IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 31, NO.

1, FEBRUARY 2016 87

Fault Current Interruption in

Multiterminal HVDC Networks
Matthias K. Bucher, Student Member, IEEE, and Christian M. Franck, Senior Member, IEEE

Abstract—In a multiterminal HVDC network, dc circuit An HVDC CB has to be able to create a current zero, to dis-
breakers (CBs) are required at each end of a line to selectively sipate the energy stored in the circuit, and to establish the di-
isolate a fault. Several CB concepts have been implemented in electric strength. While dc CBs are available for low- (LV) and
PSCAD, which differ significantly in their structure and perfor-
mance. This paper describes the interaction of the different HVDC medium-voltage (MV) applications, only transfer and load cur-
CB topologies with a meshed four-terminal network and assesses rent switches are in use in HVDC systems [1]. Numerous con-
their performance in terms of maximum currents and voltages. cepts for such an HVDC CB have been presented up to now
The tradeoff between network parameters and CB requirements in patents and articles, all of which show a similar arrangement
is analyzed. with a switching element in the nominal path to build the voltage
Index Terms—HVDC circuit breaker, HVDC transmission, withstand capability, a commutation path to create the current
power system faults, power system simulation, power system zero, and an absorber path to dissipate the stored energy. The
transients, PSCAD.
main switching element can either be an arc between the con-
tacts of a mechanical CB, a solid-state based semiconducting de-
I. INTRODUCTION vice, or the combination of both. Each proposed CB concept has
advantages and drawbacks either in on-state losses or speed [4].
H VDC CIRCUIT breakers (CB) are widely considered as a
key enabling technology for future multiterminal HVDC
(MTDC) networks [1], which have been envisioned in various
The design of the HVDC CB has to be chosen according to the
expected maximum fault current in an MTDC network, which
depends on various factors. These prospective fault current in-
studies and projects from academia and industry. The avail-
fluencing factors include among others: the dc capacitor size
ability of HVDC CBs will be critical for the reliability of these
and fault resistance [5], the transmission-line technology [6],
networks. Point-to-point HVDC connections can be adequately
the grounding scheme [7], and the layout of the MTDC network
protected by conventional CBs on the ac side of the converter,
[8], [9].
even if this results in the de-energization of the entire dc system
It has been shown that the expected fault current exceeds the
[2]. A real MTDC grid, however, requires dc CBs at each end
breaking capability of most of the proposed HVDC CB concepts
of a line section to selectively isolate a fault by quickly and re-
or their construction costs are not economical. Therefore, the
liably breaking the fault current.
MTDC network design has to be adapted to the CBs capability
The interruption of an HVDC circuit generally requires
and additional fault clearing support options have to be chosen
the following [3]: a current zero has to be produced, the
as addressed in [4].
magnetic energy that is stored in the system inductance has
Recent publications either present detailed HVDC CB mod-
to be dissipated, and sufficient dielectric strength has to be
eling concepts, but without specific implementations [10], or
established to withstand the transient recovery voltage (TRV).
demonstrate their functionality based on simplified simulations
The first and second requirement are fulfilled rather easily in ac
with a single voltage source [11], or are applied with a focus on
systems given the natural current zero crossing. In dc systems,
the differences between HVDC system configurations in a ra-
however, the CB has to produce the current zero itself. This
dial MTDC grid [12]. This paper aims to combine the transient
can be done either by inserting a counteracting voltage or by
network study in a meshed MTDC layout with detailed HVDC
injecting current with opposite polarity. For practical purposes,
CB models to analyze the interaction between the CB and net-
the interruption process has to be completed within a certain
work. Detailed simulations in EMTDC-PSCAD of the fault cur-
time and the resulting switching surge has to be within the
rent interruption process are presented and the influence of the
insulation's withstand capability.
different components is analyzed based on parameter variations.
The modeling of four different HVDC CB concepts in
Manuscript received November 03, 2014; revised January 30, 2015 and April PSCAD is presented and their performance is studied in a
22, 2015; accepted June 14, 2015. Date of publication June 23, 2015; date of
current version January 21, 2016. This work was supported in part by ABB meshed four-terminal MTDC cable network during the current
Switzerland Ltd., in part by Siemens AG, Alstom Grid, in part by the Swiss interruption process of a pole-to-ground fault. The influence
Federal Office of Energy (BfE), and in part by CTI (SCCER-FURIES). Paper of the CB itself, the network components, and the protection
no. TPWRD-01328-2014.
The authors are with the Power Systems and High Voltage Laboratories, ETH system properties on the maximum fault currents, as well as
Zurich, Zurich 8092, Switzerland (e-mail: bucher@eeh.ee.ethz.ch; franck@eeh. maximum and minimum voltages in the system are shown.
ee.ethz.ch). The considered CB concepts include the passive and active
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. resonance CB, the solid-state based CB, as well as a hybrid
Digital Object Identifier 10.1109/TPWRD.2015.2448761 CB concept. Results are presented for a cable system, since ca-

