250 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 32, NO.

1, JANUARY 2017

Distributed Frequency Control Through MTDC

Transmission Systems
Martin Andreasson, Student Member, IEEE, Roger Wiget, Student Member, IEEE,
Dimos V. Dimarogonas, Member, IEEE, Karl H. Johansson, Fellow, IEEE, and Göran Andersson, Fellow, IEEE

Abstract—In this paper, we propose distributed dynamic con- by HVDC transmission lines, so called multi-terminal HVDC
trollers for sharing both frequency containment and restoration (MTDC) systems [2].
reserves of asynchronous ac systems connected through a multi- The fast operation of the DC converters enables frequency
terminal HVDC (MTDC) grid. The communication structure of
the controller is distributed in the sense that only local and neigh- regulation of one of the AC grids connected to the HVDC link.
boring state information is needed, rather than the complete state. One such example is the frequency regulation of the island of
We derive sufficient stability conditions, which guarantee that the Gotland in Sweden, which is connected to the much stronger
ac frequencies converge to the nominal frequency. Simultaneously, Nordic grid through an HVDC cable [3]. By connecting multi-
a global quadratic power generation cost function is minimized. ple AC grids by an MTDC system, enables frequency regulation
The proposed controller also regulates the voltages of the MTDC of one or more of the AC grids connected. Traditional AC fre-
grid, asymptotically minimizing a quadratic cost function of the
deviations from the nominal dc voltages. The results are valid quency controllers and HVDC voltage controllers do however
for distributed cable models of the HVDC grid (e.g., π-links), as not take advantage which the increased connectivity of the grids
well as ac systems of arbitrary number of synchronous machines, brings. Rather than sharing control reserves, each AC area is
each modeled by the swing equation. We also propose a decentral- responsible for maintaining its own frequency in an acceptable
ized communication-free version of the controller. The proposed range [4], which reduces the need for frequency regulation re-
controllers are tested on a high-order dynamic model of a power serves in the individual AC systems [5], [6]. A challenge is to
system consisting of asynchronous ac grids, modeled as IEEE 14
bus networks, connected through a six-terminal HVDC grid. The bring back the HVDC grid, e.g., the DC voltages, to a normal
performance of the controller is successfully evaluated through operation state after a contingency have happens.
simulation. Stability analysis of combined AC and MTDC systems was
performed in [7]. In [8] and [9], decentralized controllers are
Index Terms—Distributed control, HVDC transmission, power
system dynamics, power system stability. employed to share frequency control reserves. In [9] no sta-
bility analysis is performed, whereas [8] guarantees stability
provided that the connected AC areas have identical parameters
I. INTRODUCTION and the voltage dynamics of the HVDC system are neglected.
[10] considers an optimal decentralized controller for AC grids
OWER transmission over long distances with low losses
P is an important challenge as the distances between gen-
eration and consumption increase. As the share of fluctuating
connected by HVDC systems.
By connecting the AC areas with a communication network
supporting the frequency controllers, the performance can be
renewables rises, so does the need to balance generation and con-
further improved, compared to a decentralized controller struc-
sumption mismatches, often over large geographical areas, for
ture without such communication. In this paper, we seek to
which high-voltage direct current (HVDC) power transmission
explore controllers which improve performance of existing con-
is a commonly used technology. In addition to offering lower
trollers. For this, we first propose a controller performance mea-
cost solutions for longer overhead lines and cable transmission
sure.
[1], the controllability of the HVDC converters offers flexibility
Several distributed and decentralized controllers for sharing
and means to mitigate problems due to power fluctuations from
frequency control reserves have been proposed in the literature.
renewables. Increased use of HVDC technologies for electrical
In [11], a distributed controller, relying on a communication
power transmission suggests that future HVDC transmission
network, was developed to share frequency control reserves of
systems are likely to consist of multiple terminals connected
asynchronous AC transmission systems connected through an
MTDC system. However, the controller requires a slack bus to
Manuscript received July 7, 2015; revised November 17, 2015 and February control the DC voltage, and is thus only able to share the gen-
10, 2016; accepted April 4, 2016. Date of publication April 21, 2016; date of eration reserves of the non-slack AC areas. Another distributed
current version December 20, 2016. This work was supported in part by the
European Commission, the Swedish Research Council (VR), and the Knut and
controller is proposed in [12]. Stability is guaranteed, and the
Alice Wallenberg Foundation. Paper no. TPWRS-00969-2015. need for a slack bus is eliminated. The voltage dynamics of
M. Andreasson, D. V. Dimarogonas, and K. H. Johansson are with the AC- the MTDC system are however neglected. Moreover the imple-
CESS Linnaeus Centre, KTH Royal Institute of Technology, Stockholm 114 28, mentation of the controller requires every controller to access
Sweden (e-mail: [email protected]; [email protected]; [email protected]).
R. Wiget and G. Andersson are with the Power Systems Laboratory, ETH measurements of the DC voltages of all MTDC terminals. In
Zurich, Zurich 8092, Switzerland (e-mail: [email protected]; andersson@ [13], [14] distributed secondary generation controllers are pro-
eeh.ee.ethz.ch). posed, where the MTDC dynamics are explicitly modeled, and
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. the DC voltages are controlled in addition to the frequencies.
Digital Object Identifier 10.1109/TPWRS.2016.2555939 The controller does not rely on a slack bus for controlling the
0885-8950 © 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGYstandards/publications/rights/index.html
See http://www.ieee.org/publications ROORKEE. Downloaded on May 06,2024for
at more
05:49:48 UTC from IEEE Xplore. Restrictions apply.
information.
 ANDREASSON et al.: DISTRIBUTED FREQUENCY CONTROL THROUGH MTDC TRANSMISSION SYSTEMS 251

DC voltages. The distributed control architecture is more scal- be connected. Note that the converter model (1) of the MTDC
able than a centralized architecture where information from all system does not take the dynamics of the HVDC lines into ac-
controllers has to be processed simultaneously. By using lo- count, caused by the inductance and capacitance of the lines. In
cal and neighboring state information, we propose controllers, Section V, however, we show that the model (1) can be gener-
which can be implemented even when communication is un- alized to a π-link model, where each HVDC line consists of an
available. This paper builds on the results in [13]–[15], but arbitrary number of resistors, inductors, and capacitors in series.
significantly generalizes the models of the power system. The Only HVDC nodes which are connected to a converter are con-
proposed controllers are part of a hierarchical control system for sidered in our model (1). This implies that intermediate nodes
a combined AC and DC grid. At the highest level, a centralized are not captured by the model. Modelling intermediate nodes
grid controller will coordinate the overall DC grid operations, would result in differential-algebraic equations, resulting in a far
and the set points of this controller give the nominal DC voltages more complex analysis. While systems with intermediate nodes
and power flows of the proposed controllers. The minimization can be transformed into systems without intermediate nodes by
of the operation costs considering all constraints including, e.g., Kron reduction [18], this is beyond the scope of this paper. Each
the N-1 security criterion, is handled by the grid controller [16]. AC system is assumed to consist of a single generator which is
The remainder of this paper is organized as follows. In Section connected to a DC converter, representing an aggregated model
II-B, the system model and the control objectives are defined. of an AC grid. The dynamics of the AC system are given by
In Section III, a distributed secondary frequency controller for [19]:
sharing frequency control and restoration reserves is presented,
mi ω̇i = Pigen + Pim − Piinj , (2)
and is shown to satisfy the control objectives. In Section IV and
V, the results are generalized to general AC networks and π-link where mi > 0 is its moment of inertia. The constant is the Pigen
models of the HVDC lines, respectively. In Section VI, simu- m inj
generated power, Pi is the power load, and Pi is the power
lations of the distributed controller on a six-terminal MTDC injected to the DC system through converter i, respectively.
test system are provided, showing the effectiveness of the pro- While the model (2) is restricted to single-generator AC systems,
posed controller. The paper ends with concluding remarks in we show in Section IV that this model can be generalized to a
Section VII. network of arbitrary many generators.
The control objective can now be stated as follows.
II. MODEL AND PROBLEM SETUP Objective 1: The frequency deviations are asymptotically
equal to zero, i.e.,
A. Notation
lim ωi (t) − ω ref = 0 i = 1, . . . , n, (3)
Let G be a static, undirected graph. Denote by V and E the t→∞
set of vertices and edges of G, respectively. Let Ni be the set
where ω ref is the nominal frequency. The total quadratic cost
of neighboring vertices to i ∈ V. Denote by LW the weighted
of the power generation is minimized asymptotically, i.e.,
Laplacian matrix of G, with edge-weights given by the elements ∗

