This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2017.2749450, IEEE
Transactions on Smart Grid
1

A Virtual Multi-Terminal Current Differential

Protection Scheme for Distribution Networks
With Inverter-Interfaced Distributed Generators
Bowen Han, Student Member, IEEE, Haifeng Li, Member, IEEE, Gang Wang,
Dehui Zeng, and Yuansheng Liang

 However, grid-connected DGs can also adversely affect

Abstract--The deployments of inverter-interfaced distributed power grids, particularly in terms of relay protection [5]-[7].
generators (IIDGs) in medium voltage distribution networks may The integration of decentralized DGs causes a radial MV dis-
cause undesirable trips or tripping failures of conventional two- tribution network to become a scattered multi-source system,
terminal current differential protective relays. It is economically
in which the short-circuit levels and fault current paths change
and technically unfeasible to reconfigure large numbers of time-
synchronized sampling clocks and communication channels for much more than those in radial distribution networks. In addi-
each terminal inside the protected zone. In this paper, a more tion, most renewable DGs are connected to MV distribution
accurate IIDG mathematical equivalent model is established that networks using power electronics interfaces, i.e., IIDGs. Due
considers the control strategies of IIDG, and the profile charac- to the fast responses of their controllers, IIDGs behave differ-
teristics of the positive-sequence voltage at each point of common ently from conventional synchronous generators, particularly
coupling are investigated. Then, a real-time estimation algorithm
during grid fault conditions [8], [9]. Moreover, the deployment
is introduced for the fault current contribution of grid-connected
IIDGs. On this basis, this paper proposes a novel virtual multi- of intermittent energy IIDGs in distribution networks leads to
terminal current differential protection scheme. This scheme only a certain reduction in the sensitivity and selectivity of the con-
necessitates exchanging the electrical quantities between the pro- ventional overcurrent protective relays that rely on the magni-
tective relays at both terminals via an existing two-terminal pilot tude of fault current, and thus these grid-connected IIDGs can
channel, thus reducing the cost and requirements for communi- even cause protection blinding or sympathetic tripping (also
cations. The feasibility of the proposed protection is demonstrat-
known as false tripping) of these relays [10]-[13]. Therefore,
ed in both the PSCAD/EMTDC simulation platform and the
RTDS hardware experimental laboratory. A comparison of the improved overcurrent protection schemes for distribution net-
proposed protection alongside the conventional protection is also works in the presence of DGs have been conducted [14]-[17].
presented. Test results indicate that the proposed protection is These methodologies, however, necessitate frequent modifica-
widely applicable to various fault conditions, thereby improving tions of the relay current settings when DG operation switches
the sensitivity and selectivity of conventional protection schemes. between grid-connected and islanded modes.
Additionally, to mitigate the effects of IIDGs on the relay
Index Terms--Current differential relays, current estimation,
protection system, Yazdanpanahi et al. [18] developed a new
inverter-interfaced distributed generators (IIDGs), medium volt-
age (MV) distribution network, virtual multi-terminal protection. control strategy to limit the fault current contribution of IIDGs
based on the voltage sag detected by each IIDG. However, the
I. INTRODUCTION ability of IIDGs to support the grid voltage would be deprived
because the IIDG current should be decreased to zero in the
T HE research and development of renewable clean energy
technologies has been universally acknowledged as one
of the most effective measures to alleviate the energy resource
case of severe voltage sag. Similarly, Margossian et al. [19]
explored the possibility of controlling the fault current levels
of IIDGs by setting the grid code parameters for inverters. The
crisis and improve the environment. Distributed generators
primary shortcoming of this scheme is the frequent necessary
(DGs), particularly inverter-interfaced DGs (IIDGs) that are
modifications of these parameters. Zhan et al. [20] presented
represented primarily by solar photovoltaic (PV) plants and
an optimal placement and sizing method for IIDGs to prevent
doubly-fed induction generators (DFIGs), play an increasingly
large-scale changes in original protection systems, but the
important role in modern power grids [1]-[3]. For instance,
constraints on the locations and capacities of IIDGs should be
many solar PV plants of more than 1.0 MW power in capacity
taken into consideration.
(which are usually integrated into medium voltage, MV, dis-
Furthermore, to shorten circuit outages and increase power
tribution networks) have been widely installed in Europe and
supply reliability, most feeders in modern distribution systems
many other countries and regions around the world [4].
are often divided into several smaller segments to isolate the
faulted section, and ensure other healthy sections remain ener-
This work was supported by the National Natural Science Foundation of gized [21], [22]. In this case, however, the existing conven-
China (No. 51477057). The authors are with the School of Electric Power En-
tional overcurrent protection or other improved adaptive over-
gineering, South China University of Technology, Guangzhou 510641, China
(email: [email protected]; [email protected]; [email protected]; da current protection schemes mentioned above may not be suit-
[email protected]; [email protected]). Corresponding author: Haifeng Li.

1949-3053 (c) 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2017.2749450, IEEE
Transactions on Smart Grid
2

able for the reasons that the time settings of these protective the protected line, only one communications channel is re-
relays must be delayed longer and protection settings coordi- quired for the proposed scheme. This scheme is therefore con-
nation is more challenging to implement. By contrast, current ducive to reducing the communication requirements in com-
differential protection appears to be a preferred choice for the parison with the physical multi-terminal current differential
MV distribution networks integrated with IIDGs because of its protection system, enabling the need for protective device
high sensitivity, adequate selectivity, high-speed tripping, and reconfigurations and protection system modifications to be
the wide applications of microprocessors and communication minimized. Additionally, the installation constraints associated
technology in modern distribution grids and microgrids [23]- with the IIDG placements and capacities can be eased, at least
[25]. In practical engineering, medium-scale IIDG plants can from the perspective of protection. This advantage contributes
be incorporated into power grids at the distribution level via to improving the flexibility of IIDG integration because IIDGs
substations or directly tapped to the grid using teed feeders [3]. equipped with various capacities are allowed to be installed at
The latter is also called the T-connected mode. In fact, with any location inside the protected zone.
the increasing density of distribution lines and the growing The remainder of this paper is organized as follows. The in-
shortage of overhead corridors in China, most medium-scale fluence of IIDGs on current differential protection are briefly
IIDG plants rated in the range of 1.0-10.0 MW, particularly analyzed in Section II. A notable estimation algorithm for the
those solar PVs installed in rural areas, are directly tied to MV fault current contribution of IIDG is presented in Section III.
distribution networks (10-35 kV) by teed feeders (Fig. 1) to On this basis, a virtual multi-terminal current differential pro-
reduce the cost of building substations at connecting points. tection scheme is proposed in Section IV. Then, the proposed
The conventional two-terminal current differential protec- protection scheme is tested in the PSCAD/EMTDC simulation
tion scheme, however, may also cause poor performance when platform and the real-time digital simulator (RTDS) hardware
IIDGs are connected to power grids using teed feeders. The experimental laboratory, and the test results are presented in
most critical reason is that the remote relays installed at both Section V. Finally, Section VI concludes this paper.
terminals of the protected line cannot directly extract the cur-
rent information from IIDGs. The point of common coupling II. IMPACTS OF THE IIDG ON CONVENTIONAL CURRENT
(PCC) is thus an isolated point for current differential relays. DIFFERENTIAL PROTECTION
Clearly, T-connected IIDGs have a further adverse impact on The limitations of the conventional two-terminal current
the protective relays, inevitably resulting in a challenging is- differential protection scheme, in the presence of IIDGs, are
sue for conventional protection schemes. briefly reviewed in the following subsections.
A seemingly feasible solution to this issue may be to estab-
lish a complete synchronous communication system and to A. Current Differential Protection Without Restraint
exchange the time-synchronized current data between multiple In the distribution network without IIDGs, the operation
terminals inside the protected zone, and multi-terminal current threshold current I∗K.set for current differential protection can be
differential relays can function adequately in this system [26]. set as a small value that is greater than the maximum errone-
Communication requirements, however, are extremely harsh. ous differential current IEDC.max, if the measuring errors from
For an N-terminal system, the escalated need for the synchro- ratio mismatch and unequal saturation of current transformers
nization of numerous sampling clocks and simultaneous data (CTs) are accounted for [27], [28]. In this case, the differential
poses a major challenge. Additionally, the number of commu- currents are much higher than the operation threshold current
nications ports and channels required greatly expand with in- I∗K.set during internal fault conditions, and the protective relays
creasing N. This expansion significantly increases costs and can reliably trip for these faults with the required sensitivity.
can even affect the communication reliability [27]. Moreover, However, as the typical schematic diagram shown in Fig. 1,
an established communication topology, if there is a new IIDG the current İIIDGk supplied by the k-th IIDGk (1 ≤ k ≤ n) is an
installation, must be updated accordingly. It is clear that this unknown current for the relays R1 and R2, with the result that
protection scheme is thus not a technically and economically the sum of these currents (i.e., | ∑ İIIDGk |) is similar to IEDC from
feasible solution for those MV distribution networks incorpo- the perspective of current differential relays. In this paper, the
rating scattered multiple IIDGs. direction that current flows from the bus into the protected line
This paper proposes a novel virtual multi-terminal current is regarded as the reference direction, and thus the differential
differential protection scheme for teed feeders used in multi- current value |İM +İN | is equal to | ∑ İIIDGk | under normal oper-
terminal distribution network along with IIDGs. The proposed
ating conditions or external faults.
protection operates on the sum of the currents from all termi-
R1 PCC1 f PCCn R2 R3 PCCn+1
nals. Fortunately, this protection does not require extra chan-
nels to directly transmit the local current information of each GRID M
IM
I IIDG1 I IIDGn
IN N
I IIDGn+1
IIDG to the remote protective relays that are installed at both Normal Current
Load
Fault Current
terminals of the protected line, and therefore this methodology Relay
...
Transformer IIDG1 IIDGn IIDGn+1
can be easily implemented by exchanging the simultaneous Fig. 1. Single-line diagram of distribution network with IIDGs using teed
electrical quantities between the local and remote terminals feeders and its current distribution during both normal and fault conditions.
via an already existing two-terminal communications channel.
Note that IIDG output current İIIDGk is constantly changing
This means that no matter how many IIDGs are connected to
in real time due to various meteorological conditions; however,

