Flatness-based feed-forward control of an HVDe
power transmission network
c. Schmuck
Max-Planck-Institute for Dynamics of Complex
Technical Systems,
39106 Magdeburg, Germany
e-mail: schmuck@mpi-magdeburg.mpg.de
A. Gensior
Elektrotechnisches Institut,
Technische Universitat Dresden
01062 Dresden, Germany
e-mail: albrecht.gensior@tu-dresden.de
Abstract-An efficient and well-established technology for power
transmission across long distances is high voltage direct current
transmission (HVDC). However, HVDe is up to now almost com
pletely limited to peer-to-peer connections or networks with peers
situated closely to each other. This contribution introduces the
flatness-based design of a feedforward control of tree-like, i.e.
cycle-free, HVDe transmission networks comprising two or more
converter stations. The resulting control concept allows a flexible
determination of the power distribution within the network. Fur
thermore, effects like power losses and delays due to wave prop
agation, which are related especially to long transmission lines,
can be easily taken into account. Numerical simulations for an
example network are included to prove the value of the results.
Power grid; Multi-Terminal HVDe; Travelling waves on
transmission lines; Flatness-based control;
I. INTRODUCTION
Electric power transmission by means of alternating current
(AC) is not feasible for transmission distances larger than 1000
km due to high reactive currents and undesired wave reflections.
High voltage direct current transmission (HVDC) is an effi
cient alternative to overcome these limitations [1,2]. The well
established standard configuration of an HVDC system is a
peer-to-peer link connecting two conventional AC networks as
depicted in Fig. 1. The AC network and the DC link are coupled
by a converter terminal equipped with a power converter [3] ,
which works as inverter or rectifier depending on the direction
of the power flow.
DC transmission line
converter tenninal converter terminal
Fig. 1. Peer-to-peer HVDC link with two converter terminals and a DC
transmission line connecting them
Although, up to now the vast majority of all implemented
HVDC systems are in standard peer-to-peer configuration,there
has been increasing interest in HVDC networks with more than
two converter terminals, the so called Multi-terminal HVDC
[1,4-6]. As a result of the evolving technology for power con
verters and the increasing exploitation of renewable energy
978-1-4577-1250-0/11/$26.00 2011 IEEE
F. Woittennek
Institut flir Regelungs- und Steuerungstheorie,
Technische Universitat Dresden,
01062 Dresden, Germany
e-mail: frank.woittennek@tu-dresden.de
J. Rudolph
Lehrstuhl flir Systemtheorie und Regelungstechnik,
Universitat des Saarlandes,
66041 Saarbriicken, Germany
e-mail: j.rudolph@lsr.uni-saarland.de
resources such networks have been put into practice, e.g. for
offshore wind farms [7-9]. A central goal for the control of
an HVDC Multi-Terminal network is to keep the power bal
ance between the electrical power fed into and taken from
the DC network by the connected converter stations. At the
same time one desires to adjust the power distribution between
the converter terminals flexibly during the operation of the
system. Furthermore, time delays due to travelling waves can
become considerable for long transmission distances [6] and
should then be taken into account. This work proposes a control
method that reaches these goals taking a flatness-based ap
proach. For the discussed transmission system, the description
of which involves partial differential equations (PDEs), this
means that the solution of the system equations is parametrized
by the trajectories of a special set of system variables, called a
flat output of the system [10-12]. The number of the compo
nents of the flat output equals the number of the control inputs.
The remaining part of the paper is structured as follows.
Section II describes the mathematical model of the HVDC
transmission network. Thereafter, a flat output of this model
is derived and the flatness-based control design is explained
in section III. The results are illustrated by the numerical
example of section IV. Finally, section V gives some remarks
on practical issues and on potential extensions to be considered
in future work.
II. MODEL OF THE HVDC NETWORK
This section introduces the mathematical model of the
HVDC network,which the control design is based on.
A. General network structure
A general transmission network is assumed to consist of np
uniquely numbered nodes P/1' J1 E 'P where 'P is the set of
all node indices existing in the network. Two arbitrary nodes
P/1 and Pv can be connected by an electric transmission line
denoted by L where the notations Land If., coincide, see
Fig. 2. Then L is the set of the index pairs of all nL
existing
lines. Regarding the transmission line L z
the notation is used
for the spatial coordinate at node P/1 and z:; is used at node Pv
respectively.
 Every node J1 E tp can be connected to a converter ter
P/1' C. Converter terminals
minal which is called in this case. The set tpa C;;; tp comprises
C/1
the indices of all network nodes equipped with a converter ter For each converter terminal VJ1 E tpa the voltage is
C/1'
minal, called active nodes, whereas tpp tp \ tpa comprises the = U/1 and the current is Iw
Thanks to modern semiconductor
indices of all nodes without converter terminal, called passive technology recent power converters allow to generate almost
nodes. Every node which is connected to only one line is called arbitrary current or voltage trajectories at their DC side [3].
terminating node. Obviously, a passive terminating node would Hence, each converter can be seen either as an ideal current
be useless in practice, which is why all terminating nodes are source with a freely adjustable current or as an ideal I/1(t)
assumed to be active. voltage source with a freely adjustable voltage The AC U/1(t).
/1, v E 'P part of the converter is therefore neglected. The converters are
P" (J1, v) E.[ P"
the actuators of the transmission system. If is chosen to C/1
0 0
be a current source then 1/1
is a control input of the network.