of the diagonal matrix W [17]. Let ei denote the ith Cartesian limt→∞ Pigen = Pigen , ∀i = 1, . . . , n, where
unit vector. Let C − denote the open left half complex plane, and 1  P  gen 2
n
∗ ∗
C̄ − its closure. We denote by cn ×m an n × m-matrix, whose [P1gen , . . . , Pngen ] = argmin f Pi (4)
P 1 ,...,P n 2 i=1 i
elements are all equal to c. For simplifying notation, we write
cn for cn ×1 . subject to (3), i.e., Pigen + Pim − Piinj = 0, ∀i = 1, . . . , n and
n inj
i=1 Pi = 0, i.e., power balance both in the AC grids and
B. Model and Objective in the MTDC grid. The positive constants fiP represent the
We will give here a unified model for an MTDC system inter- local cost of generating power. Finally, the DC voltages are
connected with several asynchronous AC systems. We consider such that the a quadratic cost function of the voltage devia-
an MTDC transmission system consisting of n converters, de- tions is minimized asymptotically, i.e., limt→∞ Vi = Vi∗ , ∀i =
noted i = 1, . . . , n, each connected to an AC system, i.e., there 1, . . . , n, where
are no pure DC nodes of the MTDC grid. The converters are as-
1 V
n
sumed to be connected by an MTDC transmission grid, i.e., there [V1∗ , . . . , Vn∗ ] = argmin f (Vi − Viref )2 (5)
exist only one connected MTDC grid and not several MTDC V 1 ,...,V n 2 i=1 i
grids. The node connected to converter i is modeled by
subject to (3)–(4), and where the fiV is a positive constant re-
 1 flecting the local cost of DC voltage deviations and Viref is the
Ci V̇i = − (Vi − Vj ) + Iiinj , (1)
R ij nominal DC voltage of converter i.
j ∈Ni
Remark 1: Note that the order in which the optimization
where Vi is the DC voltage of converter node i, Ci > 0 the to- problems (4)–(5) are solved is crucial, as (3) and the optimal
tal capacitance of the converter and the HVDC line connected solution of (4) are constraints of (5).
to the considered converter, and Iiinj the injected current from Remark 2: The minimization of (4) is equivalent to power
the DC converter to the DC node. The constant Rij denotes sharing, where the generated power of AC area i is asymptot-
the resistance of the HVDC transmission line connecting the ically inverse proportional to the cost fiP . The cost fiP can be
converters i and j. The MTDC transmission grid is assumed to chosen to reflect the available generation capacity of area i.

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY ROORKEE. Downloaded on May 06,2024 at 05:49:48 UTC from IEEE Xplore. Restrictions apply.
 252 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 32, NO. 1, JANUARY 2017

Remark 3: It is in general not possible that limt→∞ Vi (t) = where KiV , Kiω and γ ≥ 0 are positive controller parameters,
Viref ∀i = 1, . . . , n, since this does not allow for the currents and the controller variables cφij satisfy cφij = cφji > 0, i.e., the
between the HVDC converters to change by (1). Note that the communication graph is supposed to be undirected. Further-
optimal solution to (4) fixes the relative DC voltages, leaving more the communication graph is assumed to be connected.
only the ground voltage as a decision variable of (5). Note also The constant Piinj,nom is the nominal injected power. If γ = 0,
that the reference DC voltages Viref , i = 1, . . . , n, are generally the converter controller (8) can be interpreted as an emulation of
not uniform, as is the reference frequency ω ref . an AC network between the isolated AC areas, as it resembles
Remark 4: Note that Objective 1 does not include constraints the swing equation. The controller states φi are then equiva-
of, e.g., generation and line capacities. This requires that the lent to the phase angles of AC area i, whose differences govern
perturbations from the operating point are sufficiently small, to the power transfer between the areas. Larger magnitudes of cφij
guarantee that these constraints are not violated. Incorporating correspond to higher conductances of the AC lines, and thus
these constraints will be considered in future work. stronger coupling and faster synchronization of the frequencies.
If γ > 0, damping is added to the dynamics of φi . Damping gen-
III. DISTRIBUTED FREQUENCY CONTROL erally improves stability margins, and turns out to be very useful
A. Controller Structure in the stability analysis. However, a nonzero γ also implies that
the AC dynamics are not emulated perfectly. This implies that
In this section we propose a distributed secondary frequency exact frequency synchronization might not be possible in gen-
controller. In addition to the generation controller proposed in eral. In contrast to a connection with AC lines, the power is fed
[13], we also propose a secondary controller for the power in- into the MTDC grid and then transfered to the other AC areas
jections into the HVDC grid. We implement the controllers for through the MTDC grid rather than through an AC grid. Also
single AC generators. In Section IV, we generalize the controller the converter controller can be implemented in a decentralized
for AC grids of arbitrary size. version by dropping the states φi , resulting in the following
The distributed generation controller of the AC systems is controller
given by
Piinj = Kiω (ωi − ω ref ) + KiV (Viref − Vi ). (9)
KV
Pigen = −Kidroop (ωi− ω ref ) − iω Kidroop,I ηi
Ki The HVDC converter response is assumed to be instantaneous,
 i.e., injected power on the AC side is immediately and losslessly
η̇i = Kidroop,I (ωi − ω ref ) − cηij (ηi − ηj ), (6) converted to DC power. This assumption is reasonable due to the
j ∈Ni
dynamics of the converter typically being orders of magnitudes
where Kidroop , Kidroop,I , KiV and Kiω are positive controller pa- faster than the primary frequency control dynamics of the AC
rameters. Moreover, the controller variables cηij satisfy cηij = system [4]. The relation between the injected HVDC current
and the injected AC power is thus given by
cηji > 0, i.e., the communication graph is supposed to be undi-
rected. Furthermore the communication graph is assumed to be
Vi Iiinj = Piinj . (10)
connected. The above controller can be interpreted as a dis-
tributed PI-controller, with a distributed averaging filter acting By assuming Vi = V nom i = 1, . . . , n, where V nom is a global
on the integral states ηi . The first line of Equation (6) resem- nominal DC voltage, we obtain
bles a decentralized droop controller with a setpoint given by
ηi . The second line of Equation (6) updates the variable ηi in a V nom Iiinj = Piinj . (11)
distributed fashion by a distributed averaging integral controller.
The magnitudes of the variables cηij determine how fast the gen- Assumption (11) relies on the assumption that the voltages Vi do
erated power levels converge. While a larger magnitude of cηij not deviate significantly from the nominal voltage V nom . Since
could lead to faster convergence of the generated power, it can for most HVDC converters the acceptable deviation from the
also induce oscillations. It is possible to implement a decentral- nominal voltage is less than 5% [20], the approximation (11)
ized version of (6) by dropping the states ηi . This results in the would result in a relative error smaller than 5%. To summarize,
following controller the combined generation and power injection controllers are
given by either Equations (6)–(8), Equations (6)–(9) or Equa-
Pigen = −Kidroop (ωi − ω ref ). (7) tions (7)–(8).
The proposed converter controllers governing the power injec-
tions from the AC systems into the HVDC grid are given by B. Stability Analysis

