You are on page 1of 9

channel Link: http://telegram.


Protection of Low-Voltage DC Microgrids

Daniel Salomonsson, Student Member, IEEE, Lennart Sder, Member, IEEE, and Ambra Sannino, Member, IEEE

AbstractIn this paper, a low-voltage (LV) dc microgrid pro- lizing the power-electronic interfaces of the DRs, together with
tection system design is proposed. The LV dc microgrid is used to fast protection systems. To operate the microgrid in island mode,
interconnect distributed resources and sensitive electronic loads. an islanding detection system is necessary, which safely discon-
When designing an LV dc microgrid protection system, knowl-
edge from existing dc power systems can be used. However, in most nects the microgrid when an ac grid outage occurs to prevent
cases, these systems use grid-connected rectifiers with current-lim- energizing the ac grid [4].
iting capability during dc faults. In contrast, an LV dc microgrid A low-voltage (LV) dc microgrid is most suitable to use where
must be connected to an ac grid through converters with bidirec- most of the loads are sensitive electronic equipment. The ad-
tional power flow and, therefore, a different protection-system de- vantage of an LV dc microgrid compared to an LV ac micro-
sign is needed. In this paper, the operating principles and tech-
nical data of LV dc protection devices, both available and in the grid is that loads, sources, and energy storage can be connected
research stage, are presented. Furthermore, different fault-detec- through simpler and more efficient power-electronic interfaces
tion and grounding methods are discussed. The influence of the se- [5]. So far, LV dc microgrids have been used in telecom power
lected protection devices and grounding method on an LV dc mi- systems, and power-system control and protection systems [6].
crogrid is studied through simulations. The results show that it is To ensure reliable operation of the LV dc microgrid, it is
possible to use available devices to protect such a system. Problems
may arise with high-impedance ground faults which can be diffi- important to have a well-functioning protection system. As a
cult to detect. starting point, knowledge from existing protection systems for
high-power LV dc power systems, for example, in generating
Index TermsCircuit transient analysis, DC power systems,
power distribution faults, power distribution protection, power stations and traction power systems [6], [7] can be used. How-
electronics. ever, these systems utilize grid-connected rectifiers with cur-
rent-limiting capability during dc faults. In contrast, an LV dc
microgrid must be connected to an ac grid through converters
I. INTRODUCTION with bidirectional power flow and, therefore, a different protec-
tion system design is needed. Short-circuit current calculations
SE of distributed resources (DRs) in the electric power
U system at the distribution level opens new possibilities.
A part of the distribution system with its sources and loads can
for LV dc systems have been treated in [8] and fault detection in
[9]. However, the protection devices have not been considered.
So far, the influence of protection devices on the system perfor-
form an isolated electric power systema microgrid [1]. During mance has only been considered in studies of high-voltage (HV)
normal operating conditions, the microgrid is connected to the dc applications, such as electric ships and HV dc transmission
ac grid at the point of common coupling (PCC), and the loads systems [10][12].
are supplied from the local sources and, if necessary, also from In this paper, a protection-system design for LV dc microgrids
the ac grid. If the load power is less than the power produced will be proposed. Different LV dc protection devices, which
by the local sources, the excess power can be exported to the today are in the research stage or commercially available, will
ac grid. The sources used in a microgrid are often small ( 500 be presented. Finally, the influence of the protection system on
kW) and are based on renewable energy, for example, PV arrays, the LV dc microgrid during faults will be studied by using the
fuel cells, and microturbines. These sources produce power with software package PSCAD/EMTDC [13].
different voltage amplitude and frequency than those used in the
microgrid and, therefore, need to be interfaced through power- II. LOW-VOLTAGE MICROGRID
electronic converters [2]. An LV dc microgrid is well suited for naturally demarcated
A microgrid is well suited to protecting sensitive loads from power systems, for example, office buildings with sensitive
power outages and, in some cases, also disturbances, for ex- computer loads or rural power systems, but also electric ve-
ample, voltage dips [3]. High reliability can be obtained by uti- hicles and ships. Since ac distribution is widespread and not
all sources and loads benefit from having a connection to dc,
Manuscript received March 12, 2008; revised November 07, 2008. First pub- it is reasonable to consider a mixed ac/dc microgrid as in the
lished April 28, 2009; current version published June 24, 2009. This work was
supported in part by ELFORSK, Sweden; in part by ABB Corporate Research;
example in Fig. 1. The dc microgrid is denoted as Zone 1, the
and in part by the Swedish Energy Administration within the Elektra program. ac microgrid is Zone 2, and the ac grid is denoted as Zone 3.
Paper no. TPWRD-00183-2008. A mixed ac/dc microgrid can typically be used in systems up to
D. Salomonsson and L. Sder are with the Electric Power Systems Lab,
Royal Institute of Technology, Stockholm SE-100 44, Sweden (e-mail:
a few megawatts, and the issue regarding how to interconnect; different sources, loads, and energy storage with the ac grid has
A. Sannino is with the ABB Corporate Research, Power Technologies, earlier been treated in [14]. However, the resulting design was
Vsters SE-721 78, Sweden (e-mail: partially based on the assumption that no loads were connected
Color versions of one or more of the figures in this paper are available online
at directly to the dc bus. The main components used in an LV
Digital Object Identifier 10.1109/TPWRD.2009.2016622 microgrid are: sources, converters, energy storage, and loads.
0885-8977/$25.00 2009 IEEE
channel Link:

connected to the dc bus, but flywheels are connected through a

machine and a converter.

D. Loads
The LV microgrid can be used to supply loads which require
a power supply with high availability. Examples of such loads
are lighting systems, data and communication systems, control
systems, safety systems, and equipment for heat, ventilation,
and air conditioning (HVAC) [19]. Many of these loads can be
supplied today with dc without any modifications, especially
electronic loads, such as computers and lighting appliances [20].
Less important ac loads, which can handle short interruptions
can instead be connected to the ac microgrid.


A well-designed protection system is necessary to ensure reli-
able operation of the LV microgrid. The protection system con-
sists of protection (current interrupting) devices, protective re-
lays, measurement equipment, and grounding. The LV micro-
Fig. 1. Example of a small low-voltage ac/dc microgrid. grid protection system can be divided into a dc protection system
for Zone 1 and an ac protection system for Zone 2. Zone 2 can
be described as a utility-consumer interconnected bus with gen-
A. Sources eration [21], and its protection system can be divided into five
Sources used in an LV microgrid can be of various types. Pho- subsystems: 1) ac grid protection, 2) ac/dc converter protection,
tovoltaic arrays and fuel cells produce dc voltage and, there- 3) diesel-generator protection, 4) feeder protection, and 5) bus
fore, are suitable to connect to a dc power system via a dc/dc protection. These subsystems will not be further treated in this
converter. Microturbines are also preferably connected to a dc paper. Instead, a dc protection system for Zone 1 will be pro-
power system due to their high-frequency output voltage which posed and analyzed. A general model of Zone 1, based on a
requires conversion. Similarly, wind-turbine generators produce data center power system, is shown in Fig. 2 [19]. The ac/dc
voltage with varying frequency. To connect the wind turbine converter is modeled as a three-phase, two-level voltage-source
to a dc bus, only one converter is required. Internal combus- converter (VSC), as described in [22]. The battery is modeled as
tion engines (ICE), for example, diesel engines, commonly used a dc source with a resistive-inductive impedance [23]. Finally,
for standby-power generation, are preferably connected to an ac the loads are modeled as constant power (CP) loads, together
power system. with an input filter, and are connected between the two poles
[20]. Parameters for each component are listed in Table I.
B. Converters
A. Grounding
Both ac/dc and dc/dc converters are used in the LV micro- Grounding is a complex issue and there are many different
grid, where ac/dc converters are used to interconnect the ac mi- approaches to designing grounding in an electric power system,
crogrid and the dc microgrid. These converters need to generate and different solutions result in different performance [24],
sinusoidal ac voltages and currents, and be able to control the [25]. Grounding is used for the detection of ground faults and
bidirectional power flow. Furthermore, the converters must have for personnel and equipment safety [6]. An LV dc microgrid
galvanic isolation, and be able to handle grid disturbances, such can be ungrounded, high resistance grounded, or low resistance
as voltage dips with unsymmetrical voltages [15]. Finally, the grounded. Moreover, the ground can be connected either to
converters should have high efficiency. one of the poles or to the middle point of the converter and the
Different dc/dc converters will be used to connect different battery. The two alternatives are shown in Fig. 3.
sources and loads to the dc microgrid. DC/DC converters can Fig. 3(a) shows a TN-S dc system. It has the middle point
be built simpler compared with ac/dc converters, which results of the converter and the battery connected to ground (T), and
in lower cost and higher efficiency [16]. separate (S) wires are used throughout the system for neutral (N)
and protective earth (PE). The alternative in Fig. 3(b) is an IT
C. Energy Storage
dc system. It has the positive pole connected to ground through
The availability or the transient response of some DRs, such an impedance (I). The positive pole is preferably connected to
as wind power, solar cells, and fuel cells require them to be ground compared with the negative one to reduce the impact of
combined with other energy sources or energy storage. Fur- corrosion.
thermore, energy storage can be used for power-quality (PQ) Using alternative (a) in Fig. 3 results in a large ground cur-
improvement, load leveling, or emergency power supply [17]. rent and a large dc-link voltage transient in case of a low-resis-
Commonly used storage techniques are batteries, (super)capac- tance ground fault. The large voltage transient may affect other
itors, and flywheels [18]. Batteries and capacitors can be directly loads connected to the faulted pole, but not loads connected to
channel Link:

Fig. 2. Detailed scheme of Zone 1.


load can be increased to limit the voltage transient. However,

this results in increased losses. Alternative (b) is commonly used
in telecom power systems [26].