Z = -p
,.11 Z = z:.
Otherwise, if C/1
is chosen to be a voltage source then is U/1
L"" a control input. The control design is independent from this
choice which thus can be made according to technical aspects.
Fig. 2. Notation for nodes and transmission lines in a general network

D. Transmission line equations and boundary conditions for

A node P/1
is called neighbor of node Pv
if there exists a line tree-like networks
L , (;1, v) E L connecting them. The indices of all neigbors mv
of Pv form the set Nv. The model of the transmission lines should allow to take
effects like wave propagation, related delays and transmission
B. Tree-like networks losses into account. Therefore, the voltage profile and the u/1
current profile on each transmission line L ' J1 E tp \ {a} are
i/1 /1
The considerations in this paper are restricted to the special described by the hyperbolic system of PDEs
case of tree-like, i.e. cycle-free, networks sketched in Fig. 3. aU/1 ai/1
The abscence of transmission line cycles implies that the path a; (z,t) + Lfjf(z,t) + Ri/1(z, t) = 0 (Ia)

ai/1
C aU/1
az (z,t) + at (z,t) + GU/1(z,t)
0 (1b) =

with z E (0,/1)' t > 0 and the constant positive, line-specific

parameters R, G, Land C [13]. The current i/1 is considered
positive in the direction assigned to line L/1 which is from Pp{J1)
to Pw

The electrical interconnection of the transmission lines at

the network nodes leads to a coupling of the corresponding line
PDEs at their boundaries, see Fig. 5 and Fig. 6. The boundary
'
Fig. 3. Network in tree-like structure with an arbitrarily chosen initial node P (f,
conditions at each node P/1
result from Kirchhoff s current and
voltage laws. The current law yields
{a(t)'
one of the remaining nodes PI' and the path between them
if a E tpa
between any two converter terminals through the network is - I ik(O,t) =

' if a E tpp
(2a)
unique. Because of this property every node in a circuit-free kENa
network has a unique predecessor with respect to a certain for the boundary values of the line currents at the initial node
initial node. Chosing an arbitrary node a, a E tp as this initial
P Paand
node, the predecessor p{J1) of
the node which precedes
J1 E tp \ {a} is definded as
P
P/1
P/1'
on the unique path from a to P . .
li1" t) - L....J {
lk(O,t) - I/1(t),
_ VJ1 E tpa \ {a}
(2b)
Pw Note that the predecessor index p(;1) belonging to node P/1 0, VJ1 E tpp \ {a}
kEN"
depends on the choice of the initial node a although this is not P k*p{J1)
represented by the notation for the sake of simplicity. for the remaining nodes. The converter current I/1(t) is defined
In the case of tree-like networks the notation for transmis to be positive if it is directed away from node Pw Kirchhoff's
sion lines and their node coordinates may be simplified to voltage law leads to
/1 - LP/1{J1)
/1 ''-- Lp{J1) - 0, P{J1) uk(O,t) u/1(t), V k E N/1 \ {p(;1) }, VJ1 E tp (3a)
L - ' .J1
;(;p{J1) - .
;(;/1 - /1' =

