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/ACCESS.2019.2952934, IEEE Access

Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000.
Digital Object Identifier 00.0000/ACCESS.2019.DOI

Protection of MV Closed-Loop
Distribution Networks with Bi-Directional
Overcurrent Relays and GOOSE
Communications
BOŠTJAN POLAJŽER1 , (Member, IEEE), MATEJ PINTARIČ1 , MIRAN ROŠER2 , and
GORAZD ŠTUMBERGER.1 , (Member, IEEE)
1 Faculty of Electrical Engineering and Computer Science, University of Maribor, Koroška 46, 2000 Maribor, Slovenia
2 Elektro Celje, d.d., Vrunčeva 2a, 3000 Celje, Slovenia
Corresponding author: Boštjan Polajžer (e-mail: [email protected]).
This work was supported by ARRS under projects: P2-0115, L2-5489, L2-7556. A part of the smart-grid project is financed by a
partnership between the Japanese agency NEDO and the Slovenian transmission system operator ELES.

ABSTRACT This paper proposes measures to improve the protection of MV distribution networks
operating with feeders in a closed-loop arrangement. Bi-directional overcurrent relays (OCRs) are discussed,
the selectivity of which is achieved through the timing coordination of their operation. The classic approach
is formulated as a minimization of the operating times of all the OCRs. The proposed approach enhances the
selectivity by considering the maximum operating time of substation OCRs and the unwanted trips of in-
loop OCRs. Moreover, the sensitivity is also increased by introducing an objective function that minimizes
the pickup-current settings of all the OCRs together with their operating times. Furthermore, to fulfill
the demanding requirements for operating times, variable penalties are introduced. Thus, the optimization
procedure is forced towards the region with viable solutions for the optimization problem. Two variants of
self-adaptive differential evolution have been used that both show better convergence when compared to the
classic differential evolution. Moreover, ten mutation strategies were tested, where “rand/1/bin” showed the
best results. A comparison with other methods for timing coordination shows that the proposed optimization
results in a comparable value for the OCRs’ operating times. In order to further reduce the operating times,
GOOSE communications between the OCRs are adopted. The proposed measures for improved protection
operation are fully confirmed through dynamic simulations of the faults in the discussed 20-kV network.
Moreover, the proposed protection design is already implemented and permanently operates in a 20-kV
network with more than 5000 customers, whereas the field results show selective and reliable protection
operation.

INDEX TERMS Closed-loop operation, overcurrent relay, selectivity, sensitivity, communications.

I. INTRODUCTION However, the key challenge in the realization of such a net-

One of the priorities of modern distribution networks is im- work operation is a proper protection system design, where
proved power-supply reliability. This is normally evaluated all the modern solutions, such as the exchange of GOOSE
by the system indices SAIFI (System Average Interruption (Generic Object Oriented Substation Event) messages among
Frequency Index) and SAIDI (System Average Interruption the relays, are applied in order to reach the goal of improved
Duration Index). A well-known solution for the improve- supply reliability.
ment of system indices is fault location, isolation, and self- The basic performance requirements for a protection relay,
restoration procedures [1]. Feeders in radial or closed-loop or a relaying system, are reliability, sensitivity, selectivity
arrangements, equipped with properly placed and adjusted and fault-clearing time [2], [3]. The selectivity of overcur-
protection Relays with Switching Capability (RSCs), rep- rent relays (OCRs) is achieved by the timing coordination
resent another approach to improving the supply reliability. of their operation. This can be performed using different

VOLUME 4, 2016 1

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 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/ACCESS.2019.2952934, IEEE Access

B. Polajžer et al.: Protection of MV Closed-Loop Distribution Networks with Bi-Directional Overcurrent Relays and GOOSE ...

optimization approaches [4]–[17]. In [4] an interval linear

R1 R2
programming problem was formulated, considering changes
in the network topologies. A seeker algorithm was pro- MV substation
R6 R8 DG
posed in [5] to solve a mixed-integer nonlinear programming RTR
(MINLP) problem. The complexity of the non-linear pro-
graming problem was reduced in [6] through reformulation Source R5 R9
Load
as an equivalent, quadratically constrained, quadratic pro- R4 R7
gramming (QCQP) model. In [7] a local-fit method was pro- R3
Load
posed by defining reference marks based on the OCR’s char-
acteristics. Furthermore, several metaheuristic algorithms Relay pairs (primary, back-up): (R2,R1)1, (R3,R2)2, (R3,R9)3,
were applied, like enhanced differential evolution [8], a com- (R5,R4)4, (R6,R5)5, (R6,R9)6, (R7,R4)7, (R7,R3)8, (R8,R2)9, (R8,R5)10
bination of the differential evolution and linear programming FIGURE 1. Illustrative example with relay pairs.
problem [9], a genetic algorithm [10], a combination of the
genetic algorithm and the linear programming problem [11],
or a combination of a specialized genetic algorithm and GOOSE communications is discussed. The parameterization
an efficient heuristic algorithm [12]. Moreover, a symbiotic of the RSCs and substation OCRs was performed according
organism search algorithm [13], teaching learning-based op- to the proposed methodology. Furthermore, a detailed net-
timization [14], particle-swarm optimization [15], ant-colony work model was built, together with a dynamic model of the
optimization [16], and whale optimization algorithm [17] discussed RSC and GOOSE communications, which is not
were also applied. Lately, distributed generation (DG) was reported in the literature. The results are given in Section VI.
also considered within the timing coordination of the OCR’s First, two different SADE variants and ten mutation strategies
operation [18]–[21]. are analyzed through the best objective function value and the
The timing coordination inherently introduces a time de- convergence. Next, the selectivity is tested using theoretical
lay to the OCRs’ operation, which can be improved using calculations of the operating times of the primary and back-
a peer-to-peer communication by applying GOOSE mes- up OCRs. Furthermore, based on dynamic simulations of the
sages [22], [23]. Typically, the GOOSE-based communica- faults, the fault-clearing times are given for all the RSCs and
tion is used in substation-area protection [24], [25]; however, substation OCRs, which is not discussed in the literature.
inter-substation communication was already applied, e.g., for Moreover, a comparison is also made for a protection sys-
accelerated distance-protection operation [26]. When using a tem with and without GOOSE-based communications. The
proper communication network, GOOSE messages can also proposed protection design has already been implemented
be sent between all the OCRs in a MV distribution network. and since January 2019 has been in permanent operation in
The required reliability and security of GOOSE messages can a 20-kV network of Distribution Company (DisCo) Elektro
be achieved using recovery protocols [27] as well as digital Celje with more than 5000 customers, which represents an
signature algorithms [28]. additional contribution of this paper. Section VII concludes
The organization of the paper is as follows. In Section II, the paper.
the background to the parametrization of bi-directional OCRs
is presented. Section III proposes measures for the improved II. BACKGROUND
operation of bi-directional OCRs. The sensitivity is increased A. RELAY NOTATION AND RELAY PAIRS
by minimizing the pickup-current settings, together with the Most of the research about the timing coordination of the
operating times, which is an original contribution of this operation of OCRs assumes that the OCRs are located at
paper. Furthermore, a simple approach is proposed to avoid the network’s buses, which is not feasible in MV distribution
unwanted trips of the in-loop OCRs. Moreover, the selectivity networks. The OCRs are located at the substation feeders,
of the substation OCRs is achieved by limiting their operating whereas along the feeders the RSCs are mounted directly on
times, which is not discussed in the literature. A self-adaptive a tower construction. Furthermore, MV distribution networks
differential evolution (SADE) algorithm [29] is applied to can also operate in a closed-loop, whereas between two
minimize the operating times and the pickup-current settings successive OCRs there might be a considerable in-feed due
of all the OCRs, while introducing variable penalties. In order to the DG. One of the most favorable protection solutions is
to further reduce the operating times of the OCRs, a peer-to- the bi-directional OCR [18].
peer communication is proposed, according to the GOOSE An arbitrary relay notation will be used, as shown in Fig. 1
model. The proposed methodology for the parametrization for an illustrative example. Note that an individual bi-
of the OCRs is given and validated in Section IV. directional relay is separately denoted for each direction, e.g.,
In order to confirm the discussed measures for the im- R2 , R6 . Furthermore, the relays located at substation R1 and
proved protection operation, a case study is considered in R4 are non-directional, as well as R7 , since no source is
Section V. A 20-kV network with three possible network assumed on that side branch. Moreover, the incorporation of
topologies (two different loops and a radial type) that is R9 depends on the DG capacity, which should be checked
equipped with properly placed bi-directional RSCs with using short-circuit calculations.
2 VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 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/ACCESS.2019.2952934, IEEE Access