0885-8977 © 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
88 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 31, NO. 1, FEBRUARY 2016

Fig. 2. Electrical equivalent scheme of the converter model with blocked

IGBTs ( is the ac voltage, is the ac resistance, is the ac inductance,
is the transformer reactance, is the phase reactor, is the dc
Fig. 1. MTDC network layout (CBs and inductors only shown for line 12). capacitor, and is the line reactor).

bles yield higher fault current levels than overhead lines (OHL) in the model, and the tripped converters remain blocked
[6]. Pole-to-ground faults are regarded as significantly more fre- throughout the simulation time. Equal ac network parameters
quent compared to pole-to-pole faults [13], particularly light- are assumed at all terminals, so there is no power flow if all
ning-related faults in OHL systems, although the latter fault converters are blocked. This is a valid assumption, as only the
would lead to more severe conditions [14]. transient period is of interest.
The ac network adjacent to the converter terminal is mod-
II. MODELING OF MTDC eled by its equivalent short-circuit impedance consisting of
The performance of the different HVDC CBs is tested in a and , and a voltage source . The converter transformer
meshed four-terminal MTDC network as shown in Fig. 1 using with reactance has a grounded star point on the high-voltage
PSCAD simulations. The pole-to-ground fault occurs at a dis- (HV) side and delta windings on the secondary side. An addi-
tance of 50 km to terminal 1 and 150 km to terminal 2. CBs are tional phase reactor is installed between the converter bridge
installed at all cable ends, but only CB1 and CB2 are tripped and transformer for harmonic filtering of the ac currents. In an
assuming differential relaying [15]. The chosen network layout MMC, the phase reactor represents the arm reactors.
and cable lengths result in higher fault currents in CB1 than in
CB2 in order to demonstrate their different behavior. III. HVDC CB CONCEPTS AND MODELING
Four different HVDC CB concepts and their modeling in
A. Cable Model PSCAD will be presented as following: the passive (P-RCB)
The cross-section of the frequency-dependent, distributed- and active resonance CB (A-RCB), the hybrid CB (HCB), and
parameter cable model is derived from a real 150-kV XLPE the full solid-state CB (SCB) concept.
VSC-HVDC submarine cable [16], [17]. The cross section has A simple fault detection mechanism is implemented at CB1
been scaled up to a 320-kV cable respecting the diameter of the and CB2 (cf. Fig. 1) with a level comparator to simulate differ-
copper conductor [18], while keeping the electric-field stress in ential protection [15] and selective opening of the CBs at the
similar cold conditions. The same material properties and cable faulted line ends. All other CBs remain closed. A very opti-
cross-section dimensions as in [5] are applied. The cable sheath mistic detection delay of 1 ms [15] is chosen for all
is assumed to have ground potential over the entire cable length CB concepts. The detection delay is defined as the time from
and is, therefore, mathematically eliminated in the simulations. occurrence of the fault to sending the trip signal to the CB and
accounts for all signal-processing and communication delays as
B. Converter Model required in differential protection schemes [15].
The converters are modeled as a generic 320-kV
bipolar half-bridge VSC topology with concentrated mid- A. P-RCB and A-RCB
point-grounded dc capacitors at each terminal and line The passive and active resonance CBs, as illustrated in Fig. 3,
reactors at each cable end in series to the CBs as illus- have a mechanical breaker in the nominal path, which is usually
trated in Fig. 2. The value of depends on the converter an ac air-blast CB [21] or an SF puffer CB [22] for the P-RCB
technology and can be zero in case of the modular multi- and a vacuum CB [23] for the A-RCB. In the passive scheme,
level converter (MMC). The converter's local overcurrent the components in the red boxes in Fig. 3 do not exist, whereas
protection blocks the insulated-gate bipolar transistor (IGBT) in the A-RCB, the capacitor is precharged by closing “ ”
modules above a threshold value of about twice the nominal and the active current injection is initiated through closing of
current [19] to protect them from overcurrents, making the “ ”. Both auxiliary switches and are modelled as ideal
half-bridge-based VSC an uncontrolled rectifier [20]. There- switches.
fore, the converter model to be implemented for the transient The temporal development of the current and voltage during
period can be simplified to the diode rectifier shown in Fig. 2. the interruption process in the P-RCB [21], [22], [24] is illus-
No postfault control strategy for the converters is implemented trated in Fig. 4 with example curves for the current in the nom-
BUCHER AND FRANCK: FAULT CURRENT INTERRUPTION IN MULTITERMINAL HVDC NETWORKS 89

Fig. 3. P-RCB and A-RCB model with additional components in red boxes.

