Designing for

High-Voltage dc
Grid Protection

©istockphoto.com/iskrinka74
Fault Clearing
Strategies
and Protection
Algorithms

P
Protecting HigH-VoLtAge (HV) Dc griDs requires A Differ-
ent approach compared with that of ac power system protection and poses one
of the major challenges that must be resolved before the realization of large-
scale HVdc grids that use equipment from multiple vendors. HVdc grid pro-
tection, which is essential for safe and reliable HVdc grid operation, entails the
appropriate detection and fault clearing of dc-side short circuit faults (i.e., dc
faults). in this context, fault clearing refers to interrupting the dc fault current
and isolating the faulted component. fault current interruption is much more

By Willem Leterme, Ilka Jahn, Philipp Ruffing,

Kamran Sharifabadi, and Dirk Van Hertem
Digital Object Identifier 10.1109/MPE.2019.2897188
Date of publication: 17 April 2019

may/june 2019 1540-7977/19©2019IEEE ieee power & energy magazine 73


Udc
+Udc/2
–Udc/2 0 0
(a) (b)
figure 1. Three possible configurations for HVdc systems that indicate voltages during normal operation and prospec-
tive fault voltages and currents during pole-to-ground faults. (a) and (d) A symmetric monopole, (b) and (e) an asymmetric
monopole with metallic return, and (c) and (f) a bipole with metallic return.
complex in HVdc grids compared to ac systems because dc
fault currents have no naturally recurring zero crossings and,
without countermeasures, quickly increase to values that are
unacceptable for power electronic components. conversely,
the technologies used within HVdc grids offer options for
fault clearing beyond the well-known approach of using cir-
cuit breakers in existing ac systems.
Functional Requirements
for HVdc Grid Protection
to enable a cost-effective and reliable future power system
with HVdc grids, the design of HVdc grid protection must
begin with determining the likelihood of faults occurring
in the system, their impact on the system, and the type of
equipment available for clearing them. the HVdc grid-
protection design is, therefore, closely related to the overall
design of the HVdc grid itself and with the selection of com-
ponents within the HVdc grid. in this context, we refer to the
overall design of the HVdc grid as the choices made in the
HVdc grid structure, i.e., the number of terminals, a radial or
Fault Impact
Category 2: Unacceptable Impact
Category 1: Highly Unlikely
Category 3: Unacceptable Risk
Category 4: Acceptable Risk
Fault Probability
figure 2. An example classification of power system faults
and impacts based on RTE’s “Mémento de la sûreté du
système électrique.”
74 ieee power & energy magazine
+Udc If,ss >> U≈0
If,ss >> U≈0
If,ss = 0 0
0 0
–Udc
–Udc –Udc
(c) (d) (e) (f)
meshed connection of terminals, the power rating of ac–dc
nodes, and the choice of grid configuration and grounding,
such as high-impedance grounded symmetric monopole or
low-impedance grounded asymmetric monopole or bipole.
for the selection of its components, we refer to the tech-
nology used for transmission lines, converters, and fault-
clearing equipment. the combination of all of these aspects
determines the probability of faults in the system and their
impact without HVdc grid protection.
the impact of the fault on the system in terms of over-
current or overvoltage depends on the combination of fault
type and the grounding and configuration of the HVdc grid.
for pole-to-pole faults, irrespective of the grounding and
configuration, dc fault currents quickly increase and reach
a high steady-state value in the absence of any protection.
the impact of pole-to-ground faults depends on the type of
grounding: for a low-impedance grounded grid, prospec-
tive fault currents are high, whereas for a high-impedance
grounded system, prospective overvoltages on the healthy
poles are high (figure 1).
the faults occurring in the system can be grouped into
several categories indicating the acceptability or unaccept-
ability of their probability and impact, which are combined
to determine risk (figure 2). the resulting design of HVdc
grid protection will largely depend on the outcome of this risk
assessment and determine the final impact of faults on the
system. for instance, it is conceivable that the protection of
a symmetric monopolar cable grid differs considerably from
that of a bipolar grid that makes use of overhead lines. in the
first case, HVdc grid protection may use HVdc circuit break-
ers with low current interruption capability or may not use
HVdc circuit breakers at all to deal with rare fault occurrences
and the high likelihood of permanent faults. in the second
case, it may make sense to provide HVdc circuit breakers
at both ends of all the lines to quickly isolate the more fre-
quently recurring faults and provide for automatic reclosing of
the HVdc circuit breakers in the event of nonpermanent faults.
Although a set of specific functional requirements for
HVdc grid protection can only be compiled once the HVdc
may/june 2019
 It is conceivable that the desired protection philosophy
provides a higher operation speed and fault-clearing
selectivity as the grid grows in size.

grid design has been determined, the overall objectives for they may be required to absorb a large amount of energy
every HVdc grid-protection scheme are the same: HVdc during fault current interruption.
grid protection should ensure continuity of secure power
system operation and avoid damage to system components Technology That Assists With
in case faults occur. for short circuit faults, this means that Clearing dc Faults
the HVdc grid protection should detect the faults, interrupt several means of interrupting or limiting dc fault currents
the short circuit currents, isolate the faulted component, and are available. Applied technologies include converters, ac
restore the system to a secure operation state. the protec- circuit breakers, dc circuit breakers, and combinations of
tion scheme should perform these functions without causing converter and circuit breakers.
unwanted ac or dc system instabilities. in addition, it should
avoid damage to equipment and property and guarantee AC Circuit Breakers
safety for personnel and the public. in current Vsc HVdc point-to-point schemes, ac circuit
A set of functional requirements for interacting with the breakers are installed in a converter station as the primary
connected ac system can be defined based on the constraints dc fault-interrupting device given that the ac–dc converter
for its secure operation and its components. each ac system itself, i.e., either a two-level or half-bridge modular multi-
grid code defines a maximum power imbalance that imposes level converter (MMc), has no dc fault-blocking capability.
a limit on the secure operation of the system. for instance, in for these types of converters, a fault current path from the ac
the entso-e (european) grid code, this maximum power to the dc side always exists via the antiparallel diodes within
imbalance is specified as 3 gW in continental europe for the converter’s submodules (figure 3). therefore, the ac cir-
an unspecified amount of time and in the united Kingdom cuit breakers eliminate the ac contribution to the dc fault
as 1.8 gW, to avoid a frequency deviation outside the range and typically interrupt the ac current associated with the dc
of 49.5–50.5 Hz for 60 s. HVdc grid protection should thus fault within a few cycles of the fundamental frequency. After
prevent loss of the HVDc line from causing a power imbal- eliminating the ac contribution, the dc fault current passively
ance in the connected ac system that exceeds that maximum decays to zero in a time interval that depends on converter
value. in addition to the maximum power imbalance, the topology and system parameters.
possible functional requirements for HVdc grid protection
relate to the transient stability of connected ac systems or to HVdc Circuit Breakers
the damage to its components. HVdc circuit breakers are the dc counterparts of ac circuit
Within the HVdc grid, functional requirements for HVdc breakers and can interrupt both nominal and short circuit
grid protection can be based on the continuity of the HVdc currents up to their short circuit rating. typically, HVdc cir-
grid operation itself (if required) or from the limits of its cuit breakers must interrupt currents in the 5–15-kA range
components. At present, there is no HVdc grid code that
defines the constraints within which secure operation is
achieved, and the following suggestions are given for illus-
tration purposes only. HVdc grid protection is required to SM SM SM
keep the voltage at converter terminals within certain limits Half Bridge
so they can remain connected. such a “converter dc fault SM SM SM +
ac Side dc Side –
ride through,” however, may not necessarily imply that every Full Bridge
converter remains in an active switching state or retains the SM +
capability of supporting the ac system during the entire fault –
duration. such grid protection should also avoid damage to
SM SM SM
the power electronic components and other equipment. At
present, for voltage source converter (Vsc)-based HVdc SM SM SM
point-to-point connections, the most critical components
are the converter power electronic switches, i.e., insulated-
gate bipolar transistors. in future HVdc grids, HVdc circuit figure 3. The schematic of an MMC comprising half- or
breakers (if any) could also become a limiting factor because full-bridge submodules. SM: submodule.