1949-3053 (c) 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2017.2749450, IEEE
Transactions on Smart Grid
3

the IIDG output current is typically limited to a maximum grid-connected interfaces for most commercial IIDG applica-
current allowed to prevent damage to inverter switches under tions [30], [31]. The schematic diagram of a grid-connected
grid fault conditions including internal and external faults [29], IIDG is depicted in Fig. 2. The IIDG system typically consists
[30]. Meanwhile, | ∑ İ IIDGk | must be taken into account in the of a dc voltage source, a three-phase pulse width modulated
operation threshold current to prevent the relays from unnec- (PWM) VSI, a three-phase output filter connected at the ac
essarily operating for external faults, and thus I∗K.set should be side of the VSI, and a transformer coupled at PCC for the in-
greater than the maximum value of | ∑ İIIDGk |+IEDC . Hence, the terconnection of the IIDG and power grid.
threshold current setting and the relay operating criterion can I d*. f , I q*. f switch time
≤ 20ms

GATE DRIVER
CONTROLLER
LVRT PCC
2

CURRENT
CONTROLLER dq
be given as R L

PWM
∑

I d* , I q* +
n POWER 1 -
CONTROLLER abc VSI GRID
I∗op ≥ I∗K.set = | ∑ İmax_IIDGk | +IEDC.max (1) Low-voltage
δ Vdc
detected IIIDG.d , IIIDG.q iabc

CALCULA-
k=1 Qref abc
-

POWER
+ dq
∑ δ

TION
+ - PLL
is the operating current equal to |İM +İN |; Imax_IIDGk is
Upcc.d , Upcc.q
I∗op
∑
abc uabc
where Pref dq

the maximum current of the k-th IIDG during external faults. Fig. 2. Control diagram of IIDG for both normal and LVRT conditions.
This threshold setting criterion can reliably prevent relays The fault characteristics of a VSI-based IIDG are mainly
false tripping and adequately guarantee their selectivity during determined by its control strategy. Thus, the impacts of control
normal operating condition and external faults. However, with strategy on the fault current contribution of IIDG must be
increasing IIDG penetration, I∗K.set also increases, and the sen- carefully taken into consideration [9], [32]. In addition, the PQ
sitivity of the relays is thus decreased, with the result that the (i.e., the active and reactive power) decoupled control strategy
relays employing this operating criterion may fail to trip for has been widely adopted by most grid-connected IIDGs to
internal faults in the presence of high IIDG penetration. supply the available natural power that is captured by the max-
B. Current Differential Protection with Restraint imum power point tracking (MPPT) controller, thereby realiz-
ing the full utilization of renewable clean energy [33]-[37].
Current differential protection with restraint can also be
For the controllers shown in Fig. 2, a Park transformation is
employed as a means of mitigating the effects caused by erro-
used to convert the ac voltages and currents in the abc natural
neous differential currents. The typical operating criterion is
frame to dc electrical quantities in the d-q frame that synchro-
I∗op ≥ kIres (2) nously rotates with the angular frequency of the grid voltage.
The grid voltage in the d-q reference frame, where the q-axis
where Ires is the restraint current equal to |İM −İN |, and k is the leads the d-axis, is defined as
restraint coefficient normally lower than 1.0. Upcc cos(δ)
It should be noted that during normal operation or external 2π 2π
Upcc.d 2 cos(θ) cos (θ - ) cos (θ + ) 2π
faults, the growing penetration of IIDGs tends to increase the [ ]= [ 3 3 ] Upcc cos (δ - )
Upcc.q 3 2π 2π 3
operating current (i.e., I∗op = |İ M +İN | = | ∑ İIIDGk |) while undesira- -sin(θ) -sin (θ - ) -sin (θ + ) 2π
[Upcc cos (δ + 3 )]
3 3
bly decreasing the restraint current (i.e., Ires = | İM −İN | = |2İM −
∑ İ IIDGk |). Thus, the increasing adoption of T-connected IIDGs (3)
throughout the distribution system may result in an unintended where Upcc.d and Upcc.q denote the d- and q-axis components of
trip for external faults. On the other hand, increasing the value the voltage at PCC, respectively; Upcc and δ denote the magni-
of k helps mitigate the impacts of IIDG during external faults tude and angle of the voltage at PCC, respectively; and θ is
but decreases the protection sensitivity and may even cause Park’s reference angle.
protection blinding for internal faults. The PQ decoupled control strategy involves transforming a
Briefly, the conventional two-terminal current differential three-phase abc natural reference frame to a d-q synchronous
protection is also not well suited for those complex distribu- rotating reference frame and aligning the d-axis with the volt-
tion networks incorporating multiple T-connected IIDGs since age at PCC for decoupled control [29], [31], [37], [38]. Thus,
the fault current contribution of these IIDGs is not taken into Upcc.q is equal to zero, which is easily implemented using a
account in the current differential relays. The following dis- phase-locked loop (PLL) to track the PCC voltage angle δ, i.e.,
cussion focuses on the fault current contribution of IIDGs. the Park’s reference angle θ in (3) equals δ. Consequently, the
Subsequently, a novel current differential protection scheme grid-side inverter respectively tracks the reference active and
coverage for a multi-terminal line is proposed. reactive currents to output the reference powers:

III. REAL-TIME ESTIMATION FOR IIDG OUTPUT CURRENT Pout = Pref = 3⁄2 ∙ Upcc.d I∗d
{ (4)
This section establishes a mathematical equivalent model Qout = Qref = − 3⁄2 ∙ Upcc.d I∗q
that considers IIDG’s control strategies and then proposes an
estimation algorithm for the IIDG output current. where Pref and Qref denote the reference active and reactive
power, respectively; Pout and Qout denote the active and reactive
A. Basic Modeling of a IIDG and its Equivalent Model power output from IIDG, respectively; and I∗d and I∗q denote the
Voltage source inverters (VSIs) are commonly applied as reference active and reactive currents which are generated by

1949-3053 (c) 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2017.2749450, IEEE
Transactions on Smart Grid
4

the power controller shown in Fig. 2 during normal operation. Iq.f = I∗q.f , ( I∗q.f ≤ Kmax IN_IIDG )
As shown in Fig. 2, a current controller is also applied in
{ 2 2 (7)
the VSI to generate modulating signals for the PWM modula- Id.f = min {I∗d.f , √(Kmax IN_IIDG ) − I∗q.f }
tor. Both the VSI output current and its corresponding refer-
ence current in d-q reference frame are fed into a differential where Id.f and Iq.f are the output active and reactive currents of
element, and the difference current is then delivered into a IIDG, respectively; IN_IIDG is the rated current of the inverter;
proportional-integral (PI) regulator embedded in the current and Kmax is the allowed maximum current coefficient.
controller, to regulate the VSI output current with zero steady- As previously mentioned, the PQ control strategy is usually
state error [29], [38]. At the same time, the VSI output voltage implemented by aligning the d-axis with the voltage at PCC.
described in the d-q reference frame can be obtained using the The relation between the current of IIDG (İIIDG ) and the volt-
following equation: age at PCC (U̇ pcc ) can thus be depicted in Fig. 3, in which α
did and δ denote the angles of U̇ pcc and İIIDG , respectively.
Vd Upcc.d R -ωL id
[V ] = [ ] +L [ dt ] + [ ][ ] (5) As can be seen from Fig. 3, an interesting relationship exists
q Upcc.q diq ωL R iq between İIIDG and U̇ pcc , as shown in the following equation:
dt
where Vd, Vq, id, and iq denote the d- and q-axis components of İIIDG = (IIIDG.d cosδ − IIIDG.q sinδ )+j(IIIDG.d sinδ +IIIDG.q cosδ ) (8)
the inverter’s voltage and current, respectively; R and L are the where IIIDG.d and IIIDG.q represent the d- and q-axis current com-
resistance and inductance of the filter, respectively; and ω is ponents, respectively.
the angular frequency of the grid, i.e., ω=dθ/dt.
A current control loop, by which the inverter’s d- and q-axis q-axis Im
currents are both fed back, is executed using the transfer func- U pcc d-axis
tion that is obtained by the Laplace transformation of (5). Fol-  I IIDG .d
lowing the inverse Park transformation of the output signals, a