B. Polajžer et al.: Protection of MV Closed-Loop Distribution Networks with Bi-Directional Overcurrent Relays and GOOSE ...

load current and lower than the minimum short-circuit cur-

t
(back-up) (primary) rent with a reasonable security margin. Furthermore, the
Ri Rj measurement error of the used current transformers (CTs)
should also be considered. Thus, the pickup-current limits of
t the j-th OCR can be determined as
(back-up) eCT% 

tr >CTI r tr >CTI r
IP j min > 1 + KL I j,L max
100
(primary) (2)
 eCT% 
fault IP j max < 1 − KSC I j,SC min
a) FL j location b) I Pj I Pi I SC I 100
FIGURE 2. Timing coordination for a fundamental relay pair (R j , Ri )r :
where I j,SC min is the minimum short-circuit current, i.e., for
time-fault location characteristics (a), and time-current characteristics (b). a Phase-to-Phase (Ph-Ph) fault located at a remote end of the
discussed line. I j,L max is the maximum value of the transient
load current. Furthermore, eCT% denotes the % measurement
In order to achieve the timing coordination of the OCRs’ error of the used CTs, whereas KSC >1 and KL >1 are
operation, relay pairs should be determined. An individual security factors [34].
relay pair consists of a primary and a back-up OCR. The
following notation will be used for the r-th relay pair, i.e., D. TIMING COORDINATION
(R j , Ri )r , where R j and Ri , respectively, denote the primary Selectivity is achieved by timing coordination, where the
and back-up OCRs. The relay pairs for the discussed example OCR intended to operate (primary) operates faster than the
are given in Fig. 1. Note that the relay pairs (R3 , R9 )3 and other OCRs (back-up), as shown in Fig. 2. Selectivity is
(R6 , R9 )6 might not be necessary, depending on the DG’s ensured when ∆tr > CTIr , where ∆tr = (t(b,r) − t(p,r) ). CTIr
capacity. Furthermore, the transformer relay RTR is not con- is a coordination time interval that is determined according
sidered in the timing coordination through relay pairs, since to the specified time delays of both OCRs and the operating
its settings are fixed and predefined. Moreover, a definite times of the switch gears.
time-current characteristic is typically used for the relay RTR , The pickup-current settings and the time-dial settings of
while other relays use an inverse time-current characteristic. all the OCRs can be determined in different ways [4]–[17].
Therefore, the timing coordination is achieved by limiting The classic approach is formulated as a minimization of an
the operating times of the relays R1 and R4 , i.e., they should objective function that is given as a sum of the operating
operate faster than the relay RTR . times of all the primary and back-up OCRs, i.e.,
NRP

B. TIME-CURRENT CHARACTERISTICS T= ∑ t(p,r) + t(b,r) + p (3)
Different types of time-current characteristics can be used, r=1
i.e., definite-time, inverse-time or custom-based, e.g., piece- where NRP is the number of all the relay pairs. The operating
wise linear [31]. In the following, only the inverse-time times t(p,r) and t(b,r) should be calculated using (1), where
characteristic will be discussed, which is given as currents that correspond to the fault location given by the
  primary OCRs should be used (Fig. 2a). Note that short-
A circuit currents passing through the primary and back-up
t j,k = TD j   +C (1)
 
I j,k
B relays have the same magnitude only for the fundamental
IP j −1 relay pair, which is in Fig. 2b denoted by ISC . However, in
the case of a side branch between both relays, e.g., R5 and R4
where t j,k and I j,k denote the operating time and the current in Fig. 1, the magnitude of the currents passing through both
passing through the j-th OCR for a fault at location FLk . relays will not be equal. The penalties p are generally applied
IP j and TD j denote the pickup-current setting and the time- when ∆tr ≤ CTIr , or when violating the limits of the pickup-
dial setting, respectively. The constants A, B and C are given current or time-dial settings. Moreover, penalties should also
according to the characteristic type, which is defined by the be applied when the operating times of substation OCRs,
IEC [32] or IEEE [33] standards. such as R1 and R4 in Fig. 1, exceed a permissible value given
The notation introduced in (1) refers to a j-th OCR and a by the operating time of the relay RTR .
fault at location FLk . However, when considering an r-th re-
lay pair, then the notation (•)(p,r) and (•)(b,r) will refer to the E. COMMUNICATIONS
primary and back-up OCRs, respectively, while considering The operating times of the OCRs can be considerably re-
the fault location at a primary OCR (Fig. 2a). duced using the communications between the OCRs. Gen-
erally, two different communication schemes can be used,
C. PICKUP-CURRENT LIMITS i.e., release or block [35]. They are shown in Fig. 3 for
In order to ensure the reliable operation of an OCR, the a line supplied from both ends. When using the release
pickup current should be greater than the maximum-possible scheme (Fig. 3a) then each OCR that picks-up according to
VOLUME 4, 2016 3

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 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/ACCESS.2019.2952934, IEEE Access

B. Polajžer et al.: Protection of MV Closed-Loop Distribution Networks with Bi-Directional Overcurrent Relays and GOOSE ...

large number of relays along the feeder, loop or a side

release block
Rm Rm Rn branch. Another is the maximum permissible operating times
of substation OCRs, like R1 and R4 in Fig. 1. In order to
Rj Ri Rj
release block enhance the selectivity, an extremely inverse characteristic is
a) b)
proposed.
FIGURE 3. Communications between OCRs on a line supplied from both
ends: release (a), and block (b). 2) Avoiding Unwanted Trips of In-loop OCRs
Generally, unwanted trips occur due to the changes in topol-
the direction criterion releases its neighboring OCR in the ogy, location or level of the fault. Such situations are possible
direction of the fault. Furthermore, when using the block for in-loop OCRs, i.e., when the current flowing through a
scheme (Fig. 3b), then each OCR that picks-up according designated relay pair is increased due to the faster operation
to the direction criterion blocks its neighboring OCR in the of the remote OCR in the direction of a fault. Let us consider
opposite direction of the fault. However, in cases of com- a fault in a section between R2 and R5 (Fig. 1). When R5
munication failure or delay in communications the operation and R9 operate first, then the current flowing through R2
of the OCRs should meet the performance requirements on and R1 will increase, which will result in a decreased ∆t
reliability, sensitivity and selectivity. between R2 and R1 . Consequently, R1 might operate faster
than R2 . In order to avoid such unwanted trips the short-
III. MEASURES TO IMPROVE AN OCR’S OPERATION IN circuit calculations should be performed for two different
MV CLOSED-LOOP DISTRIBUTION NETWORKS states, defined by the operation of the in-loop OCRs. The
A. INCREASING SENSITIVITY first state is given by the normal loop topology. The second
The pickup-current setting IP j can be chosen within the limits state considers the changed topology, i.e., the loop is opened
given by (2). A high setting of IP j will inherently reduce at the location of a remote OCR in the direction of a fault.
the sensitivity. Consequently, high-resistance faults might not The largest value between the so-obtained currents is then
be "seen" by the j-th OCR, or its operating time might be used for the timing coordination of the designated relay pair,
delayed due to the inverse time-current characteristic. It is, i.e., (R2 , R1 )1 in the discussed case (Fig. 1). This simple
therefore, preferable to set IP j as low as possible; moreover, procedure increases the selectivity and reliability.
a low setting of IP j will also reduce the operating time.
3) Substation OCRs
However, low settings of IP j must not affect the selectivity.
This can be achieved through a simultaneous minimization of The selectivity of the substation OCRs (R1 and R4 in Fig. 1)
the OCRs’ operating times and pickup-current settings. Thus, is achieved through a timing coordination with the trans-
a normalized objective function is proposed as former relay RTR . Typically, an OCR with a definite time
characteristic is used for the relay RTR , where the pickup
1 NRP t(p,r) + t(b,r) current is set for a fault at the beginning of the feeder.
q = α ∑ tEV +
NRP r=1 Consequently, the selectivity of the j-th substation OCR is
(4) achieved by limiting its operating time for a fault at the
beginning of the feeder, which is denoted as t ∗j,k . The typical
!
NR
1 IP j − IP j min
β 1+
NR ∑ IP j max − IP j min +p limit for MV networks is given as t ∗j,k < 300 ms.
j=1