may/june 2019 ieee power & energy magazine 75

 within a time interval of several milliseconds. compared
to ac circuit breakers, they are more complex because they
must deal with dc currents, which have no naturally recur-
ring zero crossings. to interrupt these nonzero currents,
HVdc circuit breakers consist of several parallel paths that
perform the functions of current interruption themselves
and absorb the energy present in the inductive elements
of the circuit. figure 4 illustrates these paths, where the
upper path typically comprises a mechanical switch and
forms the path for normal conduction, the middle path is
an auxiliary path to aid in interrupting fault currents, and
the lower path is an auxiliary path for energy absorption,
typically consisting of metal–oxide varistors. the HVdc
circuit breaker’s operation typically relies on a succession
of actions, which depend on the technology used. reso-
nant circuit breakers make use of mechanical breakers and
resonant circuits that induce current-zero crossings [fig-
ure 4(a) and (b)]. Because of their slow clearing time, pas-
sive resonant breakers are not appropriate for interrupting
short circuit currents in Vsc-based HVdc systems. HVdc
circuit breakers, based on a combination of power elec-
tronics and mechanical breakers in the normal conduction
path, e.g., so-called hybrid HVdc circuit breakers, typi-
cally involve using power electronics to avoid arcs in the
mechanical breakers [figure 4(c)]. each of the technologies
recommended by manufacturers has advantages and disad-
vantages in terms of speed, on-state losses, cost of compo-
nents, physical dimensions, maximum current interruption
capability, or overall reliability and simplicity of the com-
ponents used. Although HVdc circuit breakers are not yet
commercially available, prototypes have been tested and
found to perform in the range of expected requirements.
note that these prototypes all require fault-current-lim-
iting equipment.
Fault-Blocking Converters
in contrast to converters without fault-blocking capability, con- At present, HVdc grid protection invariably requires equip-
verters with this capability are able to block the fault current
contribution from the ac system to the dc fault. these convert-
ers rely on converter submodules that are able to insert volt-
ages of positive and negative polarity, unlike the half-bridge
submodule, which can only insert voltages of positive polar-
ity (see figure 3). As such, they can insert a voltage opposing
the ac system voltage contribution in the fault current path that fault-current limiters. inductors in series with the circuit only
(a) (b)
figure 4. Conceptual dc switchgear designs: (a) a passive resonant circuit breaker,
(b) an active resonant circuit breaker, and (c) a hybrid circuit breaker.
76 ieee power & energy magazine
connects the ac to the dc system. this process enables them to
block the fault current contribution from the ac to the dc side
without needing ac or dc circuit breakers. the most established
converter with fault-blocking capability is the full-bridge
MMc (figure 3), which will see its first application in ger-
many’s ultranet project. there are other promising converter
configurations that allow for blocking and controlling dc fault
currents while having reduced losses during normal opera-
tion, when compared with a full-bridge MMc. examples of
this are hybrid MMc configurations that use a combination of
half- and full-bridge submodules or alternative topologies such
as the alternate arm converter. fault-blocking capability may
be achieved in a controlled or uncontrolled manner. A major
advantage of controlled blocking is the uninterrupted reactive
power controllability of the converter, which allows the con-
verter to operate in a static synchronous compensator mode
during the entire fault-clearing process.
DC–DC Converters
Proposed for transforming dc voltage from one level to
another, dc–dc converters may also limit or even interrupt
fault current contribution from one side to the other. the
dc–dc converter’s ability to block the transfer of fault currents
depends on its topology. All the topologies that use an inter-
mediate ac stage, thereby providing galvanic isolation between
both sides, can block the propagation of fault current from one
side to the other, irrespective of the fault-blocking capability
of the ac–dc converters used. in direct dc–dc conversion, the
topologies are generally able to block fault-current propaga-
tion in one direction, whereas only certain types of topologies
are able to block it in both directions. for instance, to achieve
bidirectional fault-blocking capability with an MMc-type
dc–dc converter, submodules with fault-blocking capability
must be used.
Fault Current Limiters
ment that either limits the rate of rise or the magnitude of the
dc fault current. fault-current-limiting equipment may take
various forms, such as inductors in series with the circuit
(typically air-core), active fault-current limiting through con-
verters with fault-blocking capability, HVdc circuit breakers
with fault-current-limiting capability, or superconducting
limit the rate of rise of the fault
current and act as a limiter only in
conjunction with equipment that
interrupts the dc fault current, such
as HVdc circuit breakers. these
inductors may be placed directly in
series with the dc circuit breakers
(c) or can also be found in the arms
of the ac–dc converters. the ac–dc
converters with controlled fault-
blocking capability can control
may/june 2019
 HVdc grid protection does not necessarily
implement the same conventional approach used
for ac system protection.
the fault current and limit it to a desired level instead of
reducing the dc fault current entirely to zero. certain types
of HVdc circuit breakers, e.g., the ones making use of con-
trollable power electronic modules, are capable of actively
limiting the current. superconducting fault-current limiters
use a component that is in superconducting mode during
normal operation. therefore, it presents a low impedance to
the circuit. During faults, the superconducting component is
driven out of superconducting mode, resulting in the loss of
superconducting capabilities and, thereby, presents a high
impedance to the circuit. fault-current-limiting equipment
can be installed in series with the transmission lines. in
this manner, they are able to limit both the transmission
line discharge currents as well as the contributions from
the ac–dc converters. fault current limiters that are located
in series with the ac/dc converters do not affect the line
discharge currents.
High-Speed Switches
Dc high-speed switches (Hsss) can be used to quickly iso-
late a faulted line from the remaining dc network and oper-
ate only under near-zero voltage and current conditions;
therefore, these Hsss are not required to interrupt fault cur-
rents. However, depending on the application, these switches
may be required to interrupt small residual currents in the
grid, which, e.g., result from passive discharge of capacitive
or inductive grid components. in case residual currents have
zero crossings, these Hsss may make use of traditional ac
circuit breaker technology. Without current-zero crossings,
the Hsss must provide a sufficiently high countervoltage or
have an auxiliary circuit that creates a zero crossing, e.g.,
a passive resonant circuit used in load transfer breakers in
classic point-to-point HVdc connections.
Classification and Characterization of
Fault-Clearing Strategies for HVdc Grids
HVdc grid protection does not necessarily implement the
same conventional approach used for ac system protection.
in the conventional approach to ac system protection, circuit
breakers are placed throughout the system and used to simul-
taneously interrupt the fault current and isolate the faulted
component. this has led to a fault-clearing strategy that
divides the power system into zones containing grid elements
such as transformers or transmission lines. in case of a fault,
the protection scheme disconnects and de-energizes just the
zone containing the fault. given the different types of equip-
ment available for HVdc grid protection and the characteris-
may/june 2019
tics of this equipment, alternatives to the selective fault-clear-
ing strategy exist. regarding these alternatives, the protection
zones used for fault current interruption do not necessarily
coincide with the components that should be isolated. these
fault-clearing strategies can be classified in terms of “extent
of the HVdc grid which is deenergized,” an approach followed
in the cigre technical brochure (tB) 739, or described in
terms of “action at the protection zone point-of-connection,”
an approach followed by the european committee for elec-
trotechnical standardization (ceneLec).
in cigre tB 739, fault-clearing strategies are divided
into three main philosophies. in the first philosophy, nonse-
lective fault clearing, the entire HVdc grid is completely
de-energized prior to isolation of the faulted component
under near-zero voltage and current conditions. the faulted
component should be identified during or after grid deenergi-
zation and may be automatically isolated using Hsss. After
the faulted component is isolated, the remaining part of the
HVdc grid is reenergized before power flow can resume. in
the second philosophy, partially selective fault clearing, the
HVdc grid is subdivided into several protection zones. Here,
the faulted zone is first quickly isolated from the healthy
zones of the grid. this requires that equipment capable of
interrupting dc fault currents is present at all boundaries of
each protection zone. thereafter, the faulted element within
the faulted zone is isolated, as in that of a nonselective phi-
losophy. then the remaining portion of the initially faulted
zone is reenergized and reconnected to the healthy parts of
the grid. the third philosophy, fully selective fault clearing,
adopts an approach to dc fault clearing similar to the conven-
tional approach in ac systems.
in the approach followed by ceneLec, three main con-
cepts applied within a certain protection zone are defined
based on the impact of dc faults within a protection zone on
all ac and dc points of connection, i.e., 1) continuous opera-
tion, 2) temporary stop, and 3) permanent stop. for each of
these concepts, a typical “fault separation time,” i.e., the
time at which recovery of the active power flow can start, is
defined. these three main concepts are extended to a total
of five when considering the availability of reactive power
during dc fault separation. for a continuous operation, the
exchange of active power with dc systems (connected at
a c point of connection) and active or reactive power with
ac systems (connected at an ac point of connection) must
remain controllable during the entire fault separation pro-
cess, resulting in fault separation times of a few millisec-
onds. in the temporary stop concept, the disruption of active
ieee power & energy magazine 77
 power exchange with dc systems (connected at a dc point of protection concepts, all of the converter stations adopt the
of connection) or active and reactive power exchange with “permanent stop” concept, as defined by ceneLec. As soon
ac systems (connected at an ac point of connection) must as a fault is detected using the sensors at the converter ter-
remain short such that the ac or dc system at a point of con- minals, a trip order is sent to the ac circuit breakers to inter-
nection does not enter into an abnormal operating mode. As rupt the fault current. When the ac circuit breakers of all ac/
such, the temporary stop concept is expected to have a fault dc converters in the grid have opened, the ac fault-current
separation time of a few hundred milliseconds. for a per- contribution to the dc fault current is eliminated, and the dc
manent stop, the disruption of active or active and reactive fault current and dc voltage passively decay to a value close
power may cause the transmission systems at the point of to zero. thereafter, the faulted component is isolated using
connection to enter an abnormal mode of operation, result- disconnect switches or Hsss connected at each end of the dc
ing in fault separation times of seconds. lines. reenergizing the ac–dc converters in the new postfault
each fault-clearing strategy results in different conse- situation can be achieved by reclosing the ac circuit break-
quences for the dc system and the connected ac systems, ers, restoring the dc voltage to within the normal operation
and, consequently, the functional requirements for HVdc grid range, and restoring the power flow to securely operate the
protection depend on adopted strategy. in the fully selective HVdc grid. this strategy is illustrated in the dc1 system in
philosophy, or in those zones making use of the “continuous- figure  5(a), where the ac circuit breakers of the half-bridge
operation” concept, HVdc grid protection should primarily MMcs are used to interrupt the fault current, and Hsss or
ensure that the dc voltage remains within acceptable limits, disconnect switches are used to isolate the faulted equipment.
which keeps the ac–dc converters connected. in nonselec-
tive strategies, the dc voltage inherently collapses, and HVdc Nonselective With Fault-Blocking Converters
grid protection is mainly required to clear the fault and allow in this strategy, converters with fault-blocking capability are
for later system restoration that does not endanger the secure used to quickly stop the fault current contribution from the
operation of the connected ac system. in this respect, options ac system to the HVdc grid through a joint action by all of
for the protection concepts are either a permanent or a tempo- the converters. consequently, the HVdc grid quickly deen-
rary stop. in partially selective strategies, HVdc grid protec- ergizes, and the fault currents decay to values close to zero.
tion must keep the dc voltage in the healthy zones of the sys- thereafter, the faulted line is isolated under near-zero volt-
tem and not endanger the secure operation of the ac system. age and current conditions with a dc Hss that has residual
therefore, the healthy zones of the system should adopt the dc interruption capability. once the faulted line is isolated,
continuous operation concept. the faulted zone takes either the converters restore the dc voltage and, subsequently, the
the permanent or temporary stop, depending on the constraints power flow. this concept is illustrated in application to the
imposed by the grid codes of the connected ac system(s). dc2 system shown in figure 5(a). Actively restoring the dc
voltage from zero requires using converters with controlled
Fault-Clearing Strategies for HVdc Grids fault-blocking capability. the active voltage restoration can
then occur faster compared with converters without con-
Nonselective With ac Circuit Breakers trolled fault-blocking capability or without fault-blocking
this strategy relies on the ac circuit breakers on the ac side of capability at all. Hence, with a nonselective strategy that
the ac–dc converters to interrupt the dc fault current. in terms uses converters with controlled fault-blocking capabil-
ity, dc voltage recovery can occur
faster compared with the nonse-
ac1 dc1 ac2 ac1 dc1 ac2 ac1 dc1∗ ac2 lective strategy that uses ac cir-
H H H H H H cuit breakers in conjunction with
converters without fault-blocking
H H H H H H capability. With this strategy, the
actions of the converters with con-
dc2 ac3 dc2 ac3 ac3
F F F
trolled fault-blocking capability
F F F
fall under the protection concept
F F F F F F known as temporary stop.
(a) (b) (c)
Partially Selective With
H F Half-/Full-Bridge Converter Circuit Breaker High-Speed Switch HVdc Circuit Breakers or
dc–dc Converters
figure 5. Example fault-clearing strategies in an extended HVdc grid: (a) a nonselec- this strategy is similar to the
tive fault clearing using ac circuit breakers (dc1) or fault-blocking converters (dc2), (b) nonselective fault-clearing strate-
a partially selective fault clearing using HVdc circuit breakers or a dc–dc converter, gies in that fault currents are inter-
and (c) a fully selective fault clearing using HVdc circuit breakers. rupted by deenergizing the HVdc