Piinj = Kiω (ωi − ω ref ) + KiV (Viref − Vi ) We now analyze the stability of the closed-loop sys-
 φ tem. Define the state vectors ω̂ = ω − ω ref 1n and V̂ = V −
+ cij (φi − φj ) V ref , where ω = [ω1 , . . . , ωn ]T , V = [V1 , . . . , Vn ]T , V ref =
j ∈Ni [V1ref , . . . , Vnref ]T , η = [η1 , . . . , ηn ]T , and φ = [φ1 , . . . , φn ].
Combining the MTDC (1), the AC dynamics (2) with the
Kiω
φ̇i = ωi − γφi , (8) generation control (6), the converter controller (8) with the
KiV power-current relationship (11), we obtain the closed-loop

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY ROORKEE. Downloaded on May 06,2024 at 05:49:48 UTC from IEEE Xplore. Restrictions apply.
 ANDREASSON et al.: DISTRIBUTED FREQUENCY CONTROL THROUGH MTDC TRANSMISSION SYSTEMS 253

dynamics Assumption 2 lower bounds for the damping coefficient of the

 converter controllers. Note that the bound on γ is independent of
ω̂˙ = M −(K droop + K ω )ω̂ + K V V̂ the topology of the communication network. This is particularly
 desirable in a plug-and-play setting, where new nodes can be
− K V (K ω )−1 K droop,I η − Lφ φ + P m added to the system, without having to change γ.
  Theorem 1: If Assumptions 1 and 2 hold, the equilibrium of
˙ 1 KV 1 (14) is globally asymptotically stable.
V̂ = nom
EK ω
ω̂ − E L R + nom
V̂ + nom ELφ φ Proof: The proof follows from Theorem 4, and is thus
V V V
omitted. 
η̇ = K droop,I ω̂ − Lη η Corollary 2: Let Assumption 1 hold and let γ, kφ be given
such that Assumption 2 holds. Let K V , K ω and K droop,I
φ̇ = (K V )−1 K ω ω̂ − γφ, (12)
be such that (F P )−1 = K V (K ω )−1 K droop,I and F V = K V ,
where M = diag(m1 −1 , . . . , mn −1 ) is a matrix of inverse gen- where F P = diag(f1P , . . . , fnP ) and F V = diag(f1V , . . . , fnV ).
erator inertia, E = diag(C1−1 , . . . , Cn−1 ) is a matrix of electrical Then the dynamics (14) satisfy Objective 1 in the limit when
elastances, LR is the weighted Laplacian matrix of the MTDC (K ω )−1 K V ∞ → 0, provided that the disturbance Pim is
grid with edge-weights 1/Rij , Lη and Lφ are the weighted constant.
Laplacian matrices of the communication graphs with edge- Proof: By Theorem 1, (14) has a unique and stable equi-
weights cηij and cφij , respectively, and P m = [P1m , . . . , Pnm ]T . librium. Letting φ̇ = 0n −1 implies S T (K V )−1 K ω ω̂ − γφ =
We define the diagonal matrices of the controller gains as 0n −1 . Now (K ω )−1 K V ∞ → 0 implies that S T ω̂ = 0 ⇔ ω̂ =
K ω = diag(K1ω , . . . , Knω ), etc. k1 1n for some k1 ∈ R. Letting η̇ = 0n in (14) yields
Let y = [ω̂ T , V̂ T ]T define the output of (12). Clearly the K droop,I ω̂ − Lη η = 0n .
linear combination 1Tn φ is unobservable and marginally stable
with respect to the dynamics (12), as it lies in the nullspace of By inserting ω̂ = k1 1n and premultiplying the above equation
Lφ . In order to facilitate the stability analysis, we will perform with 1Tn , we obtain that k1 = 0, so ω̂ = 0n so Equation (3) of
a state-transformation to this unobservable mode. Consider the Objective 1 is thus satisfied. Thus η = k2 1n for some k2 ∈ R.
following state-transformation: ˙
Finally, we let V̂ = 0n in (14):
⎡ ⎤  
1  
√ 1Tn K ω ω̂ − V nom LR + K V V̂ + Lφ Sφ = 0n . (15)
 ⎢ n ⎥ 1
φ =⎣ ⎦φ φ = √ 1n S φ (13)
n
S T Inserting ω̂ = 0n and premultiplying (15) with 1Tn yield
  1Tn K V V̂ = 0. (16)
where S is an n × (n − 1) matrix such that √1n 1n S is or-
thonormal. By applying the state-transformation (13) to (12), Inserting ω̂ = 0n and η = k2 1n in (6) yields
we obtain dynamics where it can be shown that the state φ1 is
unobservable with respect to the defined output. Hence, omit- P gen = −k2 K V (K ω )−1 K droop,I 1n , (17)
ting φ1 does not affect the output dynamics. Thus, we define
where P gen = [P1gen , . . . , Pngen ]T . It now remains to show that
φ = [φ2 , . . . , φn ], and obtain the dynamics
the equilibrium of (14) minimizes the cost functions (4) and (5)
 of Objective 1. Consider first (4), with the constraints Pigen +
ω̂˙ = M −(K droop + K ω )ω̂ + K V V̂ n
Pim − Piinj = 0, i = 1, . . . , n and n i=1gen Piinj = 
0. By summing
 the first constraints we obtain i=1 Pi = − ni=1 Pim . The
− K V (K ω )−1 K droop,I η − Lφ Sφ + P m KKT condition of (4) is
 
˙ 1 KV 1 F P P gen = −k3 1n . (18)
V̂ = EK ω̂ − E LR + nom V̂ + nom ELφ Sφ
ω
V nom V V
Since (F P )−1 = K V (K ω )−1 K droop,I , (17) and (18) are
η̇ = K droop,I ω̂ − Lη η identical for k2 = k3 . We conclude that (4) is minimized. Since
φ̇ = S T (K V )−1 K ω ω̂ − γφ . (14) P gen = −K V (K ω )−1 K droop,I η = −k2 K V (K ω )−1 K droop,I 1n
and ω̂ = 0n , premultiplying the first n rows of the equilibrium
We are now ready to show the main stability result of this of (14) with M −1 , and adding to the (n + 1)th to 2nth rows
section. The following assumptions are later used as sufficient premultiplied with V nom E −1 yields
conditions for closed-loop stability.
−V nom LR V̂ − k2 K V (K ω )−1 K droop,I 1n = P m .
Assumption 1: The Laplacian matrix satisfies Lφ = kφ LR .
Assumption 1 can be interpreted as the emulated AC dynam- Premultiplying the above equation with 1Tn yields k2 =
ics of (8) having the same susceptance ratios as the conductance   Kω
− ni=1 Pim ni=1 K V Ki droop . Additionally, LR V̂ is uniquely de-
ratios of the HVDC lines. Assumption 1 can always be satisfied i i

by appropriate choices of the constants cij in (8). termined. Now consider (5). Note that Piinj and hence Iiinj ,
Assumption 2: The gain γ satisfies γ > kφ /(4V nom ). are uniquely determined by (4). By the equilibrium of (1),