Fig. 5. Fault current interruption in A-RCB.

the absorber path (cyan) and the voltage across the CB (ma-
genta curve) rises rapidly. The fault current (blue) starts to de-
crease when the voltage across the CB is larger than the system
voltage after about 30 ms. The PSCAD built-in MOA is used for
the energy dissipation in the absorber path. The nonlinear V-I
characteristic curve for dc applications is taken from [7]. The
reference voltage is chosen in order to have a leakage current
of 1 mA under continuous operation voltage. Consequently, the
current is 1 kA at a voltage of 1.32 p.u. and a residual voltage
of 1.85 p.u. results in a discharge current of 10 kA. The residual
Fig. 4. Fault current interruption in P-RCB. current disconnector “RD” as shown in Fig. 3 is opened after
current zero at around 43 ms.
The current interruption process in the active scheme
inal path (red), in the commutation path (green), in the absorber (A-RCB) [23], [25] is shown in Fig. 5. A CB opening delay
path (cyan), the sum of all currents (blue), and the CB voltage of 4 ms is again assumed, but can possibly be reduced, if a
(magenta). Note that the CB parameters in Fig. 4 are different vacuum CB with a special electromagnetic drive was used
from the ones presented in Section IV-A-2 to improve the visi- [26]. Through closing of “ ” after contact separation, the
bility of the oscillating currents. precharged capacitor injects a negative current (green
After the current has exceeded the CB threshold at 0.6 ms and curve in Fig. 5) into the nominal path (red), which forces the
the differential protection system has given the trip signal selec- arc current to zero. This considerably reduces the time to the
tively to the CB at 1.6 ms, the CB is tripped and the drive starts first current zero crossing compared to the passive scheme. It
to move the contacts. An optimistic estimate of requires, however, an additional charging unit with a dc source.
4 ms is assumed for the delay between the trip signal and con- It is assumed that the CB's required voltage withstand capability
tact separation. The arc voltage forces the current to commutate is reached within the rise time of the capacitor voltage of 2 ms.
into the commutation path, which consists of a resonance cir- 1) Arc Modeling: In contrast to the A-RCB with active cur-
cuit with an inductance , a capacitor , and a parasitic rent injection, the passive concept requires a negative V-I char-
resistance as indicated in Fig. 3. After another acteristic of the arc conductance to achieve an unstable, growing
5 ms, the moving contact cleared the nozzle, and the CB is oscillation in the resonance circuit. Therefore, accurate mod-
able to clear the arc. Also, the chosen is an optimistic eling of the arc burning and extinction process is needed.
lower estimate for fast gas CBs. Once the nozzle is cleared, Black-box arc models are widely used to simulate the dy-
the arc has a negative V-I characteristic. Together with a suf- namic arc behavior due to their computational efficiency. Their
ficiently small parasitic resistance, the oscillations of the cur- accuracy, however, depends on the exact description of the arc
rent in the commutation path start to grow at 10.6 ms (green parameter functions. The determination of these functions is dif-
curve in Fig. 4). This oscillating current is superimposed on the ficult and they are usually only valid under specific conditions
current through the nominal path (red), which results in a cur- [27]. Nevertheless, the dynamic arc behavior in P-RCB can suc-
rent zero crossing at around 29 ms and extinction of the arc. At cessfully be predicted as presented in [22] and [27].
that time, the total current (blue) is still high due to the stored The entire process of arc elongation during contact opening,
energy in the system inductance and resonance inductor. After dynamic behavior during cooling, and arc behavior near cur-
arc extinction, the capacitor is charged by the commu- rent zero cannot be described by a single black-box arc model.
tated current, until the threshold voltage of the metal–oxide ar- Therefore, an approximation of two submodels is chosen for the
rester (MOA) is exceeded. Then, the current commutates into study at hand. One model is for the arc elongation until nozzle
90 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 31, NO. 1, FEBRUARY 2016

opening assuming clogged gas flow and one model is for the
dynamic behavior under forced cooling in the high-current and
low-current range near current zero.
Arc elongation: During the arc elongation period, arc resis-
tance exhibits a positive V-I characteristic

(1)

where is the arc length at the nozzle opening, is the

arc voltage gradient, is the instantaneous arc current, and
is the required time for nozzle opening.
Arc dynamics: After opening of the nozzle, the gas flow in- Fig. 6. HCB model.
creases and the arc characteristic is dominated by the heating
and cooling dynamics modeled as a black-box model.
The majority of the available black-box models are modifi-
cations of Mayr's [28] and Cassie's equation [29] that are based
on the energy balance equation, which describes the change in
the arc column's energy content, resulting from the imbalance
between ohmic heating and forced cooling.
Mayr's equation originally assumed constant arc cooling
power and arc time constant . A modification of it can be
found in [30]

(2)

with the arc parameters and depending on the arc

conductance . These two arc parameters are described by
Fig. 7. Fault current interruption in HCB.
power functions

(3)
(4)

The constant cooling power factor depends linearly on the

blow pressure resulting in [24], [27].
To implement the dynamic arc model into PSCAD, (2) has to
be transformed into an integral form
Fig. 8. SCB model.