u/1 (/1' t) u/1(t),

= VJ1 E tp \ {a} (3b)
VJ1 E tp \ {a}, where /1 isthe length of LJ1" This notation
illustrated in Fig. 4 will be used throughout the remaining parts
for the boundary values of the line voltages at with P/1 u/1(t)
of this work.
denoting the node voltage at Pw
Accordingly, the voltages at
the converters connected to the active nodes of the network are
/1E'P (3c)
Pp{J,) P"
0 0
Altogether, the 2nL PDEs (1) and the boundary conditions
z = !p(p) = 0 Lp z = ') = (p (2) and (3) constitute a linear distributed parameter model of the
transmission network. The currents 1/1
or the voltages J1 E U/1'
Fig. 4. Simplified notation for nodes and transmission lines in tree-like networks tpa of the converter terminals are the concentrated control inputs
located at the boundaries of the transmission lines.
 III. FLATNESS-BASED CONTROL DESIGN

",' Lk
/

This section deals with the design of a flatness-based feed

forward control for the transmission network model described . ik(O, I) (T(t)i,,(ep' I),
=

in section II. It is shown that the trajectories of all system \ {pC!l))
variables can be calculated from prescribed trajectories of the
current Iaand the voltage Vaof the converter at an arbitrarily
chosen initial node Paand some current allocation parameters
=
lI,,(O,I)
iip(p)(1)
jO VO,I)
: Vk E HI'

(CAP), which are introduced additionally at each network node. Pp(p)

Therefore, the mentioned variables form a fiat output of the
system. Once the trajectories for the flat output are chosen the Fig. 6. Currents and voltages at one of the remaining nodes P/1' Jl E 'Pa \ {a:}
remaining system trajectories can be conveniently calculated and the connected line L/1
from node to node. Finally, this yields the desired control input
trajectories for the converter currents or voltages.
Til

A. Derivation of a flat output + I g(l/l' f) U/l(O,t - f) dl

Initial node The first step is to choose an arbitrary node
with converter terminal as the initial node Pa, a E 'Pa. This
determines the simplified notation for the network, which is
Til
clarified by Fig. 5 and Fig. 6. In the following the system
variables shall be calculated from some prescribed current IaCt)
and voltage Va(t) at Ca. Because of (3a) and (3c), the converter
- I hu(l/l' f) i/l(O,t - f) dl, (7a)
-T
"
voltage Vadirectly gives the line voltages
uk(O,t) == Va(t), Vk E Na (4)
of the lines connected at Pa' In order to determine the currents Til
at Paone introduces ma real, time-varying current allocation
parameters (CAP) , Vk
E Na, such that
+ I g(lll' f) i/l(O,t - f) dl
ik(O,t) -(F(t)IaCt), Vk E Na,
== (5)
where
.L: (t) == 1 (6) Til

has to be guaranteed to avoid the violation of the current law

- I hi(lll' f) U/l(O,t - f) dl (7b)
(2a). This means that the trajectories for (ma 1) of the ma -
new parameters can be chosen freely to determine the desired
fraction (5) of IaCt)
for each line Lk, Vk
E Na at node Pa' with T/l YLEl/l and the functions
==

-
The vector comprising these (ma 1) chosen parameters as
components is denoted by (Fain the following. The equations hu(z,t) Rf(z,t) + L afat (z,t),
==

-
(4) (6) give a complete parametrization of the line voltages
uk(O,t) and currents ik(O,t),Vk
E Na at node Pain terms of hi(z,t) G f ez,t) + C af (z,t),
==

Va,Iaand (Fa' at -

g (z,t) Of (z,t), f ez,t) -:10(f3 ../LCZ2 - t2)

