This article has been accepted for inclusion in a future issue of this journal.

Content is final as presented, with the exception of pagination.

Microgrid Protection
By A l i H o osh ya r , Member, IEEE and R e z a I r ava n i Fellow, IEEE

ABSTRACT  | The proliferation of distributed energy resources energy resource (DER) units with conventional “passive”
is setting the stage for modern distribution systems to operate distribution systems. This resulted in the emergence of
as microgrids, which can avoid power disruptions and serve “active” distribution systems [1], which can evolve into
as resources for fast recovery during macrogrid disturbances. microgrids [2]. A microgrid is defined as a coordinated
Microgrids are, therefore, major assets to improve the group of DER units that service a set of loads through a
grid resilience. However, the offered resilience is seriously distribution system with the capability to 1) operate con-
undermined if microgrids are not properly protected in the nected to the conventional power grid (i.e., the macrogrid);
event of faults within their own boundaries. Distribution 2) ­operate islanded from the macrogrid; and 3) provide
protective devices cannot reliably protect microgrids due to the smooth transition between the macrogrid-connected and
variable and often limited short-circuit capacities of microgrids. islanded modes [3].
Moreover, the research on microgrid protection has not led to A distribution system that operates as a microgrid brings
a commercially available microgrid relay to date and has little sufficient generation close to load and so can maintain the
prospect of reaching that level in the near future. As a result, power supply in the event of macrogrid faults. In addition,
the existing options for reliable microgrid protection remain if a microgrid has excess generation capacity, it can provide
effectively the subtransmission and transmission system the macrogrid with system recovery resources, thus decreas-
protective devices, e.g., directional overcurrent, distance, and ing the frequency and duration of outages. Such features add
differential relays. Although years of operation in macrogrids flexibility to the aforementioned architecture and make a
support these relays, their performance for microgrids is yet to
microgrid a tremendous asset to improve the grid resilience
be analyzed. This paper presents such analysis for different relay
to macrogrid failures.
types by considering various fault and generation conditions in
However, the resilience offered by a microgrid will be
a microgrid. Time-domain simulations are used to identify the
in jeopardy if the microgrid is not properly protected during
scenarios where the relays function correctly as well as the
the short-circuit faults that occur within its own boundaries.
problematic conditions, on which future research should focus.
Although microgrids are erected at the distribution level of
This paper also presents a short review on direct current (dc)
the legacy grid, the available distribution protective devices,
microgrids and their protection requirements.
such as fuse and recloser [4], are not suitable candidates for
KEYWORDS  |  Direct current (dc) microgrid protection; microgrid protection due to various reasons, an important
distributed energy resources (DERs); distribution protection one of which is the complete dependence of these devices
system; electronically coupled distributed generation (ECDG), on local current magnitude. The short-circuit capacity of the
induction-machine-based distributed generation (IMDG); macrogrid can be larger than that of the relatively small DER
macrogrid; microgrid; microgrid protection; mesogrid units of a microgrid by more than one order of magnitude,
making the fault current levels over the macrogrid-connected
I.  I N T RODUC T ION and islanded modes of a microgrid sharply different. Such a
The legacy power grid has a top-down architecture from big difference makes setting and coordination of existing dis-
the generation to transmission to distribution systems. This tribution protective devices difficult and often unattainable.
structure offers limited resilience to faults and failures in A variety of methods have been published in the last
the transmission system because they can potentially result few years to address microgrid protection issues [5]–[9].
in loss of many distribution systems. Within the past dec- Meanwhile, given the sensitivity of reliable protection,
ade, however, technological advancements and government extensive testing and verification is needed for a protec-
incentives facilitated significant integration of distributed tion scheme to reach a commercial level, and the published
research on microgrid protection has not come close to this
level yet. In the absence of an off-the-shelf microgrid relay,
Manuscript received September 1, 2016; revised December 12, 2016; accepted
February 7, 2017.
protective devices of subtransmission and transmission sys-
A. Hooshyar is with the Electrical Engineering and Computer Science Department, tems, e.g., directional overcurrent, distance, and differential
York University, Toronto, ON M3J 1P3 Canada (e-mail: hooshyar@yorku.ca).
R. Iravani is with the Electrical and Computer Engineering Department, University
relays, are the only choices that can make a protection sys-
of Toronto, Toronto, ON M5S 3G4 Canada (e-mail: iravani@ecf.utoronto.ca). tem more immune to the challenging features of microgrids,
Digital Object Identifier: 10.1109/JPROC.2017.2669342

0018-9219 © 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.
Proceedings of the IEEE   1
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

Hooshyar and Iravani: Microgrid Protec tion

such as variable short-circuit current level, albeit at a higher of a microgrid that includes two parallel feeders emanating
cost. Consequently, although there are certain differences from the point of common coupling (PCC) with the macro-
between the fault characteristics and topologies of micro- grid. DER units and loads within the microgrid are equipped
grids and macrogrids, these relays are expected to be the with local controllers (LCs) and communicate with the
main options available to protect microgrids at least for the microgrid supervisory controller and energy management
next few years. system (SC-EMS). The degree of sophistication and func-
These relays are backed by decades of successful opera- tionalities of SC-EMS can vary widely. As of now there are
tion for macrogrids, and so their correct operation for no guidelines to streamline such functions, which can have
microgrids is often taken for granted [10]. However, a key significant ramifications for protection of microgrids.
assumption in the development of most of these relays is
that the generation units include directly coupled synchro- A. Types of Microgrids
nous machines (SMs)—a typically invalid assumption about
the DER units of a microgrid. Therefore, the effects of DER Based on the characteristics/properties of feeder(s) that
units as well as some unique characteristics of microgrids constitutes the backbone of a microgrid, it can be catego-
on off-the-shelf protective relays need to be inspected to rized as follows.
identify the necessary upgrades and modifications to these • Urban microgrids: In this category, the feeders are
relays. These are essential steps to ensure reliable protection in a populated or a concentrated industrial area, the
of microgrids, and in turn, the grid resilience. feeders are densely loaded, the main trunk and the
This paper sheds light on the application of commercially laterals are fairly short, and the degree of imbalance
available relays for microgrid protection. Section II presents is not high. Short-circuit ratio1 of an urban micro-
a background on the architecture and control of microgrids. grid at its PCC is normally above 25. Therefore,
The major challenges associated with microgrid protection during the macrogrid-connected mode of an urban
and a test system that exemplifies these issues are discussed microgrid, the voltage and frequency are dictated
in Section III. Section IV reviews the expected behavior of by the macrogrid, voltage profile is fairly flat, and
distributed generation (DG) units during a microgrid fault. transient instantaneous voltage is well behaved.
Section V analyzes the effects of electronically coupled DG Thus, ECDG units can be synchronized with the
(ECDG) units on protective relays in a microgrid. Induction- microgrid with relative ease, which is of importance
machine-based DG (IMDG) units are studied in Section VI. for protection.
Protection of direct current (dc) microgrids is discussed in • Rural microgrids: In this category, the feeders are
Section VII, and Section VIII concludes the paper. located in a sparsely populated area, and thus, the
load is scattered; the main trunk and the laterals
II.  BACKGROU N D are noticeably long; the short-circuit ratio is not
necessarily as high as that of the urban microgrid;
This section reviews different types and modes of microgrids
and the voltage profile is not flat. In a rural micro-
and their components. Fig. 1 shows the schematic diagram
grid, voltage imbalance and fluctuation can be sig-
nificant. Thus, DER units have impact on voltage,
and if permitted, can be controlled to assist in the
feeder voltage regulation.
• Off-grid microgrids: An off-grid microgrid is either
geographically located in a remote area with no pos-
sibility for macrogrid connection, or surrounded by
difficult terrain for transmission line connection.
By definition, an off-grid microgrid always oper-
ates in islanded mode, and thus, does not comply
with the strict definition of “microgrid,” i.e., does
not have macrogrid-connected mode of operation.
However, integration of large-size DER units with
off-grid microgrids is occurring at faster and higher
pace than urban and rural microgrids, developed/
under-development SC-EMS strategies for off-grid
microgrids are fully applicable to islanded mode
of urban and rural microgrids, and there exist con-
Fig. 1. Schematic diagram of a macrogrid-connected microgrid
(CHP: combined heat and power; EV: electric vehicle; PCC: point of 1
Short-circuit ratio is defined as the ratio of the macrogrid short-
common coupling; PV: photovoltaic; SC-EMS: supervisory controller circuit capacity at the PCC to the total DER generation capacity of the
and energy management system). microgrid.

2  Proceedings of the IEEE

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

Hooshyar and Iravani: Microgrid Protec tion

siderable social, economical, political, and envi- aspects of the microgrid. A necessary step to tackle micro-
ronmental incentives/pressure to integrate renew- grid control/protection issues is to identify its modes of
able energy-based sources with off-grid microgrids. operation as follows.
Therefore, off-grid microgrids will potentially be 1) Macrogrid-Connected Mode: In the macrogrid-con-
the fastest growing microgrid type and remain as nected mode, each DER unit includes an LC that operates
a member of microgrid family in terms of require- the unit based on locally measured current and voltage,
ments and challenges. and their byproducts. The microgrid’s SC-EMS generates
DER reference signals, which are needed to coordinate the
operation of DER units. In the macrogrid-connected mode,
B. DER Units
SC-EMS also has the option to receive information from the
DER units are divided into electronically coupled and macrogrid, e.g., control commands, to adjust the microgrid
rotating-machine-coupled units. The main reason for this internal operation.
categorization is that the interface medium exhibits the 2) Islanded Mode: In this mode of operation, LCs and
dominant characteristics of a unit in terms of protection SC-EMSs are responsible only for the internal operation
and control. of the microgrid. Control strategies for islanded and mac-
1) ECDG Units: In this group, an alternating current/ rogrid-connected modes of microgrid are often different,
direct current (ac–dc) voltage–sourced converter (VSC) e.g., in macrogrid-connected mode all DER units can pos-
provides the interface medium between the source and sibly operate in PQ-control mode while after transition to
the microgrid [11], [12]. Depending on the power genera- islanded mode, a specific set of DER units need to operate in
tion capacity of the source, the VSC is either three phase or voltage-frequency control mode, and such a control transi-
single phase. The adopted DG technologies that use VSC- tion must be determined and managed by SC-EMS.
based coupling include solar-photovoltaic (PV) systems, 3) Mesogrid Mode: We define a mesogrid as a cluster of
Type-4 wind-turbine-generator systems, and medium- and microgrids, each including its LCs and SC-EMSs, which are
high-speed gas-turbine-generator systems. An ECDG unit is connected to a host macrogrid and operate in a prespeci-
either a nondispatchable unit or a dispatchable unit. In a fied coordinated manner. Each microgrid, within the mes-
nondispatchable unit, the primary source is often controlled ogrid, operates as a virtual power plant based on the control
and operated under a maximum-power-point-tracking ­commands that are generated by its SC-EMS and commu-
(MPPT) strategy [13]. nicated to LCs within the microgrid. The second (higher)
2) Rotating-Machine-Coupled DG Units: This category layer of the SC-EMS, i.e., the mesogrid SC-EMS, receives
adopts rotating machines as the coupling medium and information from the macrogrid and from each individual
includes the following: microgrid and, on that basis, specifies control/operation
• type-1 wind-turbine-generator units, which utilize responsibilities of each microgrid with respect to its PCC.
squirrel cage induction machines (SCIMs);
• type-3 wind-turbine-generator units, which incor-
D. DC Microgrid
porate doubly-fed induction machines (DFIM);
• diesel-generator units, which adopt filed-controlled Although high-voltage dc (HVDC) technologies have
SMs; been extensively utilized, the concept of medium-voltage
•  conventional (low-speed) gas-turbine-generator (MV) and low-voltage (LV) dc microgrid and its protection
units, which use field-controlled or permanent- requirements are still in infancy. In contrast to ac micro-
magnet SMs. grids, which have been contemplated for a wide range of
existing ac systems, dc microgrids have been considered and
investigated only for specific applications, including: 1) LV
C. Microgrid Control (up to 1500-V dc) data centers [15]; 2) MV (up to 35-kV dc)
Control strategies for microgrids include distributed, systems at electric ships [16]; and 3) MV collector system
centralized, and hierarchical controls [14]. Some inherent of offshore wind power plants. Recent technical literature
characteristics of the microgrid—such as high degree of proposes LVDC distribution systems as dc microgrids for
imbalance and the diversity of DER units—impose chal- building applications [17]. Protection issues are among the
lenges to devise a viable control for all operating scenarios. main challenges that have hindered widespread deployment
Furthermore, accommodating the two distinct operat- of dc microgrids.
ing modes (that is, the islanded and macrogrid-connected
modes) and the requirement for transition between these III.   M IC RO GR I D PRO T EC T ION
modes further adds to the complexity of the microgrid con- CH A R AC T ER IST ICS A N D A D OP T ED
trol. Since controls of electronically coupled units have large T E ST M IC RO GR I D
impact on the microgrid transient behavior, particularly Although a significant amount of research has been carried
subsequent to faults, they significantly affect protection out on microgrid protection, a protective relay that takes

