1
Abstract—The paper deals with harmonic analysis of HVDC
submarine cables. The longitudinal series impedance matrix is
computed with reference to armoured cables. The results are also
compared with Finite Element models showing a very good
agreement. This investigation can be considered the first step of
more detailed further analysis about the propagation of
harmonics on the d.c. and a.c. sides. Heretofore the topic seems
to be almost neglected in technical literature so this paper
investigates the harmonic behaviour of HVDC cable lines.
Index Terms— HVDC cables, Harmonic analysis, Armoured
cables, Impedance calculation.
I. INTRODUCTION
Projects and installations of off-shore wind farms are
becoming more and more widespread, drawing the Power
Electric Society attention to the HVDC option. The possible
resonances and propagation of harmonic into the system need
a very accurate modeling of submarine (also land) HVDC
cables. The HVDC submarine cables are rather complex
systems (multilayered structures) with, beyond the core
conductor, a metallic shield and armour which is in contact
with the sea water. Since the sheath and armour are multi-
point bonded and earthed, this implies that induced currents
circulate into them. For mass-impregnated paper insulated
cables, also the presence of a metallic reinforcement must be
taken into account.
The following configurations have been tackled:
¾ Monopolar configuration with sea return;
¾ Monopolar one with metallic return (placed at great
distance or in bundle configuration);
¾ Bipolar one.
These models can be usefully applied also to single-core a.c.
submarine cables (50 or 60 Hz). For the three-core a.c. cables
the procedures must be suitably adapted and will be subject of
further investigations.
In section II the typologies of installation and the HVDC cable
characteristics are defined; in section III the theoretical
backgrounds of the matrix procedure are presented including
the matrix passage from loop quantities to the usual series
impedences and the elimination of grounded conductors; in
section IV the computation results are shown and compared
with a FE model.
R. Benato and M. Forzan are with Department of Electrical Engineering,
University of Padova, Italy Via Gradenigo, 6/A 35131.
(e-mail: roberto.benato@unipd.it)
M. Marelli, A. Orini and E. Zaccone are with Prysmian S.r.l., Milan, Italy,
Via Chiese, 72 20126.
II. HVDC CABLE INSTALLATION FEATURES
A. HVDC Submarine installations
The installation features of submarine HVDC cables are rather
diverse. In general, the cables are buried at about 1,5 m under
the seabed surface. The metallic return (if any) can be placed
either in contact with the pole cable (bundle configuration) or
very distant (up to 300 m) in order to avoid that trawling or
anchoring activities can damage both. The sea water electrical
resistivity can range between 0,2 Ω⋅m and 0,5 Ω⋅m depending
upon the salt content. A possible range for the seabed
electrical resistivity is 1÷10 Ω⋅m.
Figure 1 Submarine HVDC cables (not to scale) in monopolar configuration
with metallic return
B. HVDC cables characteristics
There are three great categories of HVDC cables:
¾ Mass-Impregnated (MI): Insulated with special
paper, impregnated with high viscosity compound;
¾ Self-Contained Fluid Filled (SCFF): Insulated with
special paper, impregnated with low viscosity oil;
¾ Extruded: Insulated with extruded polyethylene-
based compound.
MI Cables are the most used; they have been in service for
more than 40 years and have been proven to be highly
reliable. At present they are used for voltage levels up to 500
kV d.c. and conductor sizes are typically up to 2500 mm2.
Self Contained Fluid-Filled Cables are used for very high
voltages (qualified for 600 kV d.c.), for short connections and
for voltage levels up to 500 kV d.c. Conductor sizes are up to
3000 mm2.
Extruded Cables for HVDC applications are still under
development; at present they are used for relatively low
voltage levels (up to 200 kV d.c.), mainly associated with
Voltage Source Converters (VSC), that could reverse the
power flow without reversing the polarity of the cable.
Harmonic Behaviour of HVDC cables
Roberto Benato Member IEEE, Michele Forzan,
Marco Marelli, Ambrogio Orini, Ernesto Zaccone Member IEEE
≈1,5 m ≈0,15÷300 m
ρsea=0,2÷0,5 Ω⋅m
ρseabed=1÷10 Ω⋅m
978-1-4244-6547-7/10/$26.00 © 2010 IEEE
2
Fig. 2 shows an HVDC armoured MI-cable: it is composed of
a core conductor, a lead-alloy sheath, a steel reinforcement
and a steel-wire armour. Consequently, it is a four metallic
layered cable. In the monopolar configuration with metallic
return, the return cable is usually XLPE-insulated: it does not
need any metallic reinforcement so that it is a three metallic
layered cable.