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY ROORKEE. Downloaded on May 06,2024 at 05:49:48 UTC from IEEE Xplore. Restrictions apply.
 254 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 32, NO. 1, JANUARY 2017

LR V̂ = I inj , where I inj = [I1inj , . . . , Ininj ]T . Thus, the KKT con- are assumed to be given by the linearized swing equation [19],
dition of (5) is where the voltages are assumed to be constant. As before, we
consider the incremental states with respect to their reference
F V V̂ = LR r, (19) values:
where r ∈ Rn . Since LR V̂ is uniquely determined, we premul- δ̇i k = ω̂i k
tiply (19) with 1Tn and obtain the equivalent condition 
mi k ω̂˙ i k = −(Kidroop
k
+ Kiωk )ω̂i k − ki k j (δi k − δj )
1Tn F V V̂ = 0. (20) j ∈Ni k

Since F V = K V , (16) and (20) are equivalent. Hence (5) is + Pigen

k
+ Pimk − Piinj
k
, (21)
minimized, so Objective 1 is satisfied. 
Remark 5: Corollary 2 provides insight in choosing the con- where δi k is the phase angle and ω̂i k = ωi k − ω ref is the incre-
troller gains of (6) and (8), to satisfy Objective 1. mental frequency at bus ik , mi k > 0 is the inertia of bus ik ,
While the generation controller (6) and the converter con- ki k j = |Vi k ||Vj |bi k j , where Vi is the constant voltage of bus i,
troller (8) offer good performance in terms of satisfying Ob- and bi k j is the susceptance of the power line (ik , j). Moreover
jective 1, it may not be possible to implement these distributed Kidroop
k
= 0 for k = 1, since power injection through the HVDC
controllers, e.g., due to lack of communication infrastructure. converter only takes place at bus i1 . The constant Pigen k
is the
For such MTDC systems where appropriate communication is generated power by the generation control, Pimk is the uncon-
lacking, it may be desirable to instead implement decentralized trolled deviation from the nominal generated power at generator
generation and converter controllers. In other situations it might ik , respectively. The variable Piinj k
= 0 for k = 1 is the power
be possible to implement the distributed generation controller injected to the DC system through converter i. We assume that
(6), while it is more desirable to have the HVDC converters the AC voltages are constant, thus implying that kij is constant.
operating independently with decentralized controllers. In the In order to account for the additional generators, we need to
following corollary, we show that the decentralized generation slightly modify Objective 1.
and converter controllers (7) and (9) also globally asymptoti- Objective 2: The frequency deviations converge to zero, i.e.,
cally stabilize the combined MTDC and AC system.
Corollary 3: Let Assumption 1 hold and let γ, kφ be given lim ωi k (t) − ω ref = 0 k = 1, . . . ni , i = 1, . . . , n. (22)
t→∞
such that Assumption 2 holds. Consider the dynamics of the The total cost of the power generation is minimized asymptoti-
MTDC dynamics (1) and the AC dynamics (2) with the gener- ∗
cally, i.e., limt→∞ Pigen = Pigen , i = 1, . . . , n, where
ation controller (7) or (6), respectively, and the converter con-
1 T P
troller (9). The equilibria of the resulting closed-loop systems n
∗ ∗
are globally asymptotically stable. [P1gen , . . . , Pngen ] = argmin P f Pi (23)
P 1 ,...,P n 2 i=1 i i
Proof: The proof is in line with the proof of Theorem 4,
where we discard the variables η and φ. subject to 1Tn i (Pigen + Pim − Piinj ) = 0, i = 1, . . . , n and
While the optimality results of Corollary 2 do not hold for n inj
any other controller combinations than (6) and (8), the following i=1 Pi 1 = 0, i.e., power balance both in the AC grids and in
remark can be made about the average frequency errors. the MTDC grid. Here Pigen = [Pigen 1
, . . . , Pigen
ni
]T , i = 1, . . . , n.
Lemma 1: Consider the dynamics of the MTDC dynamics Finally, the DC voltages are such that a quadratic cost func-
(1) and the AC dynamics (2) with the generation controller (6) tion of the voltage deviations is minimized asymptotically, i.e.,
and the converter controller (8). Any equilibrium of the resulting limt→∞ Vi = Vi∗ , i = 1, . . . , n, where

closed-loop system satisfies ni=1 Kidroop,I (ωi − ω ref ) = 0, i.e.,
1 V
n
the average frequency errors are zero. [V1∗ , . . . , Vn∗ ] = argmin f (Vi − Viref )2 (24)
Proof: Consider the closed-loop dynamics (12). Letting V 1 ,...,V n 2 i=1 i
η̇ = 0n and premultiplying this equation with 1Tn yields 0n =
 subject to (3)–(4). Here fiP and fiV are positive constants.
1Tn K droop,I ω̂ − 1Tn Lη η = ni=1 Kidroop,I (ωi − ω ref ).
B. Controller Structure
IV. GENERALISATION TO AC GENERATION NETWORK
In this section we generalize the distributed secondary fre-
In this section we generalize the single-generator model of quency controller (6) and the converter controller (8) to the full
Section II-B to an AC grid with arbitrary size. AC network. The distributed generation controllers of the AC
network i are in this case given by
A. Objective
KiV droop,I
Consider the AC transmission grid connected to converter i, Pigen = −Kidroop ω̂i k − K ηi , k = 1, . . . , ni
k k
Kiω1 i k
and suppose it consists of ni generator buses. Without loss of
generality, we may assume that converter i of the MTDC grid 
ni 
is connected to generator i1 of the AC system i. Let δi k be η̇i = Kidroop,I
k
ω̂i k − cηij (ηi − ηj ). (25)
the phase angle of bus ik . The dynamics of the power system k =1 j ∈Ni

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY ROORKEE. Downloaded on May 06,2024 at 05:49:48 UTC from IEEE Xplore. Restrictions apply.
 ANDREASSON et al.: DISTRIBUTED FREQUENCY CONTROL THROUGH MTDC TRANSMISSION SYSTEMS 255

where Kidroop , KiV , Kiω1 and Kidroop,I are positive controller pa- φ = [φ2 , . . . , φn ] we obtain the dynamics
k
rameters, and cηij = cηji > 0. Compare Equation (6). The above
δ̇i = SiT ω̂i , i = 1, . . . , n
controller can be interpreted as a distributed PI-controller, with 
a distributed consensus filter acting on the integral states ηi . The ω̂˙ i = Mi −(Kidroop + Kiω )ω̂i + e1 KiV V̂i − LAC 
i Si δi
converter controller governing the power injections from bus i1
of the AC system i into the HVDC grid is given by KiV droop,I 

− K 1 η
ni i − e eT
1 i Lφ Sφ + P i
m
, i = 1, . . . , n
Kiω1 i
Piinj = Piinj,nom + Kiω1 (ωi 1 − ω ref ) + KiV (Viref − Vi )  
˙ 1 KV 1
 V̂ = nom E K̃ ω ω̃ − E LR + nom V̂ + nom ELφ Sφ
+ cφij (φi − φj ) V V V
j ∈Ni 
n
η̇ = ei 1Tn Kidroop,I ω̂i − Lη η
Kiω1 i=1
φ̇i = (ωi 1 − ω ref ) − γφi , (26)
KiV φ̇ = S T (K V )−1 K̃ ω ω̃ − γIn −1 φ ,

(29)

where γ > 0 and cφij = cφji > 0. Compare Equation (8). In where
vector-form (21) becomes ω̃ = [ω̂1 1 , . . . , ω̂n 1 ]T , Kidroop = diag(Kidroop , . . . , Kidroop ),
1 n i

δ̇i = ω̂i Kidroop,I = diag(Kidroop,I

1
, . . . , Kidroop,I
ni
), K̃ ω = diag(K1ω1 , . . . , Knω1 ).