uZ 2 vLC
== ==

employing the bessel function

constants
10 of the first kind and the

Y ==
2 (!!.L + )
C '
see [14]. Equations (7) reflect the wave propagation process
taking place on line L/l'
since they involve distributed delays
and predictions. This means the values and at u/l/l' t) i/l(l/l' t)
t
a certain time instant are determined by the trajectories of
Fig. 5. Currents and voltages at the initial node PrI, a: E 'Pa u/O, f) and Va, f) on the complete time intervall lE + [t - T/l' t
T/l]'The delay T/l
can be interpreted as the time that a voltage
LJl'
and current wave needs to travel between the ends of
Remaining nodes Fig. 6 shows one of the remaining network
nodes P/l' J1 E 'P \ {a}, its unique predecessor PpC!l) and the Analogously to the procedure at the initial node the distri
line L/l connecting them. The solution of the line equations bution of the currents between the lines and a possibly con
(1) allows for the direct calculation of the voltage and current nected converter at P/l shall be determined by (m/l - 1) CAPs
trajectories of line L/l at P/l from some known voltage and cr;" Vk E N/l \ (P(J1) } and an additional CAP i5'/l if P/l is active.
current trajectories at the preceeding end at PpC!l) by This means
 ik(O,t)= (t) i,.8fJ, t), Vk E NfJ \ (p(jJ)} (8a) transition between two states of rest which is relevant for the ap
plication of this work, too. Most of the time the HYDC system
ffJ(t)= O"fJ(t) ifJ(efJ' t), if/1 E 'Fa. (8b)
will be operated in a balanced state of rest with constant voltage
'
Again Kirchoff s current law (2b) requires
{
and current values which meet all operational requirements. If
'" cr': 1' if/1 E 'Fp these requirements change the transition to a suitable new state
L.J it)= 1- O"fJ(t), if/1 E 'Fa.
(9) of rest will be desired.
k
k*EN"
p(JJ) In a state of rest every system variable remains constant
Due to (9) one can freely chose the trajectories for only - 2) (mfJ over time by definition. Since for flat systems all trajectories
out of - 1) CAPs if /1 E 'Fp or
(mfJ - 1) out of (mfJ
CAPs mfJ are parametrized by the flat output and its derivatives each state
if /1 E 'Fa. The vector comprising these chosen parameters as of rest of the system is completely characterized by constant
components is denoted by O"w
According to (3) the voltages at values Yk for the ny
components Yb k =
1,2, . . . of the flat ,ny
PfJ are output:
uk(O,t)= ufJ(t)= ufJ(efJ' t), Vk E NfJ \ (p(jJ)} (10a)
j= 1,2,. . . (12)
UfJ(t)= UfJ(efJ' t), if/1 E 'Fa. (lOb)
The equations (7)--(10) give a parametrization of all voltages In order to implement the transition from an initial state of
and currents at
t)
PfJ
in terms of the voltage
VO, at the preceeding node Pp(JJ)
and the CAPs at
ufJ(O,t)
and the current
O"fJ PfJ"
rest with Yk(t) = Y to a new final state of rest with Yk(t) = y{
These equations hold for all remaining nodes /1 E 'F \ { a} . PfJ,
one can chose polynomials Pk(t)
to connect the constant parts

Therefore, it is now possible to calculate all voltage and current

trajectories at each node in the network from the trajectories of
Ua and fa
at the initial node and the freely determined CAP
of the trajectories, such that
{
Y-ki '
Yk(t)= (t),
if t < ti
if ti:os; t:os; tf (13)
trajectories V/1 E 'F at the network nodes. Hence, these
O"fJ' YJk ' if t > tf
variables form the flat output
with
y = (Ua,la'(O"fJ)VfJEP) (11) (l4a)
of the transmission network. k= 1,2, . . . ,ny. (14b)
Clearly, it is convenient to perform the calculations of the The obtained trajectory for one component Yk is depicted in
system trajectories stepwise from node to node beginning at Fig. 7. The desired transition time !-"t = tf - ti can be chosen
the initial node Pa.
At first (4) and (5) are used to determine the freely.
voltage and current trajectories at node Pa
from the determined
trajectories of Ua, fa
and the chosen CAPs 0" After that a' The conditions (l4b) assure continuous differentiability
the neighbors of Pa
are considered. Their predecessor is Pa. with respect to time for the trajectories of the flat output. This
Employing (7)-(10) with p(jJ) =
a together with the CAP smoothness property is maintained during the computations in
trajectories for each node /1 E PfJ, Na
gives all voltage and section III.A. because derivations with respect to time do not
current trajectories at these nodes. In the next step, these nodes occure. Thus, continuous differentiability is obtained for all
PfJ serve as predecessors for all their neighbors (except Pa) system trajectories. The four requirements of (14) can be met
and the equations (7)-(10) can be applied again. One follows with the third order polynomials
t - ti
this procedure until finally all terminating nodes are reached. A
Pk(t)= yi + (y{ - yi) (3 - 2l) fl, t= tf - t/
_