Figure 2 MI-paper insulated HVDC cable
III. THEORETICAL BACKGROUNDS FOR THE COMPUTATION OF
LONGITUDINAL SERIES IMPEDANCE MATRIX
The theory of coaxial cylindrical conductors (see fig. 3) is
well known in the technical literature thanks to the
fundamental contribution of Schelkunhoff [1].
Its application allows computing the longitudinal series
impedance matrix of the loop circuits shown in fig. 4. These
loop circuits are formed of:
¾ Loop 1 (L1): core conductor c and sheath s as return;
¾ Loop 2 (L2): sheath s and armour a as return;
¾ Loop 3 (L3): armour a and sea water as return.
The voltages (uL1, uL2 and uL3) and currents (iL1, iL2 and iL3) of
these loops are clearly shown in fig. 4: it considers the loop
currents as flowing in the xth conductor and returning into the
(x+1)th conductor (if x=armour, x+1 is the sea water).
Figure 4 also shows the electrical quantities (currents and
voltages) pertaining to the classical, usual phase-to-ground
voltages (uc, us and ua) and currents injected into the
conductor (ic, is and ia).
Figure 3 Composition of the three-layered single-core cable: core, sheath,
armour and insulating materials
Figure 4 Specification of the loop (grey) and "usual" (black) currents and
voltages for the three-conductor cable
In the example of fig. 4, the single-core cable has three
conductors (four if considering the sea water), three
insulations and three loops.
For generality purpose, the exposition can be referred to n
loops (and n+1 conductors included the sea return path).
The longitudinal series impedance matrix of the loops ZL
(order n×n) has a tridiagonal structure with the self
impedances of the loops laying in its main diagonal and the
mutual impedances between the loops located in the off-
diagonal elements as shown in fig. 5.
Figure 5 The tridiagonal loop matrix ZL
In general, the xth loop has the following self-impedance:
1x,inc1x/x,insx,outcL,L zzzz xx +−+− ++= (1)
For the last loop (i.e. x ≡ n), it is:
1n,inc1n/n,insn,outcL,L zzzz nn +−+− ++= (2)
where n+1 conductor is the sea water. The definitions of the
abovementioned impedances are:
Copper core
d
Semiconducting
MI-paper insulation
Semiconducting
Lead-alloy sheath
PE sheath
Steel reinforcement
Bedding
Steel wire armour
Serving PP yarn
iL1
iL2
Insulation c/s
Insulation s/a
CORE c
SHEATH s
uL1
uL2
ic
is
ia
uc
us
LOOP 1
LOOP 2
axis of the cable
iL3 LOOP 3 ua
SEA WATER
Insulation a/sea
ARMOUR a
uL3
zLx,Lx zLx,Lx+1 0 0 0
zLx+1,Lx ... … 0 0
0 … … 0
0 0 … … zLn-1,Ln
0 0 zLn,Ln-1 zLn,Ln
Lx
Lx+1
.
.
.
Ln-1
Ln
Lx Lx+1 . . . Ln-1 Ln
0
ZL=
x
WATER
core
insulation
core/sheath
sheath
insulation
sheath/armour
armour
insulation
armour/sea (if any)
3
zc-out,x = per unit length internal impedance of the xth coaxial
conductor (subscript c-) with the current returning in
the (x+1)th conductor (outer conductor whence the
subscript out);
zins,x/x+1 = per unit length impedance of the insulation
(subscript ins) between the conductor xth and (x+1)th;
zc-in,x+1 = per unit length internal impedance of the (x+1)th
coaxial conductor (subscript c-) with the current
returning in the xth conductor (inner conductor whence
the subscript in);
zc-in,n+1 = zself_sea = per unit length self sea-return impedance of
the nth conductor.
As is well known, the skin effect is considered by means of
the expression of zc-out,x whose real part is the a.c. conductor
resistance.
For the off-diagonal mutual impedances between the loop Lx
and the Lx-1 one of the matrix ZL the following relations are
valid:
x,mcL,LL,L
zzz
x1x1xx −−== −− (3)
where the minus sign takes into account the opposing
directions of the loop currents (e.g. iL2 is negative in the loop
L1).
Between Lx and Lx+2 loops there are no common branches so
that the coupling mutual impedance is zero.