 Theorem 4: The equilibrium of the dynamics (29) is globally

ω̂˙ i = Mi −(Kidroop + Kiω )ω̂i − LAC
i δi asymptotically stable under Assumptions 1 and 2.
 Corollary 5: Let Assumption 1 hold and let γ, kφ be
+ Pigen + Pim − Piinj , (27) given such that Assumption 2 holds. Let KiV , Kiω and
Kidroop be such that (FiP )−1 = KiV (Kiω1 )−1 Kidroop , i = 1, . . . ,
where δi = [δi 1 , . . . , δi n i ]T , ωi = [ωi 1 , . . . , ωi n i ]T , Mi = n and F V = K V , where F P = diag(f1P , . . . , fnP ) and F V =
diag(m−1 −1
i 1 , . . . , mi n i ), Li
AC
is the Laplacian matrix of the diag(f1V , . . . , fnV ). Then Objective 2 is satisfied in the limit
graph corresponding to the AC transmission system, with edge- when (K ω )−1 K V ∞ → 0, provided that the disturbance Pim
weights given by ki k j , Kidroop = diag(Kidroop , . . . , Kidroop ), is constant.
1 ni
Proof: Consider (29). Letting δi = 0n i i = 1, . . . , n, yields
Kiω = diag(Kiω1 , 0, . . . , 0), and Pigen = [Pigen , . . . , P gen T
] , etc.
1 in i ω̂i = ki 1n i i = 1, . . . , n. Letting φ̇ = 0n yields
Consider the output yi = ω̂i of (27). With respect to yi = ω̂i , the
dynamics have a marginally stable unobservable mode. Thus, S T (K V )−1 K̃ ω ω̃ − γIn −1 = 0n .
similar to Section III-B we consider the state transformation
Now (K ω )−1 K V ∞ → 0 in the above equation implies
⎡ ⎤
  1 T S T ω̃ = 0 ⇔ ω̃ = k1n , k ∈ R. This implies ω̂i = k1n i , i =
1 ⎢ √ 1 n ⎥ 1, . . . , n. Letting η̇ = 0n , inserting ω̂i = k1n i i = 1, . . . , n,
δi = √ 1n Si δi δi = ⎣ in ⎦δi ,
in and premultiplying the equation with 1Tn finally yields k = 0,
SiT and thus ω̂i = 0n i , i = 1, . . . , n, i.e., (22) is satisfied. Letting
η̇ = 0n and inserting ω̂i = 0n i i = 1, . . . , n yields η = k1 1n ,
where Si is an in × (in − 1)-matrix such that [ √1i 1n , Si ] is which inserted in (25) yields
n
orthonormal. It can be shown that δi 1 is unobservable, and can
KiV droop,I
be omitted by introducing the state δi = [δi 2 , . . . , δi n ]T . This Pigen =k K , k = 1, . . . , ni . (30)
i k
Kiω1 i k
state-transformation results in the dynamics
˙
Finally we let V̂ = 0n , insert ω̃ = 0n and premultiply the equa-
δ̇i = SiT ω̂i tion with 1Tn C and obtain

ω̂˙ i = Mi −(Kidroop + Kiω )ω̂i − LAC  1Tn K V V̂ = 0. (31)
i Si δi

 By similar arguments as in the proof of Corollary 2, we can

+ Pigen + Pim − Piinj . (28) show that (30) and (31) are equivalent to the KKT conditions of
(23) and (24), respectively. This concludes the proof. 
Since the input-output dynamics of (27) and (28) are identi-
cal, we henceforth only consider the dynamics (28). By com- V. GENERALIZATION TO π-LINK HVDC MODEL
bining the dynamics (1) and (28) with the controllers (25) In this section we extend the HVDC line model to consider the
and (26), and considering the change of coordinates (13) and inductance and capacitance of the HVDC lines. We model the

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY ROORKEE. Downloaded on May 06,2024 at 05:49:48 UTC from IEEE Xplore. Restrictions apply.
 256 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 32, NO. 1, JANUARY 2017

HVDC lines as series of  π-links consisting of resistors, induc-

tors, and capacitors. The dynamics of HVDC line k, connecting
converters i and j, are given by

Ci V̇i = Iiinj − Ik ,1
k ∈Niin

Lk I˙k ,1 = −Rk Ik ,1 + Vi − Vk ,1
Ckline V̇k ,1 = Ik ,1 − Ik ,2
Lk I˙k ,2 = −Rk Ik ,2 + Vk ,1 − Vk ,2
..
.
Ckline V̇k ,−1 = Ik ,−1 − Ik ,
Lk I˙k , = −Rk Ik , + Vk ,−1 − Vj
˙ 
Cj V̂j = Ijinj + Ik , , (32)
k ∈Niout

where Vk ,q and Ik ,q denote the DC voltage and current of line

segment q, respectively, and Ckline , Rk , and Lk are the capaci- Fig. 1. Test grid consisting of 6 asynchronous IEEE 14 AC grids connected
tance, resistance and inductance of each line segment of line k, through an MTDC grid.
respectively. The sets Niin and Niout denote the incoming and
outgoing HVDC lines to converter i, respectively. To simplify
TABLE I
the derivations we only consider AC areas consisting of single HVDC GRID LINE PARAMETERS
generators, but the results can be generalized to also include
the AC generator network model of Section IV. Combining the (i, j ) R i j [p.u.] L i j [10 −3 p.u.] C i j [p.u.]
DC voltage dynamics (32), the frequency dynamics (2) with
the generation control (6), the converter controller (8) with the (1, 2), (1, 3), (2, 4), (3, 4) 0.0586 0.2560 0.0085
power-current relationship (11), with φ defined as in Section (2, 3)
(2, 5), (4, 5)
0.0878
0.0732
0.3840
0.3200
0.0127
0.0106
III we obtain the closed-loop dynamics (2, 6), (3, 5), (5, 6) 0.1464 0.6400 0.0212