where α and β are the weights (α +β ≤ 1), whereas NRP C. OPTIMIZATION

and NR denote the number of relay pairs and the number of 1) Self-Adaptive Differential Evolution
OCRs, respectively. The first term in (4) represents the mean A stochastic search algorithm called differential evolution
value of the operating times of all the primary and back- (DE), capable of solving nonlinear and bounded optimization
up OCRs normalized to the expected value tEV . The second problems, has been applied as an optimization tool [30].
term represents the mean value of the pickup-current settings, It has been chosen because of its simplicity and proven
which are normalized using a min-max approach. suitability for solving real-life engineering problems. DE
incorporates a scheme that generates trial parameter vectors,
B. ENHANCING SELECTIVITY which involves three operations: mutation, crossover, and
Selectivity is achieved by the timing coordination of the selection. There are several mutation strategies; however, the
OCR’s operation, as described in Section II-D, i.e., by fulfill- best strategy for solving a particular optimization problem
ing the condition ∆tr > CTIr for the r-th relay pair. In order depends on the problem itself. The control parameters of the
to enhance the selectivity of the OCRs in MV distribution DE algorithm are the population size NP, the step size F and
networks, the following topics should be addressed. the crossover-probability constant CR. The suggested choices
of [30] are F ∈ [0.5, 1], CR ∈ [0.8, 1], and NP = 10D. Here,
1) Inverse Time-Current Characteristic Type D denotes a dimension of a parameter vector. In the given
The timing coordination of the OCRs in MV distribution case D=2NR , since the j-th OCR has two setting parameters,
networks faces two competing factors. One factor is the i.e., IP j and TD j . Furthermore, choosing suitable values for F
4 VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 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/ACCESS.2019.2952934, IEEE Access

B. Polajžer et al.: Protection of MV Closed-Loop Distribution Networks with Bi-Directional Overcurrent Relays and GOOSE ...

and CR is a problem-dependent task. Therefore, self-adaptive mechanism. This allows, e.g., the OCR to deliver the infor-
differential evolution (SADE) has been used, where F and mation simultaneously to a predefined group of destination
CR are adapted as proposed in [29]. Note that when using a OCRs.
classic DE, then approximately one million iterations were In the given case the pick-up GOOSE messages will be
needed for the case study discussed in this paper, whereas sent between the RSCs and the substation OCRs. However,
when using SADE the number of iterations was considerably an instantaneous trip will be enabled only when the OCR
reduced. receives the message and when the condition given by a
logical equation is fulfilled. Thus, e.g., the instantaneous trip
2) Variable Penalties of R3 and R4 shown in Fig. 1 should be enabled only when
A very important part of the performed optimization is the they receive a pick-up message from each other, and when
penalties p in the objective function (4). Note that when using they do not receive a pick-up message from R7 . In this way
fixed penalties the optimization process was not converging the faults on the line between R3 and R4 can be cleared faster,
for the case study discussed in this paper. Therefore, the whereas for the faults on the side branch only R7 will operate
penalties were introduced in such a way that they force according to the set time-current characteristic.
the optimization procedure towards the region with viable
IV. METHODOLOGY FOR THE PARAMETERIZATION OF
solutions of the optimization problem. This means that the
OCRS
penalties are dependent on the difference between the actual
This section summarizes the proposed measures for improv-
values of the individual parameters, and their minimum- or
ing the OCRs’ operation in MV distribution networks. Thus,
maximum-allowed values, thus ensuring the convergence of
a methodology for the parameterization of OCRs is given by
the optimization procedure.
the following steps:
The penalties are calculated for ∆tr ≤ CTIr (typical CTI
value is within the range 100 to 300 ms), t ∗j,k ≥ 300 ms, Step 1:Relay points are determined that require bi-
TD j < TD j min , IP j < IP j min and IP j > IP j max . For all the afore- directional OCRs. In cases where larger DGs are
mentioned parameters (∆tr ,t ∗j,k , TD j , IP j ), the values for which connected, the short-circuit calculations are needed
V are bounded with the minimum Vmin and the maximum to check their impact on the current directions (R9
Vmax allowed values, the variable penalties pV are introduced in Fig. 1).
according to the pseudo code: Step 2:Short-circuit calculations are performed for Ph-Ph
faults at all the relay points, as well as at the ends of
if V < Vmin then
all the lines. For all the in-loop relays, short-circuit
pV = ((|Vmin −V | + 1)K1 )K2
calculations should be performed for two different
end if
states (Section III-B2).
if V > Vmax then
Step 3:Pickup-current limits are determined for all the
pV = ((|V −Vmax | + 1)K1 )K2
OCRs (Section II-C).
end if
Step 4:Pickup-current settings and time-dial settings are
where K1 = 10 and K2 = 4 are the constants, the values of determined through the minimization of the pro-
which were determined based on experiences. The values of posed objective function using the SADE algo-
all the penalties pV determined in this way for all relay pairs rithm and the mutation strategy “rand/1/bin” (Sec-
are summed up and added as a penalty p in (4). The described tions III-A and III-C).
procedure forces (4) to fulfill all the optimization bounds Step 5:Communications are applied according to the
before the actual operating times and pickup-current settings GOOSE model, where a release communication
are minimized. The authors are not aware of a publication scheme is adopted for all the in-loop OCRs (Sec-
where the proposed approach is applied to the operating-time tions II-E and III-D).
minimization of OCRs that provides the required selectivity The proposed timing coordination is validated for a three-
and sensitivity in a MV closed-loop distribution network. bus system through a comparison with the MINLP using
a standard branch-and-bound (SBB) and a seeker algo-
D. PEER-TO-PEER COMMUNICATIONS rithm [5], and through the comparison with the QCQP-
The IEC standard 61850 [22] enables peer-to-peer commu- based algorithm [6]. To make a clear comparison the same
nications between different devices. Protection applications assumptions and limits were used as in [5], [6], i.e., IP j and
require high-speed peer-to-peer-communications that should TD j were limited to the minimum values of 1.5 A (secondary)
guarantee the total transfer time below the order of a quarter and 100 ms, respectively, whereas the IEC standard inverse
of a cycle. GOOSE messages have the shortest maximum- characteristic was used. Furthermore, one-directional OCRs
allowed transfer time among all the IEC 61850 messages, were used, and thus the relay pairs were defined as in Fig. 4.
corresponding to the required 3 ms. They can be sent over Note that timing coordination with the protection relays of
TCP/IP or substation local area networks using high-speed the generators G1 , G2 and G3 was not considered. Moreover,
switched Ethernet. Furthermore, the exchange of GOOSE CTIr = 0.2 s was used when comparing the results with [5],
messages between devices is based on a publisher-subscriber whereas CTIr = 0.3 s was used when comparing the results
VOLUME 4, 2016 5

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 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/ACCESS.2019.2952934, IEEE Access

B. Polajžer et al.: Protection of MV Closed-Loop Distribution Networks with Bi-Directional Overcurrent Relays and GOOSE ...

TABLE 1. Relay settings for the three-bus system – IEC Standard Inverse (A = 0.14, B = 0.02, C = 0).