In order to particularize this general exposition, the single-
core cable of fig. 3 can be considered so that x=c, s, a where
c=core, s=sheath and a=armour. It has:
321s,incs/c,insc,outcL,L zzzzzzz 11 ++=++= −− (4)
765a,inca/sheath,inss,outcL,L zzzzzzz 22 ++=++= −− (5)
11109sea_selfsea/a,insa,outcL,L zzzzzzz 33 ++=++= − (6)
where the abovementioned impedances have been re-named
z1,..,z11. The impedance zc-out (i.e. z1, z5 and z9 in (4)÷(6)) can
be computed as:
[ ])r(I)r(K)r(K)r(I
Hr2
z in1ext0in1ext0
ext
outc ⋅⋅⋅+⋅⋅⋅⋅⋅
⋅=− σσσσπ
σρ ; (7)
whereas the impedance (z3 and z7) are given by:
[ ])r(I)r(K)r(K)r(I
Hr2
z ext1in0ext1in0
in
inc ⋅⋅⋅+⋅⋅⋅⋅⋅
⋅=− σσσσπ
σρ ; (8)
where:
ω = 2πf = angular frequency [rad/s];
ρωμσ j= = reciprocal of the complex depth of penetration
[1/m] of the conductor with absolute magnetic
permeability μ [H/m] and electric resistivity ρ
[Ω⋅m];
rin = inner radius of the considered conductor [m];
rext = outer radius of the considered conductor [m];
I0(x), I1(x) = first kind modified Bessel functions of order 0
and 1 respectively;
K0(x), K1(x) = second kind modified Bessel functions of order
0 and 1 respectively;
H = I1(σ⋅rext)⋅K1(σ⋅rin) - I1(σ⋅rin)⋅K1(σ⋅rext).
If the core conductor is not hollow ( 0rin = ), eq. (7) becomes:
)r(Ir2
)r(Iz
ext1ext
ext0
1 ⋅⋅⋅
⋅⋅⋅= σπ
σσρ .
The impedances of insulating layers zins (z2, z6 and z10) are
computed by:
⎟⎟⎠
⎞
⎜⎜⎝
⎛⋅=
in_ins
ext_insins
ins r
r
ln
2
jz π
μω (9)
where:
μins = absolute magnetic permeability of insulating material
(usually ≅ 4⋅π 10-7[H/m]);
rins_ext = outer radius of the insulating material [m];
rins_in = inner radius of the insulating material [m].
It is worth reminding that the insulating layer between armour
and sea is usually missing, and the armour is in contact with
the sea water so that z10=0.
With regard to the off-diagonal elements of the matrix ZL, the
per unit length mutual impedance between L1 and L2 is given
by;
4m_sL,LL,L zzzz 1221 =−==
whereas for L2 and L3 the mutual impedance is:
8m_aL,LL,L zzzz 2332 =−==
where:
Hrr2
z
extin
x,mc ⋅=− π
ρ . (10)
As aforementioned zL1,L3=zL3,L1=0.
The computation of the self-impedance taking into account the
sea water as return (z11) would be rather complex also because
the cables are laid in the sea-bed having a resistivity different
from that of the sea itself (see fig. 1). It is worth remembering
that exact formulae of the self earth-return impedance have
been first derived from Pollaczek [2] in the hypothesis of
cable buried in a semi-infinite earth. In the following, as in
[3], the hypothesis of "Infinite Sea Model" (or Infinite Earth
Model) is assumed. As it will be demonstrated also with the
comparison of FEM, it is a really strong assumption. Attention
4
must be paid, since the cable can be considered as surrounded
by an infinite sea in all directions around it, when the
penetration depth is much less than the burial depth (measured
by the sea surface):
]m[depthburial
f
5032d
)Hz(
)m(sea
sea
sea <<≅= Ω
ρ
σ
This approximation could not be valid when the cable is
installed in very swallow water.
The impedance z11 can be inferred by (12) under the
hypothesis that the sea is a tubular conductor with infinite
radius (rext→∞) and that the inner radius is equal to the cable
external one (see fig. 6). If the armour is in contact with the
water, the cable outer radius coincides with the armour
external radius (rout_cable=re_armor).