Proceedings of the IEEE   3

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

Hooshyar and Iravani: Microgrid Protec tion

microgrid characteristics into account has not been manu- and 60 Hz. This solidly grounded system is made up of two
factured to date. The following issues make protection of microgrids—namely Microgrid 1 and Microgrid 2—that are
microgrids rather uniquely challenging. radial, but meshed topologies also can be obtained by clos-
1) The fault current levels of microgrids are hugely ing switches S1, S2, and S3. The detailed specifications and
different in the macrogrid-connected and islanded the power flow results of the system can be found in [19]. In
modes. this paper, ​Rij​denotes the relay next to bus i​​looking toward
2) The presence of generation units and complex bus j​.​ The relays’ sampling rate is 5 kHz, and the full-cycle
operating scenarios in macrogrids together with discrete Fourier transform is used for phasor measurement.
the possibility of widespread blackouts as a result Three DG units are added to the benchmark system.
of macrogrid protection failures justify major 1) DG1: This is either an ECDG or an IMDG unit rated
investment to develop sophisticated protection at 4 MW and connected to bus 6. The specifications
systems for macrogrids. Depending on its size of the ECDG and IMDG units are as follows.
and configuration, a microgrid also can raise pro-   – ECDG: The full-scale VSC of this unit is controlled
tection challenges similar to those observed for by the rotating reference frame control system [11],
a macrogrid. Some microgrid features—such as and its fault current is limited to 1.2 pu. The inter-
potentially large phase imbalance—can make its face transformer is rated at 5 MVA, 4.16/12.47 kV,​
protection even more difficult. However, the fail- X =​0.07 pu, with dYG winding connection. The
ure of a microgrid during a fault does not impact source and the converter that provides the dc volt-
a wide area, thereby limiting the investment in age are modeled by a controllable current source
microgrid protection. because the dynamics on the source and grid sides
3) A significant portion of generation units are of an ECDG unit are effectively decoupled through
expected to be interfaced with a microgrid through the dc link capacitor during the short span of a
VSCs, which exhibit unconventional fault behavior, fault.
including small short-circuit currents.   – IMDG: This wind unit includes either an SCIM or
The North American version of the CIGRE MV bench- a DFIM. The controllers of the DFIM are similar
mark system, shown in Fig. 2, exemplifies the above features. to the system developed in [20], unless otherwise
1) Its short-circuit ratio can exceed 20. stated. The transformer of the unit is rated at 5
2) The combination of different switch statuses can cre- MVA, 0.575/12.47 kV, ​X =​0.07 pu, with dYG wind-
ate topologies similar to those of transmission systems. ing configuration. The turbine model is similar to
Moreover, the current imbalance factor (defined by a commercial wind turbine [21], and the nominal
[18]) can reach 20% due to single-phase laterals. power is achieved at wind speed ​​Vw​ ​ =​13 m/s.
3) More than half of the load can be supplied by VSC- 2) DG2: This unit includes two 4.3 MVA SMs connected
based units added to the original system. to bus 8 through a transformer rated at 10  MVA,
Therefore, this paper uses the PSCAD/EMTDC simula- 4.16/12.47 kV, ​X =​ 0.07 pu, with dYG winding con-
tion of this mesogrid, which has been derived from a real nection. The subtransient reactances of the SMs are​​
MV network that covers a small town and its neighboring X​​d″  ​ ​ =​ 0.18 and ​​X​q​ ″  ​ ​ =​ 0.21 pu. DG2 can supply the
rural area. The rated voltage and frequency are 12.47 kV entire load, making the system capable of operating
off-grid. DG2 is disconnected during the macrogrid-
connected mode.
3)  DG3: This unit is similar to DG2, but has only
one SM.

I V.   D G OPER AT ION DU R I NG FAU LTS

From a protection perspective, a microgrid has a mix
of transmission and distribution system features, such
as bidirectional fault currents and the need for low-
cost relays. Therefore, a review of DG units’ response
to faults—irrespective of the DG type—in transmission
and distribution systems casts light on the expected fault
behavior of DG units in microgrids.

A. DG Fault Ride-Through
The integration of large renewable energy power plants
Fig. 2. CIGRE benchmark MV system [19]. with transmission systems has led to grid codes mandating that

4  Proceedings of the IEEE

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

Hooshyar and Iravani: Microgrid Protec tion

time curve defining the fault ride-through requirement

needs to be revisited for microgrid applications because the
response time of protective relays at the distribution level
is normally longer than that in transmission systems, and
the stability margins of a VSC-dominated microgrid is not
necessarily the same as those of bulk power grids.

B. DG Power Factor
Unlike conventional SM-based sources, the fault cur-
rent angle of a VSC-interfaced source—either ECDG or
Fig. 3. Fault ride-through curve for renewable sources in the
DFIM unit—is determined by the VSC’s control system.
European Union [22].
Therefore, besides the fault ride-through capability, some
transmission grid codes lay down specific requirements
renewable energy sources remain connected to the grid during for the power factor (PF) of renewable energy plants dur-
faults. This requirement, commonly referred to as fault ride- ing faults. In order to boost the voltage stability of system,
through, is defined through a stepwise/linear voltage versus European grid codes typically require generation of reac-
time-after-fault curve similar to the one shown in Fig. 3 [22]. As tive current following the onset of a fault. For instance, the
long as the operating point at the point of connection (PC) of a reactive current versus PC voltage of a renewable energy
renewable energy plant is above this curve, the plant must not plant in Spain must be situated inside the shaded area of
be tripped. The parameters along the voltage and time axes in Fig. 4 [27]. On the other hand, in terms of the reactive
Fig. 3 vary among different grid codes and are identified based current generated by a renewable energy plant, North
on various factors, including the characteristics of the protec-
American grid codes do not differentiate between fault
tion system. For example, ​​tc​ lear​ is normally around 150  ms,
and normal condition, during which a 0.95 plus positive-
which is about the maximum time taken by the primary protec-
sequence PF at the PC is typically required [28].
tion in transmission systems to clear a fault. During this period,
At the distribution level, a close-to-unity PF is usually
the renewable energy plant must be able to withstand close-to-
zero voltages. maintained by DG units during faults, since the voltage is
In contrast, distribution systems have generally fol- supported by the grid [24]. This practice can be followed
lowed the guidelines of the IEEE Standard 1547 in terms in a macrogrid-connected microgrid as well. However, reac-
of DG integration, which prohibits the formation of unin- tive current support by the DGs may be required to maintain
tentional islands energized by DG units [23]. Among a voltage stability during the islanded mode, resulting in low
variety of anti-islanding protection techniques, the most PF at the PC of a DG unit over faults. As a result, the protec-
commonly adopted is the active approach in which a DG tion system of a microgrid might face different PFs during
unit checks the response of the grid to a small disturbance macrogrid-connected and islanded modes.
created by the interface VSC. The disturbance signal is It is to be noted that the PF of a DG unit during a fault
effectively nulled if the grid fully supports the voltage at may vary due to the response time of the unit’s control
the DG’s PC, and is amplified otherwise. A fault in the system to regulate the current angle. For instance, the
vicinity of a DG unit prevents such voltage support by renewable energy plants that follow the curve of Fig. 4 are
the grid, hence tripping of the unit. A recent study shows allowed to have high PFs or even absorb reactive power in
that commercially available VSCs by six different vendors the first 150 ms of a fault. Thus, the protection system of
trip within ten cycles after the inception of a fault in their a microgrid must be able to cope with a wide PF range for
proximity [24]. DG units.
Meanwhile, if a fault-resilient microgrid of a decent size
and complexity is to be materialized, then DG units must
remain connected to the microgrid during disturbances in
the same way that large power plants satisfy the fault ride-
through requirement of transmission grid codes. DG units
with full-scale VSCs can ride through faults using a breaking
chopper circuit that restricts the VSC’s dc link voltage [25].
In addition to a chopper circuit, the fault ride-through capa-
bility for type-3 wind turbines is made possible through a
crowbar circuit that shorts the rotor winding of the DFIM
to limit the overcurrents passing through the rotor-side Fig. 4. Reactive current versus voltage for renewable sources in
VSC [26]. It should be mentioned that the voltage versus Spain [27].

Proceedings of the IEEE   5

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

Hooshyar and Iravani: Microgrid Protec tion

V.   EC D G U N I TS fault resistance is high, a positive-sequence directional ele-

This section discusses the effect of ECDG units on commer- ment is known to be reliable for all fault types [32]. Suppose,
cially available relay. DG1 is an ECDG unit unless otherwise for instance, that DG1 in Microgrid 1 of the test system is
stated. replaced by an SM-based DG unit generating 4 MW at unity PF.
A symmetrical fault in the islanded mode with 2​Ω​ fault
resistance at ​t =​ 2 s at bus 10 is reverse for both R45 and
A. Directional Overcurrent Relays R83 when switches S1–S3 are open. Therefore, the positive-
1) Directional Feature: The capability to identify the sequence torque angles of both relays lie correctly inside the
fault direction is a basic prerequisite for selective protection second quadrant in Fig. 5(a), identifying a reverse fault and
in a nonunit relaying scheme for a microgrid with bidirec- preventing unnecessary tripping of the feeder that includes
tional fault currents. A diversity of techniques can be used bus 6. However, for exactly the same fault and microgrid
to detect fault direction. The following inspects the per- condition, if DG1 is an ECDG unit, the positive-sequence
formance of the most common directional elements in the torque angles of the two relays change to the curves shown
presence of an ECDG unit. in Fig. 5(b). ∠​ ​T​ +​of R45 is inside the second quadrant and a
a) Positive-sequence directional element: This ele- reverse fault is correctly detected. For R83, however, ​∠​T​ +​​
ment is used either as the main directional element of a remains in the first quadrant, hence erroneous detection of
relay or in conjunction with other directional elements a forward fault and relay tripping.
[29], [30]. One approach to realize this element, referred to The contrast between the operation of R45 and R83 for
as the torque-based approach, is through the SM-based and ECDG units can be explained in terms
of the positive-sequence power generated by the DG unit
​ T​​ +​  =  ​V​ +​ I​ ​ +​ cos (∠​V​ +​− (∠​I​ +​+∠​Z​ +​ ) )​ (1) and consumed by loads during the fault. Before the fault,
the active power of the load connected downstream from
in which “+” superscript denotes positive-sequence quan- bus 8 is less than the 4-MW rated power of the DG units.
tities,​T​is torque, and ​Z​is the impedance of the protected Thus, the positive-sequence active power measured by both
element. The argument of the cosine term is denoted by R45 and R83 is negative before the fault. Meanwhile, since
​∠​T​ +​​and is termed the positive-sequence torque angle. the DG at bus 6 (either the SM-based or the ECDG unit)
A fault is in the forward direction if ​​T​ +​ is positive [ 31 ]​. There operates at unity PF, the positive-sequence reactive power
are impedance-based implementations of positive-sequence measured by the relays is positive.
directional element as well [29], but the underlying logic After the fault inception, the large current of the SM
behind both the torque- and impedance-based approaches results in the SM’s elevated active and reactive powers,
is the positive-sequence torque angle: ∠​ ​T​ + ​is in the first or shown in Fig. 6(a). In the meantime, the reduced voltage
fourth quadrants during a forward fault, and is inside the drop decreases the power consumed by loads. Therefore, the
second or third quadrants for a reverse fault. active power measured by R45 and R83 moves deeper into
Unless the angle between the two end voltages of a line is the negative territory, while the reactive power at the relay
excessively large—which is unlikely for a microgrid—or the location approximates to zero. Thus, the angle between the