cluster of three-phase sinusoidal PWM signals is generated to   Re (A-axis)
drive the power electronic switches. 
 I IIDG
Additionally, to improve the voltage stability of the power
grids tied with IIDGs, many grid codes worldwide (including I IIDG .q
Germany, Denmark, and China) have demanded that priority
Fig. 3. Relationship between the output current of IIDG and the voltage at
should be set on the reactive current being provided by medi- PCC shown in phasor form.
um- or large-scale grid-connected IIDGs (e.g., wind parks and
PV plants) to support the grid voltage when the voltage at Moreover, in response to most grid codes, grid-connected
PCC is lower than the predetermined value [39]-[41]. Hence, IIDGs are forced to ride through grid disturbances, including
the low voltage ride through (LVRT) control strategy must be unsymmetrical faults. It has been reported that the existence of
taken into consideration. Fig. 2 shows that the VSI should be negative-sequence voltages at PCC can cause severe negative-
switched to LVRT control within 20 ms once the low voltage sequence currents and oscillations in the system, with the re-
at PCC is detected. Taking the LVRT requirements in Germa- sult that double-frequency ripples appear in the output power
ny as an example [41]-[43], the reference active and reactive and the dc-link voltage [31], [34], [36]. Such a ripple, if it ex-
currents during a grid fault are typically defined as follows: ceeds the maximum voltage at dc bus, can lead to unexpected
trips and massive IIDG disconnections [44], [45]. Therefore, a
K(Upcc(0) − Upcc.f ) balanced control strategy based on the positive-sequence volt-
I∗q.f = Iq(0) +
Upcc(0) age at PCC has been widely employed in real-life IIDG appli-
(6) cations to mitigate the impacts of unbalanced voltage sags that
2 Pref.f
I∗d.f = ∙ are caused by unsymmetrical faults [30], [31], [44]-[46]. As a
{ 3 Upcc.f
consequence, the voltage described by (8) denotes its positive-
where I∗d.f
and I∗q.f denote the reference active and reactive cur- sequence component, and only a symmetrical current is theo-
rents during the fault condition, respectively; Upcc(0) and Upcc.f retically supplied by IIDG in the situation of unsymmetrical
are the voltages at PCC before and during the fault occurrence, grid disturbances. This fault characteristic is considerably dif-
respectively; Pref.f denotes the reference active power of the ferent from that presented by synchronous generators.
IIDG during the fault; Iq(0) is the reference reactive current iabc uabc
(which is typically taken as 0 during normal operation); and K 
Upcc ∠δ0 + Zeq
PCC
is the voltage support coefficient greater than 2 if the voltage IIIDG∠α0 PCC
Eeq I IIDG
PCC -
is lower than 0.9 p.u.; otherwise, it is set to 0. PLL
I IIDG
Meanwhile, the maximum fault current of the IIDG is usu- PWM δ +
 
U PCC I IIDG =f (U PCC )
ally limited to an acceptable value for the protection of invert- PQ control strategy
IIDG -
ers. Thus, the current limiter must also be considered in its Primary System Control Scheme Equivalent Circuit
equivalent model, and the fault current contribution of IIDG
Fig. 4. Equivalent circuit diagram of grid-connected IIDG, depicted as a
can be given as current source controlled by the positive-sequence voltage at PCC.

1949-3053 (c) 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2017.2749450, IEEE
Transactions on Smart Grid
5

A mathematical equivalent model considering the IIDG’s C. Calculation for Positive-sequence Voltage at PCC
control strategies has been presented in (6)-(8), in which the In this subsection, we conduct a detailed analysis on the
IIDG is viewed as a current source controlled by the positive- positive-sequence voltage profiles of a general scenario (as
+
sequence voltage at PCC, namely, İIIDG =f (U̇ pcc ) as depicted indepicted in Fig. 5) in which multiple IIDGs are incorporated to
Fig. 4. Consequently, in addition to the intermittent renewable line MN, and we propose a strategy to extract the voltage at
energy, the positive-sequence voltage variation at PCC is also each PCC.
a key factor affecting the fault contribution of IIDGs [44]. As shown in Fig. 5, there are n IIDGs embedded to line MN.
All IIDGs are sequentially named as IIDG1, IIDG2, … , and
B. Estimation for Reference Active Power
IIDGn, and their corresponding PCCs are sequentially named
Equation (6) indicates that I∗d.f is associated with the refer- as PCC1, PCC2, … , and PCCn. Z0 denotes the line impedance
ence active power Pref.f during LVRT operation, i.e., the power between bus-M and PCC1, while Zn denotes the line imped-
captured by MPPT controllers at the fault moment. In addition, ance between PCCn and bus-N. Zi (1 ≤ i < n) denotes the line
natural energy (e.g., wind and solar energy) normally changes impedance between PCCi and PCCi+1.
more slowly than the fast switching of LVRT controllers and
the fast tripping of relays. Hence, the reference power can be R PCC1 PCC2 PCCi PCCi+1 PCCn-1 PCCn R
1 2
 
approximated as a constant value within a very short time be- f
System1 M N System2
fore and after the fault occurrence, i.e.,  
Pref.f =Pref. (9) IIDG1 IIDG2 IIDGi IIDGi+1 IIDGn-1 IIDGn
In this subsection, we introduce a method to estimate each Fig. 5. Single-line diagram of a distribution network incorporating multiple
IIDGs, where System1 and System2 represent the utility grid and other remain-
IIDG’s reference active power (Pref.IIDGi) based on the propor-
ing IIDGs located at the downstream of R2, respectively.
tional allocation principle. In the distribution network shown
in Fig. 1, for instance, the active power provided by the T- Assume that an internal fault occurs at point f, which is lo-
connected IIDGs under normal operating conditions can be cated between PCCi and PCCi+1. The positive-sequence net-
expressed as work is presented in Fig. 6, in which the line impedance Zi is
n divided into two segments, namely, Zi1 and Zi2.
∑ PIIDGk = PN − PM + ∆Ploss (10) UM IM Z0 Z1 Z2  Zi-1 Zi1 Zi2 Zi+1 Zn-2 Zn-1 Zn IN UN
k=1  
M N
where PIIDGk is the output active power of the k-th IIDG; PM ZS1 I DG1 I DG2 I DGi U f I DGi 1 I DGn 1 I DGn ZS2
+ +
and PN are the active power flowing through bus-M and bus-N, Es1   Es 2
respectively; and ΔPloss is the power loss in line MN. - -
With multiple IIDGs tied to line MN, the current in this Fig. 6. Positive-sequence network of the distribution network when an inter-
feeder segment is unevenly distributed. Nevertheless, the cur- nal fault occurs at point f.
rents produced by IIDGs are normally less than that supplied Let U̇ M_pcc0 = U̇ M and U̇ N_pccn+1 = U̇ N ; and based on the positive-
by the utility grid because of the constraints on IIDG penetra-
sequence network shown in Fig. 6, the calculated values of the
tion and the IIDG maximum current. Moreover, ΔPloss is actu-
positive-sequence voltages at PCCs derived from bus-M and
ally also far less than the delivered power (i.e., PM and PN).
bus-N can be described as follows:
The line power loss ΔPloss can thus be reasonably simplified as
2 k−1
|V̇ M − V̇ N |
∆Ploss ≈ ∙RMN (11) U̇ M_pcck = U̇ M_pcck−1 − (∑ İIIDGj + İM ) Zk−1
|ZMN |2 j=1
(1 ≤ k ≤ n) (12)
where V̇ M and V̇ N denote the line-voltages at bus-M and bus-N, n

respectively, and ZMN and RMN denote the impedance and re- U̇ N_pcck = U̇ N_pcck+1 − ( ∑ İIIDGj + İN ) Zk
sistance, respectively, of line MN. { j=k+1

Hence, according to (9)-(11), the total output active power

where U̇ M_pcck and U̇ N_pcck represent the calculated values of the
ΣPIIDGk at the fault moment can be extracted by the measuring
positive-sequence voltage at the k-th PCC derived from bus-M
devices installed at both terminals of line MN.
and bus-N, respectively; U̇ M and U̇ N are the actual values of the
Moreover, it is reasonable to suppose that the reference ac-
positive-sequence voltages at bus-M and bus-N, respectively;
tive power (Pref.IIDGi) captured by the per-rated installed IIDG
İM and İN are the currents seen by R1 and R2, respectively; and
capacity is approximately identical since the natural meteoro-
İIIDGj is the output current of the j-th IIDG obtained by (8).
logical conditions (e.g., the light intensity and temperature) of
line MN, to which those IIDGs are connected, are similar. On Equation (12) shows that U̇ M_pcc1 , for instance, can be dedu-
this basis, ΣPIIDGk can be allocated according to the proportion ced by U̇ M_pcc0 ; then, U̇ M_pcc2 is deduced by U̇ M_pcc1 , and so on.
of the rated installed capacity of each IIDG unit. The reference Similarly, U̇ N_pcck can also be sequentially obtained by (12) one
active power of each IIDG unit (i.e., Pref.IIDG1, Pref.IIDG2, … , and by one. Note that the voltage used in (8) is the calculated value
Pref.IIDGn) at the fault moment can, therefore, be calculated re- derived from (12) instead of the actual voltage. According to
spectively. (12), the profile characteristics of the positive-sequence volt-

1949-3053 (c) 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2017.2749450, IEEE
Transactions on Smart Grid
6

1.6 1.8
Simulated current
age at each PCC can be depicted in Fig. 7. Simulated current

Id in IIDG1 (p.u.)

Id in IIDG2 (p.u.)
1.4 1.6
1.4
Positive-sequence Voltage (p.u.)

Positive-sequence Voltage (p.u.)