ω̂˙ = M −(K droop + K ω )ω̂ + K V V̂

− K V (K ω )−1 K droop,I η − Lφ Sφ + P m the adjacency relations between the in- and outgoing HVDC
lines and the converters. Element (i, j) of Din or Dout is 1
˙ E if line j originates or terminates at converter i, and 0 other-
V̂ = (−K V V̂ − V nom Din I1 + V nom Dout I wise. The variables Iq = [I1,q − I1nom , . . . , Im ,q − Im
nom T
] and
V nom
Vq = [V1,q − V1,q , . . . , Vm ,q − Vm ,q ] define the incremental
nom nom T
+ K ω ω̂ + Lφ Sφ )
currents and voltages of the q:th line segment for all HVDC
I˙1 = L−1 (−RI1 + Din
T
V̂ − V1 ) lines 1, . . . , m. The constant Iknom denotes the nominal current
of each line segment of line k, and Vknom ,q denotes the nominal
V̇1 = E line (I1 − I2 ) voltage of segment q of line k.
I˙2 = L−1 (−RI2 + V1 − V2 ) Theorem 6. The equilibrium of the dynamics (33) is globally
asymptotically stable under Assumptions 1 and 2.
..
.
VI. SIMULATIONS
V̇−1 = E line (I−1 − I )
In this section, simulations are conducted on a test system
I˙ = L−1 (−RI + V−1 − Dout V̂ ) to validate the performance of the proposed controllers. The
simulation was performed in Matlab, using a dynamic phasor
η̇ = K droop,I ω̂ − Lη η
approach based on [21]. The test system is illustrated in Fig. 1.
φ̇ = S T (K V )−1 K ω ω̂ − γIn −1 φ , (33) We model the HVDC line as a single π-link, with param-
eters given in Table I. The terminal capacitances are Ci =
where M , E K droop , K ω , K V , K droop,I , Lφ , LR , Lη , 0.375 × 10−3 p.u. The AC grid parameters were obtained from
P m , ω̂, V̂ are defined as in Section III, E line = diag [22]. The generators are modeled as a 6th order synchronous
((C1line )−1 , . . . , (Cm
line −1
) ), L = diag(L1 , . . . , Lm ), and R = machine model controlled by an automatic voltage controller
diag(R1 , . . . , Rm ). The n × m-matrices Din and Dout describe and a governor [4]. The consumers in all AC grids are modeled

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY ROORKEE. Downloaded on May 06,2024 at 05:49:48 UTC from IEEE Xplore. Restrictions apply.
 ANDREASSON et al.: DISTRIBUTED FREQUENCY CONTROL THROUGH MTDC TRANSMISSION SYSTEMS 257

TABLE II
CONTROLLER PARAMETERS

K iω K iV K idroop K idroop,I c ηij c φij γ

1501 80 9 3.35 5/R i j 15/R i j 0

Fig. 3. DC converter voltages. After a reduced generation of 0.2 p.u. at t = 1 s,

an immediate voltage drop is followed by a restoration of the average voltage
errors to zero when the distributed generator controller (6) is employed together
with either converter controller (8) or (9). This corresponds to the minimization
of (5). However, when (7) is employed, the average voltage error does not
converge to zero.

Fig. 2. Average AC area frequencies. After a reduced generation of 0.2 p.u.

at t = 1 s, the frequencies synchronize fast, and are subsequently restored to
the nominal frequency when the distributed generator and converter controllers
(6) and (8) are employed. For the controllers (6) and (9) or (7) and (9), a static
control error remains present.

as constant power loads. It is assumed the linear power-current

relation (10) holds.
The three different controllers proposed in this paper are ap-
plied to the test grid, i.e., (7) and (9), (6) and (9), and (6) and (8),
with parameters given in Table II. The communication network
of (6) and (8) is illustrated by the dashed lines in Fig. 1. Fig. 4. Total increase of generated power in the AC areas. After a reduced
We set γ = 0, so Theorem 4 does not guarantee stability generation of 0.2 p.u. at t = 1 s. The generated power is increased for all AC
areas regardless of the controllers employed. When the distributed generator
of the equilibrium. However, the closed-loop system matrix can and converter controllers (6) and (8) are employed, all generators generate to
easily be verified to be Hurwitz. The initial states and voltage set the same power and have the same marginal generation costs asymptotically,
points of the combined grid are calculated by using an security thus minimizing (4) asymptotically. For (6) and (9) or (7) and (9), the marginal
generation costs do not converge to the same value.
constrained optimal power flow [23]. At time t = 1 the output
of generator 2 in AC area 1 was reduced by 0.2 p.u. Fig. 2 shows
the average frequencies of the AC grids for all three controllers.
Immediately after the fault the frequencies of the AC area of the (8), the average voltage deviation is restored to zero as predicted
increased load drop. The frequency drop is followed by a DC by (31).
voltage drop in all converter nodes, due to (8) or (9). Fig. 4 shows the total increase of the generated power within
We note that the frequencies are restored to the nominal fre- each AC area. Initially all generators have a similar oscillating
quency for (6) and (8) as predicted by (22). However, for the behavior, due to the AC voltage oscillations. For the controllers
controller combination (6) and (9) the frequencies are not re- (6) and (8) the total increase in the generated power of the AC
stored to the nominal values, despite that a secondary frequency areas converge to the same value, thus minimizing (23), since
controller is employed. An intuitive explanation to this is that the controller parameters are chosen uniformly for all convert-
the distributed frequency controller (6) requires that the frequen- ers i. For the other two controller combinations, the increase
cies in the different areas synchronize. This synchronization is in the generated power of the AC areas does not converge to
achieved by (8), but not by (9). the same value. This implies that (23) is not minimized, and
Fig. 3 shows the DC voltages of the terminals. A similar we conclude that the simulations confirm that the controllers
behavior as from the frequencies can be seen. As foreseen, the (6) and (8) have superior power sharing properties. Clearly,
voltage deviations can not be controlled back to zero, since this the communication in the controllers (6) and (8) is essential
would result in zero current flows. For the controllers (6) and to eliminate the frequency error and sharing frequency control

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY ROORKEE. Downloaded on May 06,2024 at 05:49:48 UTC from IEEE Xplore. Restrictions apply.
 258 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 32, NO. 1, JANUARY 2017

which is positive definite and radially unbounded:

 
 n
Kiω1  T T AC  −1
W = (δi ) Si Li Si δi + ω̂i Mi ω̂i
T

i=1
2KiV
V nom T 1 1 T
+ V̂ C V̂ + η T η + φ S T Lφ Sφ .
2 2 2
Differentiating W along trajectories of (29) yields

n
Kiω1
Ẇ = V
(δi )T SiT LACi Si Si ω̂i
T

i=1
K i


+ ω̂iT −(Kidroop + Kiω )ω̂i + e1 KiV V̂i − LAC i Si δi

KiV droop,I 

Fig. 5. Converter power set-points for all 6 areas. After a reduced generation − ω Ki 1n i ηi − e1 ei Lφ Sφ
T
of 0.2 p.u. at t = 1 s. For all controllers the area with the disturbance starts Ki 1
to extract power from the MTDC grid and all others inject power. When the
distributed generator and converter controllers (6) and (8) are employed, the
   
converter set-points stabilizing after 6 s to a point with perfect sharing of the + V̂ T K̃ ω ω̃ − V nom LR + K V V̂ + Lφ Sφ
disturbance. For (6) and (9) or (7) and (9), it takes some more time to find a  n 
stable point and the converters differ from each other.  droop,I
+η T T
ei 1n Ki ω̂i − Lη η
i=1
 