CTIr = 0.2 s CTIr =0.3 s

SBB [5] Seeker [5] DE–Proposed QCQP [6] DE–Proposed
IP j [A] TD j [ms] IP j [A] TD j [ms] IP j [A] TD j [ms] IP j [A] TD j [ms] IP j [A] TD j [ms]
R1 1.5 151 2.5 107 1.958 100 2.862 100 2.863 100
R2 2.0 100 1.5 112 1.500 100 1.720 100 1.720 100
R3 1.5 128 2.0 108 1.500 100 1.500 100 1.500 100
R4 2.0 130 3.0 100 1.795 100 2.482 100 2.482 100
R5 2.5 106 2.5 100 1.500 100 1.500 100 1.500 100
R6 2.5 104 2.5 100 1.690 100 2.342 100 2.323 100
T [s] 6.362 6.324 4.534 5.032 5.033
T 0 [s] 6.403 6.359 4.550 5.031 5.027

R1 R3 Legend: 110/20 kV substation, RMU location, 20 kV branch

feeder 2
feeder 3

feeder 1
R2 R4
G1 G2 loop 1 loop 2

R5 R6

G3
a) b) c)
Relay pairs (primary, back-up): (R1,R5)1,
(R2,R3)2, (R3,R6)3, (R4,R1)4, (R5,R4)5, (R6,R2)6 FIGURE 5. Discussed topologies of a 20-kV network: Loop 1 (a), Loop 2 (b),
and Radial (c), where RMU is a Ring Main Unit.
FIGURE 4. The three-bus system with relay pairs.

Source
with [6]. The objective function (4) was minimized, where RTR
α = 0.3, β = 0.7 and tEV = 0.3 s. The obtained settings of the 110/20 kV
OCRs are given in Table 1 together with the value of T . Note 31.5 MVA
that these settings cannot be applied. Therefore, the value of
T 0 is also given for the settings rounded to 10 mA (secondary) FL13
R8
FL10
R6
FL1
R1 FL
14

and 10 ms. The proposed timing coordination gives a better

FL2 R9
value of T and T 0 when compared to the SBB and seeker FL15
R20 R2
algorithms, whereas the difference with the result obtained
FL3
using the QCQP-based algorithm is negligible. Furthermore,
FL12 R7 FL29
a comparison with a recent paper [17] was also made, where
FL4
only the value for ∑ t(p,r) is given, i.e., 1.526 s. The result FL11
R21 R19 R3
obtained using the proposed timing coordination was 1.513 s. FL9
FL8
R17 R18 FL5
R16
FL16
V. CASE STUDY
R5 FL7 R4 FL
A 20-kV network of DisCo Elektro Celje is discussed. Three FL18 6
R10
feeders, connected to a 110/20-kV transformer substation, FL17 FL19
R11
supply altogether 109 MV/LV distribution transformers with FL21 FL22
over 5000 customers, along with 30 DG units. The entire 20- FL20 FL28 SG 0.4 kV
R12 450 kW
kV network contains 296 nodes and 297 branches, forming
two loops that can be remotely opened or closed by ring main R13 R14 R15
units (RMUs). Furthermore, compact RSCs [36] are used at
FL25 FL26 FL27 FL23 FL24
13 relay points that were determined by the DisCo, whereas
ordinary OCRs are adopted at the substation. Moreover, all FIGURE 6. Discussed 20-kV network with denoted OCRs and fault locations.
the OCRs enable GOOSE communications.

A. DISCUSSED NETWORK TOPOLOGIES DGs are micro solar power plants and are not considered
Fig. 5 shows three network topologies, i.e., Loop 1, Loop 2 because of their negligible contribution to the fault currents.
and Radial, that will be further discussed. The entire 20-kV Relay pairs, fault locations, as well as communication pairs
network is shown in Fig. 6. Only a small hydro-power plant are given in this section for all three network topologies.
with a synchronous generator (SG) is shown, whereas other
6 VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 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/ACCESS.2019.2952934, IEEE Access

B. Polajžer et al.: Protection of MV Closed-Loop Distribution Networks with Bi-Directional Overcurrent Relays and GOOSE ...

TABLE 2. Relay pairs (∗ denotes substation OCR) for all three network TABLE 4. Relays for all three network topologies with fault locations used for
topologies with fault locations used for timing coordination. the determination of the pickup-current limits.

Loop 1 Loop 2 Radial Loop 1 Loop 2 Radial

(R2 , R∗1 )1 FL3 (R2 , R∗1 )1 FL3 (R2 , R∗1 )1 FL3 R1 FL2,14 R1 FL2,14 R1 FL2,14
(R3 , R2 )2 FL5 (R17 , R6 )2 FL8 (R9 , R∗1 )2 FL14 R2 FL4 R2 FL4 R2 FL4
(R4 , R3 )3 FL7 (R17 , R7 )3 FL8 (R17 , R∗6 )3 FL8 R3 FL6 – – – –
(R5 , R4 )4 FL9 (R18 , R17 )4 FL6 (R18 , R17 )4 FL6 R4 FL8 – – – –
(R17 , R∗6 )5 FL8 (R7 , R∗8 )5 FL11 (R10 , R∗6 )5 FL19 R5 FL18,10 – – – –
(R18 , R17 )6 FL6 (R21 , R∗6 )6 FL12 (R11 , R10 )6 FL28 R6 FL9,18 R6 FL9,18,11 R6 FL9,18
(R19 , R18 )7 FL4 (R9 , R∗1 )7 FL14 (R12 , R10 )7 FL21 R7 FL29 R7 FL10,9,18 R7 FL29
(R20 , R19 )8 FL2 (R10 , R7 )8 FL19 (R13 , R10 )8 FL26 R8 FL12,16 R8 FL12,16 R8 FL12,16
(R7 , R∗8 )9 FL11 (R10 , R∗6 )9 FL19 (R14 , R10 )9 FL27 R9 FL15 R9 FL15 R9 FL15
(R9 , R∗1 )10 FL14 (R11 , R10 )10 FL28 (R15 , R14 )10 FL23 R10 FL28,21,26,27 R10 FL28,21,26,27 R10 FL28,21,26,27
(R9 , R20 )11 FL14 (R12 , R10 )11 FL21 (R7 , R∗8 )11 FL11 R11 FL20 R11 FL20 R11 FL20
(R10 , R5 )12 FL19 (R13 , R10 )12 FL26 (R16 , R∗8 )12 FL16 R12 FL22 R12 FL22 R12 FL22
(R10 , R∗6 )13 FL19 (R14 , R10 )13 FL27 – – R13 FL25 R13 FL25 R13 FL25
(R11 , R10 )14 FL28 (R15 , R14 )14 FL23 – – R14 FL23 R14 FL23 R14 FL23
(R12 , R10 )15 FL21 (R16 , R∗8 )15 FL16 – – R15 FL24 R15 FL24 R15 FL24
(R13 , R10 )16 FL26 (R16 , R21 )16 FL16 – – R16 FL17 R16 FL17 R16 FL17
(R14 , R10 )17 FL27 – – – – R17 FL7 R17 FL7 R17 FL7
(R15 , R14 )18 FL23 – – – – R18 FL5 R18 FL5 R18 FL5
(R16 , R∗8 )19 FL16 – – – – R19 FL3 – – – –
R20 FL1,14 – – – –
– – R21 FL13,16 – –
TABLE 3. Relay pairs with relay locations (RL) where a loop should be
opened to avoid in-loop unwanted trips.
TABLE 5. GOOSE message subscriptions and logic equations.
Loop 1 Loop 2
(R2 , R1 )1 RL3 (R7 , R8 )5 RL6 Subscribed to GOOSE Logical equation for
(R3 , R2 )2 RL4 (R21 , R6 )6 RL8 message send by instantaneous trip
(R4 , R3 )3 RL5 – – R1 R20 , R9 y1 · x20 · x̄9
(R5 , R4 )4 RL6 – – R2 R19 y2 · x19
(R17 , R6 )5 RL4 – – R3 R18 y3 · x18
(R18 , R17 )6 RL3 – – R4 R17 y4 · x17
(R19 , R18 )7 RL2 – – R5 R6 , R10 y5 · x6 · x̄10
(R20 , R19 )8 RL1 – – R6 R5 , R7 , R10 y6 · (x5 + x7 ) · x̄10
Note: RLx corresponds to the location of a relay Rx . R7 R6 , R10 y7 · x6 · x̄10
R8 R21 , R16 y8 · x21 · x̄16
R9 – –
R10 R11 , R12 , R13 , R14 y10 · x̄11 · x̄12 · x̄13 · x̄14
1) Relay Pairs and Unwanted Trips R11 – –
Protection is applied at 16 relay locations, as shown in Fig. 6. R12 – –
R13 – –
The relays are numbered R1 to R21 , whereas the fault lo- R14 R15 y14 · x̄15
cations are denoted as FL1 to FL29 . All the in-loop OCRs, R15 – –
except the substation relays R1 , R6 and R8 , are bi-directional. R16 – –
R17 R4 y17 · x4
Furthermore, R12 and R10 are non-directional, since a small R18 R3 y18 · x3
SG does not change the fault current directions. Relay pairs R19 R2 y19 · x2
are given in Table 2. Moreover, relay locations at which a R20 R1 , R 9 y20 · x1 · x̄9
R21 R8 , R16 y21 · x8 · x̄16
loop should be opened to avoid unwanted trips are given
Legend: y j . . . local pick-up, xi . . . received pick-up message, x̄i . . . not
in Table 3, as proposed in Section III-B2. received pick-up message, “·” logical AND, “+” logical OR.