(5)
disconnector “D” is opened. The current commutates into the
snubber circuit and into the commutation path (green curve)
B. HCB about 10 s later as illustrated in Fig. 7. The current in the
There are numerous different hybrid HVDC CB arrange- nominal path (red curve) is reduced to zero immediately. After
ments [31], [32]. The description and modeling of all of them is full opening of the mechanical disconnector, the IGBTs in the
beyond the scope of this study, and one representative concept commutation path are blocked at around 4.6 ms and the current
similar to [33] as illustrated in Fig. 6 is chosen. commutates in their snubber circuits. In contrast to the SCB, the
To reduce the losses during normal operation, the nominal HCB cannot interrupt the current immediately after commuta-
path consists only of a few IGBTs and a fast mechanical dis- tion, but has to wait until the disconnector “D” has fully estab-
connector in series. Series-connected IGBTs with full voltage lished its dielectric strength to be able to withstand the TRV.
withstand capability are installed in the commutation path and This delay is named commutation delay and is assumed to be 3
an MOA in the absorber path. For zero voltage switching and ms. Afterwards, the TRV rises rapidly and the current commu-
loss reduction, all IGBTs are protected by RCD snubber circuits tates into the absorber path (cyan curve), where the remaining
consisting of a snubber capacitor , a resistor , and a diode. energy is dissipated. Again, the fault current starts to decrease
Note that Fig. 6 shows only the IGBTs for the positive current as soon as the voltage across the CB (magenta curve) exceeds
direction. For reverse blocking ability of the HCB, the same the system voltage (yellow curve).
number of IGBTs with opposite polarity has to be installed in
series. C. SCB
After receiving the trip signal from the protection at around The SCB configuration as illustrated in Fig. 8 has been
1.6 ms, the IGBTs in the nominal path are blocked and the chosen among different possible concepts [32]–[35]. This SCB
BUCHER AND FRANCK: FAULT CURRENT INTERRUPTION IN MULTITERMINAL HVDC NETWORKS 91

TABLE II
CB PARAMETERS

Fig. 9. Fault current interruption in SCB.

TABLE I
BASE-CASE SYSTEM PARAMETERS

TABLE III
INTERRUPTION TIME AND MAXIMUM CB CURRENT IN THE BASE CASE

has RCD snubber circuits parallel to each IGBT as explained

in Section III-B and the MOA is parallel to the nominal path.
2) Base-Case CB Parameters: The parameters for all CB
The current interruption process of the SCB is illustrated in
types are given in Table II.
Fig. 9. At around 1.6 ms, the IGBTs are blocked and the current
A snubber capacitor of 85 F per IGBT is chosen to limit the
commutates into the snubber circuits (green curve). Meanwhile,
rate of rise of the voltage to 300 s [36].
the voltage across the CB (magenta dashed curve) rises and the
current commutates finally into the absorber path (cyan curve) B. Comparison of CB Concepts in the Base Case
at around 2.5 ms. The different HVDC CB concepts presented in Section III
are compared in terms of interruption time, maximum CB fault
IV. RESULTS AND DISCUSSION current, and maximum and minimum voltages at the terminals
during the fault current interruption process.
The simulations are performed in PSCAD using a time step
The interruption time is defined as the time from fault detec-
of 1 s. Results for the CB concept comparison in the base case
tion at the CB to current zero (CZ) in the faulted line, that is,
and parameter variations are presented and discussed in the fol-
when the CB's total current becomes zero. In contrast to
lowing paragraphs.
the breaking time that considers only the time until current zero
in the nominal path, the interruption time also takes the energy
A. Parameters
dissipation process into account. Another performance indicator
1) Base-Case System Parameters: The default system pa- is the time, when the of the total current becomes nega-
rameters for the interaction study are summarized in Table I. In tive for the first time. Table III presents the comparison of the
the base case, a line reactor of 100 mH is installed between the different CB concepts in terms of time to CZ and time to nega-
cable end and CB (cf. Fig. 2) to limit the of the fault cur- tive in the base case. Table III also summarizes the cor-
rent [36]. The short-circuit power ratio (SCR) is defined as the responding maximum fault currents through CB1 and CB2 at
ratio of the short-circuit capacity at the point of common cou- the ends of the faulted cable for different CB concepts. Fig. 10
pling (PCC) and the rated power of the converter. illustrates the corresponding curves of the total CB current for
92 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 31, NO. 1, FEBRUARY 2016

Fig. 10. Curves of total CB current in SCB, HCB, and A-RCB for the base
case. Fig. 12. Terminal voltages in the MTDC network with slow P-RCB.