78 ieee power & energy magazine may/june 2019

 In the faulted zones, any approach belonging to the nonselective
protection philosophies can be used to interrupt the dc
fault current and isolate the faulted component.
grid as well as the faulted component isolated under voltages
and currents close to zero; however, it differs significantly
from those strategies in that only a part of the HVdc grid
is deenergized. these strategies must rely on HVdc circuit
breakers or dc–dc converters placed in between the protec-
tion zones to isolate the faulted zone from the healthy zones.
in the healthy zones, the dc voltage must be kept within the
boundaries of the ac–dc converters to ensure power flow
continuity. in the faulted zones, any approach belonging
to the nonselective protection philosophies can be used to
interrupt the dc fault current and isolate the faulted com-
ponent. After fault clearing, the faulted zone is reenergized
and reconnected to the healthy zones. in figure 5(b), a par-
tially selective fault-clearing strategy is adopted to achieve
connection of systems dc1 and dc2 if an outage of the entire
system, i.e., (dc1 + dc2), is, e.g., unacceptable to the con-
nected ac grid, ac1.
Fully Selective With HVdc Circuit Breakers
for a selective fault-clearing strategy, HVdc circuit breakers
are located at the end of each line to interrupt the fault current
and simultaneously isolate the faulted line [(figure 5(c)]. this
strategy is similar to the conventional approach to ac system
protection, in which all of the converters adopt the continuous
operation concept (as defined by ceneLec), which implies
that the fault is cleared sufficiently quickly to avoid collapse
of the dc voltage. if the symmetric monopole configuration
is used, HVdc grid protection must also rebalance the pole
voltages when there are pole-to-ground faults. for pole rebal-
ancing, either special dc chopper circuits or converters that
enable the injection of zero-sequence ac currents into the dc
side can be used.
Discussion
HVdc grid protection can be designed using a wide variety
of fault-clearing strategies, as described previously. choosing
a fault-clearing strategy is not straightforward and depends
on many factors, such as the desired overall reliability of the
power system, the relative power rating of the HVdc grid
compared to the connected ac system, the cost of fault-clear-
ing equipment, or adaptability of the adopted strategy to sys-
tem expansion. the nonselective strategies only consider the
secure operation of connected ac systems and offer the lowest
cost of all of the components at the dc side. these strategies
require considerable effort to restore the HVdc grid, and grid
restoration times should be adapted to meet the ac system
constraints. the selective strategies consider the protection
may/june 2019
of the HVdc grid itself by keeping the dc voltage within an
acceptable range. these strategies require a higher invest-
ment cost in terms of fault-clearing equipment on the dc side
(i.e., the HVdc circuit breakers at the end of each transmis-
sion line) but may require less effort in restoring the power
flow, when compared with nonselective strategies. the par-
tially selective strategies face a tradeoff in limiting the extent
of the HVdc grid disconnected against the investment in dc-
side fault-clearing equipment. the final choice for a certain
strategy will depend on the investment costs associated with
the required fault-clearing equipment, the probability of each
type of fault, and the desired impact of fault clearing on the
HVdc grid itself and the connected ac systems. it is conceiv-
able that the desired protection philosophy provides a higher
operation speed and fault-clearing selectivity as the grid
grows in size. While each fault-clearing strategy has specific
consequences for designing and operating HVdc grids, it is
plausible that systems using multiple strategies will arise,
e.g., for backup and primary protection or for different sec-
tions of a single grid.
HVdc Grid-Protection Algorithms
the functional requirements for protection algorithms used
to detect faults and identify their location in HVdc grids are
largely the same as for their ac counterparts. requirements
can be set for parameters such as speed, reliability, or sen-
sitivity. the main challenge is to achieve the desired reli-
ability, in terms of security and dependability, within a very
small amount of time. in particular, protection algorithms
must make the correct decisions on whether to trip fault-
clearing equipment within the first milliseconds after fault
detection. these decisions are based on the measured volt-
ages and currents. increased operating time of the protec-
tion algorithm may lead to a more reliable decision but
also to an increased overall fault-clearing time, which
is unwanted given the quickly increasing dc fault cur-
rent. When choosing and tuning protection algorithms,
the HVdc grid topology, converter technology, system
grounding, measurements, and operating points are impor-
tant considerations.
this section focuses on algorithms for primary line pro-
tection in HVdc grids. fault detection for selective fault-
clearing strategies is more challenging than for nonselective
strategies, so the focus is on algorithms for selective fault
detection. However, this does not prevent using the same
algorithms for nonselective fault-clearing strategies. the
protection algorithms can be implemented using principles
ieee power & energy magazine 79
 The main challenge is to achieve the desired reliability,
in terms of security and dependability, within a
very small amount of time.
based only on local measurements or on the communication
of measurements from distinct locations in the grid.
Protection algorithms without communications rely only
on locally measured voltages and currents, as shown in fig-
ure 6(a). these algorithms can be classified as those detect-
ing a fault during a transient and those monitoring a signal
for a certain amount of time.
Many of the proposed algorithms without communica-
tions detect faults during the transient following the fault
event. Among these are voltage-derivative, current-deriva-
tive, combinations of voltage- and current-derivative,
and traveling-wave algorithms. the algorithms determine
whether the fault is inside the protection zone by using the
shape of the wavefront following a fault. Differentiating
wave shapes for internal and external faults requires know-
ing the electrical distance. this electrical distance is typi-
cally driven by the series inductors required for the HVdc
circuit breakers, which act as low-pass filters on the wave-
fronts generated by external faults. remote faults on long
cables are more difficult to detect than closer faults because
of wavefront distortion and attenuation during propagation
on the transmission line. the basic fault-detection meth-
ods based on transient voltages and currents can be further
Sharp Transient, Dampened Transient,
Easy Detection More Difficult Detection
Protection Zone: Cable
(Local)
dc Line
Protection
(a)
Largest Smallest
Communication Delay Communication Delay
Protection Zone: Cable
(Local)
dc Line
Protection
Communication
(b)
figure 6. A comparison of protection algorithms (a) without and (b) with communi-
cation. (Note that the comments refer to local dc line protection.)
80 ieee power & energy magazine
refined using signal processing methods to improve their
function in a noisy environment. communicationless protec-
tion algorithms, based on undervoltage and overcurrent, are
more suited to complement main or backup protection. the
respective undervoltage or overcurrent criteria have a fixed
time requirement. in contrast with the fastest algorithms,
these two criteria do not rely on the shape of the waveforms
during the fault transient.
communication-based protection algorithms combine
information from both line ends to make a tripping deci-
sion and, therefore, require a communication channel to
exchange logic signals or measured quantities from both line
ends, as shown in figure 6(b). the communication chan-
nel must allow for high-speed communication and most
likely would rely on fiber optic cables. the algorithms can
be categorized as those communicating a status (directional
comparison) and actual measurement data (differential). the
protection zones are specific for communication-based pro-
tection algorithms and do not need dc line inductors to sepa-
rate the protection zones.
A directional comparison can be implemented as either a
tripping or a blocking scheme. in both cases, a fault in the for-
ward direction is detected at the local line end, which needs to
be confirmed by information from
the remote line end. in a tripping
scheme, the presence of the fault is
confirmed by a tripping signal from
the remote line end. A tripping
signal is issued by the remote line
end if the measurements at that end