2) Fault Locations
In order to perform the timing coordination of the OCRs, the 3) GOOSE-based Communications
fault locations were determined according to the rule given Table 5 shows the subscriptions of all the OCRs to the
in Section II-D. They are given in Table 2, together with the GOOSE messages sent by the other OCRs. The logical
relay pairs. Furthermore, in order to determine the maximum equations that should be fulfilled for starting an instantaneous
pickup-current limits, the fault locations were determined trip that is based on received pick-up GOOSE messages xi
according to the rule given in Section II-C. They are given and the local pick-up y j are also given. Pick-up GOOSE
in Table 4, where FLx,y denotes two possible fault locations, messages are sent between the neighboring in-loop OCRs,
i.e., FLx and FLy . In such cases only a fault location that while the OCRs on the side branches are also included to
results in a smaller current measured by the corresponding achieve selectivity, e.g., R1 is subscribed to the messages sent
OCR is considered. However, the considered current value by R20 and R9 . For the discussed case the instantaneous trip
should be larger than the maximum load current. of R1 is possible when receiving a pick-up message from R20
and not receiving it from R9 , which is denoted in Table 5 as
VOLUME 4, 2016 7

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 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/ACCESS.2019.2952934, IEEE Access

B. Polajžer et al.: Protection of MV Closed-Loop Distribution Networks with Bi-Directional Overcurrent Relays and GOOSE ...

y1 ·x20 · x̄9 . Furthermore, the operating time of R10 can also be

R19 R20
decreased, i.e., when not receiving the pick-up message from
one of R11 , R12 , R13 , R14 . The operating time of R14 can be
decreased in a similar way, i.e., when not receiving the pick- R3
RA
test location
R2 CTs { RA
up message from R15 . Note that in cases of communication
Q0
failure or delay in communications the OCRs should operate connection line
according to the set time-current characteristic. Q1

RB
B. MODELING a) neighboring DisCo b) cable

The model of the discussed 20-kV network and the FIGURE 7. Field testing on a section between relays R19 and R2 : test location
model of a bi-directional RSC were both built for MAT- (a) and a Ph-Ph fault (b).
LAB/Simulink/Simscape Power Systems. It should be
pointed out that they were verified through extensive labo-
simulation measurement
ratory and field testing. 1000

ISC [A]
500
1) Network Model
A 110-kV source was modeled as a constant impedance 0
1 2, 20 3, 19 4, 18 5, 17 6
with a maximum short-circuit capacity of 6.679 kA. Further- relay no.
more, the discussed 20-kV network incorporates π models
FIGURE 8. Measured and simulated rms values of short-circuit currents
of underground cables and overhead lines, as well as generic during a Ph-Ph test for relays in Loop 1, where bi-directional relays are
models of transformers and a SG. According to the measured denoted by two numbers.
load and generating profiles for a period of 15 months a max-
imum load was determined for all the MV/LV transformer
locations. Thus, a constant power-load model was incorpo- All the relays in the Loop 1 picked-up without operation
rated on the LV side of each corresponding transformer. The and recorded the time responses of the line currents. Note
developed 20-kV network model can be used for static short- that the directional criterion was not activated during the
circuit calculations as well as for dynamic simulations of the test. Furthermore, the distance between the test location and
relay operation. The step size for the dynamic simulations R19 is 0.6 km. Since the length of the section between R19
was set to 0.0625 ms, whereas a step size of 2.5 ms was used and R2 is 10.9 km, the fault location FL4 was assumed
for the calculation of the phasors. when comparing the field testing and simulation results. The
Field testing was performed to verify the discussed MV comparison of the measured and simulated rms values of the
network model. A Ph-Ph fault was generated in a controlled currents is shown in Fig. 8, where the simulated values agree
way, where special care was taken not to disconnect any cus- with the measured ones within a range of ±3%.
tomers. The fault was located near R19 , i.e., at the beginning
of a line that can be used to connect the discussed network 2) Bi-Directional RSC Model
to a network of the neighboring DisCo, as shown in Fig. 7a. A generic model of a switch gear was used with a specified
During normal operation this line is only connected from side time delay of 50 ms. Furthermore, a dynamic model of a bi-
A; however, it can be connected from both sides in cases of directional overcurrent function was built according to the
a total power loss in one of the DisCos. Note that an OCR specifications given by the manufacturer [36]. It is composed
is located on both sides of the line, i.e., RA and RB . During of the following units. The model of the input unit incorpo-
the test, the discussed line was connected only from side A. rates filtering and sampling. The line currents and voltages
No special equipment was needed to perform the test, except are filtered by low-pass, second-order Butterworth filters with
a cable used for the Ph-Ph connection, as shown in Fig. 7b. a cut-off frequency of 1.6 kHz, while the sampling frequency
Only the Loop-1 topology was tested, while the following is 3.2 kHz. Next, a discrete Fourier filtering is applied using
procedure was followed: a full-cycle data-window. The resulting phasors are refreshed
Step 1:All the in-loop relays were set only to pick-up every 2.5 ms. The model of the main unit describes the
without operation, whereas the relay RA was set to overcurrent function with IEC inverse-time pickup character-
pick-up and operate with no delay. istics and additional definite-time reset characteristics. Fur-
Step 2:Switch gear Q1 was closed first, then the closed thermore, a bi-directional functionality was modeled with a
poles were visually inspected. standard directional criterion, which is based on the angle
Step 3:Switch gear Q0 was closed next to generate a Ph-Ph between the current phasor and the line-voltage phasor. The
fault. model of the logic unit provides the pickup and trip signals.
Step 4:Relay RA operated, and Q0 was opened after ap- Moreover, a GOOSE model was used [23], where a transfer
proximately 60 ms. Note that none of the customers time of 5 ms was considered. Note, that the same dynamic
were disconnected. model was used for all the discussed relay locations.
8 VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 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/ACCESS.2019.2952934, IEEE Access

B. Polajžer et al.: Protection of MV Closed-Loop Distribution Networks with Bi-Directional Overcurrent Relays and GOOSE ...

TABLE 6. Minimum, mean and standard deviation (Std. Dev.) of best q values
Schneider controller for 50 independent runs – radial network topology.
ADVC3
Omicron CMC 256-6
SADE1 SADE2
Secondary current Strategy Min Mean Std. Dev. Min Mean Std. Dev.
injection test set best/1/exp 1.051 1.144 0.117 1.033 1.117 0.069
rand/1/exp 1.028 1.039 0.003 1.023 1.038 0.009
rand-to-best/1/exp 1.024 1.057 0.042 1.029 1.053 0.025
best/2/exp 1.028 1.053 0.026 1.025 1.052 0.019
rand/2/exp 1.036 1.048 0.018 1.054 1.139 0.035
best/1/bin 1.034 1.187 0.131 1.027 1.133 0.083
Secondary voltage
rand/1/bin 1.018 1.021 0.004 1.018 1.027 0.005
injection interface set rand-to-best/1/bin 1.027 1.059 0.084 1.021 1.056 0.016
best/2/bin 1.020 1.052 0.025 1.027 1.053 0.032
FIGURE 9. Laboratory set-up for secondary testing of a protection unit rand/2/bin 1.042 1.074 0.026 1.105 1.151 0.023
(Schneider controller ADVC3).