 T
reserves. We can note relatively high frequency oscillations in +φ S T Lφ S S T (K V )−1 K̃ ω ω̃ − γIn −1 φ
the transient response of the generated power. These oscillations
originate from the averaging term in the dynamics of the internal Since SS T = In − n1 1n ×n and Si SiT = In i − n1i 1n i ×n i , we
controller variables ηi . By reducing the constants cηij , these os- have Lφ SS T = Lφ and LACi Si Si = Li [24]. Furthermore
T AC
cillations can be reduced at the expense of slower convergence.
Fig. 5 shows the power set-points of the converters. For con- n
Kiω1 T
ω̂ e eT Lφ Sφ = ω̃ T K̃ ω (K V )−1 Lφ Sφ (34)
V i 1 i
troller (6) and (8) the control in area 1 overcompensate the dis- K i
i=1
turbance at the beginning, also the other areas have some minor
oscillations. A stable operation point is found after 6 s. Area 2-6 n
Kiω1 T
ω̂ e1 KiV V̂i = ω̃ T K̃ ω V̂ (35)
have all an equal share of the disturbance. It takes longer time
i=1
KiV i
for the other controllers to find a stable operation point, where
controllers (6) and (9) are significantly faster than (7) and (9). 
n 
n

The sharing of the disturbance are for both controller not equal, η T
ei 1Tn Kidroop,I ω̂i = ηi 1Tn Kidroop,I ω̂i . (36)
i=1 i=1
indicated by the different power set-points of the converters.
By defining
VII. CONCLUSIONS  T  
V̄  = √1 1n
n
S V̄ V̄ = √1 1n
n
S V̄  ,
In this paper we have studied distributed secondary controllers
for sharing frequency control reserves of asynchronous AC sys- we obtain V̄ T LR V̄ = V̄ T S T LR S V̄  and V̄ T Lφ S φ̄ = V̄ T
tems connected through an MTDC system. The proposed con- S T Lφ S φ̄, where V̄  = [V̄2 , . . . , V̄n ]T . We thus obtain
trollers were shown to stabilize the interconnected AC systems  T  ω V −1 ω 1 droop  
and the MTDC grid. The AC grid frequencies were shown to ω̃ K̃ (K ) (K̃ + 2 K̃ ) −K̃ ω ω̃
converge to the nominal frequency. Furthermore, quadratic cost Ẇ ≤ −
V̂ −K̃ ω KV V̂
functions of the voltage deviations of the MTDC terminals and   
of the generated power was minimized asymptotically. The re- Q 1
sults were first derived for single-generator AC grids and purely   
 T kφ  
resistive MTDC grids, and later generalized to ACac grids of V̄ V nom S T LR S − 2 S T LR S V̄ 
arbitrary size, and HVDC links modelled as π-links. Finally, −
φ −
kφ
S T LR S γkφ S T LR S φ
the results were validated on a six-terminal MTDC system with  2
 
connected IEEE 14 bus AC grids. Future work will focus on Q 2
stability results under communication delays.
1  T droop
n
− ω̂ K ω̂i − η T Lη η
2 i=1 i i
APPENDIX
Proof of Theorem 4: Without loss of generality, let Pim = where K̃ droop = diag(K1droop
1
, . . . , Kndroop
1 ). By applying the
0n i , i = 1, . . . , n. Consider the Lyapunov function candidate, Schur complement condition for positive definiteness, we see

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY ROORKEE. Downloaded on May 06,2024 at 05:49:48 UTC from IEEE Xplore. Restrictions apply.
 ANDREASSON et al.: DISTRIBUTED FREQUENCY CONTROL THROUGH MTDC TRANSMISSION SYSTEMS 259

that Q1 is positive definite, since K ω (K V )−1 (K ω + K droop ) − [13] M. Andreasson, R. Wiget, D. V. Dimarogonas, K. H. Johansson, and
K ω (KV )−1 K ω = K ω (K V )−1 K droop > 0. By similar argu- G. Andersson, “Distributed secondary frequency control through multi-
terminal HVDC transmission systems,” arXiv:1503.07358, 2015.
ments Q2 is positive definite iff (γkφ − kφ2 /(4V nom ))S T LR S > [14] M. Andreasson, R. Wiget, D. V. Dimarogonas, K. H. Johansson, and
0. Clearly the above matrix inequality holds under Assump- G. Andersson, “Coordinated frequency control through MTDC transmis-
sion systems,” arXiv:1505.07402, 2015.
tion 2, since S T LR S ≥ 0, and Sx = k1n for k = 0. Thus Ẇ ≤ [15] M. Andreasson, R. Wiget, D. V. Dimarogonas, K. H. Johansson, and
0 under Assumption 2, and the set where W is non-decreasing G. Andersson, “Distributed primary frequency control through multi-
is given by G = {δi ∈ Rn i , i = 1, . . . , n, η = k1n }, for any terminal HVDC transmission systems,” in Proc. Amer. Control Conf.,
pp. 5029–5034, 2015.
k ∈ R. The largest invariant set in G with respect to (29) [16] R. Wiget, “Combined AC and multi-terminal HVDC grids—Optimal
 K iV droop,I
is the origin, since LAC
i Si δi + K iω Ki 1n i ηi = 0n i implies power flow formulations and dynamic control,” Ph.D. dissertation, ETH
Zurich, Zurich, Switzerland, 2015.
δi = 0n i and ηi = 0. By LaSalle’s theorem, the origin of (29) [17] N. Biggs, Algebraic Graph Theory. Cambridge, U.K.: Cambridge Univ.
is globally asymptotically stable. Press, 1993.
Proof sketch of Theorem 6: Without loss of generality, let [18] F. Dörfler and F. Bullo, “Kron reduction of graphs with applications to
electrical networks,” IEEE Trans. Circuits Syst. I, Reg. Papers, vol. 60,
P m = 0n in (33). One can verify that no. 1, pp. 150–163, Jan. 2013.
[19] J. Machowski, J. W. Bialek, and J. R. Bumby, Power System Dynamics:
  Stability and Control. Hoboken, NJ, USA: Wiley, 2008.
V nom l 
−1
[20] D. Jovcic and K. Ahmed, High Voltage Direct Current Transmission:
T T −1 T line
W = V̂ C V̂ + Ii L Ii + Vi C Vi Converters, Systems and DC Grids. Hoboken, NJ, USA: Wiley, 2015.
2 i=1 i=1 [21] T. Demiray, “Simulation of power system dynamics using dynamic phasor
models,” Ph.D. dissertation, ETH Zurich, Zurich, Switzerland, 2008.
(37) [22] F. Milano, Power System Modelling and Scripting. New York, NY, USA:
1 1 1 Springer, 2010.
+ ω̄ T K ω (K V )−1 M −1 ω̄ + η̄ T η̄ + φ̄T S T Lφ S φ̄ [23] R. Wiget, E. Iggland, and G. Andersson, “Security constrained optimal
2 2 2 power flow for HVAC and HVDC Grids,” in Proc. 18th Power Syst.
(38) Comput. Conf., 2014, pp. 1–7.
[24] M. Andreasson, D. Dimarogonas, H. Sandberg, and K. Johansson,
“Distributed control of networked dynamical systems: Static feedback,
is a Lyapunov function for (33), where C line = diag integral action and consensus,” IEEE Trans. Autom. Control, vol. 59,
no. 7, pp. 1750–1764, Jul. 2014.
(C1line , . . . , Cm
line
). Differentiating W along trajectories of (33)
m
where P = 0n yields By LaSalle’s theorem, the equilibrium
of (33) is globally asymptotically stable.
Martin Andreasson (S’12) received the M.Sc. degree in engineering physics
from the KTH Royal Institute of Technology, Stockholm, Sweden, in 2011
REFERENCES where he is currently pursuing the Ph.D. degree. His research interests include
distributed control of multiagent systems, and control of power systems.
[1] Z. Melhem, Electricity Transmission, Distribution and Storage Systems
(Woodhead Publishing Series in Energy). Amsterdam, The Netherlands:
Elsevier, 2013.
[2] T. M. Haileselassie and K. Uhlen, “Power system security in a meshed
north sea HVDC grid,” Proc. IEEE, vol. 101, no. 4, pp. 978–990, Apr.
2013. Roger Wiget (S’11) was born in Lucerne, Switzerland. He received a Bachelor’s
[3] J. Arrillaga, Y. H. Liu, and N. R. Watson, Flexible Power Transmission: degree in electrical engineering from the ETH Zurich, Switzerland, in 2009 and
The HVDC Options. Hoboken, NJ, USA: Wiley, 2007. a Master’s degree in Energy Science and Technology from the same institution
[4] P. Kundur, Power System Stability and Control (The EPRI Power System in 2011. He joined the Power Systems Laboratory of ETH Zurich, Switzerland,
Engineering Series). New York, NY, USA: McGraw-Hill, 1994. in 2011 where he completed his Ph.D. in 2016. His research is dedicated to
[5] R. Li, S. Bozhko, and G. Asher, “Frequency control design for offshore HVDC networks.
wind farm grid with LCC-HVDC link connection,” IEEE Trans. Power
Electron., vol. 23, no. 3, pp. 1085–1092, May 2008.
[6] T. M. Haileselassie and K. Uhlen, “Primary frequency control of remote
grids connected by multi-terminal HVDC,” in Proc. IEEE Power & Energy
Soc. General Meeting, Jul. 2010, pp. 1–6.
Dimos V. Dimarogonas (M’10) was born in Athens, Greece, in 1978. He re-
[7] N. R. Chaudhuri, R. Majumder, B. Chaudhuri, and J. Pan, “Stability
ceived the Diploma in electrical and computer engineering and the Ph.D. in me-
analysis of VSC MTDC grids connected to multimachine AC systems,”
chanical engineering from the National Technical University of Athens (NTUA),
IEEE Trans. Power Del., vol. 26, no. 4, pp. 2774–2784, Oct. 2011.
Greece, in 2001 and 2006, respectively. Between May 2007 and February 2009,
[8] J. Dai, Y. Phulpin, A. Sarlette, and D. Ernst, “Voltage control in an HVDC
he was a Postdoctoral Researcher at the Automatic Control Laboratory, School
system to share primary frequency reserves between non-synchronous of Electrical Engineering, ACCESS Linnaeus Center, Royal Institute of Tech-
areas,” in Proc. 17th Power Syst. Comput. Conf., 2011.
nology (KTH), Stockholm, Sweden. Between February 2009 and March 2010,
[9] B. Silva, C. L. Moreira, L. Seca, Y. Phulpin, and J. A. Peas Lopes, “Pro-
he was a Postdoctoral Associate at the Laboratory for Information and Decision
vision of inertial and primary frequency control services using offshore
Systems (LIDS) at the Massachusetts Institute of Technology (MIT), Boston.
multiterminal HVDC networks,” IEEE Trans. Sustain. Energy, vol. 3, He is currently an Assistant Professor at the Automatic Control Laboratory,
no. 4, pp. 800–808, Oct. 2012.
ACCESS Linnaeus Center, Royal Institute of Technology (KTH), Stockholm.
[10] J. A. Taylor and L. Scardovi, “Decentralized control of DC-segmented
His current research interests include multiagent systems, hybrid systems and
power systems,” in Proc. 52nd Annu. Allerton Conf. Commun., Control,
control, robot navigation and networked control. He received a Docent in Au-
Comput., 2014, pp. 1046–1050. tomatic Control from KTH in 2012.
[11] J. Dai, Y. Phulpin, A. Sarlette, and D. Ernst, “Impact of delays on a
Dr. Dimarogonas serves in the Editorial Board of Automatica, the IEEE
consensus-based primary frequency control scheme for AC systems con-
TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING and the IET Con-
nected by a multi-terminal HVDC grid,” in Proc. IEEE Bulk Power Syst.
trol Theory and Applications and is a member of the Technical Chamber of
Dyn. Control, 2010, pp. 1–9. Greece.
[12] J. Dai and G. Damm, “An improved control law using HVDC systems for
frequency control,” in Proc. Power Syst. Comput. Conf., 2011.