1.2 1.2 1.2
Derived from bus-M Derived from bus-N 1.2 Estimated current
1 Estimated current
1.0 1.0 1
Intersection 0.8 0.8
0.8 Band 0.8 0.1 0.15 0.2 0.25 0.1 0.15 0.2 0.25
Time (s) Time (s)
0.6 0.6 (a)
ΔUMN_pcc 1

ΔUMN_pcc 2

ΔUMN_pccn-1

ΔUMN_pcc n
ΔUMN_pcc i ΔUMN_pcci+1 0.2
Simulated current
0.5
Simulated current
0.4 0.4

Iq in IIDG1 (p.u.)

Iq in IIDG2 (p.u.)
0
0
0.2 0.2 Estimated current Estimated current
-0.2
0 0 -0.5
M PCC1 PCC2  PCCi PCCi+1  PCCn-1 PCCn N -0.4
-0.2 -0.2
-0.6 -1
0.1 0.15 0.2 0.25 0.1 0.15 0.2 0.25
Time (s) Time (s)
Fig. 7. Positive-sequence voltage profiles at PCCs that are derived from bus- (b)
0.24 0.35

Current of IIDG1 (kA)

Current of IIDG2 (kA)

M and bus-N.
0.22 Estimated current 0.3
Estimated current
Fig. 7 explicitly illustrates that the U̇ M_pcck of those PCCs lo- 0.2 Simulated current 0.25
Simulated current
0.18 0.2
cated between bus-M and the intersection band are equal to
their actual values U̇ pcck , while the U̇ N_pcck are equal to U̇ pcck for
0.16 0.15
0.1 0.15 0.2 0.25 0.1 0.15 0.2 0.25
Time (s) Time (s)
(c)
those PCCs located between the intersection band and bus-N. 6 4

Estimation error of

Estimation error of
Based on the voltage profile characteristics shown in Fig. 7, 3

IIDG1 (%)

IIDG2 (%)
4

this intersection band can be easily determined by 2

2
1

∆UMN_pcci∗ = min {∆UMN_pcck = |U̇ M_pcck − U̇ N_pcck | , (1 ≤ k ≤ n)} (13) 0

0.1 0.15 0.2 0.25
0
0.1 0.15 0.2 0.25
Time (s) Time (s)
(d)
*
where the subscript i is the location number corresponding to Fig. 8. Comparison between the estimated and simulated currents of two
the PCC with minimum voltage difference ΔUMN_pcc. IIDGs during an “A-g” fault at f1. (a) Estimated and simulated active current Id
Clearly, the positive-sequence voltage at the k-th PCC can in p.u. (b) Estimated and simulated reactive current Iq in p.u. (c) Estimated and
simulated currents of IIDGs. (d) Estimation errors of IIDG currents.
be obtained by the following equation:
Compared with those in high-voltage transmission systems,
U̇ M_pcck , 1 ≤ k < i*
the distribution lines are much shorter, and the distributed ca-
U̇ pcck = max (U̇ M_pcck , U̇ N_pcck ) , k = i* . (14) pacitance is far smaller in MV levels; thus, the improved cur-
rent differential protection scheme applied in distribution net-
U̇ N_pcck , *
i <k≤n
{ works does not require to consider restraint current. The oper-
ating criterion for the improved protection can be expressed as
At this point, all the actual positive-sequence voltages at
PCCs have been solved, and the fault current contribution of n

each IIDG can be estimated by (6)-(14). The flowchart depict- Iop = |İM + İN + ∑ İIIDGj | ≥ IK.set (15)
ing this estimation procedure is shown in Section IV (Fig. 9). j=1
To demonstrate the proposed current estimation algorithm,
where Iop is the operating current; İM and İN are the currents
we simulated a case corresponding to a single-phase ground-
seen by the relays installed at both terminals of the protected
ing fault “A-g” that is described later in Section V. The test
line; İIIDGj is the estimated current of the j-th IIDG terminal;
results are plotted in Fig. 8. As shown in Fig. 8(a) and (b), the
proposed algorithm is capable of accurately tracking the active and IK.set is the operation threshold current.
and reactive currents supplied by these two grid-connected It can be clearly seen that not only the İM and İN but also the
IIDGs before and during the fault occurrence. Fig. 8(c) shows currents provided by IIDGs, are considered for the operating
that the estimated currents and simulated currents are nearly current Iop in (15), with the result that the proposed relays can
identical. Furthermore, Fig. 8(d) indicates that the maximum operate on the sum of the currents from all terminals inside the
estimation errors in this case are less than 5%, which illus- protected zone. This case is considerably different from that of
trates the high accuracy of the estimation algorithm. the conventional protection shown in (1). Moreover, the pro-
posed protection does not depend on a physical multi-terminal
IV. A NOVEL VIRTUAL MULTI-TERMINAL CURRENT synchronous communication system but only necessitates an
DIFFERENTIAL PROTECTION SCHEME exchange of electrical information between both terminals of
the protected line via an already existing two-terminal com-
A. Protection Criterion munications channel. This is the meaning of “virtual”. A flow
A novel virtual multi-terminal current differential protection chart for the implementation of this methodology in real-life
scheme, based on the current estimation algorithm in Section protective devices is shown in Fig. 9. As can be seen, an extra
III, is proposed for distribution networks with multiple IIDGs. computational time ∆T, which is examined later in Section V-
The estimation algorithm embedded in the scheme is activated C, is required during the current estimation procedure.
once the protective relay is picked up, and thus the differential The proposed methodology can be easily and rapidly im-
current coverage for a multi-terminal line can be extracted us- plemented by modern microprocessor-based current differen-
ing an existing two-terminal pilot communications channel. tial relays. Furthermore, the operating criterion in (15) strictly

1949-3053 (c) 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2017.2749450, IEEE
Transactions on Smart Grid
7

Pickup
Pre-set Parameters
Store impedance parameters of
each line section Z0, Z1,…, Zn Start

Store rated capacity of each IIDG Calculate the reference active power of each
PN_IIDG1, PN_IIDG2 ,…, PN_IIDGn IIDG, Pref_IIDG1, Pref_IIDG2 ,…, Pref_IIDGn

Threshold current setting IK.set Calculate the positive-sequence

. .
voltages at
PCCk , i.e., UM_ pcck and U N_ pcc k that are
respectively derived from bus-M and bus-N
Extract and exchange electrical
using (6)-(8) and (12)
data between bus-M and bus-N

Estimate the output current of Detailed Obtain two arrays of the calculated
each IIDG Flowchart values U M_pcc and U N_pcc

Calculate the multi-terminal Obtain the actual voltage Upcck at each

differential current Iop PCC using (13) and (14)

No Estimate the output current of each

Iop ≥ IK.set ?
IIDG using (6)-(8)
Yes
Send a tripping signal and End
output a fault report
Current estimation procedure, which requires a
Protective certain amount of extra computational time
Return Execution

Fig. 9. Flow chart for the virtual multi-terminal current differential protection and a detailed flowchart for the proposed current estimation algorithm.

obeys Kirchhoff's current law, which ensures that the protec- using the PSCAD/EMTDC program and the RTDS hardware
tive relays can trip for internal faults with high sensitivity and experimental platform to verify both the accuracy of the IIDG
adequate selectivity. current estimation algorithm and the feasibility of the virtual
multi-terminal current differential protection scheme proposed
B. Threshold Current Setting Principle
in this paper.
The performance of the virtual multi-terminal differential
current protection is related to the accuracy of the current es- R4 R5 R6 R7

timation for IIDGs. Hence, current estimation errors should be Feeder2 Load
Load
Tie switch
carefully accounted for in the operation threshold settings for
this protection scheme. GRID
R1 PCC1 PCC2 R2 R3

In the theoretical analysis described in Section III-A, only a Feeder1

f1 f2 f3
Load

symmetrical current is provided by the IIDG, even when an M N

unsymmetrical fault occurs. In practical applications, however, PV PV DFIG

the fault currents provided by IIDGs might still contain very

small negative- or zero-sequence component currents during IIDG1 IIDG2 IIDG3

unsymmetrical faults. This flaw leads to a certain estimation Fig. 10. Single-line diagram of a simplified 10 kV distribution network with
multiple IIDGs offered by a local power supply bureau.
error for the IIDG estimated current. In addition, estimation
errors also originate from the reference active power estima- TABLE I.
LINE AND LOAD DATA
tion for each IIDG, as presented in Section III-B. Therefore,
the threshold current setting IK.set in (15) should consider the Parameter Value
effects of current estimation errors on the proposed protection Line Impedance 0.13+j0.356 Ω/km
scheme, namely, Length of Line Section M-PCC1 2.0 km
Length of Line Section PCC1-PCC2 2.5 km
n
Length of Line Section PCC2-N 3.0 km
IK.set = Ktol ∙ | ∑ İmax_IIDGk | +IEDC.max (16) Total Maximum Load 17.36 MVA
k=1 Load Power Factor 0.9

where Ktol is the tolerable current error coefficient determined TABLE II.
by the accuracy of the IIDG current estimation algorithm, and IIDG SYSTEM PARAMETERS
IEDC.max is the erroneous differential current that is the same as NO. IIDG1 IIDG2 IIDG3
the current in (1). Note that |İmax_IIDG | is equal to KmaxIN_IIDG. IIDG Type PV plant PV plant DFIG
Rated Capacity Range 1-5 MW 3 MW 1.5 MW
V. CASE STUDY Rated Voltage 10 kV 10 kV 10 kV

In this section, a 10 kV rural distribution network tapped The transmission grid is represented by “GRID”, and its
with multiple IIDGs, as depicted in Fig. 10, was simulated short-circuit power capacity is set as 426 MVA. Feeder1 and

1949-3053 (c) 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2017.2749450, IEEE
Transactions on Smart Grid
8

Feeder2 are connected by a tie switch that is normally open. The virtual four-terminal differential currents for ABC are
Both the line data and load data of the test system are given in respectively denoted by Iop.A, Iop.B, and Iop.C, which consist of
Table I, and the IIDG parameters are presented in Table II. the currents seen by R1 and R2, and the estimated currents of
The maximum current of these three IIDGs are all typically IIDG1 and IIDG2. The conventional two-terminal differential
limited to twice their rated current [7], [25], [29]. The voltage currents that consist only of the currents seen by R1 and R2 are
support coefficients K1, K2, and K3 are all set to 2.0 [42]. denoted by I∗op.A , I∗op.B , and I∗op.C for each phase, respectively.
A faulted point f1 is placed between PCC1 and PCC2, while Additionally, both the threshold current settings, i.e., I∗K.set and
f2 and f3 are located at PCC2 and the external zone of line MN, IK.set, are closely related to the total installed capacity of IIDG1
respectively. The behaviors of the relays R1 and R2 installed at and IIDG2, and should thus be set to various values as listed in
both terminals of line MN are the study focus of this section. Appendix according to (1) and (16), respectively.