Fig. 11. Line currents in the MTDC network with slow P-RCB. Fig. 13. Maximum and minimum terminal voltages in the base case.

the SCB, HCB, and A-RCB. The current through the P-RCB direction and all four terminals feed the fault current in CB1.
has a larger time scale and is not shown here to provide better This results in a higher current load in CB1 (blue solid curve)
readability of the plot. As expected, the SCB has the lowest in- compared to the case if CB2 would not open at all (blue dashed
terruption time below 5 ms and, therefore, also the lowest max- curve) and the interruption time increased by about 142%. In
imum CB current of about 5 kA, whereas the P-RCB concept this simulation, due to the slow reaction of the P-RCB, the cur-
reveals the highest interruption time of up to 94 ms and 27-kA rent through all converters exceeds the threshold value and all
maximum current in CB1. The P-RCB concept also exhibits the converters are blocked. Thus, even after the fault is cleared by
highest difference between CB1 and CB2 in terms of interrup- CB1 and CB2, the current through all lines is zero (also see the
tion time and maximum fault current. It takes more than three explanation in Section II-B).
times longer to interrupt the current in CB1 than in CB2. Fast The corresponding terminal voltages of all four terminals are
CBs with interruption times below 10 ms act within the capac- illustrated in Fig. 12 for the P-RCB. After fault inception at 0 ms,
itor-dominated period of the transient fault current [5] and the the voltages at all four terminals drop to about 0.4 p.u. within
loading of the CBs is almost equal given the same dc capacitor 10 ms in the network with P-RCBs. In contrast, the terminal
size at both terminals. Slow acting-CBs, such as the P-RCB, voltages decrease to only about 0.7 p.u. using SCBs as shown
however, interrupt the current in the later ac infeed-dominated in Fig. 13.
period, during which the fault location is decisive. CB1 is lo- The SCB reveals better performance in terms of minimum ter-
cated closer to the fault compared to CB2 and has more feeders minal voltages, because the fault is cleared before the terminal
at the adjacent busbar, through which terminals 3 and 4 feed the voltages reach their lowest possible level as in the slow-acting
ground fault (cf. Fig. 1). P-RCB. The lower limit of the terminal voltage depends on the
The fact that CB2 opens earlier than CB1 even worsens CB1's line reactor and dc capacitor size as will be explained in Sec-
situation due to re-routing of the fault currents in the network as tions IV-C and D. It also depends on the fault resistance and the
depicted in Fig. 11 for a system with P-RCBs. After CB2 has distance to the fault as described in [5].
interrupted the fault current at around 28 ms (green curve), line The P-RCB exhibits not only lower minimum voltages, but
currents (cyan curve) and (yellow curve) change their also higher overvoltages after fault clearing compared to the
BUCHER AND FRANCK: FAULT CURRENT INTERRUPTION IN MULTITERMINAL HVDC NETWORKS 93

Fig. 14. Influence of the line reactor size on SCB performance Fig. 15. Influence of dc capacitor size on SCB performance 10 mH).
.

0 and 50 mH. Above this value, the line reactor size has much
SCB as can be seen in Fig. 12. After the current is interrupted less influence on the terminal overvoltage.
in CB2 at 28 ms, the voltage at terminal 2 (green curve) rises In general, large line reactors should be avoided, because they
to about 1.3 p.u. and the voltage at terminal 1 decreases even deteriorate the control performance due to the increased system
further to about 0.1 p.u.. The voltage at terminal 1 later increases time constant. Also, the volume and mass of the converter sta-
to 1.5 p.u. after fault clearance in CB1. The higher overvoltages tion increase with increasing reactor size, which is critical in
using P-RCBs are due to the higher currents in the MOAs and, offshore applications [40].