particular result of these calculations are the trajectories of the

--

(15)
converter currents ffJ
and voltages /1 E 'Fa obtained in (8b)
UfJ, k= 1,2, . . . ,ny.
and (lOb). Together with the trajectories for and fa
these are Ua
the feedforward control trajectories that will lead to the system The subsequent computation of the remaining system tra
behaviour defined by the previously chosen trajectories for the jectories according to the stepwise procedure of section III
flat output. comprises the repeated use of (7) including predictions and
Note that the initial node plays a special role for the opera delays. This entails that the chosen trajectories (13) are involved
tion of the network. In contrast to the other nodes the converter on some larger time intervall1 [ti
- Tmax, tf
+ Tmax] rather that

current and voltage trajectories at this node can be chosen freely only on [ti,tf)
as indicated in Fig. 7. The resulting system
since they are included in the flat outputy. Therefore, if a direct trajectories and the control input trajectories in particular will
determination of the current or voltage trajectories at a certain leave their initial constant values already up to Tmax before
node is desired for some operational maneuver this node should t = ti and will reach their final constant values up to Tmax after
be chosen as the initial node. It might be useful to choose t= tf. Again, this reflects the wave propagation process taking
different initial nodes for different maneuvers. The CAPs being place on the transmission lines. In practice this means that an
the remaining components of the flat output y determine the operational maneuver,which intends to change the values of the
current fractions on the transmission lines at each node. This is variables belonging to the flat outputy within ti:os; t:os; tf,
has to
why they can be used to adjust the power distribution between start at t= ti
- Tmax and will not end before + Tmax't= tf
the converter stations within the network. J The maximum delay time Tmax can be computed by Tmax = max h where
kE'Pt
'P, is the set of the indices of all terminating nodes of the network and TJ1 the
B. Trajectory planning: Transition between two states of rest sum of all delays Tk of the lines Lk forming the path from the initial node PlY to
the terminating node PJ1, J1 E 'P,. This means TJ1 is the time that a current and
Several control tasks can be solved much easier if a flat voltage wave needs to travel from the initial node PlY to the terminating node
output of the system is known. A particular example is the Pj1"
 Yk components of the flat outputy are V] and h. The currents and
y{ ----------------------...---r--- voltages at p]
u2(0,t)= V] (t), i2(0,t)= -I] (t) (16)
are obtained using (4) and (5). No CAP is introduced at p]
because only one transmission line is connected = 1). (m]
Y -----r-./
Proceeding to the only neighbor of PI,
which is P2,
equations
-

/:J.I (7) with J1 = 2 and p{J1) = 1 allow to compute u2(2,f), i2(2' t)

Ii Tmax I; If If + Tmax
from u2(0,t),i2(0,t).
Because m2
=3 lines are connected to the
m2
-

passive node P2 one needs to introduce - 1 = 2 CAPs cr ,

Fig. 7. Polynomial trajectory for one component Yk of the fiat output y for the i
cr to determine the currents and voltages
transition between two states of rest
uk(O,t)=u2(2' t), ik(O,t)= (t)i2(2' t), k=3,4 (17)
according to (8a) and (lOa). The trajectory for only m2 - 2=1
of the two CAPs i can be chosen freely. For this example
cr , cr

cr is selected and is therefore included in the flat output as the

third component. After that, the other CAP is determined by
(9):
(18)
Fig. 8. Example of a circuit-free network with three converters Now the last two nodes can be considered. Employing (7) with
J1 = 3,p{J1) = 2 for line L3and again with J1 = 4,p{J1) = 2
for line L4
yields uk(k' t), ik(k' t), k
= 3,4. Finally the voltages
and currents at the converters C3 C4
and follow from (8b):
Vk(t)=uk(k' t), h(t)=ik(k' t) k=3,4. (19)
At the nodes P3 and P4 again no new CAPs are introduced
since there is only one line per node (m3 = m4 = 1). Finally,
all system variables are calculated and the flat output of the
example network is
(20)