TABLE III.
TEST RESULTS FOR A-PHASE GROUNDING FAULT UNDER DIFFERENT INSTALLED CAPACITY OF IIDG1 AND FAULT RESISTANCE (KA)
Rated Fault
Capacity Resistance İIIDG1 İIIDG1 İIIDG2 İIIDG2 Proposed Protection Conventional Protection
(MW) (Ω) Estimated Value Simulated Value Estimated Value Simulated Value Iop.A IK.set Relay I∗op.A I∗K.set Relay
1 0.0675∠-19.7° 0.0681∠-19.6° 0.2278∠-30.7° 0.2297∠-30.5° 2.7724 T 2.4910 T
1 10 0.0592∠-3.62° 0.0598∠-3.53° 0.1834∠-6.87° 0.1851∠-6.88° 0.5225 0.1072 T 0.2816 0.5460 NT
40 0.0587∠-2.59° 0.0592∠-2.56° 0.1805∠-5.23° 0.1821∠-5.24° 0.1353 T 0.1041 NT
1 0.1991∠-17.8° 0.2014∠-17.7° 0.2245∠-29.2° 0.2268∠-29.0° 2.8015 T 2.4081 T
3 10 0.1763∠-2.55° 0.1783∠-2.54° 0.1820∠-5.81° 0.1841∠-5.83° 0.5250 0.1187 T 0.1725 0.7769 NT
40 0.1747∠-1.52° 0.1766∠-1.57° 0.1792∠-4.17° 0.1811∠-4.13° 0.1354 T 0.2188 NT
1 0.3301∠15.9° 0.3311∠-15.8° 0.2235∠-27.6° 0.2241∠-27.5° 2.8301 T 2.3269 T
5 10 0.2945∠-1.48° 0.2953∠-1.49° 0.1824∠-4.78° 0.1831∠-4.80° 0.5294 0.1303 T 0.0754 1.0078 NT
40 0.2918∠-0.46° 0.2926∠-0.43° 0.1797∠-3.13° 0.1803∠-3.16° 0.1399 T 0.3317 NT

TABLE IV.
TEST RESULTS FOR BC PHASE-TO-PHASE FAULT UNDER DIFFERENT FAULT LOCATION AND FAULT RESISTANCE (KA)
Fault
Fault Resistance IIIDG1 IIIDG1 IIIDG2 IIIDG2 Proposed Protection Conventional Protection
Location (Ω) Estimation Simulation Estimation Simulation Iop.B Iop.C Relay I∗op.B ∗
Iop.C Relay
1 0.2363 0.2387 0.2984 0.3011 3.0166 3.0176 T 2.4946 2.7101 T
f1
10 0.1812 0.1830 0.1899 0.1919 0.8536 0.8520 T 0.5262 0.6111 NT
1 0.2217 0.2240 0.3118 0.3148 2.5314 2.5330 T 2.0079 2.2606 T
f2
10 0.1807 0.1828 0.1953 0.1974 0.8187 0.8175 T 0.5032 0.5630 NT
1 0.2010 0.2034 0.2504 0.2545 0.0177 0.0131 NT 0.4481 0.4586 NT
f3
10 0.1812 0.1831 0.1973 0.1994 0.0013 0.0066 NT 0.3746 0.3793 NT

TABLE V. TABLE VI.

TEST RESULTS FOR INTERNAL FAULT AT f1 UNDER DIFFERENT FAULT TYPES TEST RESULTS FOR EXTERNAL FAULT AT f3 UNDER DIFFERENT FAULT TYPES

Fault Proposed Protection Conventional Protection Fault Proposed Protection Conventional Protection
Types Iop.A Iop.B Iop.C Relay I∗op.A I∗op.B I∗op.C Relay Types Iop.A Iop.B Iop.C Relay I∗op.A I∗op.B I∗op.C Relay
A-g 0.2695 0.0074 0.0039 T 0.0878 0.3625 0.3575 NT A-g 0.0162 0.0236 0.0185 NT 0.4211 0.4229 0.4154 NT
BC-g 0.0077 0.2725 0.2668 T 0.3645 0.0857 0.0921 NT BC-g 0.0013 0.0101 0.0103 NT 0.5292 0.5365 0.5375 NT
BC 0.0091 0.4577 0.4557 T 0.3694 0.2047 0.2647 NT BC 0.0062 0.0039 0.0085 NT 0.4692 0.4638 0.4739 NT
ABC 0.2688 0.2696 0.2701 T 0.0907 0.0898 0.0894 NT ABC 0.0429 0.0546 0.0422 NT 0.6365 0.6512 0.0422 NT

TABLE VII.
TEST RESULTS FOR THE SITUATION CONSIDERING THE CALCULATION ERRORS OF EACH IIDG REFERENCE ACTIVE POWER Pref (KA)
IIDG1 IIDG2 |İ IIDG1 +İIIDG2 | Simulation Proposed Protection Conventional Protection
Output Output Fault |İIIDG1 +İIIDG2 |
Power Power Location Estimation
(MW) (MW) A B C Iop.A Iop.B Iop.C Relay I∗op.A I∗op.B I∗op.C Relay

Scenario f1 0.2491 0.2490 0.2524 0.2525 0.1366 0.0038 0.0039 T 0.1128 0.2524 0.2525 NT
1 2.8 1.4
f3 0.4914 0.5025 0.5017 0.5004 0.0299 0.0315 0.0294 NT 0.5025 0.5017 0.5004 NT
Scenario f1 0.2505 0.2480 0.2502 0.2508 0.1387 0.0067 0.0038 T 0.1127 0.2502 0.2508 NT
4.0 0.2
2 f3 0.4911 0.4458 0.4462 0.4473 0.0719 0.0704 0.0706 NT 0.4458 0.4462 0.4473 NT
Scenario f1 0.2491 0.2497 0.2531 0.2537 0.1361 0.0039 0.0060 T 0.1131 0.2531 0.2537 NT
3 2.2 2.0
f3 0.4911 0.5381 0.5173 0.5221 0.0617 0.0528 0.0399 NT 0.5381 0.5173 0.5221 NT

1949-3053 (c) 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2017.2749450, IEEE
Transactions on Smart Grid
9