Fig. 5. Positive-sequence torque angle of R45 and R83 for a

balanced fault at bus 10 when (a) DG1 is replaced by an SM; and (b) Fig. 6. Positive-sequence power of the DG at bus 6 for the fault of
DG1 is an ECDG unit. (a) Fig. 5(a); and (b) Fig. 5(b)

6  Proceedings of the IEEE

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

Hooshyar and Iravani: Microgrid Protec tion

positive-sequence voltage and current is close to 180°. Given

the feeders’ 68.2° impedance angle, ​∠​T​ + ​is placed in the sec-
ond quadrant. The above analysis is supported by the corre-
lation between the ​∠​T​ +​and the power waveshapes shown in
Figs. 5(a) and 6(a), respectively.
The ECDG unit maintains the unity PF operation dur-
ing the fault. On the other hand, as opposed to the SM-based
unit, the active power of the ECDG unit decreases during fault
due to the limited fault current of the converter and the voltage Fig. 8. Negative-sequence torque angle of R45 and R83 if the fault
of Fig. 5(a) is changed from balanced to ABG.
drop caused by the fault [Fig. 6(b)]. The amount of load con-
nected to the feeder between R45 and the ECDG unit is small,
so the power measured by R45 is similar to that of the ECDG in which “−” superscript denotes negative-sequence quanti-
unit (with a negative sign). As a result, the angle between the ties. The argument of the cosine term is the negative-sequence
positive-sequence voltage and current of R45 is only slightly torque angle, denoted by ​∠​T​ −​. A forward fault results in posi-
less than 180°, and ​∠​T​ +​of R45 remains in the second quadrant tive ​∠​T​ −​​[31]. A negative-sequence directional element also
in Fig. 5(b). However, as the amount of load between the relay can be implemented using the measured negative-sequence
and the ECDG unit increases, a higher share of the ECDG impedance [33]. Both the torque- and impedance-based
unit’s active power is absorbed by the load, before it reaches negative-sequence directional elements rely on ​∠​T​ −​​, which
the relay. Therefore, the negative active power measured by remains within the [−90°, +90°] range during forward faults.
R83 is significantly lower than the power of the ECDG unit, A negative-sequence directional element is reliable
shifting the angle between the positive-sequence voltage and during unbalanced faults in microgrids with SM-based DG
current toward the middle of the second quadrant. As a result,​ units. For example, suppose the fault of Fig. 5(a) is changed
+
∠​T​  ​is in the first quadrant in Fig. 5(b). from balanced to phase A to phase B to ground (ABG). The
The performance of positive-sequence directional ele- fault is reverse for R45 and R83, so ​∠​T​ −​of both relays, plot-
ments of R45 and R83 is exacerbated if the ECDG unit ted in Fig. 8, stays outside the [−90°, +90°] region during
generates less than its rated power—which is not unlikely the fault. The constant ∠​ ​T​ −​​in Fig. 8, as opposed to the
if the ECDG unit is energized by an intermittent renewable nonuniform ​∠​T​ +​​ in Fig. 5(a), stems from the fact that an
source. Suppose, for example, that the ECDG unit in the SM includes only an impedance in the negative-sequence
above scenario generates 10% of its rated power when the circuit, making ​∠​T​ −​ independent of the DG power. Thus,
fault occurs. Due to the small DG power, both the active if a microgrid is energized by only SM-based units, then a
and reactive power of R45 and R83 are positive, so the phase negative-sequence directional element generally exhibits a
difference between the positive-sequence voltage and cur- superior performance during unbalanced faults.
rent of the each relay is in the first quadrant, and the fourth- Correct performance of a negative-sequence direc-
quadrant ∠​ ​T​ +​ in Fig. 7 leads to false detection of a forward tional element in the presence of ECDG units requires
fault. As a result, a positive-sequence directional element that the current angles of such units be similar to those
is not reliable in the presence of ECDG units, particularly of SMs. These angles are controlled by the ECDG unit’s
when the ECDG unit is energized by an intermittent source. interface VSC and its current control loop, depicted in
b) Negative-sequence directional element: This sec- Fig. 9. The current of the ECDG unit is determined by the
tion reviews the effect of ECDG units on a negative-sequence voltage across the VSC’s filter, i.e., ​V​t​ − ​V​g​in Fig. 9, where​
directional element. The torque-based implementation of a V​g​ is the grid voltage and ​​Vt​​ is the VSC’s ac voltage. ​​Vt​​​ is
directional element operates based on controlled by the PI controllers, the outputs of which are
compensated by the ​d​- and ​q​-axis components of ​​Vg​ ​.​ The
​ T​​ −​  =  ​V​ −​ ​I​ −​ cos (∠​−​V​ −​− (∠​I​ −​+∠​Z​ +​ ) )​ (2) feedforward compensation of the grid voltage is a basic fea-
ture of an ECDG unit’s control system because it decouples
active and reactive power generation, and also improves

Fig. 7. Positive-sequence torque angle of R45 and R83 for the fault
of Fig. 5(b) when the ECDG unit operates at 10% of its rated power. Fig. 9. Current control loop and the interface of an ECDG unit.

Proceedings of the IEEE   7

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

Hooshyar and Iravani: Microgrid Protec tion

Fig. 12. Negative-sequence torque angle at the ECDG unit’s PC

during the fault of Fig. 10.

negative-sequence torque angle for the relay at the unit’s PC,

shown Fig. 12, fluctuates widely during the initial transients
of the fault, and is finally settled outside the [−90°, +90°]
range. Thus, a reverse fault is incorrectly identified.
As a result of the negligible negative-sequence current of
Fig. 10. VSC voltages when the fault of Fig. 5(b) is changed from the ECDG unit at bus 6, ​I​ −​measured by many relays across
​​
balanced to ABG: (a) ​V​g​; and (b) ​Vt​ ​.​ the system, e.g., R45, for the above fault consists primar-
ily of: 1) the negative-sequence current resulting from load
imbalance, which exists during both normal and fault condi-
the transient response of a VSC, for example, during the tions, and can be significant for a microgrid [34]; and 2) the
startup process and grid disturbances [11]. negative-sequence current that flows as a result of the fault’s
To investigate the effect of the feedforward compen- unbalanced voltage applied to the loads. None of these cur-
sation of the grid voltage on the sequence components of rents provides useful information on fault direction, and
an ECDG unit, the balanced fault of Fig. 5(b) is changed their effect on the performance of a negative-sequence
to ABG. Fig. 10(a) shows ​​Vg​ ​​whose imbalance factor dur- directional element depends on fault and load conditions,
ing the fault is about 18%. As a result of the feedforward particularly the fault type and load PF.
compensation of the grid voltage, the power frequency con- The effect of fault type depends on how the load imbalance
tent of ​​Vt​​, displayed in Fig. 10(b), exhibits broadly similar is distributed among the three phases, and is thus, system spe-
imbalance. Therefore, the difference between ​​Vg​ ​​ and ​​Vt​​​, and cific. For Microgrid 1, the above ABG fault is reverse for R45,
subsequently the current on the LV side of the transformer, but ∠​ ​T​ −​​of this relay is at the center of the [−90°, +90°] range
shown in Fig. 11, is imbalance free. In contrast with the volt- in Fig. 13, incorrectly indicating a forward fault. If the fault
age, the current’s imbalance factor is less than 1%. is changed to phase-C-to-ground (CG), however, the negative-
This tiny unregulated negative-sequence component of sequence torque angle of the same relay lies outside the shaded
the current in Fig. 11 cannot be used to identify the fault area of Fig. 13, signifying a reverse fault.
direction. While the fault is in front of the ECDG unit, the PF of the load impacts the negative-sequence current
that flows through the load because of the fault’s unbal-
anced voltage. If the PF is lagging, then the load acts like
an SM in the negative-sequence domain. The lower is the
PF, the more similar is the load to an SM’s highly induc-
tive negative-sequence impedance. For instance, the PF of
the load connected downstream of R83 for the above ABG
fault is 0.88 lagging, and the load’s negative-sequence cur-
rent generated by the unbalanced fault voltage exceeds

Fig. 11. VSC current for the fault of Fig. 10: (a) instantaneous Fig. 13. Negative-sequence torque angle of R45 for ABG and
currents; and (b) magnitude of sequence currents. CG faults at bus 10.

8  Proceedings of the IEEE

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

Hooshyar and Iravani: Microgrid Protec tion

If a VSC includes a fourth leg to control the ground cur-

rent, and the winding configuration of the interface trans-
former is ygYG, then the ground current loop involves the
VSC. As a result, a zero-sequence directional element can
be at risk of malfunctioning similar to other sequence-based
directional elements. However, a four-leg VSC along with a
ygYG transformer is not very common.
Zero-sequence directional elements are not applicable
Fig. 14. Negative-sequence torque angle of R83 during the fault of to ungrounded microgrids, but if a microgrid is grounded,
Fig. 10 for different load PFs.
the interface transformer of a DG unit normally includes
either a YG winding or a grounding transformer on the
the negative-sequence current created by the load imbal- high-voltage (HV) side [35]. Therefore, the DG unit is pre-
ance. Thus, ​∠​T​ −​​of R83 is correctly placed outside the sent in the ground fault loop, but the VSC does not affect
[−90°, +90°] range in Fig. 14. Meanwhile, a unity PF for the the zero-sequence current. Consequently, a zero-sequence
load would put ∠​ ​T​ −​​close to the boundary of the forward and directional element can correctly identify the fault direc-
reverse zones. A capacitive load PF makes ∠​ ​T​ −​​ stay within tion if the fault involves the ground. For instance, ​∠​T​ o​​ of
the [−90°, +90°] range. For instance, if a shunt capacitor R45 during the ABG fault of Fig. 10 is situated outside the
is installed at bus 5 to maintain a minimum 0.95-pu volt- shaded area of Fig. 16, correctly detecting a reverse fault.
age downstream of R83 in the islanded mode, the PF at However, if the fault does not involve the ground, then any
R83 location changes to 0.92 leading. As shown in Fig. 14, zero-sequence current measured by relay is due to load
the negative-sequence torque angle becomes 67.8°, falsely imbalance, and cannot determine the fault direction. For
detecting a forward fault. example, if the ground is removed from the above fault,
It should be noted that most relays deactivate the nega- ​∠​T​ o​ of R45 is placed inside the [−90°, +90°] range in
tive-sequence directional element if the current’s imbalance Fig. 16, falsely indicating a forward fault.
factor falls below a threshold typically around 2%–20%. As d) Phase directional elements: Phase directional ele-
shown in Fig. 15, this threshold is higher than the imbal- ments have been the main directional element of electro-
ance factor at the ECDG unit’s PC for the fault of Fig. 10. mechanical relays and are still incorporated in many digital
As a result, fault direction should be determined by other relays. The torque relations governing phase directional ele-
directional elements. However, as demonstrated in the pre- ments are
vious section, other relay elements also can be affected the
fault behavior of an ECDG unit. Moreover, the negative- ​​ T​A​  =  ​VBC ​ ​− (∠​IA​ ​+∠​Z​ +​ ) )​
​ ​ I​ A​ ​ cos (∠​VBC (4)
sequence current that flows through the load can be signifi-
cant, exceeding a negative-sequence directional element’s ​​ T​B​​  = ​
VCA ​ ​​​− (∠​IB​ ​​​+∠​Z​​  +​  )  )​
​ ​​ ​IB​ ​​  cos  (∠​VCA (5)
threshold for imbalance factor; a representative case is the
current imbalance factor of R83 during the fault of Fig. 10, ​​ T​C​​  = ​
VAB ​ ​​​− (∠​IC​ ​​​+∠​Z​​  +​  )  ).​
​ ​​ ​IC​ ​​  cos  (∠​VAB (6)
which surpasses 30% in Fig. 15. A positive torque indicates a forward fault. Sequence
c) Zero-sequence directional element: The torque- currents were proved generally unreliable for detecting
based implementation of a zero-sequence directional ele- fault direction in a microgrid with ECDG units. Since phase
ment is based on currents comprise sequence currents, phase directional ele-
​​T​​  o​  = ​ V​​  o​ ​I​​  o​  cos  (∠​−V
​​ ​​  o​​− (∠​I​​  o​​+∠​Z​​  o​  )  )​ (3) ments are not reliable either. As a representative case, the
where “o” superscript denotes zero-sequence quantities, and​​ torque angles of the three phase directional elements of R83
T​​ o​​is positive during forward faults. The impedance-based for the ABG fault of Fig. 10 are shown in Fig. 17. Phase A
version of a zero-sequence directional element is similar to element detects the fault direction correctly, but the other
that of a negative-sequence directional element. two phase elements fail to do so.