B. Trajectory planning
Apart from the variables and h the trajectory for the
V]

third component cr of y can be prescribed freely as well. It
Fig. 9. Circuit-free network with three converters and the notation for the case determines how the line current i2(2' t)
is split between line
of P J beeing chosen as initial node L3 and line L4
and can hence be exploited to set the power
distribution between the converters andC3 C4.
For the sake of
IV. A SIMPLE EXAMPLE NETWORK
simplicity, cr shall remain at a constant value of 0.3 during the
maneuver to be planned. The two states of rest corresponding to
The results from section III shall now be clarified with the the required current change at C]
with constant current voltage
help of a simple example network with =4 nodes and =3 np nL V]
and CAP cr are characterized by
transmission lines shown in Fig. 8. The network is specified by
the sets
y = V d , y=0, y =0.3,
P={1,2,3,4 }, Pa={1,3,4 }, Pp={2 }, Y-I] = Vd , -I=Id,
Y2 Y-I3 -- 0 3
. .

L={ (I,2), (2,3), (2, 4) }, If the suggested polynomial trajectories (13) with an arbitrarily
N] N3 N4 = = ={2l. N2 ={1,3,4 }. fixed transition time /I..t are now assigned to the variables of the
flat output all remaining system trajectories can be computed by
Hence, the control inputs of the system are the currents or the the procedure described in section IV.A. In (19) this yields the
voltages of the three converters. particularly desired converter current and voltage trajectories.
It is assumed that the converter current h is required to One may chose for each of the three converters either its current
change from an initial value of 0 to a new constant desired value or its voltage as control input according to the technical realities
Id while the converter voltage V]
remains at a constant value of the converter stations. If the calculated converter trajectories
Vd . Since the voltage and current values at converter shall C] are applied to the system it will show the behaviour predefined
be determined node p]
is chosen as initial node. This yields by the trajectories of the flat output.
the notation illustrated in Fig. 9 and the network specific values To illustrate the results the system trajectories have been
a=l,p(2)=l,p(3)=2 andp(4)=2. computed using a set of numerical parameter values given in
TABLE I. The resulting trajectories in Fig. 10 clarify that the
A. Flat output transition of the complete system between the two states of rest
takes longer than only the prescribed transition time =5 ms. !-"t
The stepwise procedure from section III.A. adapts to this The maneuver, which was planned to change within II ti t
example network as follows. According to (11) the first two tI, is required to start already at = t ti -
Tmax = -20.76 ms
 transmission lines short enough to neglect delays. In this case
1
-- -;2(0,/); 1,(1)
C
Land are set to zero in the line equations (1) which changes

0.5 ....
- - - -;,(e2, I) z.
them to ordinary differential equations in Their solution
;3(e3, I); 13(1)
;3(0, I)
ull(fll,t) AluiO,t) + A2VO,t)
:= (21a)
-- ;,(e"I); 1,(1)
ill(fll,t) := AilUIl(O,t) + Ai2VO,t)
(21b)
i,(O, I)

-0.5
with Al Ai2 cosh( VRGfll), A2 .yR/G sinh( VRGfll)
:= := :=

and Ail (G/R)A2 replaces (7) in the calculation procedure of

:=

-1 section lILA. Because the changed line equations do not model

-30 -20 -10 10 20 30 40 50 60
fin ms
wave propagation anymore no delays and predictions occure
in (21). As a result, the duration of a planned meneuver will
(a) Currents al the aClive nodes (solid) and line currenls at P, (dashed)
reduce to !1t.
1.15 r----.-------.--::::=::;:---.--
= -=:::::J:=
: ,,====s;::::;=::::;:!=:;:;===il
-- 11,(0,/); U ,(I) To face the problem of model uncertainties and disturbances
I
-- 113(e3, I); U3(1)
-- 114(e"I); U4(1)
the feed-forward control scheme could be extended by local
1.1 ... i;,(/) feed-back controllers at each converter. This is a potential topic
. ..;3,-7 for future work.