B. Protection Devices
Protection devices commercially available for LV dc systems
are fuses, molded-case circuit breakers (MCCB), LV power
CBs, and isolated-case CBs [27]. Some of these models are
specially designed for dc, but most can be used in ac and dc
applications. However, the ratings for ac and dc operation are
different, and must be carefully considered when designing
Fig. 3. LV dc microgrid grounding. (a) TN-S dc system. (b) IT dc system. the protection system. Examples of commercially available
protection devices for LV dc systems, together with their nom-
inal voltage and current, and short-circuit-current-interruption
the other pole. The fault is easily detected and can be quickly capability are listed in Table II.
cleared. A TN-S dc system provides a well-defined pole-to- 1) Fuses: A fuse consists of a fuse link and heat-absorbing
ground voltage and paths for leakage currents from noise filters. material inside a ceramic cartridge. The fuse link is made of
An IT dc system has only a small current and voltage transient copper or silver, and its design depends on its current-time and
in case of a ground fault. This will ensure stable operation of voltage ratings. The heat-absorbing material, used to quench the
the loads during a single ground fault. However, a ground in arc, is usually silica sand.
the system will change the pole-to-ground voltage, which may Voltage and current ratings of fuses are given in root-mean-
affect sensitive electronic loads. Due to the small ground-fault square (rms) values, and are therefore valid for both ac and dc.
current, it can be difficult to measure and detect the fault, and However, when using fuses in a dc system, it is important to
metal enclosures of loads may be energized. To further improve consider the system time constant. It determines the rise time
the system in case of a ground fault, the line impedance of each of the current transient, but not its final steady-state value. The
channel Link:

rise time affects the ability of the fuse to interrupt the current. IV. PROTECTION SYSTEM DESIGN
A small time constant ( 2.5 ms) will result in a fast current The overall function of the LV dc microgrid protection system
increase and the fuse link will quickly melt and the arc can be is to detect and isolate faults fast and accurately in order to min-
cooled by the heat absorbing material. However, if the time con- imize the effects of disturbances. The design of the protection
stant is large ( 6 ms), the current and, in turn, the temperature system depends on a number of issues which will be treated in
of the fuse will slowly increase and when an arc is eventually this section, and is illustrated on the system in Fig. 2. The design
formed, it cannot be cooled enough by the heat absorbing mate- issues are the type of faults which can occur, their consequences,
rial since its temperature has increased [34]. Furthermore, fuses the type of protection devices required, the need for backup pro-
used in dc systems must also be able to handle light overcurrents tection, detection methods, measures to prevent faults, and fi-
in order not to malfunction. nally, measures to prevent incorrect operation of the protection
2) Circuit Breakers: An MCCB consists of a contactor, a system.
quenching chamber, and a tripping device. MCCBs are usu-
ally equipped with a thermal-magnetic tripping device, but can A. Possible Fault Types
be extended with an electronic one. Furthermore, as for fuses, Possible fault types in the dc microgrid are pole-to-pole and
the voltage and current ratings are given in rms values. The pole-to-ground faults. Pole-to-pole faults often have low fault
magnetic tripping senses the instantaneous value of the current, impedance, while pole-to-ground faults can be characterized as
which means that the rated current for dc is times higher either low-impedance or high-impedance faults. The location of
than for ac. However, for the thermal tripping, the values are the faults can be either on the bus or on one of the feeders. These
the same [35]. To improve the voltage rating, multiple poles four possible faults have been indicated as F1-F4 in Fig. 2. Fault
can be connected in series. However, it is important to consider F1 is a short circuit between the positive and the negative poles
the grounding arrangement in the system to prevent having full at the bus, and F2 is a short circuit between the positive pole
system voltage across one pole during a ground fault. and ground. These faults are critical for the whole system, in
For larger LV dc systems, such as traction power systems, particular, the converter and the battery. Faults F3 and F4 are
special high-speed dc-circuit breakers are available. These CBs of similar types, but located on the feeder close to one of the
are designed to fully cope with the rated voltage and current, two loads. These faults are important to consider when investi-
even in a system with a large time constant. A high-speed CB gating how the nonfaulted load is affected, and how to design the
starts to interrupt the fault current within 0.01 s. Problems may backup protection of the feeders. Faults inside the converter, the
arise with low currents which can cause the CB contacts to weld battery, or the loads have not been considered separately. Con-
together [7]. verter or battery failure will be similar to faults F1 or F2, and a
3) Power-Electronic Protection Devices: There are some load fault to F3 or F4.
known problems associated with fuses and CBs in LV dc sys-
tems, such as large time constants and long breaker operation B. Bus Faults
time. By utilizing power-electronic switches, such as thyristors, Faults F1 and F2 will affect all sources and loads connected
the operation speed decreases and the inductive current inter- to the bus, and the loads and sources that are to be connected
ruption capability can be increased [10], [36]. However, the depend on the current operation mode. Since the sources are
related losses of a power-electronic solution are much higher connected in parallel, they can be treated separately.
compared with a mechanical switch. Therefore, a combination The battery can be located in a separate battery room, and
of a mechanical switch and a power-electronic switch has been can, consequently, be connected to the dc bus through cables.
proposed in [37]. The total fault impedance, as it appears to the battery when a
fault F1 occurs, is then the sum of the internal battery impedance
C. Protective Relays and Measurement Equipment ( and ) and the cable impedance ( and ) (the
cable capacitance can be neglected during pole-to-pole faults).
High-speed dc CBs are equipped with mechanical instanta- The battery fault current can then be calculated as
neous overcurrent tripping devices, which can be set to trip the
breaker if the current exceeds 14 p.u.. The electromagnetic (1)
force generated by the current is used to trip the CB. However,
tripping the CB due to other events requires a protective relay. where . The design of the
Protective relays use information from measured voltages and battery determines how long it can supply a short-circuit current
currents, and in some cases, information based on communica- without causing internal damage [34].
tion with other units. It is important to note that the measurement The converter is directly connected to the bus and, hence, has
equipment must be able to handle dc quantities in order to work a low impedance, mainly consisting of the series impedance of
properly. the capacitors and , where the latter can be neglected
Besides overcurrent, protective relays can calculate time [39]. Fault F1 will cause the capacitors to discharge, which re-
derivatives and step changes of currents to determine whether sults in a current with high amplitude and low rise time, but with
the dc system is in normal operation or if a fault has occurred limited duration [40]. The capacitor fault current can be calcu-
[7]. More sophisticated numerical methods (e.g., the use of lated as
neural networks) can also be used to detect faults and separate
them from normal operation [38]. (2)
channel Link:

where . However, when the dc-link voltage be- are chosen in this case study to protect the battery. However,
comes almost zero, the converter will lose its current-control ca- other solutions exist: fuse in series with a switch for example.
pability and a fault current will flow through the insulated-gate The converter protection (P1-P2) must be able to limit the
bipolar transistors (IGBTs) antiparallel diodes, limited only by fault current through the IGBT diodes. In this particular case
the grid filter ( and in Fig. 2). These diodes are sensitive study, the diodes can handle a maximum current of 7 p.u. up to
to overcurrents, and the current through them must not exceed 10 ms [41], which will give the maximum fault clearing time
a certain amplitude and have a duration longer than a certain of the converter protection. Possible choices which can be fast
time [41]. enough are power-electronic switches, ultra-fast hybrid dc CBs
or fuses.
C. Feeder Faults In both cases, it is important to have a fast detection method.
Fault F3 is located after the protection devices protecting Possible methods that were described earlier are amplitude, step
feeder L1, and will permanently affect only the loads which are change or derivative of the current, or dc-link voltage amplitude.
connected to this feeder. Before the fault is cleared, the other A fault F1 results in a large battery fault current . How-
feeder will sense a voltage drop whose magnitude depends on ever, the large time constant makes the derivative of during
the fault impedance. A larger fault impedance will result in a a fault only three times larger than a load step. Hence, it is more
lower voltage drop, but a lower fault current, which can in- convenient to use the current amplitude together with the dc-link
crease the fault-clearing time. Fault F3 results in a second-order voltage. The overcurrent protection can be set to a percentage of
-system, and guidance to analytically calculating the fault the steady-state fault current, and the undervoltage protection to
current can be found in [40]. a percentage of the dc-link voltage .
Due to the fast discharge of the capacitors , it is difficult
D. Ground Faults to use the amplitude of the converter current to detect a fault.
Faults F2 and F4 are positive-pole-to-ground faults, and in Another choice is to use the current derivative which can be
these two cases, only the faulted pole is affected. How much the calculated as
voltage decreases depends on the system-grounding method and
the fault impedance. Furthermore, the cable capacitances , (3)
, and need to be consider during the ground fault in
high-impedance-grounded systems. The absolute value is required since the capacitor discharge
during the fault results in a negative derivative. (3) can be used
V. CASE STUDY to calculate the minimum level of the derivative setting to be
When selecting which protection device (P1-P8) to use to able to detect the fault within a certain time. The suggested set-
protect the different components, in this case, the converter, the tings for the battery and converter protection are:
battery, and the two feeders, it is necessary to consider the fol- battery overcurrent level: 80% of the maximum battery
lowing issues: function, operation time, controllability, and de- fault current;
tection. This will be treated in this section as a case study. In this converter-current derivative detection time: 1 ms;
study, a TN-S dc system, shown in Fig. 3(a), has been consid- dc-link undervoltage level: 40%.
ered since it gives the possibility to investigate the ground-fault Using the suggested values from before gives the protection
currents. Furthermore, the protection system is modeled as an settings used in the test case. The battery protection is set to
ideal system: there is no delay of the fault detection due to com- 27-p.u. overcurrent and 0.4-p.u. dc-link voltage. The total op-
putation or communication, and the formation and influence of eration time of the battery protection is assumed to be 20 ms.
arcs is neglected. The converter protection is set to 6.4 p.u./ms (current deriva-
tive) and 0.4-p.u. dc-link voltage. In this case, an ultrafast hybrid
A. Converter and Battery Protection (P1-P4) dc CB is used to protect the converter. However, it would also
The primary function of the battery and converter protec- be possible to use fast dc fuses. Simulations show that the max-
tion is to clear faults which appear between the sources and imum fault clearing time for the converter protection in this case
the bus. Hence, a sustained fault F1 will be studied. The bat- must not exceed 2 ms; otherwise, the diodes will exceed 7 p.u.
tery fault-current amplitude and rise time can be calculated by (their maximum current rating). A 2-ms fault-clearing time has
using (1) as 34 p.u. and 2.96 ms, respectively. The discharge of been used in the test case shown in Fig. 4. Fig. 4(a) shows that
the converter capacitors results in a short transient. The ampli- the converter dc current has a very high but short transient, and
tude and time constant can be calculated by using (2) as 125 p.u. the battery current reaches 34 p.u. More important, the currents
and 0.112 ms, respectively. Since the dc-link voltage becomes through the converter reach 6 p.u. and become zero after 6.5 ms.
almost zero during the fault, the converter will act as a diode The diodes can handle these fault currents, and the converter is
rectifier and the current will go through the reverse diodes only. protected. The converter dc-link voltage becomes 1.2 p.u. due
The fault current through the diodes is then limited by the grid to the current transients through the diodes after the converter
filter between the ac grid and the converter. has been disconnected. A longer total interruption time could
The battery protection (P3-P4) must disconnect the bat- damage the reverse diodes and the dc-link capacitors.
teries before they are damaged due to the high fault current A low-impedance ground fault F2 will result in large ground
as described earlier. Furthermore, it can be desirable to be currents from the connected sources using the grounding
able to connect and disconnect the battery and, therefore, CBs scheme applied here. Furthermore, the dc-link voltage becomes
channel Link:

Fig. 4. Fault F1 cleared by the converter protection after 2 ms and the battery Fig. 5. Fault F3 cleared by the fuse after 1.0 ms. (a) DC currents. (b) Grid
protection after 20 ms. (a) DC currents. (b) Grid currents. (c) DC-link voltage. currents. (c) DC-link voltage.

approximately 0.5 p.u. since one of the poles is shorted. On the

other hand, a high-impedance ground fault on the bus results in
a small fault current and only a small reduction of the dc-link
voltage. Hence, it can be difficult to detect these faults. One
method to detect ground faults in a dc auxiliary power system is
to measure the ground currents and . If they go out-
side a predefined range, a positive or negative pole-to-ground
fault has occurred. Leakage currents must be considered when
setting the level.

B. Feeder Protection (P5-P8)

Each feeder must be protected against short circuits and
overloads, and for that purpose, it is possible to use MCCBs
and fuses. It is easier to obtain selectivity by using fuses than
MCCBs due to their magnetic sensing. However, the advantage
of MCCBs compared with fuses is that they open both poles in
case of a fault. Therefore, MCCBs are used only closest to the
load sometimes.
The fault clearing time for fuses and MCCBs depends on the
available short-circuit current which, in turn, is limited by the Fig. 6. Fault F3 cleared by the fuse after 5.0 ms. (a) DC currents. (b) Grid
source and feeder impedance. Furthermore, a larger time con- currents. (c) DC-link voltage.
stant can prolong the fault-clearing time. The simulation of a
fault F3 cleared after 1 ms is shown in Fig. 5. After the fault
is cleared, the converter current and the battery current start to fault current during a low-impedance ground fault F4 (Fig. 7)
oscillate against each other. These oscillations can be reduced is large enough to make the protecting device clear the fault.
by having a shorter fault clearing time or a reduced controller When the fault occurs, there is a voltage shift between the two
bandwidth. The current transients through the converter are ap- converter dc-link capacitors. The voltage across the unfaulted
proximately 2.1 p.u., and the dc-link voltage drops to 0.8 p.u. pole and ground increases, and the voltage across the faulted
for less than 5 ms. Increasing the fault-clearing time to 5 ms re- pole and ground decreases. In total, the dc-link voltage is re-
sults in a larger dc-link voltage transient, which can affect the duced. Due to the voltage change, the converter ground current
loads connected to the other feeder. This can be seen in Fig. 6. will become zero at steady state.
Pole-to-ground faults can, for example, occur after an isola- The last case analyzes a fault F4 with 5.0- impedance. In
tion failure in the system. The ground currents should be zero this case, there is only a very small fault current, which can
during normal operation and, therefore, they can be useful to be seen from Fig. 8. This can only be detected by the ground-
measure these at each source to detect a ground fault. The total current measurement of the sources and . However,
channel Link:

Fig. 7. Fault F4. (a) DC currents. (b) Grid currents. (c) DC-link voltage.

Fig. 9. Fault F1 and F3. (a) Battery fault current. (b) Converter fault current.
(c) Derivative of converter fault current. (d) DC-link voltage.

it instead. In this case study, the converter, the battery, and the
feeder protection must be coordinated.
The converter protection (P1-P2) and the battery protection
(P3-P4) are in this case study backup protection for the feeder
protections (P5-P8). If a fault F3 occurs, it is then important that
only protection P5-P6 operates. However, if P5-P6 fail, P1-P2
and P3-P4 must operate and clear the fault. Protection P1-P4
must therefore be able to distinguish fault F1 from F3. These
two faults are compared with respect to battery fault current,
converter fault current, derivative of the converter fault current,
and the dc-link voltage, and are shown in Fig. 9.
Fig. 9(a) shows the battery fault current during fault F1 and
Fig. 8. Fault F4 (high-impedance fault). (a) DC currents. (b) Grid currents. (c) F3. The amplitude and the rise time differ, and it it possible for
DC-link voltage. P3-P4 to separate these two faults. By lowering the tripping level
of the current and introducing a time delay, P3-P4 can act as
backup for P5-P6.
the sensitivity of the protection system to high-impedance faults Fig. 9(b) and (c) shows the converter fault current and its
depends on the level of stray currents in the system. Therefore, derivative. Already after 1 ms, it is difficult to distinguish a fault
in some cases, it can be better to only signal that there is or F1 from F3. However, the dc-link voltage shown in Fig. 9(d) is
has been a high-impedance ground fault, instead of tripping the a good indicator whether it is a fault F1 or F3. Consequently,
sources. it is suitable to combine the derivative of the converter current
with the dc-link voltage to obtain selectivity between P1-P2 and
C. Protection Coordination P5-P6.
So far, the converter, the battery, and the feeder protections The converter and the battery protection cannot distinguish
have been studied separately. However, to have high reliability, a fault F5, located directly after P5-P6, from a fault F1. There-
it is important that only the dedicated protection operates when fore, protection P5-P6 must in this case be faster than protec-
a fault occurs. Sometimes, the primary protection fails to clear tion P1-P4 in order to achieve selectivity. Due to the fast re-
the fault, and then the backup protection must operate and clear sponse of P1-P2, there might be a risk that it will operate even
channel Link:

though P5-P6 will clear the fault. More important, however, is [9] X. Feng, Z. Ye, C. Liu, R. Zhang, F. Lee, and D. Boroyevich, Fault
that P3-P4 do not operate to prevent a total shutdown. detection in dc distributed power systems based on impedance charac-
teristics of modules, in Proc. Conf. Rec. IEEE Ind. Appl. Soc. Annu.
Finally, fuses connected on the ac side of the converter can Meeting, Rome, Italy, Oct. 812, 2000, vol. 4, pp. 24553462.
be used as backup protection for the converter. In contrast, it [10] M. Baran and N. Mahajan, Overcurrent protection on voltage-source-
would not be possible to have backup protection for the battery. converter-based multiterminal dc distribution systems, IEEE Trans.
Power Del., vol. 22, no. 1, pp. 406412, Jan. 2007.
However, in some dc auxiliary power systems, the battery is [11] L. Tang and B. Ooi, Locating and isolating dc faults in multi-terminal
not protected at all since it is considered to have lower system dc systems, IEEE Trans. Power Del., vol. 22, no. 3, pp. 18771884,
reliability [6]. Jul. 2007.
[12] W. Litzenberger and P. Lips, Pacific hvdc intertie, IEEE Power En-
ergy Mag., vol. 5, no. 2, pp. 4551, Mar./Apr. 2007.
VI. CONCLUSION [13] 2008, PSCAD/EMTDC Manitoba HVDC Research Center. Win-
nipeg, MB, Canada. [Online]. Available:
In this paper, an LV dc microgrid protection-system design index.cfm?.
has been proposed. A small dc microgrid has been studied [14] M.-S. Kwak and S.-K. Sul, Control of utility connected gas engine
generation system, in Proc. IEEE PESC, Jeju, Korea, Jun. 1822,
during different fault events in order to investigate which type 2006, pp. 28992904.
of protection devices are required and where to install them. [15] D. Salomonsson and A. Sannino, Centralized ac/dc power conversion
The result shows that it is possible to use commercial protection for electronic loads in a low-voltage dc power system, in Proc. IEEE
PESC, Jeju, Korea, Jun. 1822, 2006, pp. 31553161.
devices, such as fuses and CBs, to protect batteries and loads. [16] N. Mohan, T. Undeland, and W. Robbins, Power Electronics: Con-
However, converters using IGBT modules are very sensitive verters, Application and Design, 2nd ed. New York: Wiley, 1995.
to overcurrents and, therefore, require fast protection. For this [17] J. Barton and D. Infield, Energy storage and its use with intermittent
renewable energy, IEEE Trans. Energy Convers., vol. 19, no. 2, pp.
case study, a prototype of an ultra-fast hybrid dc CB was used 441448, Jun. 2004.
to protect the converter. It was shown that during a bus fault, the [18] A. Kusko and J. Dedad, Stored energy, IEEE Ind. Appl. Mag., vol.
total breaker time is very critical for limiting the fault current 13, no. 4, pp. 6672, Jul./Aug. 2007.
[19] IEEE Recommended Practice for Emergency and Standby Power Sys-
and protecting the converter. tems for Industrial and Commercial Applications, IEEE Std. 446-1995,
Low-impedance ground faults in a solidly grounded system 1995.
result in large fault current and can therefore easily be detected [20] D. Salomonsson and A. Sannino, Load modelling for steady-state and
transient analysis of low-voltage dc systems, IET Electr. Power Appl.,
and cleared. However, problems can arise with high-impedance vol. 1, no. 5, pp. 690696, Sept. 2007.
ground faults since the fault current is very small. One way to [21] IEEE Guide for Protective Relaying of Utility-Consumer Interconnec-
detect a ground fault is to measure the ground current flowing tions, IEEE Std. C37.95-2002, 2002.
[22] D. Salomonsson and A. Sannino, Comparative design and analysis
from the sources. The sensitivity is dependent on the amount of dc-link-voltage controllers for grid-connected voltage-source con-
of leakage current in the system. The consequences of a high- verters, in Proc. Conf. Rec. IEEE-Ind. Appl. Soc. Annu. Meeting, New
impedance ground fault must be evaluated, and should prefer- Orleans, LA, Sep. 2327, 2007, pp. 15931600.
[23] 2007, Industrial Energy Exide. [Online]. Available: http://www.exide.
ably only be indicated, rather than disconnecting all sources. com/., Data sheet.
Finally, coordination of the protection devices was discussed. [24] Low-Voltage Electrical Installations, IEC 60 364-1, 2005.
It was shown that problems may arise with the converter protec- [25] D. Paul, DC traction power system grounding, IEEE Trans. Ind.
Appl., vol. 38, no. 3, pp. 818824, May/Jun. 2002.
tion. It can be very difficult to determine the fault location based [26] N. Tullius, Three-wire dc distribution to telecommunication equip-
on the fault current or its derivative only. Therefore, the con- ment, in Proc. IEEE INTELEC, Vancouver, BC, Canada, Oct. 30Nov.
verter fault current can be used together with the dc-link voltage 3 1994, pp. 297300.
[27] J. Brozek, DC overcurrent protectionWhere we stand, IEEE Trans.
to solve the problem and ensure a reliable protection system. Ind. Appl., vol. 29, no. 5, pp. 10291032, Sep./Oct. 1993.
[28] 2007, Ferraz Shawmut North American Fuses and Fuseholders CHS
Control. [Online]. Available:
REFERENCES [29] 2007, Fuses (in Swedish) IFO Electric. [Online]. Available:
[1] N. Hatziargyriou, H. Asano, R. Iravani, and C. Marnay, Microgrids,, Data sheet.
IEEE Power Energy Mag., vol. 5, no. 4, pp. 7894, Jul./Aug. 2007. [30] 2007, Moulded Case Circuit Breakers Tmax and Isomax ABB [On-
[2] D. Georgakis, S. Papathanassiou, N. Hatziargyriou, A. Engler, and C. line]. Available:, Data sheet.
Hardt, Operation of a prototype microgrid system based on micro- [31] 2007, Eatons Cuttler-Hammer Series G Moulded Case Cir-
sources quipped with fast-acting power electronics interfaces, in Proc. cuit Breaker, 15-630 A CHS Control. [Online]. Available:
IEEE Power Eng. Soc. Committee, Achen, Germany, Jun. 2025, 2004,
vol. 4, pp. 25212526. [32] 2007, Sentron VL System Manual Siemens. [Online]. Available: http://
[3] G. Venkataramanan and M. Illindala, Microgrids and sensitive loads,
in Proc. IEEE Power Eng. Soc. Winter Meeting, New York, Jan. 2731, [33] 2007, DC circuit breakers Scheron. [Online]. Available: http://www.
2002, vol. 1, pp. 315322., Data sheet.
[4] F. Katiraei, M. Iravani, and P. Lehn, Micro-grid autonomous opera- [34] IEEE Guide for the Protection of Stationary Battery Systems, IEEE Std.
tion during and subsequent to islanding process, IEEE Trans. Power 1375-1998, 1998.
Del., vol. 20, no. 1, pp. 248257, Jan. 2005. [35] G. Gregory, Applying low-voltage circuit breakers in direct current
[5] D. Salomonsson and A. Sannino, Low-voltage dc distribution system systems, IEEE Trans. Ind. Appl., vol. 31, no. 4, pp. 650657, Jul./Aug.
for commercial power systems with sensitive electronic loads, IEEE 1995.
Trans. Power Del., vol. 22, no. 3, pp. 16201627, Jul. 2007. [36] S. Tennakoon and P. McEwan, Short-circuit interruption performance
[6] IEEE Recommended Practice for the Design of DC Auxiliary Power of thyristor circuit breakers, in Proc. IEEE APEC, Orlando, FL, Feb.
Systems for Generating Stations, IEEE Std. 946-2004, 2004. 1317, 1994, vol. 2, pp. 832838.
[7] J. Morton, Circuit breaker and protection requirements for dc [37] J.-M. Meyer and A. Rufer, A dc hybrid circuit breaker with ultra-fast
switchgear used in rapid transit systems, IEEE Trans. Ind. Appl., vol. contact opening and integrated gate-commutated thyristors (igcts),
IA-21, no. 5, pp. 12681273, Sep./Oct. 1985. IEEE Trans. Power Del., vol. 21, no. 2, pp. 646651, Apr. 2006.
[8] P. Sutherland, DC short-circuit analysis for systems with static [38] J. Jin, J. Allan, C. Goodman, and K. Payne, Single pole-to-earth fault
sources, IEEE Trans. Ind. Appl., vol. 35, no. 1, pp. 144151, Jan./Feb. detection and location on a fourth-rail dc railway system, Proc. Inst.
1999. Elect. Eng., Electr. Power Appl., vol. 151, no. 4, pp. 498504, Jul. 2004.
channel Link:

[39] 2007, Evox Rifa Electrolytic Capacitors, EVOX RIFA. [Online]. Lennart Sder (M91) was born in Solna, Sweden,
Available:, in 1956. He received the M.Sc. and Ph.D. degrees
Data sheet. in electrical engineering from the Royal Institute of
[40] Short-circuit currents in d.c. auxiliary installations in power plants Technology (KTH), Stockholm, Sweden, in 1982 and
and substationsPart 1: calculation of short-circuit currents, IEC 61 1988, respectively.
660-1, 1997. Currently, he is a Professor in Electric Power Sys-
[41] 2007, Datasheet IGBT SEMIKRON. [Online]. Available: http://www. tems at KTH. He works on projects concerning dereg- ulated electricity markets, distribution systems, and
integration of wind power, HVDC, power system re-
liability, protections systems, hydropower, and micro
grids. He has been involved in several national com-
mittees concerning the risk of capacity deficit and handling of bottlenecks within
the deregulated market.

Ambra Sannino (S99M01) received the M.Sc.

and Ph.D. degrees in electrical engineering from the
University of Palermo, Palermo, Italy, in 1997 and
Daniel Salomonsson (S02) received the M.Sc. and 2001, respectively.
Tech.Lic. degrees in electrical engineering from From 2001 to 2004, she was an Assistant Professor
Chalmers University of Technology, Gothenburg, with the Department of Electric Power Engineering
Sweden, in 2002 and 2005, respectively, and the of Chalmers University of Technology, Gothenburg,
Ph.D. degree at the Electric Power Systems Lab, Sweden, and an Associate Professor since 2004. She
Royal Institute of Technology, Stockholm, Sweden, has been with ABB, Corporate Research, Vsters,
in 2008. Sweden, since 2004.
His research project concerns dc distribution sys- Her interests include applications of power elec-
tems. His research interests are direct current distri- tronics in power systems, distributed generation, as well as wind power and
bution systems, power electronics, and power quality. power quality.