Fig. 15. The imbalance factor during the fault of Fig. 10 at DG1 PC Fig. 16. Zero-sequence torque angle of R45 for ABG and AB faults
and R83. at bus 10.

Proceedings of the IEEE   9

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

Hooshyar and Iravani: Microgrid Protec tion

Fig. 17. Phase torque angles of R83 during the fault of Fig. 10.

2) Overcurrent Feature: A VSC limits its reference cur-

rents to protect the power electronic switches against over-
current. Therefore, low fault current is a main adverse effect
of an ECDG unit on relay sensitivity and has been exten-
sively studied in the literature [36]. Meanwhile, besides
small fault currents, ECDG units can create phase currents Fig. 19. ECDG unit current for bolted BCG fault at bus 6 during zero
whose relative magnitudes during faults are unconventional, PF operation. (a) Phase currents. (b) Sequence current angles.
potentially affecting any protection function that requires
faulted phase selection, e.g., single-pole reclosing/tripping
schemes. The following shows that the relative magnitude transformer, and is thus determined by the voltage and
of phase currents during unbalanced faults is a function of impedance of the zero-sequence circuit of the fault.
an ECDG unit’s PF during a fault and fault resistance. The angles of the sequence components of the current
Fig. 18(a) displays the current measured at the PC of measured with respect to phase A are plotted in Fig. 18(b).
DG1 for a bolted phase-B-to-phase-C-to-ground (BCG) fault The positive-sequence component of phase B current lags the
at bus 6 when the ECDG unit operates at unity PF during positive-sequence angle in Fig. 18(b) by 120°. Therefore, the
the fault. Although phase B is faulted, its current is 0.323 kA angle between the positive- and zero-sequence components of
less than phase A current. The smaller current for phase B phase B current is 151.6°; i.e., these two components partially
is due to the difference between the angles of the ECDG cancel each other out, leading to the reduced phase B current.
unit’s sequence currents. As discussed earlier, the current in For phase A, however, the angle between these two compo-
Fig. 18(a) consists mainly of the positive- and zero-sequence nents is less than 90°, and so they do not oppose each other,
components. The former is regulated by VSC’s control sys- resulting in larger current even though phase A is sound.
tem to obtain the desired positive-sequence PF, while the As shown in Fig. 19, if the ECDG unit’s PF is zero, gen-
latter flows through the ground connection of the interface erating rated reactive current during the same fault, then
the relative magnitude of the currents will change. Similar
to the fault contribution of SMs, the current is larger in
faulted phases. The angles of the zero-sequence currents
in Figs. 18(b) and 19(b) are almost the same. However, the
positive-sequence current is shifted backward by about 90°,
hence the two sequence components of phase A current
oppose each other by about 180° whereas the same angle for
phase B is less than 60°.
Reactive current generation by an ECDG unit does not
guarantee larger currents for faulted phase(s). Fault resist-
ance is known to affect the magnitude of fault current in all
three phases, but not the relative magnitude of the phase
currents. Suppose the BCG fault of Fig. 19 includes a modest
5 ​Ω​ground fault resistance. Fig. 20(a) displays that the cur-
rent of healthy phase A is 0.123 kA larger than the fault cur-
rent of phase C. The difference between the phase currents
in Figs. 19(a) and 20(a) stems from the phase difference
between the positive- and zero-sequence currents. While the
former is kept fixed by the VSC’s control system to generated
the rated reactive current during both faults, the latter varies
Fig. 18. ECDG unit current for bolted BCG fault at bus 6 during unity based on the ​X / R​ratio of the fault’s zero-sequence loop.
PF operation. (a) Phase currents. (b) Sequence current angles.

10  Proceedings of the IEEE

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

Hooshyar and Iravani: Microgrid Protec tion

ECDG units can make coordination of overcurrent devices

very difficult or impossible. Moreover, the intermittent
nature of renewable ECDG units, which causes variable
fault current levels, complicates overcurrent relay coor-
dination further. These problems have raised the need
for many microgrids to use relays that are not dependent
only on the current magnitude [38]. A common alterna-
tive to an overcurrent relay is a distance relay, which is
largely unaffected by the small fault current of ECDG units
and different short-circuit capacities during macrogrid-
connected and islanded modes of a microgrid. Therefore,
the industrial community has started contemplating uti-
lizing distance relays in active distribution networks
[39]. The infeed and outfeed currents of DGs and tapped
loads can cause distance protection to underreach and
overreach, respectively, but these problems can be
Fig. 20. ECDG unit current for BCG fault with 5 ​Ω​ground resistance addressed by increasing the number of relays in a microgrid
at bus 6 during zero PF operation. (a) Phase currents. (b) Sequence and also applying pilot protection schemes, such as direc-
current angles.
tional comparison blocking, using relatively inexpensive
power line carrier (PLC). However, the following demon-
strates that the impedance measured by a distance relay can
Fig. 21 illustrates parallel connection of the sequence cir- be adversely impacted by the unconventional fault behavior
cuits measured with respect to phase A for a BCG fault [37]. of ECDG units as well. These impacts can render existing
Kirchhoff’s voltage law (KVL) between the positive- and distance relays inapplicable even if appropriate number of
zero-sequence circuits yields relays is installed and the relays are linked through PLC.
1) Balanced Faults: This section focuses on the distance
​​ V​​ +​  =  V
​ ​ o​ − 3 ​R​g​ I​ ​ o​​ (7)
element of the relay at bus 6, R65, whose zone 1 is set to
in which ​​Rg​ ​ is the ground fault resistance. For ​​Rg​ ​  =  0​, ​​V​ +​​ cover the feeder between buses 6 and 3. Switches S2 and
and ​​V​ o​​have the same angle. The zero-sequence current S3 are maintained open, thus R65 measures the fault con-
flowing out of the ECDG unit leads the zero-sequence volt- tribution of the ECDG unit. The reactance covered by zone
age by the zero-sequence impedance angle of the interface 1 of R65, shown in Fig. 22(a) and modeled based on [40], is
transformer, which is irrespective of the fault resistance. 0.93 ​Ω​. The feeder is tripped without delay if the measured
Therefore, ​V ​ ​ o​​ advances in phase to align ​​V​ o​ − 3 ​R​g​ I​ ​ o​​ with​​ impedance enters zone 1.
V​​  ​​when R
+
​ ​g​ is not zero. That is why the zero-sequence cur- Consider a bolted balanced fault at bus 4 inside zone 1 of
rent of Fig. 20(b) leads that of Fig. 19(b) by approximately R65. The positive-sequence reactance between R65 and the
65°. As a result, the angle between the zero- and positive-
sequence components of phase C current is more than 136°,
and these two components partially cancel each other.

B. Distance Relays
In addition to insensitivity of overcurrent relays to
ECDG units’ small fault currents, the potentially huge
difference between the fault currents during the macro-
grid-connected and islanded modes of a microgrid with

Fig. 22. R65 operation during a bolted balanced fault at bus 4.

Fig. 21. Sequence circuits connection during a BCG fault. (a) Distance element characteristic. (b) PF measured by the relay.

Proceedings of the IEEE   11

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

Hooshyar and Iravani: Microgrid Protec tion

if the VSC control system changes the maximum allowed

fault current, which is not a characteristic of the fault and
the relay should be independent of.
In a similar fashion, if the ECDG unit’s PF is zero, the
generated reactive power is translated into an impedance
located along the reactance axis of R65 in Fig. 23(a). Since
the fault current of the ECDG unit is small, the reactance
seen by R65 is large and outside distance element’s first two
zones, leading to the relay’s failure to trip the feeder.
2) Unbalanced Faults: An ECDG unit’s active and reac-
tive power cannot describe distance relay performance dur-
ing unbalanced faults. This section presents an alternative
description based on the relations that determine the effect
of remote infeed currents on distance relays.
a) Phase-to-phase faults: The impedance measured
by a phase element, say the BC element, of a distance relay
Fig. 23. R65 operation during a balanced fault with 1 ​Ω​ fault for a zone-1 fault is given by

Z​BC​  =  ​Zf​​ + ​R​ph​ ( ​ ​ − ​I​CL​)

resistance at bus 4. (a) Distance element characteristic. (b) PF
​I​ ​ − ​I​ ​
measured by the relay. ​ ​ 1 + ​ ______
BR
​IBL
   CR   ​ 
​ 
​ (8)