consequently, increased TRV and terminal voltages compared
to the SCB case. D. Influence of DC Capacitor
A comparison of maximum and minimum terminal voltages In contrast to the line reactor, an increasing dc capacitor size
for all CB concepts is illustrated in Fig. 13. A trend toward lower results in an increasing maximum CB current as illustrated in
minimum voltages at all terminals can be seen for increasing Fig. 15. The line reactor is kept constant at 10 mH for all capac-
interruption time, whereas the maximum overvoltages do not itor values. A large dc capacitor might seem at a disadvantage
show a clear trend. Due to the high time difference between at first glance, since the maximum current in CB1 (blue curve)
current interruption at CB1 and CB2 in the P-RCB, the max- increases from 10 kA in case of a small dc capacitor of 1 F
imum overvoltages differ considerably among the terminals, for to about 15 kA with a large 200- F capacitor. The voltage sta-
eample, 1.1 p.u. at terminal 3 and 1.5 p.u. at terminal 1. In gen- bility is, however, considerably improved with a larger dc ca-
eral, remote terminals 3 and 4 are less affected than terminals 1 pacitor due to its voltage supporting function. The minimum
and 2 closest to the ground fault. terminal voltage (green curve) is increased from 0 to 0.8 p.u.
The insulation demands for the IGBTs in the SCB and HCB and the overvoltage at terminal 1 (magenta curve) is completely
can be drastically reduced, if the MOAs are installed in a free- suppressed with capacitors larger than 100 F. The CB voltage
wheeling path [34] instead of arrester stacks parallel to the main (cyan curve) is again only marginally affected by dc capacitor
switch. size and remains between 1.8 and 1.9 p.u.
A further disadvantage of large dc capacitors is the increased
C. Influence of Line Reactor system inertia and converter station footprint.
Fig. 14 shows the influence of the line reactor size on the max-
imum CB current and the maximum terminal and CB voltages E. Influence of Commutation Delay
in a network with SCBs. A small 10- F dc capacitor is applied In contrast to the SCB, most HCB concepts enable proac-
in all of the variations to amplify the effect of the line reactor on tive switching [15], [33], [35], that is, after local detection of
the CB performance. A larger line reactor reduces the maximum overcurrent, the current can be commutated temporarily into
current in CB1 (blue curve) and CB2 (red curve) as expected. A the commutation path without interruption, while selective pro-
200-mH reactor results in a maximum current of around 4 kA, tection is decided, which CBs have to interrupt. With proactive
whereas CB1 has to interrupt a fault current of 15 kA in case control, the performance of the HCB can be improved and the
of a small 1-mH reactor in series. A larger line reactor is ben- time delay of the mechanical disconnector can be partly com-
eficial in terms of maximum fault current, but it substantially pensated. Fig. 16 compares the required time to CZ and the time
increases the maximum terminal voltage (magenta curve) from to in an SCB and HCB for different protection delays
1.1 p.u. at 1 mH to 1.7 p.u. at 150 mH. The voltage across the between 1 and 6 ms. The base-case parameters, as summarized
CBs (cyan curve), however, is only marginally influenced by the in Section IV-A-2, are used in these simulations. For protection
reactor size and remains within 1.7 and 1.8 p.u. The reactor size delays below 4 ms, the HCB (blue line) requires 9.2 ms inde-
has no impact on the minimum terminal voltage, which stays at pendently of the protection delay, while the interruption time in
around 0.3 p.u. (not shown in Fig. 14). The increase in terminal the SCB (red line) increases with increasing protection delay.
voltage due to an increase in reactor size is maximum between The HCB achieves the same interruption times as the SCB only
94 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 31, NO. 1, FEBRUARY 2016