Since
[ ] 0)rσ(Klim ext1
rext
=⋅
∞→
the relation (8) becomes:
[ ][ ] )r(K
)r(K
r2)r(K)r(I)r(K)r(I
)r(I)r(K)r(K)r(I
r2
z
in1
in0
inext1in1in1ext1
ext1in0ext1in0
in
inc ⋅
⋅⋅⋅
⋅=⋅⋅⋅−⋅⋅⋅
⋅⋅⋅+⋅⋅⋅
⋅
⋅=− σ
σ
π
σρ
σσσσ
σσσσ
π
σρ
So that the self-sea impedance z11 is equal to:
11
cable_outsea1
cable_outsea0
cable_out
seasea
sea_self z)r(K
)r(K
r2
z =
⎥⎥⎦
⎤
⎢⎢⎣
⎡
⋅
⋅
⋅= σ
σ
π
σρ
(11)
Figure 6 Infinite sea model
Finally, the series impedance of the three loops of the single-
core cable of fig. 3 is given by:
⎥⎥
⎥
⎦
⎤
⎢⎢
⎢
⎣
⎡
++−
−++−
−++
=
⎥⎥
⎥
⎦
⎤
⎢⎢
⎢
⎣
⎡
=
111098
87654
4321
,LL,LL
,LL,LL,LL
,LL,LL
zzzz0
zzzzz
0zzzz
zz0
zzz
0zz
3323
322212
2111
LZ
.(12)
The longitudinal series impedance matrix Z of the single-core
cable of fig. 3 is given by:
⎥⎥
⎥
⎦
⎤
⎢⎢
⎢
⎣
⎡
=
a,as,ac,a
a,ss,sc,s
a,cs,cc,c
zzz
zzz
zzz
Z . (13)
In the scientific literature, there are cumbersome algebraic
formulations in order to express the elements of series
impedance matrix Z starting from those of the loop matrix ZL
i.e.
1110987654321c,c zzzz2zzzz2zzzz +++−+++−++= ;
111098765s,s zzzz2zzzz +++−++= ;
1110987654c,ss,c zzzz2zzzzzz +++−+++−== ;
111098s,aa,sc,aa,c zzzzzzzz +++−==== ;
11109a,a zzzz ++= .
In the next subsection, a novel and elegant matrix algorithm
will be developed so avoiding the abovementioned relations.
A. Matrix procedures to pass from ZL to Z
The passage from the loop impedance matrix ZL to the usual
longitudinal series impedance matrix Z can be derived by
making extensive use of the matrix formalism rather than of
the abovementioned cumbersome algebraic expressions.
In fact, between the loop voltages (uL1, uL2 and uL3) and the
voltages with respect to ground (uc, us and ua), the following
obvious relations can be written:
scL uuu 1 −= asL uuu 2 −= aL uu 3 = .
Analogously, between loop and conductor currents, it can be
written:
1Lc ii = 12 LLs iii −= 23 LLa iii −=
Both the relation sets can be elegantly synthesized in the
matrix forms:
UTU L ′= and LITI = (14)
where: ,
i
i
i
3
2
1
L
L
L
⎥⎥
⎥⎥
⎦
⎤
⎢⎢
⎢⎢
⎣
⎡
=LI ,
u
u
u
3
2
1
L
L
L
⎥⎥
⎥⎥
⎦
⎤
⎢⎢
⎢⎢
⎣
⎡
=LU ,
i
i
i
a
s
c
⎥⎥
⎥
⎦
⎤
⎢⎢
⎢
⎣
⎡
=I .
u
u
u
a
s
c
⎥⎥
⎥
⎦
⎤
⎢⎢
⎢
⎣
⎡
=U
and the real matrix T (T' is its transpose) is given by:
⎥⎥
⎥
⎦
⎤
⎢⎢
⎢
⎣
⎡
−
−=
110
011
001
T .
The longitudinal series loop impedance ZL can be expressed
by the obvious:
LL
L IZU ⋅=∂
∂−
x
(15)
By substituting eqs. (14) in (15), it yields:
ITZUT 1L
−⋅=∂
⋅′∂−
x
= ( )[ ] ITZTU 1L1 ⋅⋅′=∂∂− −−x (16)
rin
rext →∞
sea water
5
where the matrix Z can be computed by means of loop matrix
ZL.
1
L
1' TZ)(TZ −− ⋅⋅= [Ω/m] (17)
The great advantage of the matrix formalism, as expressed by
G. Kron in [4], is the possibility of generalization from one
single-core cable to n single-core ones. For example by
considering two single-core cables as shown in fig. 7, the eq.