 
​MBC
​ ​
fault is 0.91 ​Ω​. As demonstrated by Fig. 22(b), it does not
in which ​Zf​​ is the positive-sequence impedance between the
matter whether the ECDG unit attempts to operate at unity
relay and the fault, ​​Rph
​ ​is the fault resistance between phases
or zero PF; the PF at R65’s location is 0.34 lagging for both
B and C, and​L​ and ​R​subscripts denote local and remote quan-
cases. The failure of the ECDG unit to reach its target PF
tities. In a conventional system that is supplied by SMs, the
stems from the fact that the bolted balanced fault effectively
angles of local and remote fault currents are related to the
decouples the ECDG unit from the rest of the sources in the
angles of the prefault local and remote voltages [41]. Since the
microgrid. Therefore, the angle between the fault voltage
phase difference between the two-end voltages of a feeder is
and current phasors is not determined by the VSC’s c­ ontrol
small, particularly at the distribution level [42], the numera-
system, but is dictated by the passive circuit connected to
tor and the denominator of ​​MBC ​ ​ in (8) have broadly similar
the ECDG unit—that is, the feeder between bus 6 and the
angles. Thus, the effect of fault resistance is mainly along
fault. Hence, as shown in Fig. 22(a), the AG element of R65
the resistance axis, and can be ameliorated by increasing the
measures the precise fault impedance, regardless of the
resistive reach of distance element zones. The small fictitious
ECDG unit’s target PF. There is only a 0.01 Ω ​ ​overreach in
reactance added to the measured impedance as a result of
Fig. 22(a) due to the outfeed currents passing through the
the phase difference between the local and remote currents
loads at buses 5 and 6; however, this underreach is negligi-
is usually mitigated by slightly tilting the reactance element
ble and does not affect R65’s successful operation. Similar
in the clockwise or counterclockwise directions if the relay
correct performance is observed for other distance elements
at the power-sending or power-receiving terminal of the line,
of R65 as well.
respectively [43]. The tilting angle is specified by the prefault
When the same fault includes a 1 ​Ω​fault resistance, the
voltage angles.
impedance measured by R65 is expected to be the fault loop
Assume the balanced fault of Fig. 23 is changed to a BC
impedance plus the effect of the fault resistance, which
fault. Given the small 1 ​Ω​fault resistance, R65 is expected
is mainly along the resistance axis; i.e., the yellow area in
to measure an impedance within the yellow area in zone 1
Fig. 23(a). However, the fault resistance prevents the ECDG
of Fig. 24(a). However, when the ECDG unit operates at
unit from being decoupled from DG2. Since the circuit con-
unity PF, ​ZBC
​ ​ measured by R65 lies in the fourth quadrant.
nected to the ECDG unit is no longer passive, the unit is able
Therefore, the relay fails to trip the feeder. If the ECDG unit
to regulate the current angle at its PC. Thus, the target PF is
generates the rated reactive current, the measured imped-
reached in Fig. 23(b) within a couple of cycles after the fault
ance is located far from the 0.91 ​Ω​reactance between the
inception. For unity PF, the entirely active power generated
relay and the fault, outside zones 1 and 2, causing trip failure.
by the ECDG unit is translated into an impedance along the
The incorrect impedance measured by R65 stems from
resistance axis of the distance element in Fig. 23(a). The
the fact that the angle of the ECDG unit’s positive-sequence
measured resistance is inversely proportional to the fault
current, which influences the denominator of ​​MBC ​ ​for R65,
current magnitude. Since the fault current of an ECDG unit
is specified by the target PF and is independent of the pre-
is limited, this resistance is large and outside the first two
fault voltage angles. Conversely, the angles of the remote
zones of R65, so the relay fails to clear the fault. The imped-
currents of R65 are related to the prefault voltage angle,
ance measured along the resistance axis is directly affected

12  Proceedings of the IEEE

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

Hooshyar and Iravani: Microgrid Protec tion

by the considerable magnitude of ​​MBC ​ ​​due to the ECDG

unit’s limited fault current.
b) Phase-to-ground faults: The unreliable impedances
obtained by the phase element of distance relay were proved
to be caused by the appreciable and unpredictable phase
difference between the relay’s local and remote currents. This
section shows that the impedance measured by a ground dis-
tance element is affected by other factors as well. For exam-
ple, the zero-sequence current does not affect the impedance
given by a phase element, since the phase currents are sub-
tracted in M​
​​ BC​​​. However, the impedance given by a ground
element, say the AG element, during a zone-1 fault is

Z​AG​  =  ​Zf​​ + ​R​g​ ​(_______
​IA​ L​ + ​K​  ​ I​ ​ L ​)
​I​ ​ + ​I​ ​
​ ​  AL  oA  Ro   ​ ​​ 
  (9)

 
​MAG
​ ​
in which ​K​ o​is the zero-sequence compensation factor [41].
The zero-sequence component is present in both the numer-
ator and the denominator of ​​MAG ​ ​​in (9). Fig. 25(a) shows
the impedance calculated by the AG element of R65 when
the BC fault of Fig. 24 is changed to an AG fault. For unity
PF operation of the ECDG unit, the imaginary part of ​​ZAG ​ ​​ is
about 25% of the 0.9 ​Ω​reactance to the fault—pinpointed
by a yellow circle in Fig. 25(a). When the ECDG unit gen-
Fig. 24. R65 operation during a BCG fault at bus 4. (a) Distance erates the rated reactive current, ​ZAG ​ ​ moves outside zone 1
​
element characteristic. (b) Current angles for PF =1​. (c) Current
​
angles for PF ​= 0.
after the initial fault transients die out.

as they are determined by the SM-based DG2 at bus 8.

Therefore, the substantial phase difference between the
local and remote currents results in a large imaginary part
for ​​MBC
​ ​, hence a fictitious reactance added to ​​ZBC
​ ​​.
SMs generate both active and reactive power during a
fault, and so SM fault currents lag the voltage by an angle
below 90°. Thus, as depicted in Fig. 24(b), the local current
of R65 leads the remote current during unity PF operation
by a wide margin. Consequently, ​​MBC ​ ​ has a large negative
angle, adding a negative fictitious reactance to the imped-
ance shown in Fig. 24(a) that cannot be mitigated by tilting
the reactance element. Furthermore, the fourth-quadrant
impedance signifies a reverse fault; thus, a pilot protection
scheme, such as directional comparison blocking, cannot
rectify this problem either.
When the ECDG unit generates its rated reactive current
over the fault, the local current of R65 lags the remote cur-
rent in Fig. 24(c) significantly, adding a positive fictitious
reactance to the impedance plotted in Fig. 24(a). Therefore,
tilting the reactance element in the clockwise direction, i.e.,
the common mitigation for the effect of remote infeed cur-
rents when a distance relay is at the power-sending end of
a line [43] , would actually exacerbate the performance of Fig. 25. R65 operation during an AG fault at bus 4. (a) Distance
R65. In addition, for both unity and zero PF operation of an element characteristic. (b) Angles of local sequence currents for PF ​
ECDG unit, the adverse effect of fault resistance is amplified = 1​. (c) Angles of local sequence currents for PF ​= 0​.

Proceedings of the IEEE   13

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

Hooshyar and Iravani: Microgrid Protec tion

The zero-sequence current angle of the ECDG unit is not If the ECDG unit generates a lower-than-rated ​​I​ +L​ ​ dur-
affected by the VSC, and is related to the prefault voltage ing the fault, which is likely due to the renewable sources’
angle. Given that the power flow levels are not huge in a intermittency, the effect of ECDG unit’s PF on the relay is
microgrid, ​​I​ oL ​​and ​​I​ oR ​are almost in phase. On the other hand, overshadowed by ∠​ ​K​ o​. For instance, Fig. 26 also displays
all sequence components of an SM fault current (i.e., ​​IAR ​ ​​in the sizeable underreach of R65 when the ECDG unit gener-
(9)) are aligned for an AG fault. Therefore, since ​​I​ L​  is sub- + ates 10% of its rated capacity and ​∠​K​ o​is negative. It should
stantially smaller than the remote-end currents, the angle of​​ be noted that different relay performances in Figs. 25(a)
M​AG​​’s numerator is similar to that of ​​I​ oL .​ On the other hand, and 26 occur during exactly the same fault condition, for
the denominator of ​MAG ​ ​ can be written as ​I​ +L​  + (1 + ​K​ o​ ) ​I​ oL ​​. which a distance relay is expected to measure identical
Hence, the angle of the denominator of ​​MAG ​ ​​can deviate impedances.
from ​∠​I​ oL ​​based on: 1) the relative magnitude and angle 3) Effect of Intermediate Infeed: It was demonstrated
of ​​I​ oL ​​ and ​​I​ +L​ ​; and 2) the zero-sequence compensation fac- that the significant phase difference between the local and
remote currents of a distance relay in a microgrid with
tor, which depends on the conductor physics and feeder
ECDG units causes protection malfunction. Similar phase
spacing.Fig. 25(b) depicts that for unity PF operation of
differences also exist between local and intermediate infeed
the ECDG unit, ​​I​ +L​  leads ​I​ oL ​ by about 60° and causes ​​MAG ​ ​​ to
currents, and can adversely impact the performance of a dis-
have a negative imaginary part, resulting in the overreach
tance relay. The impedance measured by the BC element
of the blue graph in Fig. 25(a). In contrast, ​​I​ +L​ ​ lags ​​I​ oL ​by 30°
of a distance relay in the presence of intermediate infeed
when the ECDG unit generates the rated reactive current in currents is
Fig. 25(c), thus pushing the imaginary part of ​​MAG ​ ​​ toward
Z​BC​  =  ​Z1​ ​ + ​Z​2​ + ​Z​2​ ​(______
BL I​CL​)
​ ​ − ​I​Cif​
​IBif
the positive region and causing the slight underreach of the ​ ​ ​I​ ​ − ​
     ​ 
​ 
​ (10)
green graph in Fig. 25(a). 
 
​​K​​  o​​for the overhead lines in Microgrid 1 is ​2 . 965∠14 . 3​​°​. ​Mif​ ​
The positive angle of ​​K​​  o​​forces the imaginary part of ​​MAG ​ ​​​ in which ​​Z1​ ​​is the impedance between the relay and the
toward the negative range. That is why the overreach asso- intermediate infeed locations, Z ​ ​2​is the impedance between
ciated with unity PF operation of the ECDG unit in Fig. the infeed location and the fault, and ​if​subscript denotes
25(a) is substantial, but the underreach for zero PF opera- intermediate infeed currents. In a conventional system with
tion is relatively modest. The angle of ​​K​​  o​​is not always posi- SM-based sources, the angles of local and infeed currents
tive, particularly for cables and HV lines. For cables, for are related to the prefault voltage angles, and are thus close,
particularly for a microgrid. Therefore, the angle of ​Z2​ ​ ​Mif​ ​ in
example, ​∠​K​​  o​​can be as low as −90° [43]. A negative angle
(10) is similar to that of the line impedance angle, and inter-
for ​​K​​  o​​pushes the imaginary part of ​​ZAG ​ ​​​in the positive direc-
mediate infeed is known to cause relay underreach [44].
tion, thereby making distance relays underreach. Suppose
The above assumptions do not necessarily hold true for
the overhead lines between buses 3 and 6 of Microgrid 1 are
systems with ECDG units. Consider a bolted BCG fault at
changed to single-core aluminum TR-XLPE underground
bus 8 during the macrogrid-connected mode of the system.
cables, for which ​​K​​  o​  =  1 . 614∠− 34 . 2°​. The length of the The reactance between R65 and the fault is 1.73 ​Ω​, i.e., the
cable is chosen such that the impedance between buses 3 fault is outside the first two zones of R65, and this relay
and 6 remains the same. If the ECDG unit generates its should not trip. In addition, the intermediate infeed current
rated power at unity PF, i.e., the condition of the blue entering the fault loop at bus 3 is even expected to make R65
graph in Fig. 25(a), R65 measures the correct reactance underreach. However, Z ​ BC
​ ​of R65 enters zone 1 in Fig. 27(a)
to the fault in Fig. 26. For this case, ​​I​  +L​  ​​ leads ​​I​  oL ​​​  by about quickly after the fault inception, causing a false trip. The
56°,  forcing the relay to overreach. However, unlike incorrect tripping is due to the large phase difference
Fig. 25(a), the effect of ​​I​  +L​  ​​’s lead over ​​I​  oL ​​​  is canceled by the between the local and intermediate infeed currents at bus 3,
negative ​∠​K​​  o​​. shown in Fig. 27(b). The infeed currents lag the local cur-
rents of R65 by 82°, so the last term of (10) adds a negative
imaginary part to ​Z ​ ​BC​​. In addition, the magnitude of this
term is large due to the ECDG unit’s limited fault current,
making ​Z​BC​enter zone 1 even though ​Rg​ ​is small.