is the most important CB requirement. A larger dc capacitor,

in turn, improves dc voltage stability, that is, it mitigates over-
voltages at the terminals and considerably decreases the voltage
drops, but slightly increases the maximum CB current.
The detection delay of the network protection mainly impacts
the fast CB concepts with short interruption times. In general,
the HCB performs worse than the SCB, but the gap between the
two concepts becomes smaller with increasing detection delay,
if proactive switching control in the HCB is used. In systems
with protection schemes, which not only rely on local measure-
ments, for example, differential protection, CB concepts with
proactive switching become more attractive, since their inter-
ruption time does not increase proportionally. The choice be-
tween SCB and HCB is thus very much influenced by the chosen
protection strategy and vice-versa.
Fig. 16. Influence of protection delay on interruption time in CB1.
REFERENCES
[1] C. M. Franck, “HVDC circuit breakers: A review identifying future
for long protection delays of more than 4 ms, but the gap be- research needs,” IEEE Trans. Power Del., vol. 26, no. 2, pp. 998–1007,
Apr. 2011.
tween the two concepts is significantly reduced even for short [2] L. Tang and B.-T. Ooi, “Locating and isolating DC faults in multi-
protection delays. For a protection delay of 1 ms, the interrup- terminal DC systems,” IEEE Trans. Power Del., vol. 22, no. 3, pp.
tion in the HCB is about 2.8 times the interruption time in the 1877–1884, Jul. 2007.
[3] W. Pucher, “Fundamentals of HVDC interruption,” Electra, no. 5, pp.
SCB, whereas for a delay of 3 ms, the HCB is only about 50% 24–38, 1968.
slower than the SCB with respect to time to CZ. The time to [4] M. K. Bucher, M. M. Walter, M. Pfeiffer, and C. M. Franck, “Options
negative shows an even smaller gap of 28%. The HCB for ground fault clearance in HVDC offshore networks,” presented
at the IEEE Energy Convers. Congr. Expo., Raleigh, NC, USA, Sep.
exhibits, however, much lower steady-state losses than the SCB 2012.
and does not require constant cooling of the IGBTs. The losses [5] M. K. Bucher and C. M. Franck, “Contribution of fault current sources
in multi-terminal HVDC cable networks,” IEEE Trans. Power Del.,
in an SCB amount to about 499 kW using 192 IGBTs [39]. vol. 28, no. 3, pp. 1796–1803, Jul. 2013.
[6] M. K. Bucher and C. M. Franck, “Analysis of transient fault currents
V. CONCLUSIONS in multi-terminal HVDC networks during pole-to-ground faults,”
presented at the Int. Conf. Power Syst. Transients, Vancouver, BC,
In this paper, four different HVDC CB concepts are imple- Canada, Jul. 2013.
[7] M. K. Bucher and C. M. Franck, “Comparison of fault currents in mul-
mented into PSCAD. A comparison of the CB's performance titerminal HVDC grids with different grounding schemes,” presented
is made with respect to interruption time, maximum CB cur- at the IEEE Power Energy Soc. Gen. Meeting, Washington, DC, USA,
rent, maximum and minimum terminal voltages, and maximum Jul. 2014.
[8] M. K. Bucher, R. Wiget, G. Andersson, and C. M. Franck, “Multi-
voltage across the CB. terminal HVDC networks—What is the preferred topology?,” IEEE
This study has demonstrated the complex interaction be- Trans. Power Del., vol. 29, no. 1, pp. 406–413, Feb. 2014.
tween CB and network components and revealed the tradeoff [9] M. D. Pfeiffer, M. K. Bucher, and C. M. Franck, “The effect of grid
topology on transient fault currents in multi-terminal VSCHVDC off-
between CB requirement specifications and network parame- shore networks,” presented at the Int. Conf. Power Syst. Transients,
ters. The choice of the CB results in a mutual optimization of Vancouver, BC, Canada, Jul. 2013.
[10] CIGRÉ Working Group B4.57, Guide for the Development of Models
line reactor size and CB technology. Therefore, also slower CB for HVDC Converters in a HVDC Grid, Brochure 604, Dec. 2014.
concepts with high-current interruption capability and/or along [11] R. Sander and T. Leibfried, “Considerations on energy absorption of
with large reactors can be of interest. HVDC circuit breakers,” presented at the Univ. Power Eng. Conf.,
Cluj-Napoca, Romania, Sep. 2014.
CB concepts, in which the interruption time depends on the [12] E. Kontos, R. T. Pinto, S. Rodrigues, and P. Bauer, “Impact of HVDC
current amplitude (e.g. the passive resonance concept), are un- transmission system topology on multiterminal DC network faults,”
suitable, as the interruption current of the second breaker to clear IEEE Trans. Power Del., vol. 30, no. 2, pp. 844–852, Apr. 2015.
[13] J. Candelaria and J.-D. Park, “VSC-HVDC system protection: A re-
could be excessively high. view of current methods,” presented at the IEEE Power Syst. Conf.
Due to the high rate of rise of the fault current in anMTDC Expo., Phoenix, AZ, USA, Mar. 2011.
network, the fault has to be cleared by the CBs as fast as pos- [14] J. Yang, J. Fletcher, and J. O'Reilly, “Short-circuit and ground fault
analysis and location in VSC-based DC network cables,” IEEE Trans.
sible. As expected, the fast CB concepts, such as SCB and HCB, Ind. Electron., vol. 59, no. 10, pp. 3827–3837, Oct. 2012.
perform better than the slower P-RCB and A-RCB. They yield [15] Twenties Project, Demo 3—Protection Plan for DC Networks, Deliv-
lower interruption times below 10 ms and, consequently, lower erable no. 11.2, Dec. 2012.
[16] L. Ronström, M. Hoffstein, R. Pajo, and M. Lahtinen, “The Estlink
maximum CB currents and voltages. Also, the terminals experi- HVDC light transmission system,” presented at the CIGRE Regional
ence less overvoltages and higher minimum terminal voltages. Meet. Security Rel. Elect. Power Syst., Tallinn, Estonia, Jun. 2007.
[17] , T. Worzyk, Ed., Submarine Power Cables: Design, Installation, Re-
The performance of the CB depends not only on the CB con- pair, Environmental Aspects. New York, USA: Springer, 2009.
cept, but also on the network parameters, such as the line in- [18] ABB High Voltage Cables, HVDC light cables—Submarine and land
ductor, dc capacitor, and protection scheme. A larger inductor power cables. Oct. 2006. [Online]. Available: http://library.abb.com
[19] M. Baran and N. R. Mahajan, “Overcurrent protection on
results in lower CB currents, but much higher maximum ter- voltage-source-converter based multiterminal dc distribution sys-
minal voltages. Again, this implies that the interruption time tems,” IEEE Trans. Power Del., vol. 22, no. 1, pp. 406–412, Jan. 2007.
BUCHER AND FRANCK: FAULT CURRENT INTERRUPTION IN MULTITERMINAL HVDC NETWORKS 95