101
SADE1, rand/1/exp

q (mean best)
SADE1, rand/1/bin
The developed model of a bi-directional overcurrent pro- SADE2, rand/1/exp
tection was verified with extensive laboratory testing using SADE2, rand/1/bin

the set-up shown in Fig. 9. The test currents were varied

within the range 1.2 to 5 times the pickup-current setting.
The IEC extremely-inverse characteristic was tested, where 100
0 500 1000 1500 2000 2500
IP j = 500 A and TD j = 10 ms. Time delays were determined iteration
for the model and for an actual protection unit as the differ- FIGURE 10. Mean best curves for 50 independent runs of selected algorithms
ence between the measured operating times and the operating – radial network topology.
times obtained by (1). The time delays were shorter at higher
currents for the protection model, as well as for an actual 4000 loop 1
protection unit. The average value of the time delay was loop 2
Ij,SCmin [A]

3000 radial
approximately 25 ms. However, the obtained time delay of
the tested model and the actual unit were different within 2000

the acceptable range of ±5 ms. Moreover, the directional 1000

Ij,Lmax
function was also tested for the proposed protection model 0
and an actual protection unit. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
relay no.

VI. RESULTS FIGURE 11. Maximum-load currents and minimum short-circuit currents for all
three network topologies.
A. BEST OBJECTIVE FUNCTION VALUE AND
CONVERGENCE
SADE was extensively tested for the case with the radial B. PICKUP-CURRENT LIMITS AND OCR SETTINGS
network topology. Self-adaptation was applied in two differ- The obtained maximum-load currents I j,L max and the mini-
ent ways, i.e., for the entire population, as well as for each mum short-circuit currents I j,SC min are shown in Fig. 11. Note
population member. This is denoted as SADE1 and SADE2, that equal values of I j,L max were used for all three network
respectively. Furthermore, ten different mutation strategies topologies. The values I j,SC min were determined according to
were considered. Table 6 shows the minimum, mean and the fault locations given in Table 4 and show considerable
standard deviations of the best values of the objective func- impacts of the network topology. Pickup-current limits were
tion q for 50 independent runs, where α = 0.3, β = 0.7 determined using (2), where eCT% = 10% and KSC =KL = 1.1
and tEV = 0.3 s. The convergence was also tested for the were considered. Furthermore, at several relays the value of
mutation strategies “rand/1/bin” and “rand/1/exp”. Fig. 10 I j,L max is very small. In such cases IP j min was set to the
shows that for approximately the first 135 iterations the estimated value of 50 A, which corresponds to a possible
obtained values were penalized, whereas the minimum value increase in demand over time.
was reached before 2500 iterations. The typical computation The CTIs for the relay pairs between individual RSCs were
time for a single run of our implementation in Matlab on 100 ms, while the CTIs for the relay pairs with substation re-
a 3-GHz Intel Core i5-7400 with 8 GB of RAM was less lays R1 , R6 and R8 were 150 ms. Furthermore, the minimum
than 2 min. Furthermore, the results were also qualitatively time-dial setting was TD j min = 10 ms. Minimization of the
the same for other network topologies, as well as when objective function (4), where α = 0.3, β = 0.7, tEV = 0.3 s,
testing the objective function T . Based on the obtained results was performed using SADE1 and “rand/1/bin”. The results
given in Table 6 and Fig. 10, the SADE1 variant and the are given in Table 7, where the obtained time-dial settings
“rand/1/bin” strategy were found to be the most suitable for were rounded to 10 ms, whereas the obtained pickup-current
the discussed optimization problem. settings were rounded to 1 A.
VOLUME 4, 2016 9

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 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/ACCESS.2019.2952934, IEEE Access

B. Polajžer et al.: Protection of MV Closed-Loop Distribution Networks with Bi-Directional Overcurrent Relays and GOOSE ...

TABLE 7. Relay settings – IEC Extremely Inverse (A = 80, B = 2, C = 0).

0.4
Loop 1 Loop 2 Radial 0.3

tr [s]
IP j [A] TD j [ms] IP j [A] TD j [ms] IP j [A] TD j [ms] CTI CTI*
0.2
R1 676 50 493 110 460 80
R2 254 90 98 50 120 60 0.1
R3 217 40 – – – –
0
R4 105 100 – – – – 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
R5 118 40 – – – – relay pair no.
R6 390 60 204 80 286 70
R7 74 60 136 60 95 130 0.5 primary
R8 555 30 291 70 353 70

t(p,r), t(b,r) [s]

back-up
R9 141 60 109 50 74 130 0.4
R10 234 50 272 30 169 60 0.3
R11 75 70 72 20 63 20 0.2
R12 137 80 64 150 55 30
0.1
R13 165 20 57 50 71 40
R14 161 60 168 40 100 90 0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
R15 70 110 54 40 50 10 relay pair no.
R16 63 130 113 90 94 20
R17 252 50 146 110 161 70 FIGURE 12. ∆tr values and operating times of primary and back-up OCRs –
R18 136 80 76 30 50 10 Loop-1 topology, where (•)∗ denotes the relay pairs with substation OCRs.
R19 86 70 – – – –
R20 78 10 – – – –
R21 – – 91 100 – – 0.3

tr [s]
0.2 CTI CTI*
TABLE 8. Operating times of substation OCRs for a fault at the beginning of a
feeder.
0.1
Loop 1 Loop 2 Radial
t ∗j,k [ms] t ∗j,k [ms] t ∗j,k [ms] 0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
R1 FL1 70 80 51
relay pair no.
R6 FL10 28 10 18 primary
R8 FL13 28 18 26 0.3 back-up
t(p,r), t(b,r) [s]

0.2
C. SELECTIVITY TESTING AND OPERATING TIMES
0.1
1) Theoretical Calculations
The operating times of the primary and back-up OCRs, 0
i.e., t(p,r) and t(b,r) , were calculated using (1) for the fault 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
relay pair no.
locations given in Table 2, while using the optimum settings.
The results are shown in Figs. 12–14, together with the ∆tr FIGURE 13. ∆tr values and operating times of primary and back-up OCRs –
Loop-2 topology, where (•)∗ denotes the relay pairs with substation OCRs.
values. Note that the results denoted as (•)∗ are given for
relay pairs with substation OCRs (R1 , R6 and R8 ). Selectivity
is achieved for all the relay pairs since ∆tr > CTIr . Moreover, 0.4
selectivity is also assured for the substation OCRs, since 0.3
tr [s]

t ∗j,k < 300 ms, as is shown in Table 8. 0.2

CTI CTI*

0.1
2) Dynamic Simulations
0
The models proposed in Section V-B were used, while 1 2 3 4 5 6 7 8 9 10 11 12
using the optimum settings given in Table 7. The Ph-Ph relay pair no.
faults were simulated for all possible fault locations. The
0.4 primary
simulations were performed for the RSCs with and without
t(p,r), t(b,r) [s]

back-up
the GOOSE communications. The resulting clearing times 0.3

are given in Table 9, which include the operation of the 0.2

RSCs (relay and switch gear), as well as the 5-ms transfer 0.1
time in the case of the communications. The RSCs without 0
communications operated in a selective way according to 1 2 3 4 5 6 7 8 9 10 11 12
relay pair no.
the expected time delays. Furthermore, faults located close
to the substation cannot be picked-up with in-loop RSCs, FIGURE 14. ∆tr values and operating times of primary and back-up OCRs –
Radial topology, where (•)∗ denotes the relay pairs with substation OCRs.
e.g., for the Loop-1 topology the R20 does not pick-up for
a fault at FL1 . Consequently, the R1 operated first, which
increased the current through the R20 that was then picked-
10 VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 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/ACCESS.2019.2952934, IEEE Access

B. Polajžer et al.: Protection of MV Closed-Loop Distribution Networks with Bi-Directional Overcurrent Relays and GOOSE ...

TABLE 9. Results of dynamic simulations for all the fault locations and all the topologies, where the clearing times include the operation of the RSCs (relay and
switch gear) and communication transfer times.