C. Line Current Differential Relay

Line current differential protection, which is increas-
ingly used in transmission systems, is immune to most of the
above-discussed problems associated with directional over-
Fig. 26. Operation of the distance element of R65 during the fault current and distance relays in the presence of ECDG units.
of 25 when the overhead lines are replaced by cable. Different forms of line current differential protection also

14  Proceedings of the IEEE

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

Hooshyar and Iravani: Microgrid Protec tion

during the balanced fault of Fig. 5, assuming that DG1 is

an SCIM-based wind turbine. Due to the constant flux link-
age theorem, an SCIM operates as a voltage source in the
positive-sequence circuit immediately after the onset of a
fault [48]. Consequently, during the first few cycles of the
fault, the response of the positive-sequence directional ele-
ment of R65 in Fig. 28(a) is similar to its response to a con-
ventional SM-based unit, and the relay correctly detects a
forward fault. After a few cycles, however, the voltage drop
demagnetizes the SCIM, and the generated active power
of DG1 declines significantly in Fig. 28(b). In addition, the
voltage drop decreases the reactive power generated by the
PF correction capacitors, resulting in negative ​Q​ +​​measured
at R65 location. As a result of the small ​P​ +​ and negative ​Q​ +​
, the phase difference between the positive-sequence volt-
age and current settles in the middle of the fourth quadrant.
Fig. 27. R65 operation for a bolted BCG fault at bus 8 during unity Subtracting the line impedance angle from this phase dif-
PF operation of the ECDG unit. (a) Distance element characteristic. ference makes ​∠​T​ +​ smaller than ​− 90°​, and R65 incorrectly
(b) Angles of local and intermediate infeed currents.
declares a reverse fault in the seventh cycle of the fault.
Assume DG1 is a DFIM-based wind turbine that deploys
a conventional control system [20], and operates at unity
have been suggested for microgrids in the literature [45].
PF. The voltage drop at the DFIM terminal induces large
However, line current differential protection is a pricey
overcurrents in the rotor winding. Therefore, the crowbar
solution because of the following.
circuit shorts the rotor winding and prevents damage to the
1) For the same manufacturer, the cost of a low-end line
rotor-side VSC. Once the crowbar circuit is activated, the
current differential relay is more than double the
DFIM is conceptually similar to an SCIM. The current of a
cost of a high-end directional overcurrent relay that
single-cage SCIM during a balanced fault is given by
includes communication-assisted tripping [46], [47].
2) A relay should be installed at every node of a protec- ​I​ ​
tion zone, otherwise the line current differential ​i​f​ (t )  = ​ ___ ac −t/​τac
1 − s
  ​  ​e​  ​  ​​ cos​((1 − s ) ​ω1​  ​ t + θ)​
protection incorrectly trips during normal operation. ​ ​    ​ 
​Id​ c​ −t/​τdc
  ​​ (11)
​ + ​___
 1 − s   ​  ​e​  ​  ​​ cos (θ )
Given the typically high number of load nodes in a
distribution system, the total number of required line
current differential relays throughout the microgrid ​ 1​  ​ is the fundamental angular frequency, s​​is the
in which ω
would be significant compared to other relay types. induction machine’s slip, ​τac​  ​ and ​τdc
​  ​ are the time constants
3) Line current differential relays need high-band-
width communication and cannot be linked through
PLC or high-latency wireless communication
technologies.
4) Last, but not least, line current differential relays are
normally equipped with local backup relays, such as
distance and directional overcurrent relays. The local
backup protection is activated during communica-
tion failure, or when the GPS signal is not present/
useful due to solar storms or GPS jamming devices.

V I.  I M D G U N I TS
The discussion on the effect of IMDG units on microgrid
protection for each relay starts with SCIM and is then
extended to DFIM wind DG units. DG1 is either an SCIM-
based unit or a DFIM-based unit.

A. Directional Overcurrent Relays

1) Directional Feature: Fig. 28. Operation of directional element of R65 for a balanced
a)Balanced faults: The following discusses the opera- fault at bus 4 when DG1 is an SCIM-based wind turbine. (a) Positive-
tion of the positive-sequence directional element of R65 sequence torque angle. (b) Positive-sequence power.

Proceedings of the IEEE   15

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

Hooshyar and Iravani: Microgrid Protec tion

of the ac and dc components of the fault current, and ​θ​is the reactive impedance. Therefore, a negative-sequence direc-
fault inception angle. tional element can reliably identify the direction of unbal-
An SCIM-based wind turbine rotates close to the syn- anced faults if the system is energized by SCIM-based units.
chronous speed, so the machine slip is almost zero and the Fig. 30(a) depicts the sequence currents recorded by R65
frequency of the fault current’s ac component in (11) is if the fault of Fig. 28 is changed to a BCG fault. The sig-
close to the nominal frequency. Conversely, the rotor speed nificant negative-sequence component displayed in this fig-
of a DFIM varies within a ±30% range around the synchro- ure results in a 30% plus imbalance factor during the fault.
nous speed commensurate with the wind speed, and ​(1 − s )​ Since the angle of this negative-sequence current is similar
in (11) is between 0.7 and 1.3. Therefore, the frequency of to that of an SM, ​∠​T​ −​ is correctly situated in the middle of
the current’s ac component is between 42 and 78 Hz for a the forward zone in Fig. 30(b).
60-Hz system. The current of the DFIM during the above A conventional DFIM control system regulates only
fault for a subsynchronous and a supersynchronous wind the positive-sequence current [20], making the negative-
speed is shown in Fig. 29(a), where the current frequencies sequence circuit of a DFIM similar to that of an SCIM, and
are close to 50 and 72 Hz. In contrast, the frequency of the rendering negative-sequence directional elements applica-
voltage, which is normally a memorized voltage for a positive- ble. Meanwhile, DFIM controllers have been improved to
sequence directional element, is 60 Hz. Commercial relays cope with phase imbalance through a negative-sequence
include a frequency tracking unit that measures the voltage current control loop that suppresses ​I​ −​ at the DFIM termi-
frequency and updates the digital filters used to calculate the nal [49]. Suppose DG1 is a DFIM-based wind turbine that
voltage and current phasors accordingly. Meanwhile, since has negative-sequence current suppression capability, and
the current frequency is substantially different from the volt-
the fault of Fig. 29 is changed to a BCG fault. As opposed to
age frequency for this fault, the current phasors measured by
SCIM currents in Fig. 30, the negative-sequence current of
60-Hz digital filters are not correct, and the angles of the cur-
the DFIM in Fig. 31(a) decays to around zero within a cou-
rent phasors fluctuate. Conversely, the angle of voltage pha-
ple of cycles after the beginning of the fault. The negative-
sor is uniform, and so the ​∠​T​ +​ curves in Fig. 29(b) oscillate
sequence circuit of the DFIM is no longer represented by a
widely and are irrespective of fault direction.
single impedance, and the angle of the negative-sequence
If the voltage and current frequencies are measured
current fluctuates sharply. As a result, ​∠​T​ −​is not uniform in
separately and their respective digital filters are updated
accordingly, correct voltage and current phasors with fixed Fig. 31(b) and exits the forward zone 20 ms after the fault
angles are measured. However, since these phasors corre- inception, falsely detecting a reverse fault.
spond to different frequencies, their phase difference still 2) Overcurrent Feature: The magnitude of SCIM fault
rotates and a positive-sequence directional element fails to currents is often large enough to trigger overcurrent relays.
determine the fault direction. However, similar to ECDG units, the relative magnitude of
b) Unbalanced faults: The negative-sequence circuit DFIM fault currents depends on the DG unit’s PF and fault
of an SCIM is similar to that of an SM, and consists of a resistance. In addition, the magnitude of DFIM fault currents
is considerably impacted by the features and parameters of

Fig. 29. Operation of directional element of R65 for a balanced Fig. 30. Operation of directional element of R65 for a BCG fault
fault at bus 4 when DG1 is a DFIM-based wind turbine. (a) Phase A at bus 4 when DG1 is an SCIM-based wind turbine. (a) Sequence
current. (b) Positive-sequence torque angle. currents. (b) Negative-sequence torque angle.

16  Proceedings of the IEEE

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

Hooshyar and Iravani: Microgrid Protec tion

The faulted phase C has the largest current, while the cur-
rents of phases A and B are nearly equal. Meanwhile, if I​ ​ −​​is
not suppressed by the DFIM, both the absolute and rela-
tive magnitude of the fault ­currents change in Fig. 32(b).
The presence of the negative-sequence current and unity
PF for the ­positive-sequence circuit makes the current of
phase A larger than the currents in the other phase. For this
case, the crowbar circuit is activated if at least one of the
three phase rotor currents exceeds 1.2 pu. If the threshold
for activating the crowbar circuit is changed to 2 pu, the
magnitude of the positive-sequence current changes in Fig.
32(c), as a result of which the current declines in phase A
whereas it rises in the two faulty phases. The details of the
controllers’ effects on DFIM fault currents are outside the
scope of this paper; the objective of the above analysis is to
demonstrate that a relay that relies only on current mag-
Fig. 31. Operation of directional element of R65 for a BCG fault nitude in microgrids with DFIM-based wind DG units is
at bus 4 when DG1 is a DFIM-based wind turbine. (a) Sequence impacted by factors other than the fault condition, and is
currents. (b) Negative-sequence torque angle.
thus unreliable.

the DFIM control system, such as the negative-sequence B. Distance Relays

current suppression capability and the threshold for activat-
1) Balanced Faults: During balanced faults, the effect of
ing the crowbar circuit. Take, for instance, the DFIM phase
IMDGs on distance relays is similar to that on directional
currents for the BCG fault of Fig. 31, shown in Fig. 32(a).
relays, elaborated on earlier. For the fault of Fig. 28, the
impedance measured by the AG element of R65 is situated
at the expected location in Fig. 33 due to the voltage source
operation of an SCIM immediately after a fault. However,
once the SCIM starts absorbing reactive power, the negative​
Q​and positive ​P​are translated into impedances located in
the fourth quadrant, so R65 fails to see the fault.
For the balanced fault of Fig. 29 with the DFIM-based
DG1, the impedance measured by R65 exhibits a chaotic

Fig. 32. DFIM current for a BCG fault at bus 4. (a) With ​I​ −​​
suppression capability. (b) Without ​I​ −​suppression capability and
1.2-pu threshold for crowbar. (c) Without ​I​ −​suppression capability Fig. 33. Operation of the distance element of R65 for the fault of
and 2-pu threshold for crowbar. Fig. 23 when DG1 is (a) An SCIM; and (b) A DFIM.

Proceedings of the IEEE   17

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

Hooshyar and Iravani: Microgrid Protec tion

trajectory, which is irrespective of the fault location. The from power loss during normal operation. Solid-state break-
false impedance in this figure is caused by the incorrect cur- ers normally employ integrated gate commutated thyristor
rent phasors that are measured using 60-Hz digital filters, (IGCT) instead of insulated gate bipolar transistor (IGBT),
while the actual current frequencies are 50 and 72 Hz. If the which is common in VSCs, to reduce power loss over normal
current phasors are measured using 50- and 72-Hz digital operation. These losses can be largely eliminated through
filters, distance relay still misoperates because impedance hybrid breakers, which interrupt dc current through joint
must be measured using voltage and current phasors that operation of parallel-connected mechanical and solid-state
correspond to the same frequency. devices. Although effective, hybrid breakers are hard to eco-
2) Unbalanced Faults: Within the initial few cycles of an nomically justify for microgrids, particularly for LV systems.
unbalanced fault, i.e., when distance protection normally In general, solid-state and hybrid dc breakers have relatively
operates, an SCIM acts broadly similar to an SM. There- low technology readiness levels for LV and MV systems [52].
fore, a distance relay measures a relatively accurate imped-
ance for a zone-1 fault. However, if the fault is in zones B. DC Protective System
2 or 3, for which the relay introduces a delay, the SCIM
1) DC Faults: For dc fault and protection analysis,
starts absorbing reactive power, as a result of which SCIM
Microgrid 2 of the test mesogrid is converted to a two-wire
currents lead the remote and intermediate infeed currents.
1.5-kV dc system, with which DG3 and loads are integrated
Therefore, a distance relay that measures SCIM fault cur-
through VSCs. The following analysis includes both TN-S
rents is in danger of overreach.
and IT grounding methods. For TN-S method, the ground
A DFIM’s control system can regulate the PF at the DG
is in the middle of the dc capacitors of the VSCs. For IT
unit’s PC if an unbalanced fault is not very severe, and so
method, one of the poles is grounded. The buses are con-
the angle of DFIM fault currents is not determined by the
nected through underground cables, which are modeled
angle of prefault voltages. Therefore, distance relays are
using the frequency-dependent cable model of PSCAD, ena-
endangered by the effect of remote and intermediate infeed
bling traveling wave analysis.
currents during unbalanced faults in microgrids with DFIM
units in the same way that distance relays malfunction in There are basic differences between the features and
the presence of ECDG units. stages of dc and ac faults. The pole-to-ground (PG) fault
current measured at bus 13 of the system in the case of IT
V II. PRO T EC T ION OF D C M IC RO GR I DS grounding is shown in Fig. 34(a). The fault is bolted, and its
time and location are ​t =​1 s and 1 km downstream from bus
One of the chief impediments to the realization of dc grids
13. To illustrate how the fault evolves, it is not cleared in
arises from the protection challenges posed by the dc fault
Fig. 34(a). Unless the VSCs are modular multilevel, which is
behavior of VSCs. These challenges concern lack of cost-
not normally the case for microgrid voltage levels, the main
effective dc circuit breakers and fast, yet selective and sensi-
sources of a fault current immediately after the inception of
tive dc protective relays. Given the diversity of dc breakers
a fault are the dc capacitors of the VSCs. The capacitance
and their potential interaction with different components
of underground cables also supplies the fault, but their
of a dc system, the types and features of breakers should be
taken into account when developing the protection scheme
of a dc microgrid. Therefore, although the details of dc
breaker technologies are beyond the scope of this paper,
they are briefly reviewed in this section.