(Remote) indicate a fault in the forward direc-
dc Line
Protection tion, i.e., in the direction of the pro-
tection zone. in a blocking scheme,
the presence of a fault is confirmed
by the absence of a blocking signal
from the remote line end. A block-
ing signal would indicate detection
of a fault in the reverse direction
at the remote end of the protection
(Remote)
dc Line zone. the tripping scheme intro-
Protection duces the largest number of delays
in a directional comparison scheme;
these delays occur at the local end
of the protection zone. While the
fault is perceived almost instanta-
neously by the local measurements,
may/june 2019
 The design of a reliable, effective HVdc grid-protection
scheme depends on both system characteristics and
the strategy used for fault clearing.
the algorithm must wait for signals communicated from the
remote line end to make a tripping decision.
Differential line protection compares the currents or
traveling waves seen at both line ends. the basic principles
of traveling-wave differential protection, originally devel-
oped for ac overhead line systems, may be applied to HVdc
systems. However, the algorithm may need an adaptation
for cable systems because of the larger influence of wave
distortion and attenuation in cable systems. for these algo-
rithms, the delay caused by the wave propagation is sub-
tracted from the communication delay, which favors a high
speed of operation. Because these algorithms rely on the
exchange of measured quantities rather than logical signals,
they impose a larger communication burden compared to
directional algorithms.
communicationless algorithms have the advantage of
fast fault detection due to the absence of communication
delays but face difficulties with detecting remote faults.
By contrast, communication-based algorithms can detect
remote faults while having an acceptable communication
delay. therefore, a combination of communication-based
and communicationless algorithms may be suitable for pro-
tecting long lines.
Summary
the design of a reliable, effective HVdc grid-protection
scheme depends on both system characteristics and the strat-
egy used for fault clearing. system characteristics determine
the requirements for protection by fixing the relationship
between the probability of faults and their impact and by
setting constraints on secure system operation. the strat-
egy used for fault clearing determines the technology to be
used (e.g., whether or not to use HVdc circuit breakers) and
determines the final impact of a fault on the system and its
components. for HVdc grid protection, fault-clearing strat-
egies can be classified into three philosophies: nonselec-
tive, partially selective, and fully selective. the impact of
the fault on an ac or dc system can be determined from the
actions taken by ac–dc converters during fault clearing. in
all cases, the fault-clearing strategies benefit from fast fault
detection and, in the case of a fully selective strategy, fast
fault identification. the overview of protection algorithms
given here shows that these algorithms operate mainly on
traveling waves and may use a scheme without or with com-
munication. the algorithms not using communication must
may/june 2019
rely on an impedance between the protection zones, which,
for HVdc grids, may be provided by an inductor in series
with the HVdc circuit breaker. for algorithms using commu-
nication, the algorithms may employ a traveling-wave differ-
ential or directional-comparison protection scheme and must
rely on a fast communication channel, such as a dedicated
fiber optic cable.
Acknowledgments
the authors acknowledge contributions from the members
of Work Package 4 in the european Horizon 2020 project
Progress on Meshed HVDc offshore transmission net-
works (ProMotion). this work was funded by ProMo-
tion under grant 691714.
For Further Reading
cigre, “Local control and protection of HVDc grids,”
cigre, Paris, techn. Brochure 739, tech. rep. JWg B4/
B5.59, 2017.
ProMotion consortium, “D4.2-Broad comparison of
fault clearing strategies for Dc grids,” oct. 2017. [online].
Available: https://www.promotion-offshore.net/fileadmin/
PDfs/D4.2_Broad_comparison_of_fault_clearing_strate-
gies_for_Dc_grids.pdf
D. Van Hertem, o. gomis-Bellmunt, and J. Liang, eds.
HVDC Grids for Offshore and Supergrid of the Future.
Hoboken, nJ: Wiley, 2016.
i. Jahn, n. Johannesson, and s. norrga, “survey of meth-
ods for selective Dc fault detection in MtDc grids,” in
Proc. 13th IET Int. Conf. AC and DC (ACDC) Power Trans-
mission, 2017.
Biographies
Willem Leterme is with energyVille, the university of Leu-
ven (Katholieke universiteit Leuven), Belgium.
Ilka Jahn is with the royal institute of technology,
stockholm, sweden.
Philipp Ruffing is with rheinisch-Westfälische tech-
nische Hochschule, Aachen, germany.
Kamran Sharifabadi is with equinor, stavanger,
norway.
Dirk Van Hertem is with energyVille, the university of
Leuven (Katholieke universiteit Leuven), Belgium.
p&e
ieee power & energy magazine 81
 By Dragan Jovcic, Guangfu Tang, and Hui Pang

Adopting
Circuit Breakers
for High-Voltage
dc Networks

Digital Object Identifier 10.1109/MPE.2019.2897408
Date of publication: 17 April 2019
S
Since the late 19th century, ac power tranSmiSSion
systems have been widely used for interconnecting multiple power-pro-
ducing plants with load centers. highly meshed ac transmission grids
with high redundancy and operating flexibility now operate worldwide.
these complex ac systems have proven reliable in securely providing
high levels of power, while also accommodating multiple voltage levels
to reduce loss.
two key components essential for transmission systems, besides
the lines themselves, are circuit breakers (cBs) and transformers. Such
units for ac transmission are relatively simple, reliable, and efficient.
cBs enable operators to open a circuit (line) under load and fault cur-
rents, which is essential for regular maintenance, system expansion, and
the physical integrity of large transmission systems. transformers make
it possible to change voltage levels, which is necessary for maintain-
ing safe and reliable systems, but also for reducing losses with long-
distance transmission.
however, ac transmission has many shortcomings, which have been
publicly debated since the war of the currents between tesla and west-
inghouse (who favored ac) and edison (who favored dc). with ac trans-
mission, reactive current transfers occur, which increase losses and
cause operating difficulties. also, for underwater cable transmission at
distances of more than 100 km, reactive power issues become severe
due to excessive line charging, making ac transmission in such applica-
tions impractical.
high-voltage dc (hVdc) transmission has been used since 1954. thanks
to important technological advances, its use has taken off in the last 20 years.

82 ieee power & energy magazine 1540-7977/19©2019IEEE may/june 2019

 Appropriating the Vast Advantages
of dc Transmission Grids

those advances include high-power electronic devices, such as thyristors and, more recently, transis-
tors, which have enabled the development of more compact voltage source converters (VScs). the vast
majority of hVdc systems consist of a single line or cable and use one ac–dc converter at each end. the
direct connection of multiple dc lines in networks has not been possible, primarily because suitable dc
cBs have not been available. nevertheless, in the last five years, two multiterminal VSc hVdc systems
have been built in china, and others are being designed and developed.
Dc transmission grids, a substantial technological advance over point-to-point hVdc connections,
are being considered for many applications worldwide. in europe, researchers involved in several major
projects have studied the use of a cable dc grid in the north Sea for integrating large amounts of offshore
wind energy and strengthening interconnections among european union countries. in china, plans have
been made for building a four-terminal dc grid with overhead lines.
For dc grids to work, the market has to offer dc cBs, preferably from multiple vendors. these cBs
must provide acceptable performance and reliability, ensure low loss, and be available in practical
sizes at a reasonable cost. much has been learned from long experience with ac cB technologies,
which are standardized at high voltage and very high current ratings. however, they are not suitable
for dc applications, and the technologies for dc cBs will be substantially different from those for ac
cBs. the performance requirements for dc cBs in dc grids are more stringent. also, opening a dc
circuit is technically much more challenging than opening an ac circuit.
over the last five to 10 years, some major advances in hVdc cB tech-
nologies have been introduced. multiple dc cB topologies have been
demonstrated in laboratories and brought to market. more recently,
these have been installed for the first time.