[20] J. Yang, J. Fletcher, and J. O'Reilly, “Multiterminal DC wind farm col- [35] Y. Wang and R. Marquardt, “Future HVDC-grids employing modular
lection grid internal fault analysis and protection design,” IEEE Trans. multilevel converters and hybrid DC-breakers,” presented at the 15th
Power Del., vol. 25, no. 4, pp. 2308–2318, Oct. 2010. Eur. Conf. Power Electron. Appl., Lille, France, Sep. 2013.
[21] B. Bachmann, G. Mauthe, E. Ruoss, H. P. Lips, J. Porter, and J. Vithay- [36] M. Callavik, A. Blomberg, J. Häfner, and B. Jacobson, “The hybrid
athil, “DC current interruption in HVDC SF6 Gas MRTB by means of HVDC breaker—An innovation breakthrough enabling reliable HVDC
self-excitation,” IEEE Trans. Power App. Syst., vol. PAS-104, no. 9, grids,” ABB Grid Syst. Tech. Paper, Nov. 2012.
pp. 2460–2466, Sep. 1985. [37] High Voltage Circuit Breakers. Faculty of Electrical Engineering,
[22] H. Nakao, Y. Nakagoshi, M. Hatano, T. Koshizuka, S. N. A. M. Kapetanovic, Ed. Sarajevo, Bosnia and Herzegovina: Sarajevo,
Kobayashi, T. Murao, and S. Yanabu, “DC current interruption in 2011.
HVDC SF6 Gas MRTB by means of self-excitation,” IEEE Trans. [38] V. V. Terzija and H.-J. Koglin, “New approach to arc resistance calcu-
Power Del., vol. 16, no. 4, pp. 687–693, Oct. 2001. lation,” in Proc. IEEE Power Energy Soc. Winter Meeting, Columbus,
[23] K. Tahata, S. Ka, S. Tokoyoda, K. Kamei, K. Kikuchi, D. Yoshida, Y. OH, USA, Jan. 2001, pp. 781–787.
Kono, R. Yamamoto, and H. Ito, “HVDC circuit breakers for HVDC [39] ABB Switzerland Ltd., 5SNA 2000K451300 StakPak IGBT Module
grid applications,” in Proc. CIGRÉ AORC Tech. Meeting, Tokyo, Doc. no. 5SYA 1430-00 01-2013, 2013.
Japan, May 2014. [40] E. Kontos, S. Rodrigues, R. T. Pinto, and P. Bauer, “Optimization
[24] E. Bonin, B. Koetzold, K. Möller, G. Oberdorfer, J. Schwarz, and H. G. of limiting reactors design for DC fault protection of multi-terminal
Thiel, “A method of current interruption in HVDC networks by means HVDC networks,” presented at the IEEE Energy Convers. Congr.
of AC circuit-breaker with adapted arc characteristics and energy ab- Expo., Pittsburgh, PA, USA, Sep. 2014.
sorbers,” in Proc. CIGRÉ Session, Aug. 1970.
[25] T. Eriksson, M. Backman, and S. Halen, “A low loss mechanical
HVDC breaker for HVDC Grid applications,” in Proc. CIGRÉ Ses-
sion, Paris, France, Aug. 2014.
[26] M. Homma, M. Sakaki, E. Kaneko, and S. Yanabu, “History of vacuum Matthias K. Bucher (S'12) received the B.Sc. and
circuit breakers and recent developments in Japan,” IEEE Trans. Di- M.Sc. degrees in electrical engineering from the ETH
electr. Electr. Insul., vol. 13, no. 1, pp. 85–92, Feb. 2006. Zurich, Zurich, Switzerland, in 2009 and 2011, re-
[27] M. M. Walter, “Switching arcs in passive resonance HVDC circuit spectively, where he is currently pursuing the Ph.D.
breakers,” Ph.D. dissertation, Dept. Inf. Technol. Elect. Eng., ETH, degree in high voltage engineering.
Zurich, Switzerland, 2013. He joined the High Voltage Laboratory, ETH
[28] O. Mayr, “Beiträge zur Theorie des statischen und des dynamischen Zurich, in 2011. His research is dedicated to tran-
Lichtbogens,” Arch. Elektrotech., vol. 37, pp. 588–608, 1943. sients in multiterminal HVDC networks.
[29] A. M. Cassie, Arc rupture and circuit severity: A new theory Paris,
France, CIGRÉ Rep. no. 102, 1939.
[30] J. Schwarz, “Berechnung von Schaltvorgängen mit einer zweifach mod-
ifizierten Mayr-Gleichung,” ETZ-A, vol. 93, no. 7, pp. 386–389, 1972.
[31] A. M. S. Atmadji and J. G. J. Sloot, “Hybrid switching: A review of
current literature,” in Proc. Int. Conf. Energy Manage. Power Del.,
Mar. 1998, pp. 683–688. Christian M. Franck (M'04–SM'11) received the
[32] C. Meyer, M. Kowal, and R. W. De Doncker, “Circuit breaker concepts diploma in physics from the University of Kiel, Kiel,
for future high-power DC-applications,” in Proc. Ind. Appl. Soc. Annu. Germany, in 1999 and the Ph.D. degree in physics
Meeting, Oct. 2005, vol. 2, pp. 860–866. from the University of Greifswald, Greifswald,
[33] J. Haefner and B. Jacobson, “Proactive hybrid HVDC breakers—A key Germany, in 2003.
innovation for reliable HVDC grids,” in Proc. CIGRE Symp., Bologna, He was Scientist and Group Leader for gas
Italy, Sep. 2011. circuit breakers and high-voltage systems with the
[34] K. Sano and M. Takasaki, “A surge-less solid-state dc circuit breaker Swiss corporate research center of ABB, Dättwil,
for voltage source converter based HVDC transmission systems,” pre- Switzerland, from 2003 to 2009. Currently, he is an
sented at the IEEE Energy Convers. Congr. Expo., Raleigh, NC, USA, Assistant Professor for High Voltage Technology at
Sep. 2012. ETH Zurich, Zurich, Switzerland.