 Case 5: An “A-g” fault with a fault resistance of 40 Ω at f1

A. IIDG1 and IIDG2 Units are in Service
and an “ABC” metallic fault at f3 were simulated. The rated
When all IIDGs stayed online, various possible scenarios, installed capacity of IIDG1 and IIDG2 is set as 4 MW and 2
involving different fault conditions, were applied to test the MW, respectively. Three scenarios that consider various
performance of the proposed protection scheme: output powers of both IIDG1 and IIDG2 are analyzed.
 Case 1: A series of internal A-phase grounding faults “A-g”
In Scenario 2, for example, the energy power captured by
with a fault resistance between 1 and 40 Ω were simulated MPPT is 4 MW and 0.2 MW for IIDG1 and IIDG2, respective-
at f1 under various rated installed capacities of IIDG1 in the ly, due to different meteorological conditions. The reference
range of 1-5 MW. power Pref of IIDG1 and IIDG2, however, is still calculated as
 Case 2: A phase-to-phase fault “BC” with a fault resistance
2.782 MW and 1.391 MW (ΣPIIDGi =4.173 MW) in the estima-
range of 1-10 Ω was simulated at f1, f2, and f3. tion algorithm according to the proportional allocation princi-
 Case 3: Four internal faults with a fault resistance of 20 Ω
ple described in Section III-B. As shown in Table VII, the
under various types of faults, including “A-g”, “BC”, “BC- total estimation current of |İIIDG1 + İ IIDG2 | nevertheless retains
g”, and “ABC”, were simulated at f1. high accuracy, which is conducive to the proposed protection.
 Case 4: Four external metallic faults with the four types of
Table VII also indicates that the power estimation error of Pref
faults in Case 3 were simulated at f3. for each IIDG has little influence on the estimation accuracy
All the faults were applied at t=0.15 s. The rated installed of the total currents of IIDG1 and IIDG2, and that the proposed
capacity of IIDG1 is set as 3 MW in Cases 2-4. IK.set and I∗K.set protective relays can reliably trip the circuit breakers installed
are thus set as 0.1187 kA and 0.7769 kA, respectively (see at the faulted line.
Appendix). The PSCAD/EMTDC is applied for power system The simulated cases above assume that all IIDGs supply the
fault simulation, while both the current estimation algorithm typical maximum currents of up to twice the rated current of
and the proposed protection methodology are implemented in inverters. In real-life IIDG applications, the maximum current
MATLAB. Test results for these four cases are listed in Tables of IIDGs can vary from one to three times the normal current.
III-VI, respectively. Note that “T” and “NT” in all the tables In addition, various voltage support coefficients “K” are also
represent “Tripping” and “Non-Tripping”, respectively. likely to be demanded during LVRT operation according to
The test results listed in Table III and Table IV indicate that local grid code requirements. To verify the adaptability of the
the estimated currents of IIDG1 and IIDG2 are identical to the proposed scheme to the IIDGs with various parameters, Case
simulated currents under various conditions involving differ- 6 involving the grid code requirements in China was conduct-
ent IIDG penetration, fault location, fault resistance, and fault ed [40], and the test results are given in Table VIII.
type. A comparison of the results shown in Tables III-V clear-  Case 6: The rated capacity of 3.0 MW is installed for both
ly illustrates that the proposed protective relays can reliably IIDG1 and IIDG2. Kmax2 and Kmax3 are set as 1.2 and 3.0, re-
trip for various internal faults and provide high sensitivity for
spectively, while Kmax1 varies in the range of 1.0-2.5; and the
high resistance faults. For instance, the relays can trip for a
voltage support coefficients K1, K2, and K3 are set as 2.5,
fault with a high resistance up to 40 Ω in the simulated scenar-
1.5, and 2.0, respectively. An “A-g” metallic fault located at
ios. The behaviors of the conventional protective relays, by
f1 was simulated.
contrast, are unsatisfactory. Tripping failures are more likely
Since the IIDGs parameters have been modified, the opera-
to occur in the conventional differential protection, which can
tion threshold current IK.set that is related to the variation of
accommodate only low resistance faults in most instances. The
Kmax1 should be updated to various values as listed in the Ap-
reason is that the operation threshold current I∗K.set is forced to
be set as a higher current value to avoid undesired operation pendix.
and promise adequate selectivity in external fault conditions. TABLE VIII.
Table VI shows that both the proposed relays and conven- TEST RESULTS FOR VARIOUS KMAX OF IIDG1
tional relays can prevent false operation for the metallic faults IIIDG1 IIIDG1 IIIDG2 IIIDG2 Proposed Protection
Kmax1
outside their coverage. However, it should be noted that the Estimation Simulation Estimation Simulation Iop.A Relay
two-terminal differential currents are relatively large, and thus 1.0 0.1732 0.1732 0.2079 0.2078 4.0082 T
the conventional relays, if the maximum currents of IIDGs are 1.2 0.2079 0.2080 0.2079 0.2079 4.0152 T
not accounted for in I∗K.set , will unnecessarily operate for these 1.5 0.2276 0.2313 0.2079 0.2077 4.0179 T
external faults. 2.5 0.2279 0.2315 0.2079 0.2080 4.0180 T
In summary, the test results of the proposed methodology
are encouraging. Nevertheless, one might still argue that a The test results presented in Table VIII indicate that the cur-
non-negligible estimated current error might occur in certain rent estimation algorithm is capable of accurately characteriz-
extreme situations, particularly when a large difference exists ing the fault current contribution of those IIDGs controlled by
in meteorological conditions since the captured power Pref.IIDGi various parameters. In addition, the protective relays function
is obtained based on the assumption that the natural environ- adequately in these fault scenarios.
ments are similar in the area where those IIDGs are located. Actually, the proposed methodology is able to exhibit excel-
To investigate this issue, a group of fault scenarios in Case 5 lent performance as long as the parameters of IIDGs control
were simulated, and the test results are presented in Table VII. circuitry and the requirements of the reactive current injection

1949-3053 (c) 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2017.2749450, IEEE
Transactions on Smart Grid
10

are given. Fortunately, these parameters can be directly pro- The current estimation algorithm is embedded into the pro-
vided by inverter manufacturers as well as local grid codes. tective device shown in Fig. 11(b), which consists primarily of
Therefore, the proposed protection can be applied regardless a dual-core processor OMAPL138 with 300 MHz dominant
of the control parameters or control circuitry of VSIs. frequency, four Ethernet communication ports to support IEC
61850 protocol, a 16-channel A/D sampling model with 16-bit
B. IIDG1 and IIDG2 Units are out of Service
high precision, and a hardware counter with 150 MHz fre-
When the voltages at PCC1 and PCC2 are lower than the quency to record ∆T. Additionally, a full-cycle discrete Fouri-
predefined value determined by local grid codes within a cer- er transform (DFT) algorithm is employed to extract the elec-
tain period of time, IIDG1 and IIDG2 are disconnected. Hence, trical quantities expressed in phasor form, with the result that
we tested Case 7 to assess the performance of the protection in if the grid frequency is 50 Hz, a data window with a funda-
this situation, and the test results are given in Table IX. mental cycle duration, i.e., 20 ms, is required for the protec-
 Case 7: An “ABC” fault occurring at f1 with various fault
tion. The sampling rate of the DFT algorithm is 2000 Hz, cor-
resistance was simulated. The rated installed capacities of responding to 40 samples per cycle.
IIDG1 and IIDG2 are 3 MW, but neither are in service.
Table IX shows that the differential current calculated by
the proposed protection is identical to that calculated by the
conventional protection since IIDG1 and IIDG2 are both dis-
connected. A comparison of the results shows that the relays
employing the proposed operating criterion can correctly trip
for various faults, while the conventional relays fail to operate Power
RTDS Amplifier
in some scenarios. Hence, the proposed protection is also ap-
plicable to the situation when IIDGs are out of service. In fact, Protective
from the perspective of the proposed relays, once an IIDG is Device
installed inside their coverage, the changes in network topolo-
gy resulting from the connection or disconnection of the IIDG
are equivalent to the changes in the total fault contributions of
IIDGs that have been considered in the operating current Iop.
Thus, the frequent modifications of threshold settings caused
by DG operation switching are not necessitated for the relays. Breaker
Simulator
This feature, in contrast to those adaptive overcurrent protec-
tion schemes, is another benefit of the proposed protection. (a)
TABLE IX. A/D Sampling
TEST RESULTS FOR THE ABSENCE OF IIDGS Module

Fault Proposed Protection Conventional Protection Copper

Ethernet
Resistance I
op.A Iop.B Iop.C Relay I∗op.A I∗op.B I∗op.C Relay
1Ω 2.8141 2.8119 2.8126 T 2.7447 2.7425 2.7434 T
10 Ω 0.5229 0.5235 0.5230 T 0.5222 0.5229 0.5222 NT Optical
20 Ω 0.2691 0.2694 0.2691 T 0.2689 0.2693 0.2689 NT Fiber
Ethernet
40 Ω 0.1363 0.1365 0.1365 T 0.1361 0.1363 0.1363 NT
Processor
Module
C. Experimental Test for the Duration of the Current Estima-
tion Procedure Interface
Module
Note that the current differential scheme is a selective pro- (b)
tection system that operates instantaneously for internal faults Fig. 11. (a) Hardware experimental environment in laboratory and (b) the
[27]. However, relative to the two-terminal differential protec- detailed structure inside the protective device.
tion, the proposed methodology depicted in Fig. 9 requires an
extra time ∆T for the current estimation procedure, which is  Case 8: Various fault conditions predefined in Cases 1-3,
embedded within the protection. In this subsection, the dura- under a total of 19 scenarios, were simulated again in the
tion of the current estimation procedure is demonstrated in a RTDS hardware experimental laboratory. The experiment
hardware laboratory environment, shown in Fig. 11, to justify study for these fault scenarios focuses on both the operating
that the proposed methodology is applicable in real-time. time Top, and the extra time ∆T that reports the duration of
A RTDS platform is employed for power system simulation. the IIDG current estimation procedure.
The currents and voltages measured by the simulated trans- The experimental results are identical to those listed in Ta-
formers are amplified using a power amplifier and delivered to bles III-V, and the breaker simulator can correctly trip for var-
the real-life protective device. Tripping signals, in the pres- ious faults. Fig. 12(a) and (b) show the fault current wave-
ence of internal faults, are then sent to the breaker simulator forms that are captured by a fault recorder. As can be seen, the
via IEC61850 protocol. operating times for an “ABC” three-phase fault and a “BC”

1949-3053 (c) 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2017.2749450, IEEE
Transactions on Smart Grid
11

[1 kA/div]
the execution speed of the estimation procedure is sufficiently
ia ib ic
fast that ∆T, which is far smaller than Top, can be ignored, and
the estimation algorithm is therefore not a time burden for the
protection. In other words, the proposed methodology is appli-
cable in real-time, and can be easily and rapidly implemented
Fault in microprocessor-based protective applications.
Fault clearence
occurrence time
time 40

Operating time (ms)

35

Top Tripping signal for R1

30

25
Tripping signal for R2
Tripping time 20
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

Scenario Number
[10 ms/div] Fig. 13. Experimental results of the measured operating time for various
fault scenarios.
(a)
80