Loop 1 Loop 2 Radial

without comm. with comm. without comm. with comm. without comm. with comm.
operated t j,k [ms] t j,k [ms] operated t j,k [ms] t j,k [ms] operated t j,k [ms] t j,k [ms]
FL1 R1 & R20 134, 213 134, 213 R1 134 74 R1 114 74
FL2 R1 & R20 273, 124 72, 72 R1 302 73 R1 213 73
FL3 R2 & R19 114, 210 72, 72 R2 63 63 R2 63 63
FL4 R2 & R19 542, 137 62, 61 R2 92 71 R2 131 71
FL5 R3 & R18 267, 281 69, 66 R18 81 81 R18 63 63
FL6 R3 & R18 343, 175 73, 71 R18 61 61 R18 61 61
FL7 R4 & R17 271, 311 62, 62 R17 171 71 R17 173 73
FL8 R4 & R17 280, 191 69, 65 R17 113 73 R17 123 72
FL9 R5 & R 6 235, 352 62, 62 R6 & R7 201, 233 73, 64 R7 212 72
FL10 R5 & R 6 352, 83 352, 83 R6 & R7 64, 193 64, 193 R6 74 65
FL11 R7 63 63 R6 & R7 143, 94 73, 64 R7 83 83
FL12 R8 252 73 R8 & R21 153, 104 64, 69 R8 183 73
FL13 R8 94 74 R8 & R21 74, 191 74, 191 R8 74 66
FL14 R9 63 63 R9 63 63 R9 63 63
FL15 R9 73 73 R9 63 63 R9 63 63
FL16 R16 72 72 R16 82 82 R16 62 62
FL17 R16 72 72 R16 92 92 R16 62 62
FL18 R5 & R 6 255, 371 61, 62 R6 & R7 204, 243 73, 64 R7 221 72
FL19 R10 122 72 R10 122 72 R10 122 72
FL20 R11 71 71 R11 61 61 R11 63 63
FL21 R12 181 181 R12 111 111 R12 71 71
FL22 R12 191 191 R12 112 112 R12 71 71
FL23 R15 111 111 R15 71 71 R15 61 61
FL24 R15 123 123 R15 74 74 R15 61 61
FL25 R13 151 151 R13 81 81 R13 101 101
FL26 R13 111 111 R13 71 71 R13 83 83
FL27 R14 181 71 R14 151 73 R14 141 71
FL28 R11 71 71 R11 61 61 R11 61 61
FL29 R7 71 71 R6 & R7 114, 131 73, 64 R7 93 93

up. The same situation was observed for the fault at FL10 operation of the RSCs is due to the closed-loop operation and
and the operation of R5 and R6 . Obviously, communications an appropriate protection design.
could not be activated for the discussed fault locations FL1
and FL10 , which was confirmed with the obtained results. VII. CONCLUSION
A similar result was also noticed for the Loop-2 topology This paper proposes measures that improve the operation
for the fault locations FL10 and FL13 . Based on the obtained of OCRs in MV closed-loop distribution networks. The se-
results it can be concluded that the operating times of the lectivity of the substation OCRs is achieved by considering
communication-assisted RSCs were significantly reduced. the required maximum operating time (300 ms), while the
unwanted trips of the in-loop OCRs are avoided by a simple
D. IMPLEMENTATION procedure for short-circuit calculations. The proposed timing
The proposed protection design is already implemented in a coordination of the OCRs’ operation minimizes the operating
20-kV network of DisCo Elektro Celje using compact RSCs. times and increases the sensitivity of all the relays, i.e., RSCs
Even though the GOOSE communications are not yet estab- and substation OCRs. Preliminary results showed that, when
lished, the field results are promising. During two recent Ph- using the classic DE in combination with fixed penalties, the
Ph faults in the section between R3 and R18 (Loop 1) and one optimization process was not converging for the discussed
Ph-Ph fault in the end branch near R15 the protection operated case study. Therefore, variable penalties were introduced that
in a reliable and selective way. Fig. 15 shows photographs of force the optimization procedure towards the region with
the discussed Ph-Ph faults that were all on overhead lines due the viable solutions of the optimization problem. Further-
to falling trees. Note that before modernization, the discussed more, SADE was used, which showed considerably better
network operated in a radial arrangement of feeders, using convergence than the classic DE. Two variants of SADE
only the substation relays R1 , R6 and R8 . Consequently, were extensively tested together with ten mutation strategies.
these Ph-Ph faults would have resulted in an outage of 2124 Based on the obtained results the strategy “rand/1/bin” was
customers. However, according to the data obtained from chosen together with a SADE variant that is based on the self-
DisCo Elektro Celje, only 49 customers were out of power adaptation of the control parameters for the entire population.
during faults in the section between R3 and R18 , while for The resulting operating times of the OCRs are comparable
a fault in the end branch near R15 only 26 customers were with the operating times determined using methods known in
disconnected. It should be emphasized, that the resulting the literature. Moreover, adopting the GOOSE communica-
VOLUME 4, 2016 11

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 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/ACCESS.2019.2952934, IEEE Access

B. Polajžer et al.: Protection of MV Closed-Loop Distribution Networks with Bi-Directional Overcurrent Relays and GOOSE ...

Event on 25 March 2019 Event on 26 April 2019 Event on 14 May 2019

Section between R3 and R18 Section between R3 and R18 End branch near R15

FIGURE 15. Photographs of Ph-Ph faults due to falling trees.