(17) is still true but with new meanings of the involved
matrices i.e.
110000
011000
001000
000110
000011
000001
−
−
−
−
=T
and ZL:
where the new mutual impedance between the loop L3 of P
and the loop L3 of R have not been yet defined.
Figure 7 Two single-core cables: pole (P) and metallic return (R) or second
pole
All the other loop mutual impedances between P and R are
zeroed since the loop L1 and L2 in each cable do not generate
magnetic field external to the cables themselves (e.g. the
magnetic field generated by iL1 in the core is zeroed by the
returning current iL1 in the sheath). The computation of these
mutual impedances can be performed always with the Infinite
Sea Model (for the demonstration see [5]), namely:
zm-PR=zm-RP= )r(K)r(Krr2
)d(K
R_extsea1P_extsea1R_extP_ext
ijsea0sea
⋅⋅⋅⋅⋅
⋅⋅
σσπ
σρ
(18)
where:
dij=spacing between cable axes [m];
rext_P=outer radius of the P cable [m];
rext_R=outer radius of the R cable [m].
B. Elimination of grounded conductors by a novel technique
The technique of elimination of grounded conductors is well
known in the scientific literature. It is only valid in the
hypothesis that the voltages along the entire length of the
conductors to be eliminated are zero. It is worth remembering
that for HVDC cables, since the sheath and the armour are
bonded and earthed at both ends (bonded also discretely at
intermediate intervals every 5÷10 km or separated by a
continuous semiconducting layer), and the armour is always in
contact with the sea water, their voltages can be considered
negligible along the entire cable length. The present procedure
can be performed either on one single-core cable or n single-
core cables. Once the series impedance matrix Z has been
computed, its inverse matrix Y must be firstly obtained as
shown in fig. 8. The advantage of the matrix Y is that it can be
directly multiplied by the vector
⎥⎥
⎥⎥
⎥⎥
⎥⎥
⎦
⎤
⎢⎢
⎢⎢
⎢⎢
⎢⎢
⎣
⎡
=
0
0
u
0
0
u
R
P
c
c
U
so that it is equivalent to the element extraction of fig. 8.
Figure 8 Elimination of grounded conductors (s, a) for two single-core cables
zL1,L1-P z L1,L2-P 0
z L2,L1-P z L2,L2-P z L2,L3-P
z L3,L2-P z L3,L3-P zm-PR 0
0
0 0
0 0
0 0
0
zm-RP
0
0 0
0 0
0 0
0
zL1,L1-R zL1,L2-R 0
zL2,L1-R zL2,L2-R zL2,L3-R
zL3,L2-R zL3,L3-R 0
P R
P
R
ZL=
yc,c-P y c,s-P y c,a-P
y s,c-P y s,s-P y s,a-P
y a,s-P y a,a-P y a,c-P
ycP,cR ycP,sR ycP,aR
yc,c-R yc,s-R yc,a-R
ys,c-R ys,s-R ys,a-R
ya,s-R ya,a-R ya,c-R
P R
P
R
Y=
ysP,cR ysP,sR ysP,aR
yaP,cR yaP,sR yaP,aR
ycR,cP ycR,sP ycR,aP
ysR,cP ysR,sP ysR,aP
yaR,cP yaR,sP yaR,aP
Y=Z-1
YRED
yc,c-P ycP,cR
ycR,cP yc,c-R
c s a s a c
c
c
s
a
s
a
iL3 - P iL3 - R
P R
dij
6
Therefore, the elements not to be eliminated (in grey in fig. 8)
can be extracted from this matrix and form a new reduced
matrix YRED. In order to obtain the final reduced impedance
ZRED without the grounded conductors (but taking into
account their electrical effects), it is sufficient to invert YRED
namely:
ZRED= 1REDY
− .
For the case of one single-core cable, this procedure gives
obviously the same result of the cumbersome procedures
available in literature.
C. Steel wire armour modelling
The armour is often composed of n steel wires with a given
diameter Φ: it is also stranded with a laying pitch p. The
armour can be modelled as a tubular conductor having the
inner radius equal to that of armour and the outer radius
computed so that the cross-section of tubular conductor is
equal to that of all wires S=
4
n
2Φπ⋅ . The per unit length
resistance is given by:
⎥⎦
⎤⎢⎣
⎡⋅⋅⋅
⋅=
kmcos
1
n
4r 2
armour
dc,armour
Ω
δΦπ
ρ
where δ=laying angle (see f