A. Circuit Breaker
A typical ac circuit breaker is a mechanical device that
opens at the current zero crossing. The absence of a zero
crossing in a dc renders ac breakers inapplicable to dc
microgrids, unless the breaker is rated for higher voltage and
current levels, which is not economical [50]. Nevertheless,
electromechanical breakers have been developed for dc sys-
tems, but these breakers cannot typically meet the speed
requirements for protection of VSCs. Thus, solid-state
breakers have been proposed for dc systems. These breakers
do not rely on a natural zero crossing of current and push
the current down to zero by applying a voltage that opposes
the voltage across the breaker. Solid-state breakers have sig- Fig. 34. DC fault current at bus 13 for a bolted fault. (a) IT
nificantly shorter operating times [51], but they often suffer grounding. (b) TN-S grounding.

18  Proceedings of the IEEE

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

Hooshyar and Iravani: Microgrid Protec tion

contribution is comparatively small considering the size of a

microgrid. During the initial capacitor discharge stage, the
current can undergo a large overshoot. As the capacitor dis-
charges, the primary source of the fault current is changed
to the contribution from the ac side of the VSC. This current
flows through the VSC’s free-wheeling diodes because the
IGBT switches are turned off quickly after the rise the fault
current. This large current can severely damage the diodes
and must be interrupted promptly. Fig. 36. Current for the fault of Fig. 34(b) if the ac side of each VSC
is opened after the fault inspection.
Fig. 34(b) displays the current at bus 13 for the same PG
fault if the grounding method is TN-S. The fault current is
quite smaller than the current of Fig. 34(a), but it is still 2) DC Relaying Scheme: The main attributes of an
significantly larger than the maximum current that can be effective ac protection system, such as speed, selectivity,
tolerated by the VSC diodes. Moreover, if the fault type is and sensitivity, are expected from a dc protection system
changed to pole to pole (PP), the eventual magnitude of as well. However, fulfillment of all of these requirements
the fault current approaches the PG fault current of IT sys- is not straightforward for dc systems. The speed of an ac
tems. Besides the large fault current, a TN-S system during protection system should be sufficiently high to maintain
PG faults can be impacted by the overvoltage at the healthy the stability of SMs, allowing for combined relay and
pole as well, which can reach up to twice the rated voltage. breaker operating times that are in the order of several
This overvoltage stems from the fault-induced change in fundamental frequency cycles. For a dc system, in con-
the ground reference of the system and can be mitigated by trast, the protection must be substantially faster since
means of a dc chopper circuit that balances the voltage at a VSC can be damaged if it contributes to the fault for
the two poles. more than a few milliseconds.2 Some VSCs can block a
Besides the short-circuit capacity of the system at the
dc fault current by generating a voltage that opposes the
ac side of the VSC and the total impedance of the dc fault
ac system voltage. To do so, extra modules are required,
loop, the parameters of a VSC, which are manufacturer and
making the VSC economically less feasible for microgrid
system dependent, have profound impact on the shape and
applications. In addition, it is desired that the fault be
magnitude of dc fault currents. The dc capacitor of the VSC,​​
isolated before the IGBT switches of a VSC are blocked
C​dc​, for the PG fault of Fig. 34(b) is 40 mF. Fig. 35 shows the because if the IGBTs are blocked, the controllability of
current for the same fault if ​Cdc ​ ​ =​ 20 mF. Since the capaci- the VSC will be lost and the restoration process after the
tor contribution occurs only at the beginning of a fault, the fault will take a longer time. Prevention of IGBT blocking
eventual magnitudes of the fault currents in Figs. 34(b) and to retain control over VSCs can impose a stricter speed
35 are similar; however, the current overshoots in these two requirement for dc microgrid protection.
figures differ by almost 1 kA. As a result of the low technology readiness level of dc
To smoothen current ripples and reduce the short- circuit breakers and absence of off-the-shelf digital dc relays,
circuit level, a series reactor, denoted by ​L​ dc ​ ​​, is often the common practice in existing dc systems is to trip the
inserted at the dc side of a VSC. Fig. 35 shows that if​​ breakers on the ac side of all of the VSCs once a dc fault
L​dc​ =​0.5 mH is applied during the PG fault of Fig. 34(b), occurs. DC disconnectors are then used to isolate the fault.
the current overshoot is reduced by about 1 kA. More As depicted in Fig. 36, if the ac side of each VSC is tripped,
importantly, the reactor changes the time constant of the the fault current includes only the capacitor discharge stage
fault loop, delaying the current rise and buying time for and finally declines to zero. This approach is normally fast
protection. Such system-specific features of fault signals enough to avoid damage to VSCs, but its selectivity is mini-
make it difficult to devise generic protection methods for mal because the entire dc system is likely to be lost for any
dc microgrids. fault.
Tackling this problem requires not only dc breakers,
but also dc relays that offer selectivity. Since dc protec-
tion should operate very fast, it is highly preferable, if not
imperative, that the relays use only local measurements to
eliminate the delay caused by communication. Overcurrent
protection is not an optimal choice to meet this objective
because of the following.

2
Some VSCs can block a dc fault current by generating a voltage that
Fig. 35. Current9 during the fault of Fig. 34(b) for different VSC opposes the ac system voltage. To do so, extra modules are required,
parameters and components. making the VSC economically less feasible for microgrid applications.

Proceedings of the IEEE   19

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

Hooshyar and Iravani: Microgrid Protec tion

Fig. 37. Current for the fault of Fig. 34(b) in the presence of fault Fig. 38.Current traveling waves for the fault of Fig. 34(b).
resistance.

1) The relay should be able to differentiate between successive traveling waves is not often easy. This issue has
in-zone and out-of-zone faults well before the fault been illustrated in Fig. 38, which displays the traveling
current reaches its final value. waves associated with the PG fault current of Fig. 34(b).
2) The selectivity and sensitivity of an overcurrent dc As shown in this figure, detecting the arrival time of the sec-
relay are severely impaired by fault resistance. This ond wave is made more difficult by a modest 10 ​Ω​ resistance.
problem is especially pronounced in microgrids This fault is 1 km away from the measurement location.
since the voltage levels are relatively low whereas Closer faults and presence of noise can further obscure the
the fault resistance can be primarily determined by second traveling wave. In addition to technical problems,
the ambient conditions at the fault location, which traveling-wave-based relays are among the most expensive
are independent of the voltage level. For instance, transmission system relays, and so are not economically
the current for the PG fault of Fig. 34(b) in the pres- viable for microgrids.
ence of 1 ​Ω​ and 10 Ω ​ ​fault resistance is shown in
Fig. 37. For ​R​f​  =  1 Ω​, the current can be mistaken for V III.   CONCLUSION
a fault in a farther protection zone. For ​Rf​​  =  10 Ω​ ,
the change in the current is tiny and similar to load In order for a microgrid to continue operation when a fault
variation even though 10 Ω ​ ​ is not known as a high occurs during the islanded mode, the DG units should ride
fault resistance. Fig. 37 demonstrates that, in addi- through the fault. Consequently, the relays have to identify
tion to overcurrent protection, fault resistance can the faulty component using the fault currents and voltages
also affect a relay that operates based on the rate of of DG units, which are often different from those of con-
current rise. ventional sources. The performance of commercially avail-
A common approach for fault location utilizes the able relays was examined for a benchmark test system. It
traveling waves caused by a fault [53]. This approach has was shown that the unconventional negative-sequence
been tailored for ac fault protection as well [54]. The exist- current, dependence of the relative magnitude of phase
ing traveling-wave-based protection methods, which are currents on the VSC’s PF and fault resistance, and limited
currently applied only at HV levels, mostly use communi- positive-sequence current are among the fault signatures
cation to find the arrival times of fault-induced traveling of ECDG units that impact directional overcurrent relays.
waves at the two ends of a line. The difference between the These units also can affect distance relays because the angles
two arrival times along with the wave propagation velocity of their fault currents are independent of the prefault volt-
can identify the fault location. This method can be modi- ages. Differential relays can overcome these problems, but
fied to use the measurements at one end of a line/cable at a considerably higher cost, the details of which were
only using outlined. Besides sharing some problematic features with
ECDG units, IMDG units also can cause relay misoperation
​t​ ​​  − ​t​ ​​ by making the fault current frequency dependent on the
​​x​f​​  =  ν ​ ____
2
 ​​1 
  (12)
2
induction machine’s slip. Furthermore, different DFIG con-
in which ​​x​f​​​is the distance between the relay and the fault, ​ν​ trol systems can lead to distinct currents for the same fault
is the wave propagation velocity, and t​​​ 1​​​ and ​​t2​ ​​​ are the arrival conditions, causing different relay responses. A brief review
times of the first two traveling waves. A major problem for of dc microgrids demonstrated that the stages of a fault and
applying this method to a dc microgrid is the difficulty in protection requirements are different for dc and ac systems.
precise identification of the second wave’s arrival time ​​t​2​​​, The effect of short lines/cables and fault resistance can make
particularly if the current is polluted by noise. Since feeder/ protection of dc microgrids more challenging than that of
cable lengths in a microgrid are typically short, separating other dc grids. 