tions reduces the operating times, which was fully confirmed [11] F. B. Bottura, W. M.S. Bernardes, M. Oleskovicz, et al., “Setting di-
through the dynamic simulations of the Ph-Ph faults in the rectional overcurrent protection parameters using hybrid GA optimizer,”
Elect. Power Syst. Res., vol. 143, pp. 400–408, Feb. 2017.
discussed 20-kV network. The discussed protection design [12] A. A. Kida, and L. A. Gallego, “A high-performance hybrid algorithm to
with compact RSCs is already implemented and operates solve the optimal coordination of overcurrent relays in radial distribution
permanently in a 20-kV network of DisCo Elektro Celje networks considering several curve shapes,” Elect. Power Syst. Res.,
vol. 140, pp. 464–472, Nov. 2016.
that supplies more than 5000 customers. Field results are [13] D. Saha, A. Datta, and P. Das, “Optimal coordination of directional
promising; even though the GOOSE communications are not overcurrent relays in power systems using symbiotic organism search
yet established, the RSCs have selectively operated during optimisation technique,” IET Gener., Transm., Distrib., vol. 10, no. 11,
pp. 2681–2688, Aug. 2016.
recent Ph-Ph faults. Thus, we can expect that the SAIDI
[14] M. Singh, B. Panigrahi, and A. Abhyankar, “Optimal coordination of
and SAIFI will be improved in the future. There are natural directional over-current relays using teaching learning-based optimization
extensions for the work presented in this paper, especially (TLBO) algorithm,” Int. J. Elect. Power Energy Syst. Res., vol. 50, pp. 33–
41, Sep. 2013.
the adaptation of the proposed measures for the improved
[15] H. H. Zeineldin, E. F. El-Saadany, and M. M. A. Salama, “Optimal coordi-
operation of ground fault-protection relays in MV closed- nation of overcurrent relays using a modified particle swarm optimization,”
loop distribution networks. Elect. Power Syst. Res., vol. 76, no. 11, pp. 988–995, Jul. 2006.
[16] M. Y. Shih, C. A. C. Salazar, and A. C. Enriquez, “Adaptive directional
overcurrent relay coordination using ant colony optimisation,” IET Gener.
REFERENCES Transm. Distrib., vol. 9, no. 4, pp. 2040–2049, Nov. 2015.
[1] D. P. Le, D. M. Bui, C. C. Ngo, et al., “FLISR approach for smart [17] A. Wadood, T. Khurshaid, S. G. Farkoush, et al., “Nature-inspired whale
distribution networks using E-Terra software–A case study,” Energies, optimization algorithm for optimal coordination of directional overcurrent
vol. 11, no. 12, art. no. 3333, Dec. 2018. relays in power systems,” Energies, vol. 12, no. 12, art. no. 229, Jun. 2019.
[2] Definitions for Power Switchgear, IEEE Standard C37.100-1992 (R2001), [18] H. H. Zeineldin, H. M. Sharaf, D. K. Ibrahim, et al., “Optimal protection
2001. coordination for meshed distribution systems with DG using dual setting
[3] J. Sykes et al., “Reliability fundamentals of system protection,” NERC, directional over-current relays,” IEEE Trans. Smart Grid, vol. 6, no. 1,
Princeton, NJ, USA, Report for the Planning Committee, 2010. pp. 115–123, Jan. 2015.
[4] A. S. Noghabi, H. R. Mashhadi, and J. Sadeh, “Optimal coordination [19] L. Huchel, and H. H. Zeineldin, “Planning the coordination of directional
of directional overcurrent relays considering different network topologies overcurrent relays for distribution systems considering DG,” IEEE Trans.
using interval linear programming,” IEEE Trans. Power Del., vol. 25, no. 3, Smart Grid, vol. 7, no. 3, pp. 1642–1649, May 2016.
pp. 1348–1354, Jul. 2010. [20] K. Pereira, B. R. Pereira, J. Contreras, et al., “A multiobjective optimiza-
[5] T. Amraee, “Coordination of directional overcurrent relays using seeker tion technique to develop protection systems of distribution networks with
algorithm,” IEEE Trans. Power Del., vol. 27, no. 3, pp. 1415–1422, Jul. distributed generation,” IEEE Trans. Power Syst., vol. 33, no. 6, pp. 7064–
2012. 7075, Nov. 2018.
[6] V. A. Papaspiliotopoulos, G. N. Korres, and N. G. Maratos, “A novel [21] S. Katyara, L. Staszewski, and Z. Leonowicz, “Protection coordination of
quadratically constrained quadratic programming method for optimal co- properly sized and placed distributed generations-methods, applications
ordination of directional overcurrent relays,” IEEE Trans. Power Del., and future scope,” Energies, vol. 11, no. 10, art. no. 2627, Oct. 2018.
vol. 32, no. 1, pp. 3–10, Feb. 2017. [22] Communication Networks and Systems for Power Utility Automation, 2.0,
[7] D. L. A. Negrão, and J. C. M. Vieira, “The local fit method for coordinating Standard IEC 61850, 2013.
directional overcurrent relays,” IEEE Trans. Power Del., vol. 31, no. 4, [23] A. Apostolov, “To GOOSE or not to GOOSE? - that is the question,” in
pp. 1464–1472, Aug. 2016. 68th Annual Conference for Protective Relay Engineers, College Stn, TX,
[8] M. Y. Shih, A. C. Enríquez, T.-Y. Hsiao, et al., “Enhanced differential USA, 2015, pp. 583–596.
evolution algorithm for coordination of directional overcurrent relays,” [24] Y. Liu, H. Gao, W. Gao, et al., “Development of a substation-area backup
Elect. Power Syst. Res., vol. 143, pp. 365–375, Feb. 2017. protective relay for smart substation,” IEEE Trans. Smart Grid, vol. 8,
[9] M. H. Costa, R. R. Saldanha, M. G. Ravetti, et al., “Robust coordination of no. 6, pp. 2544–2553, Nov. 2017.
directional overcurrent relays using a matheuristic algorithm,” IET Gener., [25] S. Kumar, N. Das, and S. Islam, “High voltage substation automation
Transm., Distrib., vol. 11, no. 2, pp. 464–474, Jan. 2017. and protection system based on IEC 61850,” in AUPEC, Auckland, New
[10] F. Razavi, H. A. Abyaneh, M. Al-Dabbagh, et al., “A new comprehensive Zealand, 2018, pp. 1–6.
genetic algorithm method for optimal overcurrent relays coordination,” [26] M. A. Aftab, S. Roostaee, S. M. S. Hussain, et al., “Performance evaluation
Elect. Power Syst. Res., vol. 78, no. 4, pp. 713–720, Apr. 2008. of IEC 61850 GOOSE-based inter-substation communication for acceler-

12 VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 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/ACCESS.2019.2952934, IEEE Access

B. Polajžer et al.: Protection of MV Closed-Loop Distribution Networks with Bi-Directional Overcurrent Relays and GOOSE ...

ated distance protection scheme,” IET Gener., Transm., Distrib., vol. 12, MATEJ PINTARIČ received his M.Eng. degree in
no. 18, pp. 4089–4098, Oct. 2018. electrical engineering from the Faculty of Electri-
[27] A. Leal, and J. F. Botero, “Defining a reliable network topology in cal Engineering and Computer Science, University
software-defined power substations,” IEEE Access, vol. 7, pp. 14323– of Maribor, Maribor, Slovenia, in 2016. Since
14339, Feb. 2019. 2016 he has been an Assistant of Electrical En-
[28] S. M. Farooq, S. M. S. Hussain, and T. S. Ustun, “Performance evaluation gineering at the University of Maribor, Faculty of
and analysis of IEC 62351-6 probabilistic signature scheme for securing Electrical Engineering and Computer Science. His
GOOSE messages,” IEEE Access, vol. 7, pp. 32343–32351, Mar. 2019.
research interests include smart grids and energy
[29] J. Brest, S. Greiner, B. Bošković, et al., “Self-adapting control parameters
management.
in differential evolution: A comparative study on numerical benchmark
problems,” IEEE Trans. Evol. Computat., vol. 10, no. 6, pp. 646–657, Dec.
2006.
[30] K. Price, R. M. Storn, and J. A. Lampinen, “Differential evolution: A
practical approach to global optimization,” Berlin Heidelberg, Germany,
Springer-Verlag, 2005.
[31] M. Ojaghi, and R. Ghahremani, “Piece-wise linear characteristic for coor- MIRAN ROŠER graduated in 2004 from the
dinating numerical overcurrent relays,” IEEE Trans. Power Del., vol. 32, Department of FERI Technical Faculty of the
no. 1, pp. 145–151, Feb. 2017. University of Maribor. In 2014 he received his
[32] Measuring Relays and Protection Equipement-Part 151: Functional Re- Ph.D. from the Faculty of Electrical Engineering,
quirements for Over/Under Current Protection, Standard IEC 60255-151, Computer Science and Informatics in Maribor. He
2009. is employed by the company Elektro Celje, d. d.,
[33] Inverse-Time Characteristics Equations for Overcurrent Relays, IEEE where he deals with measurements and relay pro-
Standard C37.112-2018, 2018. tection systems as well as methods for managing
[34] Guide for Protective Relay Applications to Distribution Lines, IEEE
and analyzing the operation and development of
C37.230-2007, 2007.
the distribution network. He also deals with the
[35] V. C. Nikolaidis, E. Papanikolaou, and A. S. Safigianni, “A
communication-assisted overcurrent protection scheme for radial implementation of modern solutions within the framework of smart-grid
distribution systems with distributed generation,” IEEE Trans. Smart Grid, projects.
vol. 7, no. 1, pp. 114–123, Jan. 2016.
[36] ADVC Controller Range Operations Manual, R29, Schneider Electric,
Brisbane, QLD, Australia, 2018.

GORAZD ŠTUMBERGER (M’92) received his

B.Sc., M.Sc., and Ph.D. degrees from the Uni-
versity of Maribor, Maribor, Slovenia, in 1989,
BOŠTJAN POLAJŽER (M’13) received his 1992, and 1996, respectively, all in electrical engi-
B.Sc. and Ph.D. degrees in electrical engineering neering. Since 1989 he has been with the Faculty
from the Faculty of Electrical Engineering and of Electrical Engineering and Computer Science,
Computer Science, University of Maribor, Mari- University of Maribor, where he is currently a Pro-
bor, Slovenia, in 1997 and 2002, respectively. fessor of Electrical Engineering. In 1997 and 2001
Since 1998 he has been with the Faculty of Electri- he was a Visiting Researcher with the University
cal Engineering and Computer Science, University of Wisconsin, Madison and, in 1998 and 1999,
of Maribor, where, since 2010, he has been an with the Catholic University Leuven, Leuven, Belgium. His current research
Associate Professor. In 2000 he was a Visiting interests include power-system research and the design, modeling, analysis,
Scholar with the Catholic University Leuven, Leu- and control of electrical machines and drives.
ven, Belgium, and in 2019 with the Graz University of Technology, Graz,
Austria. His research interests include electrical machines and devices,
power quality, and power-system protection and control.

VOLUME 4, 2016 13

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.