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY ROORKEE. Downloaded on May 06,2024 at 05:49:48 UTC from IEEE Xplore. Restrictions apply.
 260 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 32, NO. 1, JANUARY 2017

Karl H. Johansson (F’13) received MSc and PhD degrees in Electrical Engi- Göran Andersson (M’86–SM’91–F’97) obtained his M.S. (1975) and Ph.D.
neering from Lund University. He is Director of the ACCESS Linnaeus Centre (1980) degrees from the University of Lund, Sweden. In 1980 he joined ASEA’s,
and Professor at the School of Electrical Engineering, KTH Royal Institute of now ABB’s, HVDC division in Ludvika, Sweden, and in 1986 he was appointed
Technology, Sweden. He heads the Stockholm Strategic Research Area ICT full professor in electric power systems at KTH (Royal Institute of Technology),
The Next Generation. He has held visiting positions at UC Berkeley, Califor- Stockholm, Sweden. Since 2000 he is full professor in electric power systems
nia Institute of Technology, Nanyang Technological University, and Institute at ETH Zurich (Swiss Federal Institute of Technology), where he also heads the
of Advanced Studies Hong Kong University of Science and Technology. His power system laboratory. His research interests include power systems dynamics
research interests are in networked control systems, cyber-physical systems, and and control, power markets, and future energy systems.
applications in transportation, energy, and automation systems. Dr. Andersson is a fellow of the Royal Swedish Academy of Sciences, and
Dr. Johansson is a member of the IEEE Control Systems Society Board of of the Royal Swedish Academy of Engineering Sciences. He was the recipient
Governors and the European Control Association Council. He is past Chair of the of the IEEE PES Outstanding Power Educator Award 2007 and of the George
IFAC Technical Committee on Networked Systems. He has been on the Edito- Montefiore International Award 2010.
rial Boards of Automatica, IEEE TRANSACTIONS ON AUTOMATIC CONTROL, and
IET Control Theory and Applications. He is currently a Senior Editor of IEEE
TRANSACTIONS ON CONTROL OF NETWORK SYSTEMS and Associate Editor of
European Journal of Control. He has been Guest Editor for two special issues
of IEEE TRANSACTIONS ON AUTOMATIC CONTROL and one of IEEE CONTROL
SYSTEMS MAGAZINE. He was the General Chair of the ACM/IEEE CyberPhys-
ical Systems Week 2010 in Stockholm and IPC Chair of many conferences. He
received the Best Paper Award of the IEEE International Conference on Mobile
Adhoc and Sensor Systems in 2009 and the Best Theory Paper Award of the
World Congress on Intelligent Control and Automation in 2014. In 2009 he was
awarded Wallenberg Scholar, as one of the first ten scholars from all sciences,
by the Knut and Alice Wallenberg Foundation. He has held a fiveyear Senior
Researcher Position with the Swedish Research Council. He was awarded Fu-
ture Research Leader from the Swedish Foundation for Strategic Research in
2005. He received the triennial Young Author Prize from IFAC in 1996 and the
Peccei Award from the International Institute of System Analysis, Austria, in
1993. He received Young Researcher Awards from Scania in 1996 and from
Ericsson in 1998 and 1999.

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY ROORKEE. Downloaded on May 06,2024 at 05:49:48 UTC from IEEE Xplore. Restrictions apply.