Computational time (µs)

ic ib [1 kA/div]
70

ia
60

50

Fault
Fault clearence 40
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
occurrence time
time Scenario Number
Fig. 14. Experimental results of the measured computational time of IIDG
current estimation procedure for various fault scenarios.
Top Tripping signal for R1
VI. CONCLUSION
Tripping signal for R2
A novel virtual multi-terminal current differential protection
Tripping time
scheme using an already existing two-terminal pilot communi-
[10 ms/div]
cations channel has been proposed in this paper for a complex
scattered multi-source distribution network. The novelty of the
(b)
proposed protection is in its estimation algorithm for the fault
Fig. 12. Experimental results of the proposed scheme performance under (a) current contribution of IIDGs, by which the protection cover-
three-phase and (b) phase-to-phase fault conditions. Three-phase currents at
bus-M are captured by a fault recorder embedded within the protective device, age for a multi-terminal line can operate on the sum of the cur-
and two tripping signals are sent to R1 and R2 by the protective device. rents from all terminals without the requirements of a physical
multi-terminal time-synchronized communication system.
phase-to-phase fault are approximately 33.14 ms and 34.33 ms, The accuracy and the real-time implementations of the cur-
respectively. In addition, the operating times for all of the fault rent estimation algorithm have been confirmed experimentally
scenarios mentioned in Case 8 are presented in Fig. 13. As can in real-life hardware protective applications. The feasibility of
be seen, the Top is approximately 34.95 ms on average, which the protection methodology is substantiated by the simulation
includes the following: [i] the relay pickup time (<2.5 ms), [ii] and experiment results. These test results are encouraging and
a data window for DFT algorithm (20 ms), [iii] the relay inter- indicate that the proposed methodology can achieve full-line
face delay (1-5 ms), [iv] the end-to-end propagation time (<10 high-speed tripping during various internal fault conditions
ms), and [v] the computation time used to calculate the sum of (including the fault locations, fault resistances, and fault types),
the currents from all terminals. This methodology is thus fully and can thus effectively mitigate the adverse impacts resulting
capable of satisfying the high-speed requirement of current from T-connected IIDGs with high penetration. Additionally,
differential protection. Furthermore, it should be noted that the IIDG deployments with higher flexibility inside the protected
first four time delays of Top in the proposed methodology are zone are allowed due to the relief of the constraints on IIDG
identical to that in the conventional protection; thus the extra locations and capacities. This advantage increases the capacity
time ∆T consumed in the proposed protection is reflected pre- of power grids to absorb additional renewable clean energy
dominantly in the fifth delay, i.e., the computation time. without changing the original protective systems. Furthermore,
Due to the extremely short duration of the current estima- the proposed methodology has favorable economic benefits in
tion procedure, ∆T should be measured using the hardware real-life engineering because it can considerably minimize the
counter. The test results are given in Fig. 14, which shows that requirements for communication channels in comparison with
the maximum value of ∆T among all the relay operation sce- the same process in a physical multi-terminal channels system.
narios is 71.78 μs (i.e., the counter reads 10767), and the min- The proposed protection is intended primarily for the most
imum value is 59.13 μs (i.e., the counter reads 8869). In addi- common DGs that are connected to grids using power elec-
tion, a comparison of Fig. 13 and Fig. 14 clearly indicates that tronics interfaces. In future distribution networks, it may be-

1949-3053 (c) 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2017.2749450, IEEE
Transactions on Smart Grid
12

come routine for multiple types of hybrid DG interfaces (e.g., [6] E. J. Coster, J. M. A. Myrzik, B. Kruimer, and W. L. Kling, “Integration
issues of distributed generation in distribution grids,” Proc. IEEE, vol. 99,
power electronic converter, and ac rotating machine interfaces, no. 1, pp. 28–39, Sep. 2010.
including synchronous-based and asynchronous-based DGs) to [7] B. Hussain, S. M. Sharkh, S. Hussain and M. A. Abusara, "Integration of
be integrated into power grids. Such a scenario is beyond the distributed generation into the grid: Protection challenges and solutions,"
10th IET Int. Conf. Develop. Power Syst. Protection. Managing the
scope of this paper but will be addressed in subsequent re-
Change, Manchester, 2010, pp. 1-5.
search. Overall, the research methodology and protection phi- [8] H. Alatrash, A. Mensah, E. Mark, G. Haddad, and J. Enslin, “Generator
losophy in this paper provide an essential theoretical basis for emulation controls for photovoltaic inverters,” IEEE Trans. Smart Grid,
further research. vol. 3, no. 2, pp. 996–1011, Jun. 2012.
[9] M. E. Baran and I. El-Markaby, “Fault analysis on distribution feeders
with distributed generators,” IEEE Trans. Power Syst., vol. 20, no. 4, pp.
APPENDIX 1757–1764, Nov. 2005.
[10] P. P. Barker and R. W. D. Mello, “Determining the impact of distributed
The rated current of IIDG IN_IIDG can be obtained by generation on power systems. I. Radial distribution systems,” in Proc.
PN IEEE Power Eng. Soc. Sum. Meeting, vol. 3. Seattle, WA, Jul. 2000, pp.
IN_IIDG = (17) 1645–1656.
√3 UN [11] F. Coffele, C. Booth, A. Dysko, and G. Burt, “Quantitative analysis of
network protection blinding for systems incorporating distributed genera-
where PN represents the rated installed capacity of IIDG, and tion,” IET Gener. Transmiss. Distrib., vol.6, no.12, pp. 1218–1224, Dec.
UN represents the grid-connected rated voltage at PCC. 2012.
Ratio errors and unequal saturation of CTs are taken into [12] C. J. Mozina, “Impact of smart grids and green power generation on
distribution systems,” IEEE Trans. Ind. Appl., vol. 49, no. 3, pp. 1079–
account in the maximum erroneous differential current IEDC.max, 1090, May 2013.
which is set by [13] K. I. Jennett, C. D. Booth, F. Coffele, and A. J. Roscoe, “Investigation of
the sympathetic tripping problem in power systems with large penetra-
IEDC.max=KerKstKnpIK.max (18) tions of distributed generation,” IET Gener. Transmiss. Distrib., vol. 9, no.
4, pp. 379–385, 2015.
where Ker is the CT ratio error coefficient; Kst is the CT same [14] P. Mahat, Z. Chen, B. Bak-Jensen, and C. L. Bak, “A simple adaptive
type coefficient; Knp is the non-periodic component coefficient; overcurrent protection of distribution systems with distributed generation,”
and IK.max is the maximum fault current flowing through the IEEE Trans. Smart Grid, vol. 2, no. 3, pp. 428–437, Sep. 2011.
[15] B. Hussain, S. M. Sharkh, S. Hussain, and M. A. Abusara, “An adaptive
CTs during an external fault [47]. relaying scheme for fuse saving in distribution networks with distributed
Ker, Kst, and Knp are respectively set as 10%, 0.5, and 1.0 generation,” IEEE Trans. Power Del., vol. 28, no. 2, pp. 669–677, Apr.
[47], [48], and IK.max in the test system is as high as 1.6817 kA. 2013.
[16] F. Coffele, C. Booth, and A. Dyśko, “An adaptive overcurrent protection
Consequently, IK.set that are given in Table X for various IIDG scheme for distribution networks,” IEEE Trans. Power Del., vol. 30, no. 2,
installed capacity and maximum current coefficient Kmax1 can pp. 561–568, Apr. 2015.
be obtained according to (16)-(18), where Ktol is set as 0.05 [17] D. S. Kumar, D. Srinivasan, and T. Reindl, “A fast and scalable protec-
tion scheme for distribution networks with distributed generation,” IEEE
based on the estimation results presented in Fig. 8 and Tables Trans. Power Del., vol. 31, no. 1, pp. 67–75, Feb. 2016.
III-IV, and similarly I∗K.set listed in Table X can be calculated [18] H. Yazdanpanahi, Y. W. Li, and W. Xu, “A new control strategy to miti-
by (1), (17), and (18). gate the impact of inverter-based DGs on protection system,” IEEE Trans.
Smart Grid, vol. 3, no. 3, pp. 1427–1436, Sep. 2012.
TABLE X. [19] H. Margossian, G. Deconinck, and J. Sachau, “Distribution network
THRESHOLD CURRENT SETTINGS FOR BOTH PROTECTION SCHEMES protection considering grid code requirements for distributed generation,”
IET Gener. Transmiss. Distrib., vol. 9, no. 12, pp. 1377–1381, Aug. 2015.
Installed Capacity of IIDG (MW) 1 2 3 4 [20] H. Zhan, C. Wang, Y. Wang, X. Yang, X. Zhang, C. Wu, and Y. Chen,
IK.set (kA) 0.0899 0.0956 0.1014 0.1072 “Relay protection coordination integrated optimal placement and sizing of
I∗K.set (kA) 0.1996 0.3150 0.4305 0.5460 distributed generation sources in distribution networks,” IEEE Trans.
Installed Capacity of IIDG (MW) 5 6 7 8 Smart Grid, vol. 7, no. 1, pp. 55–65, Jan. 2016.
[21] R. J. Yinger, “Self-healing circuits at Southern California Edison,” in
IK.set (kA) 0.1130 0.1187 0.1245 0.1303
Proc. IEEE PES Transmiss. Distrib. Conf. Expo., Orlando, Florida, May.
I∗K.set (kA) 0.6614 0.7769 0.8924 1.0078 2012, pp. 1–3.
Maximum Coefficient Kmax1 1.0 1.2 1.5 2.5 [22] A. Zidan and E. F. El-Saadany, “A cooperative multiagent framework
IK.set (kA) 0.1031 0.1049 0.1075 0.1161 for self-healing mechanisms in distribution systems,” IEEE Trans. Smart
Grid, vol. 3, no. 3, pp. 1525–1539, Sep. 2012.
[23] M. Dewadasa, A. Ghosh, and G. Ledwich, “Protection of microgrids
REFERENCES using differential relays,” in 21st Australasian Universities Power Eng.
Conf., Brisbane, QLD, Sep. 2011, pp. 1–6.
[1] M. Liserre, R. Cardenas, M. Molinas, and J. Rodriguez, “Overview of [24] T. S. Ustun, C. Ozansoy, and A. Zayegh, “Differential protection of mi-
multi-MW wind turbines and wind parks,” IEEE Trans. Ind. Electron., crogrids with central protection unit support,” in IEEE TENCON Spring
vol. 58, no. 4, pp. 1081–1095, Apr. 2011. Conf., Sydney, NSW, Apr. 2013, pp. 15–19.
[2] T. Strasser et al., “A review of architectures and concepts for intelli- [25] E. Casagrande, W. L. Woon, H. H. Zeineldin, and D. Svetinovic, “A
gence in future electric energy systems,” IEEE Trans. Ind. Electron., vol. differential sequence component protection scheme for microgrids with
62, no. 4, pp. 2424–2438, Apr. 2015. inverter-based distributed generators,” IEEE Trans. Smart Grid, vol. 5, no.
[3] T. Adefarati and R. C. Bansal, “Integration of renewable distributed 1, pp. 29–37, Jan. 2014.
generators into the distribution system: A review,” IET Renew. Power [26] R. A. Hedding and F. Mekic, "Advanced multi-terminal line current
Gener., vol. 10, no. 7, pp. 873–884, Aug. 2016. differential relaying and applications," 60th Annu. Conf. Protective Relay
[4] “Europe - Major solar PV installations in 2016. pdf” [Online]. Available: Engineers, College Station, TX, Mar. 2007, pp. 102-109.
http://www.solarpowereurope.org/media/downloads/ [27] IEEE Guide for Application of Digital Line Current Differential Relays
[5] P. N. Vovos, A. E. Kiprakis, A. R. Wallace, and G. P. Harrison, “Central- Using Digital Communication, IEEE Standard C37.243-2015, Aug. 2015.
ized and distributed voltage control: Impact on distributed generation [28] IEEE Guide for Protective Relay Applications to Transmission Lines,
penetration,” IEEE Trans. Power Syst., vol. 22, no. 1, pp. 476–483, Feb. IEEE Standard C37.113-2015, Jun. 2016.
2007.