©istockphoto.com/smartboy10
Basic Performance
Requirements for dc CBs

The Role of dc CBs

the critical functions of dc cBs include
controlled opening and closing of circuits
and, in case of faults, isolating faulted
dc lines. these functions are neces-
sary for the dc network to continue
operating while a line is under a
fault. if the faulted line were

may/june 2019 ieee power & energy magazine 83

 not isolated, grid voltage would be low, and currents would
be high under the fault condition. high current would likely
lead to component damage, while low voltage would reduce
power-transfer capability. Fast dc fault isolation is particu-
larly significant for large dc grids, which may have numer-
ous converter stations and transmission lines. the converters
being considered for the european north Sea dc grid are rated
at 1.5–2 Gw, while the total dc grid capacity may exceed
150 Gw in some scenarios.
Figure 1 shows a single-line diagram of a five-
node ±400-kV dc grid, consisting of four ac–dc bipole
converters rated 800 mVa, 2 ka per pole interconnecting
the dc grid (buses 1–4) with corresponding ac systems, one
dc bus without a converter (node 5), eight dc cables, and
16 dc cBs per pole (32 dc cBs in total). the grid topology
has been selected to illustrate how power can be transferred
securely while limiting the number of dc cBs. in comparable
ac systems, the number of ac cBs would be much higher,
according to modern grid development practices. the indi-
cated power and voltage levels are in the highest range of
demonstrated dc cB technology.
For example, a short circuit fault occurring on the line
linking buses 3 and 4 would insert a low impedance at the
point of fault and result in high currents flowing into the fault
from all parts of the network (as indicated by red arrows).
Depending on the impedances of the grid cables and on the
severity of the fault itself, the fault currents would have dif-
ferent magnitudes, but they can easily reach 10–30 times
the rated current (say 20–60 ka) at some points. these are
destructive currents that grid components cannot withstand.
all high-power ac–dc converters have self-protection
control logic, which blocks and disconnects converters in the
event of disturbances. typically, self-protection thresholds
are set at approximately 2 per unit (p.u.) of current and 0.8 p.u. of
1 2
1.6 GW 2 kA
CB12_1
CB15_1
ac System 1
CB14_1
±400 kV CB
15_
5 25_
_5
45
4 CB
1.6 GW
2 kA
CB14_4
ac System 4 CB45_4
CB34_4
±400 kV
figure 1. A diagram showing a five-node dc grid with four ac–dc converters, eight dc cables, and 16 dc CBs per pole.
84 ieee power & energy magazine
dc voltage, which, for the system under consideration, implies
current of 4 ka and voltage of 320 kV.
in conventional ac transmission systems, generators have
many self-protection controls, but the thresholds are signifi-
cantly higher relative to the nameplate-rated variables. the
line/cable impedances also have much higher values in ac than
in dc systems, because the frequency, f, contributes to the reac-
tance (the imaginary part of the impedance, 2rfL, where L is
the inductance). these factors lead to the conclusion that, in a
dc system, cBs (and protection systems) must respond much
faster than in ac systems to prevent widespread loss in capacity
(converter blocking) and destructive fault current magnitudes.
Operating Speed
in a dc fault situation, current rises and voltage falls throughout
the dc grid during the delays associated with dc protection and
dc cB operation. once the dc cBs open and isolate the dc fault,
the dc voltage recovers, and dc power flows over the recon-
figured grid at a newly established balanced operating point,
which depends on the number of remaining operating convert-
ers and their control strategies. the faster the dc cBs operate,
the less likely it is that large parts of the dc grid will be lost.
however, the blocking of nearby converters cannot be
completely avoided, even with application of the fastest avail-
able dc cBs. this is because the converters closest to the fault
point may experience high current and thus block before the
dc cBs open. the most recent dc grid studies indicate that
temporary converter blocking for 10–30 ms is possible until
the dc fault is cleared. then the converters can be unblocked
to resume normal operation. the need for blocking convert-
ers will depend on the grid topology and the fault location.
many studies indicate that dc grids of various sizes and
topologies can recover from any dc line fault if the total pro-
tection time (dc cB opening time plus protection operation
2 kA 1.6 GW
CB12_2
CB25_2
ac System 2
CB23_2
5
dc–ac
5 ±400 kV
Converter
CB
CB
35_
5 3
dc CB
400 kV
CB23_3 2 kA 1.6 GW
CB35_3
CB34_3 ac System 3
Fault
±400 kV
may/june 2019
 Because operators have so little experience with dc CBs, more
study is needed about the internal complexity of dc CBs and
their failure modes to fully understand their reliability.
time) is within approximately 10 ms. this assumes that some
converters may block by self-protection temporarily or per-
manently, depending on the grid topology and the strategy
used for protecting the network. permanent blocking after
fault clearing occurs if the converter becomes isolated. this
results in grid capacity reduction.
Ratings
in ac systems, ac cBs are rated for the largest available fault
level at their point of connection (the worst case steady-state
fault current). Dc cB prototypes recently brought to market
by manufacturers are designed for only 15–20-ka peak inter-
rupting current. these state-of-the-art dc cBs have such low
ratings because of costs and technological limitations. the
ratings are, for the foreseeable future, lower than expected
fault levels at many dc buses. thus, dc cBs must operate
fast, before dc fault current exceeds their ratings and before
it reaches destructive levels causing many converters to block
and leading to wider dc grid collapse.
Bidirectional Operation
in many applications, dc cBs will be expected to interrupt
fault current in one direction only. Depending on how pro-
tection zones and protection systems are configured, some
dc cBs may be expected to operate in both directions. also,
bidirectional operation may be beneficial as a backup pro-
tection function. application of a unidirectional dc cB may
suffice in particular projects and result in substantial cost
and size savings for some dc cB topologies.
Fast, Multiple Open/Close Operations
Dc grids with overhead lines can be exposed to frequent
faults, many of which will be transient. as with overhead ac
transmission, reclosing with multiple cB operations within
200–500 ms may reasonably be expected. with underground
or underwater dc cable systems, however, almost all faults
are permanent, and reclosing may not be included in the pro-
tection logic. the number of expected operations and timing
of dc cB duty cycles may result in cost and size implications
for some dc cB topologies.
Losses, Size, and Weight
most of the time, dc cBs operate in a closed state and con-
duct load current. losses in closed-state dc cBs cost grid
operators revenue. also, dc cBs with high losses may require
additional heat-removal equipment. Dc cBs will be notice-
ably larger and heavier than ac cBs. in europe, dc grid
may/june 2019
studies of offshore environments show that the extra size and
weight of dc cBs could add significantly to platform costs.
Cost and Reliability
the widespread application of ac cBs to ac transmission sys-
tems ensures a secure power supply and enables operational
flexibility. the simple dc system in Figure 1 may not provide
the same level of security and flexibility as in ac systems but,
nevertheless, requires 32 dc cBs. the example shows that
future grids are likely to require a considerable number of dc
cBs. these will cost more than ac cBs, and for some dc cB
topologies the costs may be substantially higher.
Because operators have so little experience with dc cBs,
more study is needed about the internal complexity of dc cBs
and their failure modes to fully understand their reliability.
Some dc cBs using semiconductor technologies have self-pro-
tection as discussed later in this article. Such features need to be
considered by grid planners and protection system developers.
Standardization of Inputs/Outputs
in large dc grids, the protection system will consist of multiple
relays (microcontroller based) and dc cBs that are intercon-
nected and use adjusted settings controlled by the grid operator.
in the development of protection logic, grid topology changes
and expandability must be considered as important factors.
achieving the desired operation of protection system with com-
ponents made by different vendors requires that interconnec-
tions be standardized and that components be interoperable.
Challenges With dc Circuit Opening
DC Current Commutation
the first developers of dc systems learned that interrupting
dc current is difficult. trying to separate cB contacts under
a current flow creates an arc, which for high-voltage systems
is self-sustaining and generates large amounts of heat caused
by the arc resistance. with ac currents, there is a natural cur-
rent-zero crossing twice every cycle (every 10 ms in 50-hz
systems). this momentary current interruption—along with
engineering devices that ensure sufficient contact separation
and an arc-extinguishing chamber—enables modern ac cBs
to reliably interrupt very large currents within 20–60 ms.
there are no natural current-zero crossings in dc systems.
at low voltages of 10–30 V (automotive applications), the arc
voltage is larger than the system voltage, and this reduces the
current to zero. at higher voltages, there are two principal
methods to interrupt dc current.
ieee power & energy magazine 85
 1) the first uses power electronics to commutate current and dc cB operating times, this inductor limits the peak
into a parallel circuit. with this method, semiconductors fault current experienced by dc cBs and other components
provide solid insulation and can interrupt large currents in the dc grid. For example, a 400-kV dc bus with a desired
very quickly and without arcing. the principal challeng- interrupt time (dc cB opening and protection time) of 4 ms
es with this method are large component costs and high and a maximum peak interrupting current of 16 ka requires
losses during normal closed-state operation. an inductor of around 100 mh. Series inductors with dc cBs
2) the second uses additional circuits, such as commonly also enable the protection system to differentiate between
employed inductor–capacitor (lc) resonant systems, to faults in inner and outer protective zones. Because of these
create current-zero crossings in mechanical switches. multiple requirements for the series inductor, it is not clear
mechanical switches allow almost no closed-state losses whether the series inductor will be considered part of future
and are generally lightweight, compact, and inexpensive dc cB units or part of the dc protection system.
at high-voltage ratings. the main challenge is arcing, on the negative side, series inductors store energy and
which suggests the need for heavier and more complex make voltage control of dc grids less precise. in addition, in
contacts, leading in turn to slower opening speeds. the event of fault interruption, the energy stored in the induc-
Both methods are viable and have been demonstrated on hV tor must be dissipated by dc cBs.
and large-current dc cB units.
Main dc CB Topologies
DC Current Reduction and Dissipation of Energy Several manufacturers have invested substantially in further
the energy dissipation in a device is the product of the cur- developing dc cBs to voltage and current levels suitable for the
rent and voltage across the device, multiplied by time. with latest dc transmission technology. these manufacturers have
ac cBs, dissipation of energy is minimal because the con- reported results from laboratory tests of dc cB prototypes
tacts open when the load current crosses zero. any energy in the range of 40–80 kV with peak fault current interruption
in the system is conveniently dissipated in the arc chamber. capabilities of approximately 15–20 ka. plans call for these
units to become standard modules through series connections to
with dc cBs, nonlinear resistors (surge arresters) are gener-
ally used to provide counter dc voltage under the dc load/fault achieve higher dc voltage levels. reports from manufacturers
current at the moment of interruption. these resistors, an inte-and data from recent installations indicate that there are no
substantial obstacles to implementing 300–400-kV dc cBs.
gral part of all dc cBs, are also called energy absorbers because
they absorb the energy from the line inductances, which is a the european union has directed substantial resources
consequence of simultaneous high voltage and high current for a into research on dc cBs, including tests of several full-scale
(70–80 kV) dc cBs at an independent laboratory. there is a
short period until the dc current is brought to zero. the expected
joint effort by multiple manufacturers, grid operators, academ-
energy dissipation at a 400-kV dc voltage level is on the order of
10–100 mJ, which affects the thermal ratings and time constants ics, and consultants that is helping the industry better under-
of the absorbers. this is especially important if repeated dc cBstand dc cBs and how best to control and test them, know what
operations are required. this degree of energy dissipation also their operating limits and failure modes are, and determine
influences the cost, size, and weight of the dc cB. how they interact with dc grid components. considerable effort
is also being directed toward achieving interoperability among
Series Inductor With dc CBs various dc cB technologies and products from different manu-
each dc cB in hVdc grids requires a series inductor on the facturers and also toward initiating standardization.
order of 50–300 mh. although this inductor cannot reduce many different dc cB topologies have been pro-
the amplitude of the dc fault current, it can limit the fault posed, with some prototypes of varying ratings under-
current’s rate of rise. Sized for a given level of protection going field tests. many technologies related to dc cBs have
recently been patented, and inten-
sive research will continue in this
Residual Breaker Main Branch field. it is possible to group most
Idc Ldc VI2 IVI1 VI1 dc Cable
dc cB designs into two main
dc
dc
Load
families: 1) mechanical dc cBs
Source V IS3 VI3a
+ dcCB I LC Fault + and 2) hybrid dc cBs that use
Vdc C1a R
– 1a – semiconductor valves.
Current L VI3b
1
Injection C1b R1b Mechanical dc CBs
Branches