20  Proceedings of the IEEE

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

Hooshyar and Iravani: Microgrid Protec tion

R EFER ENCES [19] Benchmark Systems for Network Integration [33] SEL-751A Feeder Protection Relay—Manual,
of Renewable and Distributed Energy SEL, Pullman, WA, USA, Jan. 2013. [Online].
[1] “Development and operation of active Resources, Standard C6.04.02, Apr. 2014. Available: https://www.selinc.com/SEL-751A
distribution networks,” CIGRE, Tech. Rep.
457, Apr. 2011. [20] R. Pena, J. C. Clare, and G. M. Asher, [34] E. Nasr-Azadani, C. A. Cañizares, D. E.
“Doubly fed induction generator using back- Olivares, and K. Bhattacharya, “Stability
[2] R. H. Lasseter, “Smart distribution: Coupled to-back PWM converters and its application analysis of unbalanced distribution systems
microgrids,” Proc. IEEE, vol. 99, no. 6, pp. to variable-speed wind-energy generation,” with synchronous machine and DFIG based
1074–1082, Jun. 2011. Proc. Inst. Electr. Eng.—Electr. Power Appl., distributed generators,” IEEE Trans. Smart
[3] “Microgrid-1, engineering, economics and vol. 143, no. 3, pp. 231–241, May 1996. Grid, vol. 5, no. 5, pp. 2326–2338, Sep. 2014.
experience,” CIGRE, Tech. Rep. 635, Oct. [21] K. Clark, N. W. Miller, and J. J. Sanchez- [35] Distributed Generation Technical Interconnection
2015. Gasca (Apr. 2010). Modeling of GE Wind Requirements—Interconnections at Voltages 50
[4] S. H. Horowitz and A. G. Phadke, Power Turbine-Generators for Grid Studies General kV and Below, Hydro One Netw., Toronto, ON,
System Relaying, 3rd ed. New York, NY, USA: Electric (GE) Version 4.5. [Online]. Available: Canada, Mar. 2013.
Wiley, 2008. http://ceee.hust.edu.cn/yuanxm/973/ [36] H. H. Zeineldin, Y. A.-R. I. Mohamed, V.
[5] A. A. Memon and K. Kauhaniemi, “A critical images_1/2012082054141805.pdf Khadkikar, and V. R. Pandi, “A protection
review of AC microgrid protection issues and [22] (Mar. 2013). ENTSO-E Network Code for coordination index for evaluating distributed
available solutions,” Electr. Power Syst. Res., Requirements for Grid Connection Applicable to generation impacts on protection for meshed
vol. 129, pp. 23–31, Dec. 2015. all Generators, European Network of distribution systems,” IEEE Trans. Smart
[6] S. A. Hosseini, H. A. Abyaneh, S. H. H. Transmission System Operators for Electricity Grid, vol. 4, no. 3, pp. 1523–1532, Sep. 2013.
Sadeghi, F. Razavi, and A. Nasiri, “An (ENTSO-E), Brussels, Belgium. [Online]. [37] P. M. Anderson, Analysis of Faulted Power
overview of microgrid protection methods Available: http://networkcodes.entsoe.eu/ Systems. Hoboken, NJ, USA: Wiley, 1995, ch. 6.
and the factors involved,” Renew. Sustain. wp-content/uploads/2013/08/130308_Final_
Version_NC_RfG1.pdf [38] IEEE Guide for Design, Operation, and
Energy Rev., vol. 64, pp. 174–186, Oct. 2016. Integration of Distributed Resource Island
[7] L. Che, M. Khodayar, and M. Shahidehpour, [23] IEEE Application Guide for IEEE Std Systems with Electric Power Systems,
“Adaptive protection system for microgrids: 1547(TM), IEEE Standard for Standard, 2011.
Protection practices of a functional Interconnecting Distributed Resources with
Electric Power Systems, Standard, 2009. [39] A. Sinclair, D. Finney, D. Martin, and P.
microgrid system,” IEEE Electrific. Mag., vol. Sharma, “Distance protection in distribution
2, no. 1, pp. 66–80, Mar. 2014. [24] F. Katiraei, J. Holbach, and T. Chang, systems: How it assists with integrating
[8] E. Sortomme, S. S. Venkata, and J. Mitra, “Impact and sensitivity studies of PV distributed resources,” IEEE Trans. Ind. Appl.,
“Microgrid protection using communication- inverters contribution to faults based on vol. 50, no. 3, pp. 2186–2196, May 2014.
assisted digital relays,” IEEE Trans. Power generic PV inverter models,” Ontario Grid
Connection Study, Toronto, ON, Canada, [40] Line Distance Protection REL670 Application
Delivery, vol. 25, no. 4, pp. 2789–2796, Oct. Manual, document 1MRK 506 315-UENABB,
2010. Tech. Rep., May 2012.
Vasteras, Sweden, Dec. 2012. [Online].
[9] H. H. Sharaf, H. M. Zeineldin, and E. [25] J. F. Conroy and R. Watson, “Low-voltage Available: www.05.abb.com/global/scot/
El-Saadany, “Protection coordination for ride-through of a full converter wind turbine scot387.nsf/veritydisplay/1d6bacf573307830
microgrids with grid-connected and islanded with permanent magnet generator,” IET c1257b0c0046292a/$file/1MRK506315-
capabilities using dual setting directional Renew. Power Gen., vol. 1, no. 3, pp. 182–189, UEN_C_en_Application_manual__
overcurrent relays,” IEEE Trans. Smart Grid, Sep. 2007. REL670_1.2.pdf
to be published. [26] R. Cardenas, R. Pena, S. Alepuz, and G. [41] G. Ziegler, Numerical Distance Protection:
[10] “Protection of distribution systems with Asher, “Overview of control systems for the Principles and Applications, 4th ed. Erlangen,
distributed energy resources,” CIGRE, Tech. operation of DFIGs in wind energy Germany: Publicis, 2011, ch. 6.
Rep. 613, Mar. 2015. applications,” IEEE Trans. Ind. Electron., vol.
60, no. 7, pp. 2776–2798, Jul. 2013. [42] A. Carta, N. Locci, C. Muscas, and S. Sulis,
[11] A. Yazdani and R. Iravani, Voltage-Sourced “A flexible GPS-based system for
Converters in Power Systems: Modeling, [27] (May 2011). Procedure for Verification synchronized phasor measurement in
Control, and Applications. Hoboken, NJ, USA: Validation and Certification of the electric distribution networks,” IEEE Trans.
Wiley, 2010. Requirements of the PO 12.3 on the Response of Instrum. Meas., vol. 57, no. 11, pp. 2450–
Wind Farms and Photovoltaic Plants in the 2456, Nov. 2008.
[12] R. Teodorescu, M. Liserre, and P. Rodriguez, Event of Voltage Dips, Version 9. [Online].
Grid Converters for Photovoltaic and Wind Available: http://www.aeeolica.org/uploads/ [43] Manual: SIPROTEC, Distance Protection,
Power Systems. Hoboken, NJ, USA: Wiley, documents/1306-pvvc-n9-english.pdf 7SA522-V4.7, Siemens, Germany, Sep. 2013.
2011. [Online]. Available: https://www.downloads.
[28] (Dec. 2005). Interconnection for wind energy, siemens.com/download-center/Download.
[13] N. Femia, G. Petrone, G. Spagnuolo, and M. United States of America Federal Energy
Vitelli, Power Electronics and Control aspx?pos=download&fct=getasset&mandato
Regulatory Commission, Docket No. RM05-4- r=ic_sg&id1=DLA06_1186
Techniques for Maximum Energy Harvesting in 001, Order No. 661-A. [Online]. Available:
Photovoltaic Systems. Boca Raton, FL, USA: www.ferc.gov/EventCalendar/ [44] W. A. Elmore, Protective Relaying: Theory and
CRC Press, 2012. Files/20051212171744-RM05-4-001.pdf Applications, 2nd ed.New York, NY, USA:
[14] D. E. Olivares et al., “Trends in microgrid Marcel Dekker, 2004, p. 260.
[29] F-PRO 4000—Distribution Protection &
control,” IEEE Trans. Smart Grid, vol. 5, no. 4, Management Relay User Manual, ERLPhase [45] E. Casagrande et al., “A differential sequence
pp. 1905–1919, Jul. 2014. Power Technologies, Winnipeg, Canada, component protection scheme for microgrids
[15] G. AlLee and W. Tschudi, “Edison redux: 380 2013. [Online]. Available: http://www. with inverter-based distributed generators,”
Vdc brings reliability and efficiency to erlphase.com/downloads/manuals/F_ IEEE Trans. Smart Grid, vol. 5, no. 1, pp. 29–
sustainable data centers,” IEEE Power Energy PRO_4000_manual.pdf 37, Jan. 2014.
Mag., vol. 10, no. 6, pp. 50–59, Nov. 2012. [30] IRV Integrated Protection, Control & Metering [46] [Online]. Available: https://selinc.com/
[16] G. F. Reed, B. M. Grainger, A. R. Sparacino, IED, ZIV Grid Automation, ZIV Grid products/351/
and Z. H. Mao, “Ship to grid: Medium- Automation, Bilbao, Spain, 2012. [Online]. [47] [Online]. Available: https://selinc.com/
voltage DC concepts in theory and practice,” Available: http://www.gridautomation.ziv.es/ products/311L/
IEEE Power Energy Mag., vol. 10, no. 6, pp. doc-downloads/literature/manuals/ [48] C. B. Cooper, D. MacLean, and K. Williams,
70–79, Nov. 2012. BIRV1203Av14.pdf “Application of test results to the calculation
[17] E. Rodriguez-Diaz, F. Chen, J. C. Vasquez, J. [31] J. Roberts and A. Guzman, “Directional of short-circuit levels in large industrial
M. Guerrero, R. Burgos, and D. Boroyevich, element design and evaluation,” in Proc. 21st systems with concentrated induction-motor
“Voltage-level selection of future two-level Annu. Western Protective Relay Conf., loads,” Proc. Inst. Electr. Eng., vol. 116, no. 11,
LVdc distribution grids: A compromise Spokane, WA, USA, Oct. 1994. pp. 1900–1906, Nov. 1969.
between grid compatibiliy, safety, and [32] K. Zimmerman and D. Costello, [49] L. Xu and Y. Wang, “Dynamic modeling and
efficiency,” IEEE Electrific. Mag., vol. 4, no. 2, “Fundamentals and improvements for control of DFIG-based wind turbines under
pp. 20–28, Jun. 2016. directional relays,” in Proc. 63rd Conf. unbalanced network conditions,” IEEE Trans.
[18] IEEE Recommended Practice for Monitoring Protective Relay Eng., Mar. 2010, Power Syst., vol. 22, no. 1, pp. 314–323, Feb.
Electric Power Quality, Standard, 2009. pp. 1–12. 2007.

Proceedings of the IEEE   21

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

Hooshyar and Iravani: Microgrid Protec tion

[50] H. Pugliese and M. von Kannewurff, [52] Review of DC Circuit Breakers for Submarine Networks. London, U.K.: Spriner-Verlag,
“Discovering DC: A primer on dc circuit Applications, Defence Science Technology 2010, ch. 1.
breakers, their advantages, and design,” IEEE Organisation, Victoria, Australia, [54] [Online]. Available: https://selinc.com/
Ind. Appl. Mag., vol. 19, no. 5, pp. 22–28, Sep. Mar. 2012. [Online]. Available: http://www. products/SEL-T400L/
2013. dst.defence.gov.au/sites/default/files/
[51] A. Maqsood and K. Corzine, “DC microgrid publications/documents/DSTO-TN-
protection: Using the coupled-inductor solid- 107420PR.pdf
state circuit breaker,” IEEE Electrific. Mag., [53] M. M. Saha, J. J. Izykowski, and E.
vol. 4, no. 2, pp. 58–64, Jun. 2016. Rosolowski, Fault Location on Power

A BOU T THE AU THOR S

Ali Hooshyar (Member, IEEE) received the B.Sc. Reza Iravani (Fellow, IEEE) received the B.Sc.,
degree from Isfahan University of Technology, M.Sc., and Ph.D. degrees in electrical engineering.
Isfahan, Iran, the M.Sc. degree in electrical engi- Currently, he is a Professor at the University
neering from the University of Tehran, Tehran, of Toronto, Toronto, ON, Canada. His research
Iran, and the Ph.D. degree in electrical engineer- interests include power system dynamics and
ing from the University of Waterloo, Waterloo, applications of power electronics in power
ON, Canada, in 2014. systems.
He was a Postdoctoral Fellow with the Cen-
tre for Applied Power Electronics, University of
Toronto, Toronto, ON, Canada. He joined the
Electrical Engineering and Computer Science Department, York Univer-
sity, Toronto, ON, Canada, in 2015. His research interests include protec-
tion and control of renewable energy systems and smart grids.
Prof. Hooshyar has been recognized as one of the exceptional review-
ers of the IEEE Transactions on Power Delivery and one of the best reviewers
of the IEEE Transactions on Smart Grid.

22  Proceedings of the IEEE