1949-3053 (c) 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2017.2749450, IEEE
Transactions on Smart Grid
13

[29] N. Rajaei and M. M. A. Salama, “Management of fault current contribu- [48] IEEE Guide for the Application of Current Transformers Used for Pro-
tion of synchronous DGs using inverter-based DGs,” IEEE Trans. Smart tective Relaying Purposes, IEEE Standard C37.110-2007 (Revision of Std
Grid, vol. 6, no. 6, pp. 3073–3081, Nov. 2015. C37.110-1996), pp. 1–90, Apr. 2008.
[30] Y. A. R. I. Mohamed and E. F. El-Saadany, “A Control scheme for PWM
voltage-source distributed-generation inverters for fast load-voltage regu-
lation and effective mitigation of unbalanced voltage disturbances,” IEEE BIOGRAPHIES
Trans. Ind. Electron., vol. 55, no. 5, pp. 2072–2084, May 2008.
[31] A. Junyent-Ferre, O. Gomis-Bellmunt, T. C. Green, and D. E. Soto- Bowen Han (S’17) received the B.E. degree in elec-
Sanchez, “Current control reference calculation issues for the operation of trical engineering from South China University of
renewable source grid interface VSCs under unbalanced voltage sags,” Technology, Guangzhou, China, in 2015.
IEEE Trans. Power Electron., vol. 26, no. 12, pp. 3744–3753, Dec. 2011. He is currently working toward the master's de-
[32] M. A. Haj-ahmed and M. S. Illindala, “The influence of inverter-based gree in power system and its automation at South
DGs and their controllers on distribution network protection,” IEEE China University of Technology, Guangzhou, China.
Trans. Ind. Appl., vol. 50, no. 4, pp. 2928–2937, Jul. 2014. His research interests include control and operation
[33] J. A. P. Lopes, C. L. Moreira, and A. G. Madureira, “Defining control for renewable distributed generation, and protection
strategies for microgrids islanded operation,” IEEE Trans. Power Syst., for smart distribution networks.
vol. 21, no. 2, pp. 916–924, May 2006.
[34] F. Blaabjerg, R. Teodorescu, M. Liserre, and A. V. Timbus, “Overview of
control and grid synchronization for distributed power generation sys- Haifeng Li (M’16) received the Ph.D. degree in
tems,” IEEE Trans. Ind. Electron., vol. 53, no. 5, pp. 1398–1409, Oct. electrical engineering from South China University
2006. of Technology, Guangzhou, China, in 2004.
[35] A. Timbus, M. Liserre, R. Teodorescu, P. Rodriguez, and F. Blaabjerg, He is currently an Associate Professor in the
“Evaluation of current controllers for distributed power generation sys- School of Electric Power Engineering, South China
tems,” IEEE Trans. Power Electron., vol. 24, no. 3, pp. 654–664, Mar. University of Technology. His major research inter-
2009. ests include power system protection and control,
[36] R. Cardenas, R. Pena, S. Alepuz, and G. Asher, “Overview of control smart distribution networks, and high-voltage direct
systems for the operation of DFIGs in wind energy applications,” IEEE current (HVDC) technology.
Trans. Ind. Electron., vol. 60, no. 7, pp. 2776–2798, Jul. 2013.
[37] C. Wang, Analysis and Simulation Theory of Microgrids. Beijing, CN:
Science Press, 2013, pp. 119-160, ISBN: 978-7-03-039161-2. Gang Wang received the Ph.D. degree in electrical
[38] N. Rajaei, M. H. Ahmed, M. M. A. Salama, and R. K. Varma, “Fault engineering and automation from Tianjin University,
current management using inverter-based distributed generators in smart Tianjin, China, in 1998.
grids,” IEEE Trans. Smart Grid, vol. 5, no. 5, pp. 2183–2193, Sep. 2014. Currently, he is a Professor in the School of
[39] M. Garcí a-Gracia, N. E. Halabi, H. Ajami, and M. P. Comech, “Integrat- Electric Power Engineering, South China University
ed control technique for compliance of solar photovoltaic installation grid of Technology. He is the director of Power Engi-
codes,” IEEE Trans. Energy Convers., vol. 27, no. 3, pp. 792–798, Sep. neering and Technology Research and Development
2012. Center in Guangdong Province, China. He has about
[40] Technical requirements for connecting photovoltaic power station to 20 years of experience in electric power system. His
power system, China National Standard GB/T 19964-2012, Dec. 2012. major research interests include power system pro-
[41] X. Liu, Z. Xu, and K. P. Wong, “Recent advancement on technical re- tection and control, fault analysis, power system planning and reliability, and
quirements for grid integration of wind power,” J. Mod. Power Syst. HVDC technology.
Clean Energy, vol. 1, no. 3, pp. 216–222, Dec. 2013.
[42] E. Troester, “New German grid codes for connecting PV systems to the
Dehui Zeng received the B.E. and Ph.D. degrees in
medium voltage power grid,” in Proc. 2nd Int. Workshop Concentrating
electrical engineering and automation from South
Photovoltaic Power Plants: Opt. Design, Prod., Grid Connection, 2009,
China University of Technology, Guangzhou, China,
pp. 1–4.
in 2010 and 2016, respectively.
[43] Y. Bae, T. K. Vu, and R. Y. Kim, “Implemental control strategy for grid
He is currently a Post-doctor in the School of
stabilization of grid-connected PV system based on German grid code in
Electric Power Engineering, South China University
symmetrical low-to-medium voltage network,” IEEE Trans. Energy Con-
of Technology. His research interests include fault
vers., vol. 28, no. 3, pp. 619–631, Sep. 2013.
analysis, voltage source converter, RTDS simula-
[44] P. Rodriguez, A. V. Timbus, R. Teodorescu, M. Liserre, and F. Blaabjerg,
tion, and development of protection devices.
“Flexible active power control of distributed power generation systems
during grid faults,” IEEE Trans. Ind. Electron., vol. 54, no. 5, pp. 2583–
2592, Oct. 2007.
[45] L. Xu, “Coordinated control of DFIG’s rotor and grid side converters Yuansheng Liang received the Ph.D. degree in
during network unbalance,” IEEE Trans. Power Electron., vol. 23, no. 3, electrical engineering and automation from South
pp. 1041–1049, May 2008. China University of Technology, Guangzhou, China,
[46] P. Rodriguez, A. Luna, R. Munoz-Aguilar, F. Corcoles, R. Teodorescu, in 2008.
and F. Blaabjerg, “Control of power converters in distributed generation He is currently an Associate Professor in the
applications under grid fault conditions,” in IEEE Energy Convers. Congr. School of Electric Power Engineering, South China
Expo., Phoenix, AZ, Sep. 2011, pp. 2649–2656. University of Technology. His research interests
[47] J. He, Y. Li, X. Dong, and B. Li, Power System Protection Principles, 4th include protection, fault location, and automation of
ed., Beijing, CN: China Electric Power Press, 2010, pp. 173-182, ISBN: electric power systems.
978-7-5083-9595-1.

1949-3053 (c) 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.