Energy Topology
Absorber SA Figure 2 shows a typical topology
of a mechanical hVdc cB and
figure 2. A diagram showing a mechanical dc CB. illustrates the following.

86 ieee power & energy magazine may/june 2019

 1) the main branch consists of a breaker that is able to
sustain arcing. this is usually an ac vacuum inter-
rupter (Vi) with an enhanced driving mechanism to
reduce opening time.
2) the current injection branches include precharged ca-
pacitors. this topology shows two branches. this design
enables two operations in a short time, which may be
essential considering the time it may take for capaci-
tors to charge. the second branch is not needed if the
required breaker duty cycle consists of only a single
opening. Switches Vi3a and Vi3b are similar to Vi1.
the natural resonant frequency of the lc circuit can
be increased to practical values in the range of 2–3 khz,
which would reduce the required sizes of L 1 and C 1 .
the resistors R 1 are required for capacitor charging.
3) the energy absorber consists of banks of surge arrest-
ers. these arresters are based on standard surge arresters
widely used in ac systems for overvoltage protection.
4) the residual breaker, Vi2, interrupts only a small current.
this device does not need to have high opening speed.
5) the current-limiting inductor L dc reduces the rate of
the fault current rise.
the mechanical dc cB presented is also called a current
injection dc CB, because it employs precharged capacitors
C 1a and C 1b, which substantially increase the speed of
each operation.
Operating Principles
Figure 3 illustrates an interruption of a 15-ka peak dc fault
current with a mechanical dc cB on a 400-kV dc system,

18 18
16 Arcing Time 16
14 14
Idc
12 12
ILC
Current (kA)
Current (kA)

10 10
IVI1
8 8
VI2 Fully
6 6 Open
IS3
4 Trip 4 VI3 Begins
Signal VI1 Begins to Open VI3 Fully
to Close
2 VI1 Fully Open 2 Closed
0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Time (ms) Time (ms)
(a) (b)
100
600
VdcCB
400 80
Vdc
Energy (MJ)
Voltage (kV)

200 60
ESA
0 40

Fault Current Suppression Time

–200 20

–400
0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Time (ms) Time (ms)
(c) (d)

figure 3. The four graphs show the interruption of a 15-kA peak dc fault current with a mechanical dc CB on a 400-kV
dc system, assuming a dc fault at t = 0 s. (a) Contacts require about 8 ms to fully separate. (b) The fault current is fully
extinguished after an additional 20 ms. (c) The graph indicates the arrester voltage level. (d) Energy dissipation by the ar-
resters (E SA) is about 83 MJ.

