposed control can minimize
The DC-link voltage is shown in Fig. 20. The variation in dc link VSC losses and ensure smoother transitions during contingen-
voltages has been reduced in the proposed method compared to cies. Hence, it can be concluded that the proposed control adds
a technique referred to [14] as per Fig. 20. flexibility in terms of achieving accurate power sharing.
Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY DURGAPUR. Downloaded on June 23,2023 at 13:51:40 UTC from IEEE Xplore. Restrictions apply.
AMBIA et al.: ADAPTIVE DROOP CONTROL OF MULTI-TERMINAL HVDC NETWORK 577
Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY DURGAPUR. Downloaded on June 23,2023 at 13:51:40 UTC from IEEE Xplore. Restrictions apply.
578 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 36, NO. 1, JANUARY 2021
[34] J. Zhou, S. Kim, H. Zhang, Q. Sun, and R. Han, “Consensus-based Weidong Xiao (Senior Member, IEEE) received the
distributed control for accurate reactive, harmonic, and imbalance master’s and Ph.D. degrees in electrical engineering
power sharing in microgrids,” IEEE Trans. Smart Grid, vol. 9, no. 4, from the University of British Columbia, Vancouver,
pp. 2453–2467, Jul. 2018. Canada, in 2003 and 2007, respectively.
[35] K. Meng, Z. Y. Dong, Z. Xu, Y. Zheng, and D. J. Hill, “Coordinated Dr. Xiao is an Associate Professor with the School
dispatch of virtual energy storage systems in smart distribution networks of Electrical and Information Engineering, University
for loading management,” IEEE Trans. Syst., Man, Cybern.: Syst., vol. 49, of Sydney, Australia. From 2010 to 2016, he has been
no. 4, pp. 776–786, Apr. 2019. working with the Masdar Institute of Science and
[36] V. Nasirian, A. Davoudi, F. L. Lewis, and J. M. Guerrero, “Distributed Technology, United Arab Emirates. In 2010, he was
adaptive droop control for DC distribution systems,” IEEE Trans. Energy a Visiting Scholar with the Massachusetts Institute
Convers., vol. 29, no. 4, pp. 944–956, Dec. 2014. of Technology (MIT), Cambridge, USA, where he
[37] [Online]. Available: https://www.aemo.com.au/-/media/Files/Electricity/ worked on the power interfaces for PV power systems. Before the academic
NEM/Security_and_Reliability/Power_System_Ops/Procedures/ career, he worked as a R&D engineering manager with MSR Innovations
SO_OP_3705---Dispatch.pdf Inc., Canada, focusing on integration, research, optimization, and design of
[38] J. Fang, P. Lin, H. Li, Y. Yang, and Y. Tang, “An improved virtual photovoltaic power systems. His research interest includes photovoltaic power
inertia control for three-phase voltage source converters connected to a systems, power electronics, dynamic modeling, control engineering, DC sys-
weak grid,” IEEE Trans. Power Electron., vol. 34, no. 9, pp. 8660–8670, tems, and industrial applications.
Sep. 2019. Dr. Xiao is presently an Associate Editor of the IEEE TRANSACTIONS ON
[39] L. Fan, “Modeling type-4 wind in weak grids,” IEEE Trans. Sustain. INDUSTRIAL ELECTRONICS.
Energy, vol. 10, no. 2, pp. 853–864, Apr. 2019.
[40] L. Huang, H. Xin, H. Yang, Z. Wang, and H. Xie, “Interconnecting very
weak AC systems by multiterminal VSC-HVDC links with a unified virtual
synchronous control,” IEEE J. Emerg. Sel. Topics Power Electron., vol. 6,
no. 3, pp. 1041–1053, Sep. 2018.
[41] Y. R. Rodrigues, M. Abdelaziz, and L. Wang, “D-PMU based secondary Ahmed Al-Durra (Senior Member, IEEE) received
frequency control for islanded microgrids,” IEEE Trans. Smart Grid, the Ph.D. degree in ECE from Ohio State University in
vol. 11, no. 1, pp. 857–872, Jan. 2020. 2010. He is an Associate Professor in the ECE Depart-
ment at Khalifa University, UAE. His research inter-
ests are applications of control and estimation theory
on power systems stability, micro and smart grids,
renewable energy systems and integration, and pro-
Mir Nahidul Ambia (Student Member, IEEE) re-
cess control. He has one US patent, one edited book,
ceived the B.Sc. Eng. degree in electrical engineer-
12 book chapters, and over 190 scientific articles in
ing from The Islamic University of Technology, top-tier journals and refereed international confer-
Bangladesh, in 2008 and the M.Sc. degree in electrical
ence proceedings. He has supervised/co-supervised
and electronic engineering from The Petroleum Insti-
over 25 Ph.D./Master students. He is leading the Energy Systems Control &
tute, United Arab Emirates, in 2012. He is currently
Optimization Lab under the Advanced Power & Energy Center, an Editor for
pursuing the Ph.D. degree in electrical engineering IEEE TRANSACTIONS ON SUSTAINABLE ENERGY AND IEEE POWER ENGINEER-
with the School of Electrical and Information Engi-
ING LETTERS, and Associate Editor for IEEE TRANSACTIONS ON INDUSTRY
neering, The University of Sydney, Australia.
APPLICATIONS AND FRONTIERS IN ENERGY RESEARCH – SMART GRIDS.
He was Previously with Eastern University,
Bangladesh, from 2009 to 2010 as a Lecturer. From
2012 to 2013, he worked in The Khalifa University, UAE, previously known
as The Petroleum Institute as Researcher. From 2013 to 2017, he worked
as an Electrical Engineer with ADNOC Fertilizers, Abu Dhabi, UAE, under
Abu Dhabi National Oil Company (ADNOC). His research interests include Zhao Yang Dong (Fellow, IEEE) received the Ph.D.
power system modeling, multiple microgrid systems, self-healing of network, degree from the University of Sydney, Australia in
renewable energy systems and grid integration. 1999. He is currently SHARP Professor and the Di-
rector of the University of New South Wales Digital
Grid Futures Institute, The University of New South
Wales, Australia. He is also Director for ARC Re-
search Hub for Integrated Energy Storage Solutions.
He was previously Professor and Head of School of
Ke Meng (Senior Member, IEEE) received the Ph.D.
Electrical and Information Engineering, University
degree in electrical engineering from the University of of Sydney, and Ausgrid Chair and Director of the
Queensland, Brisbane, QLD, Australia in 2009. He is
Ausgrid Centre for Intelligent Electricity Networks,
currently a Senior Lecturer in the School of Electrical
the University of Newcastle, Australia. He also held industrial positions with
Engineering and Telecommunications, The Univer-
Transend Networks (now TAS Networks), Australia. His research interests
sity of New South Wales, Sydney, NSW, Australia.
include smart grid, power system planning, power system security, renewable en-
His research interests include electric power system
ergy systems, electricity market, load modelling, and computational intelligence
modelling, stability analysis, renewable energy sys-
and its application in power engineering. He is an editor of IEEE TRANSACTIONS
tems and grid integration. ON SMART GRID, IEEE POWER ENGINEERING LETTERS, IET Renewable Power
Generation and IET Smart City.
Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY DURGAPUR. Downloaded on June 23,2023 at 13:51:40 UTC from IEEE Xplore. Restrictions apply.