1.05
Ii I
[1] J. Arrillaga.
volume 29 of
REFERENCES
High Voltage Direct Current Transmission,
Power and Energy Series.
The Institution of
-30 -20 -10 10 20 60 Electrical Engineers, 1998.
I in illS

(b) Voltages al the active nodes (solid) and node vollage at P, (dashed)
[2] E. Peschke and R. v. Olshausen. Kabelanlagen fiir Hoch
und Hochstspannung. Publicis-MCD-Verlag, 1998.
Fig. 10. Current and voltage trajectories for the three lines and converters
[3] N. Mohan, T.M. Undeland, and w.P. Robbins. Power
during the transition between two states of rest for a change of the converter
Electronics.
Wiley & Sons., 2002.
current II at converter C I from the initial value 1z = 0 to Y{ = Id via a
[4] P. Karlson. DC Distributed Power Systems - Analysis,
polynomial trajectory on 0 '5 t '5 /'I,.t = 5ms
Design and Control for a Renewable Energy System. PhD
thesis, Lund University, 2002.
at converter C4
and ends not before + Tmax t tJ
25.76 ms [5] H. Jiang and A. Ekstrom. Multiterminal HVDC systems
C4. 2
:= :=

at
The impact of the CAP (T is clarified by the dashed in urban areas of large cities. IEEE Trans. Power Delivery,
graphs in Fig. 1O(a). It can be seen that the currents and i3(0,t) 13:1278-1284,1998.
i4 t)
(0, are proportional to i2(f2,t)
according to the prescribed [6] Y.F. Lescale, A. Kumar, L.-E. Juhlin, H. Bjorklund, and
constant value (T ( ) 0.3. t := K. Nyberg. Challenges with multi-terminal UHVDC trans
TABLE 1. Numerical parameter values
missions. In Joint International Conference on Power Sys
tem Technology and IEEE Power India Conference, pow
ERCON, pages 1-7, Oct. 2008.
Parameter Value Parameter Value [7] L. Jun, O. Gomis-Bellmunt, J. Ekanayake, and N. Jenkins.
Control of multi-terminal V SC-HVDC transmission for
R 1.10-5 Om-l T2 .yy;ct2 12. 25 ms
offshore wind power. pages 1 -10, sept. 2009.
=

G 3.10-10 Sm-l T3 .yy;ct3 6. 12 ms

[8] L. Xu and L. Yao. DC voltage control and power dispatch
=

L 5.10-7 Hm-l T4 .yy;ct4

= 8. 51 ms
C 2.10-10 Fm-l Tmax T2 + 74 = 20. 76 ms
of a multi-terminal HVDC system for integrating large
t2 1000 km ti Oms offshore wind farms. Renewable Power Generation, lET,
t3 500km /'I,.t 5 ms 5(3):223 -233, May 2011.
t4 700km tf = ti + I'lt 5 ms [9] W. Lu and B.T. Ooi. Optimal acquisition and aggregation
of offshore wind power by multiterminal voltage-source
y. FURTHER REMARKS AND FUTURE WORK
HVDe. IEEE Power Engineering Review,22(8):71-72,
Aug. 2002.
[10] J. Rudolph and F. Woittennek. Motion planning and open
If an initial node and a certain trajectory form, e.g. (13),
loop control design for linear distributed parameter sys
is fixed the resulting system trajectories for different trajectory
planning procedures will differ only in a few parameters but not
tems with lumped controls. International Journal of Con
in their form. Thus, the complete calculations of section lILA.
trol,81(3):457-474,2008.
need to be performed only once for the first planning procedure.
[11] J. Rudolph.Flatness Based Control of Distributed Param
For following maneuvers only the trajectory parameters have to
eter Systems. Shaker Verlag, 2003.
be updated. This reduces the computational effort rapidly.
[12] F. Woittennek. Beitriige zum Steuerungsentwurf fiir lin
eare, ortlich verteilte Systeme mit konzentrierten Stellein
The control scheme suggested in this contribution is partic griffen. Shaker Verlag, 2007.
ularly useful for HVDC networks with long tranmission lines [13] K. Ktipfmtiller, W. Mathis, and A. Reibiger. Theoretische
since wave propagation processes and related delays are taken Elektrotechnik. Springer Verlag, 2006.
into account. However, it can be also applied to systems with [14] M. Fliess, P. Martin, N. Petit, and P. Rouchon. Active sig
2 The maximum delay time Tmax given in TABLE I is the maximum of the two nal restoration for the telegraph equation. In Proc. Confer
sums T2 + T3 and T2 + T4, which refer to the travel time of a voltage and current ence on Decision and Control, 1999.
wave from PI to P3 and to P4 respectively.