may/june 2019 ieee power & energy magazine 87

 assuming a dc fault at t = 0 s. a 240-mh series inductor (L dc)
is used. Figure 3(a) shows that it takes approximately 8 ms for
the Vi1 contacts to fully separate. at that instant, Vi3 closes,
injecting an oscillating current, I S3, which facilitates a zero
crossing of I VI1 . when the arc is interrupted in Vi1, the dc
current I dc is transferred to the energy absorber, and voltage
VdcCB abruptly rises to the clipping voltage of the surge arrest-
ers, which is commonly around 1.5 p.u. (600 kV). this high
voltage brings a negative voltage across inductor L dc (400–
600 kV), enabling fault current suppression. the fault current
is fully extinguished after an additional 20 ms [Figure 3(b)].
manufacturers commonly specify the time to voltage recov-
ery (8 ms in Figure 3) as the key indicator of dc cB performance.
the peak fault current is specified as the single indicator of inter-
rupting capability (15 ka at 8 ms), which is a key parameter for
sizing the series inductor L dc . Sizing could vary for different dc
cB technologies because of different opening speeds.
the expected energy dissipation by the arresters in this
case (E SA) is around 83 mJ, as seen in Figure 3(d). Figure 3
illustrates fault current interruption in the positive direction,
but this dc cB topology can similarly interrupt dc fault current
in the negative direction. however, because the charge of capac-
itor C 1 is unipolar, the responses in the negative direction will
be slightly different, and component stresses might be higher.
Hybrid dc CB
Topology
Figure 4 shows a representative topology for a hybrid dc cB,
which consists of the following.
1) included is a main branch that has two 400-kV semi-
conductor valves, T2A and T2B (one for each direc-
tion). For protection in only one direction, just one
valve wou ld b e ne e de d. e a ch valve is similar to
one arm of a (six-arm) VSc hVdc converter. this
branch can conduct load current in a closed state, but
losses would be high with prolonged operation.
IT1
Residual
Breaker
Idc Ldc IT2
dc
Source S2
+ Main Branch
Vdc VdcCB
– –
figure 4. A diagram of a hybrid dc CB.
88 ieee power & energy magazine
2) an auxiliary branch has low resistance and conducts full
load current in the closed state. this branch includes
✔ a low-voltage (lV) semiconductor valve T1, also
called the load commutation switch, which should
have a voltage rating comparable to the closed-
state voltage drop of the valve T2 (i.e., on the order
of 10 kV)
✔ an ultrafast disconnect switch S 1, which cannot sup-
port arcing (it can open only at zero current) but has
an extremely fast opening speed (significant tech-
nological advances have been made recently using
thomson coil drivers, and the opening speed for
320-kV units has been demonstrated as approxi-
mately 2 ms)
3) energy absorber Sa, which consists of banks of surge
arresters
4) residual breaker S 2, which interrupts only the arrester
leakage current and whose opening speed is not critical.
L dc is a current-limiting inductor.
the lV valve t1 continuously conducts load current, and it
must withstand fault current in the period before the protec-
tion relay sends the trip signal. in practice, this valve consists
of several parallel branches, with each branch having a few
insulated gate bipolar transistors (iGBts) in series. Because
of the continuous conduction stress, this valve requires a
forced-liquid cooling system. the main valve T2 conducts
the fault current for only around 2 ms. therefore, it does
not need liquid cooling. however, the whole hybrid dc cB
should be located indoors (in a valve hall) similar to any
large converter.
Operating Principles
Figure 5 illustrates a dc fault on a 400-kV dc system inter-
rupted by a hybrid dc cB that uses a 76-mh series inductor
(L dc). the fault occurs at t = 0 s , and the current peaks at
15 ka. Figure 5(b) shows that S 1 begins to open when the
current is transferred to T2 and I T1 drops to zero. it takes
S1 T1
Auxiliary Branch
dc Cable
T2A T2B dc
Load
Fault +
SA
Energy Absorber
may/june 2019
2 ms for S 1 to fully open while T2 is conducting. once S 1
is open, the current is commutated to the energy absorber by
opening T2 . as the current commutates to surge arresters,
their voltage rises to 1.5 p.u., which initiates current suppres-
sion. it takes another 5 ms for the fault current to fall to zero,
as seen in Figure 5(a). Figure 5(d) shows that the expected total
energy dissipation (in this case, E SA) equals 21 mJ.
Self-Protection
as with all high-power converters, the semiconductor valves
in hybrid dc cBs open in the event of excessive current. this
is a self-protective feature. although unlikely, a failure in
some grid protection components (relay, sensor, communica-
tion, and others) could fail to send a trip signal to the dc cB
in a dc fault situation, which would result in a rising high
dc cB current without a dc cB trip command. in practical
terms, this means that the dc cB may open on its own to
protect itself at some dc current threshold (which is much
higher than dc grid protection settings). if the dc grid protec-
tion were to fail for whatever reason, the hybrid dc cB should
open the circuit at its self-protection threshold as the last line
of defense.
Fault Current Limiting
hybrid dc cBs can also act as dc fault current limiters. By
employing the on/oFF control of the individual modules
in valve T2, the number of inserted arresters can be varied,
which can regulate current. this operating mode is limited
by the thermal capability of energy absorbers and can most
likely be achieved for only short periods of time.
this fault-limiting operating mode can be beneficial in
some applications. it can help, for example, in
1) limiting fault currents, which allows additional response
time for fault clearing by other dc cBs installed on oth-
er lines in a dc grid (in practical terms, the dc cBs or
protection systems on the other lines may have slower
response times)
2) limiting inrush currents when connecting dc lines,
which eliminates the need for preinsertion resistors on
the dc cBs.

16 16
S1 Opening
14 14 Time
12 12
Idc
10 10 IT2
Current (kA)
Current (kA)

8 8

6 6
IT1 Trip Signal
4 4
S1 Begins to Open
2 2
S1 Fully Open
0 0
0 2 4 6 8 10 0 2 4 6 8 10
Time (ms) Time (ms)
(a) (b)
800 25
VdcCB

600
20

400
Voltage (kV)

Energy (MJ)

Vdc 15
200
ESA
10
0
Fault Current
Suppression Time 5
–200

–400 0
0 2 4 6 8 10 0 2 4 6 8 10
Time (ms) Time (ms)
(c) (d)

figure 5. The four graphs show a dc fault on a 400-kV dc system interrupted by a hybrid dc CB that uses a 76-mH series induc-
tor (L dc). (a) The fault occurs at t = 0 s, and the current peaks at 15 kA. (b) S 1 begins to open when the current is transferred to
T2 . (c) The graph indicates the arrester voltage level. (d) The expected total energy dissipation (E SA) equals 21 MJ.

may/june 2019 ieee power & energy magazine 89

 In 2016, a hybrid dc CB rated 200 kV was installed at the
Dinghai Station on the Dinghai-to-Daishan line, which made
the system more reliable and easier to control.
DC CBs Installed in China
Installed Hybrid dc CB in the Zhoushan
Multiterminal System
among the islands of the Zhoushan archipelago, the largest
power-consuming islands include the main island of Zhoushan, fol-
lowed by the islands of Daishan, Qushan, yangshan, and Sijiao.
the annual load of the Zhoushan power grid reached 780 mw in
2010, and is estimated to reach 2,000 mw in 2020 as a result
of higher demand stemming from economic development.
a ±200-kV five-terminal VSc hVdc project, built by the
State Grid corporation of china, began commercial operation
in 2014. it transmits power from the mainland to the five islands.
this project uses a modular multilevel converter and connects
by submarine cables. During normal operation, the Dinghai
converter station on Zhoushan island provides dc voltage con-
trol, and the other stations operate in power control modes. as
a result, the power flows from the Dinghai Station to the other
stations. Due to the lack of hVdc cBs, clearing a dc fault was
originally achieved by slow-acting operations of ac system
breakers. the lack of rapid fault isolation and the inability to
switch circuits caused several technical problems, indicating a
need for improved system security, greater operating flexibility,
higher reliability, and faster system restoration capabilities.
in 2016, a hybrid dc cB rated 200 kV was installed at
the Dinghai Station on the Dinghai-to-Daishan line, which
Yangshan Station Sijiao Station
100 MW 100 MW
32 km
39 km
dc
Breaker
46 km 17 km
400 MW
Dinghai Station
300 MW 100 MW
Daishan Station Qushan Station
figure 6. A diagram showing the Zhoushan five-terminal
HVdc transmission system with the installed hybrid dc CB.
90 ieee power & energy magazine
made the system more reliable and easier to control (Figure 6).
Figure 7 shows the topology and a photo of a prototype for 200-kV
hybrid hVdc cB. the three branches consist of the following:
1) the auxiliary branch carries nominal current and con-
tains a series that combines an ultrafast mechanical dis-
connector (uFD) with an auxiliary commutation switch,
which consists of a full-bridge submodule (FBSm) using
high-power-rating iGBts. the FBSms appearing in the
auxiliary branch and main branch adopt the same full-
bridge structure. this consists of four press-pack
iGBts and a snubber circuit. the auxiliary branch should
have as few FBSms as possible to reduce the on-state
impedance. therefore, a 2 × 3 FBSm matrix design is
used, featuring low conduction losses and high reliabil-
ity. to make maintenance easier, each FBSm uses an
integrated modular unit connected by an inductance-less
busbar. a water-cooling system is needed because the
auxiliary branch conducts the long-term load current.
Because the uFD is opened with zero current, the con-
tacts inside take up less space, and its driving mechanism
is lightweight. the compact design is achieved by using a
Vi with a voltage rating of 40.5 kV and current rating of
2 ka rather than an SF6 cB. the maximum separation
time of the breaker must be within 2 ms to meet the fault-
clearance time requirements of the Zhoushan project. as
a result, a series connection of six Vis is adopted, and
each Vi is equipped with voltage-sharing devices.
2) the main branch interrupts fault currents and consists of
four series-connected modular units. each modular unit
contains 36 FBSms in a series connection designed for
50-kV and 15-ka breaking capability. a cooling system
is not necessary for the main branch because it only ex-
periences the fault current for several milliseconds. the
FBSm topology provides a bidirectional current path and
allows the hybrid cB to interrupt the fault current in either
direction. in addition, the capacitors in the FBSm enable
soft turn-off of the iGBts, significantly decreasing the in-
duced voltage stress and balancing the voltage distribution
across the modular unit during fault interruption.
3) the energy absorber branch is made up of surge arresters
that limit transient interruption voltages (tiVs) and dis-
sipate the magnetic energy stored in the system.
the installed hybrid hVdc cB offers the following ad-
vantages:
✔ a small footprint and low height, resulting from the
integrated and compact design (installed at a 200-kV
high-potential platform)
may/june 2019
 ✔ low power losses during normal operation, resulting
from use of fewer power semiconductor devices in the
main branch (six FBSms form a 2 × 3 matrix, cooling
system, and power adapter)
✔ use of a lightweight contact system (six Vis in series;
40.5 kV because each uFD separates without experi-
encing current)
✔ modular design (four modular units, 50 kV each; 36
FBSms in each unit), which facilitates maintenance
and replacement of parts
✔ high reliability resulting from exposure to lower volt-
ages due to a series design that distributes voltages
during switching.
Because no relevant international electrotechnical com-
mission (iec) standards currently exist, conformance tests
had to be performed to verify the hVdc cB design. these
were developed based on the operating principles and condi-
tions of the hVdc breaker as well as the present iec stan-
dard for testing VSc hVdc valves. the test was divided into
two parts: the insulation test and the operating test.
the insulation test verifies the voltage-withstanding capac-
ity and the voltage-sharing characteristic of an hVdc cB. the
operating test checks the capacity of each core component to
withstand serious stress and the operational properties under
different circumstances. table 1 shows some of the main test-
ing considerations.
the 200-kV hybrid dc cB passed the test in which a
15-ka short circuit current was cleared within 3 ms and the
transient interruption voltage across the test object exceed-
ed 320 kV.

Auxiliary Branch
iT1

Ultrafast Auxiliary
Disconnector FBSM Branch

Main Branch Energy

iT2 Absorber
Main
Branch
FBSM FBSM FBSM FBSM FBSM FBSM FBSM FBSM
Modular Unit Modular Unit Modular Unit Modular Unit
iabsorb

Energy Absorber

uDB
(a) (b)

figure 7. (a) The topology and (b) a prototype of a 200-kV hybrid HVdc CB.

table 1. Factors related to testing the insulation and operating characteristics of HVdc CBs.
Number Test Test Object Items Objective
1 Insulation test Phases to ground DC voltage withstand test To verify the voltage-withstanding
capacity and partial-discharge level of
2 Switching impulse withstand test support structure
3 Lightning impulse withstand test
4 Across open DC voltage withstand test To verify the voltage-withstanding
contacts capacity across open contacts
5 Switching impulse withstand test
6 Operating test Key components Peak withstand current test To verify the current-withstanding
capacitor of the CB
7 Overload withstand current test
8 Short-time withstand current test
9 Short-time withstand current test
10 Overall prototype Rated current-interruption test To verify the current-interruption
capacity and interruption time
11 Short circuit current-interruption test of the overall prototype and the
12 Rated current-making test program correctness of the control
and protection system
13 Short circuit-making test

may/june 2019 ieee power & energy magazine 91

 Installed Mechanical dc CB in Nan’ao
Multiterminal System
the nan’ao multiterminal hVdc project, with a ±160-kV-
rated dc voltage, was completed by the china Southern power
Grid and put into operation at the end of 2013. this proj-
ect includes three converter stations: Sucheng, Jinniu, and
Qing’ao. the project collects wind power supplied to Jinniu
and Qing’ao Stations and sends it to the mainland. the proj-
ect originally adopted the same method from the Zhoushan
project of using ac breakers to interrupt dc faults. with this
method, the lowest possible power losses were achieved, and
capital expenses were kept low. however, the project suffered
from slow dc fault-clearing times. the project was upgraded
in 2017, with the installation of three mechanical dc cBs
rated 160 kV (Figure 8).
Figure 9 illustrates the topology and shows a photo of the
installed prototype 160-kV mechanical hVdc cB. it consists
of the following elements.
Sucheng Station Jinniu Station
200 MW 100 MW
Cable
dc
Breaker
Overhead Line
50 MW
Qingao Station
figure 8. A diagram showing the Nan’ao three-terminal HVdc transmission system
with installed mechanical dc CB.
1) the main branch is made up of four 40-kV-rated dc
voltage Vis connected in series, which satisfies re-
quirements for meeting system voltage levels and fast
response times. adequate voltage sharing during dy-
namic events and static state requirements are achieved
by arming each Vi with a transient grading capacitor
cj, damping resistor rj, and static voltage-sharing re-
sistor R X .
2) the current injection branch is divided into a hV side
and a lV side. these are isolated electrically through
the use of an air-core transformer with mutual induc-
tance M, where the high coupling factor k allows for
high energy-transfer efficiencies. the air-core design
provides the most efficient method because it accom-
modates high power transfer at high frequency.
3) the energy absorber consists of banks of surge arrest-
ers directly connected to the series connection of the
four Vis.
Functioning under an operating
principle similar to that described
in the section “mechanical dc cB,”
the nan’ao mechanical hVdc cBs
use a combination of inductors and
precharged capacitors to produce
the current zero through the Vis.
the difference is that the storage
capacitor in this scheme is charged
on the lV side of the cB rather
than at the hV side. this techni-
cal solution charges the storage
capacitor C 1 to the target voltage
by introducing an uninterruptible
power system (upS) boost circuit.
the upS voltage is boosted by the
isolation transformer Tc and then
rectified into a dc voltage through

Energy Absorber
MOV

VI1 VI2 VI3 VI4

Main Branch Cj Rj Cj Rj Cj Rj Cj Rj
Rx Rx Rx Rx Main Current
Branch Injection
C2 L2 HV Side Branches
Current M
Injection L1 LV Side Energy
UPS

Branches Rc Dc Absorber
C1
Tc SCR D

(a) (b)

figure 9. (a) The topology and (b) a prototype of a 160-kV mechanical HVdc CB.

92 ieee power & energy magazine may/june 2019

 converter station is the receiv-
ing terminal used to supply re-
Fengning
Kangbao dc dc newable power for Beijing.
1,500 MW 1,500 MW
Breaker Breaker the Zhangbei project adopts
ac a half-bridge modular multilevel
WF 227 km 500 kV
technology and uses overhead
dc dc
Breaker Breaker lines. as a result, dc breakers
must be installed at both ends
66 km 126 km of the line to clear potential dc
faults, such as those induced by
Zhangbei dc Beijing
dc lightning. a 500-kV dc breaker
3,000 MW Breaker 3,000 MW
Breaker
prototype was developed in 2018,
ac
219 km and the project is scheduled for
500 kV
dc dc operation in 2020.
Breaker Breaker
Acknowledgment
the work for this article was
ac ac funded in part by the national
500 kV 1,000 kV
Key r&D program of china
(2017yFB0902400).
figure 10. A diagram of the Zhangbei four-terminal HVdc grid with dc CBs.
WF: wind farm. For Further Reading
D. Jovcic and K. ahmed, High
Voltage Direct Current Trans-
the hV silicon rectifier D c, where the resistor R c is used to mission: Converters, Systems and DC Grids. hoboken, nJ:
limit the inrush current during the charging process. wiley, 2015.
compa red to convent iona l mechanical dc cBs, this J. häfner and B. Jacobson, “proactive hybrid hVDc
mechanical hVdc cB is less costly and more compact breakers—a key innovation for reliable hVDc grids,”
thanks to its air-core coupling reactor and smaller key presented at the cigré 2011 Bologna Symp., Bologna,
components in the injection branch. experimental results italy.
show that the mechanical hVdc cB successfully inter- K. tahata et al., “hVDc circuit breakers for hVDc grid
rupted a 9.2-ka fault current, with peak tiV reaching 272 kV applications,” in Proc. IET Int. Conf. on AC and DC Power
in 3.9 ms. Transmission, Birmingham, u.K.: Feb. 2015. doi: 10.1049/
cp.2015.0018.
DC CB Installed in the Zhangbei HVdc Grid G. tang, Z. he, h. pang, X. huang, and X. Zhang, “Basic
Due to its geographical advantage, the Zhangbei region of topology and key devices of the five-terminal Dc grid,” CSEE
northern china has seen considerable development of proj- J. Power Energy Syst., vol. 1, no. 2, pp. 22–35, June 2015.
ects for harnessing wind energy. the development of wind G. tang, X wei, and w Zhou, “research and develop-
projects is expected to continue at least into 2020. however, ment of a full-bridge cascaded hybrid hVDc breaker for
the task of successfully integrating these resources presents VSc-hVDc applications,” in Proc. Cigré 2016 Sess., paris,
challenges, because the wind farms are located some distance paper a3-117.
from where the power is consumed and because wind and, progress on meshed hVDc offshore transmission net-
thus, energy generation are intermittent and highly variable. works. accessed on: 2018. [online]. available: https://www
many of these concerns can be addressed with the implemen- .promotion-offshore.net/
tation of hVdc networks equipped with hVdc cBs. thus, a
±500-kV, 3,000-mw four-terminal hVdc grid project has Biographies
been proposed. Dragan Jovcic is with the aberdeen hVDc research cen-
the project adopts a four-terminal ring, and the operating tre at the university of aberdeen, united Kingdom.
configuration is a bipolar topology with a metallic return line Guangfu Tang is with the Global energy interconnec-
(Figure 10). the Zhangbei and Kangbao converter stations tion research institute of the State Grid corporation of
are the sending terminals collecting local wind power. the china.
Fengning converter station is a regulating terminal connected Hui Pang is with the Global energy interconnection re-
to a local pumped-storage hydroplant, which can even out search institute of the State Grid corporation of china.
p&e
the fluctuations in the production of wind power. the Beijing

may/june 2019 ieee power & energy magazine 93