Electric Power Systems Research 203 (2022) 107647

Contents lists available at ScienceDirect

Electric Power Systems Research

journal homepage: www.elsevier.com/locate/epsr

Optimal reliability-centered maintenance strategy based on the failure

modes and effect analysis in power distribution systems
MohammadHosein Enjavimadar a, Mohammad Rastegar b, *
a
Departmenrt of power and Control, School of Electrical and Computer Engineering, Shiraz University, Shiraz, Iran
b
Departmenrt of power and Control, School of Electrical and Computer Engineering, Shiraz University, Shiraz, Iran

A R T I C L E I N F O A B S T R A C T

Keywords: Owing to the impressive development in the electricity distribution network, the managers of the system should
Electricity distribution network be able to handle the costs of the network effectively. The practical implementation of the reliability-centered
Failure mode and effect analysis (FMEA) maintenance is one of the ways to address this concern. In this paper, we initially use the multi-criteria deci­
Reliability-Centered Maintenance (RCM)
sion-making method of gray relative analysis to prioritize failure modes and obtain their risk factor in the
network. Various failure modes in network equipment such as insulators, jumpers, cutout fuses, switches,
transformers, and the cables are considered. Then, to select optimal strategies for each failure mode, we present
an optimization based on the energy not supplied of electrical distribution feeder, the maintenance cost, the
outage cost, the safety cost and the risk factor of failure modes. Hotline, coldline, and neutral are among the
strategies to be implemented in the current research. To show the effectiveness of the proposed method, we
utilize various scenarios to investigate the impacts of maintenance duration, hotline safety costs and practical
restrictions on the results. Findings indicate that total maintenance costs, compared to those of the business as
the usual strategy, can be reduced by at least 7% when the propose method is applied.

incorporates three major steps: the critical component identification in

1. Introduction order to perform the useful inspection, the failure mode and effect
analysis (FMEA), and the assignment of the optimal maintenance
The Failure and erosion of system components are inevitable and, strategy to all failed modes.
therefore, proposing a comprehensive structure for their maintenance is The critical component identification is considered as the first and
a crucial issue. In addition, the implementation of a structured main­ foremost step in the implementation of RCM, which can mainly meets
tenance planning not only leads to maintain the system operation status the reliability factors of the system [5]. The identification of critical
but also prevents the failures that can stop component functioning. components in power distribution system is one of the first practical
Hence, the maintenance planning and cost reduction are the main steps in this field [6]. Fuzzy analytical hierarchy process (FAHP) method
concerns in all industries and their ignorance can degrade the quality is used in [7] to identify critical components in distribution power sys­
and profitability [1–3]. These factors together cause the appearance of tems. A new method based on the reliability indices to rank critical
maintenance program based on the reliability-centered-maintenance feeders in distribution network is also proposed in [7]. In [8], the
(RCM). RCM is a technique, which defines the maintenance economi­ application of game theory is employed to identify critical components
cally in order to restore and preserve the operational ability of compo­ in power systems. Researchers in [9] utilize fuzzy best-worth method
nents. In fact, RCM with a focus on the asset management and the cost (FBWM) and Technique for Order of Preference by Similarity to Ideal
reduction is a trade-off between the preventive and corrective Solution (TOPSIS) to prioritize critical feeders in an electrical distribu­
maintenance. tion system.
RCM, known as one of the most prominent Asset Management The failure mode and effect analysis (FMEA) which is a fruitful tool
strategy among all maintenance approaches [4], focuses on the reli­ in RCM analysis is organized to assess the potential failure modes in
ability aspect of maintenance. One of its target is to achieve the optimal components. Its main goal is to reduce or eliminate the potential causes
maintenance strategy in the system components. It tries to satisfy the of failure and prioritize them based on the defined criteria. FMEA which
cost-based limitation of the facility in the whole system. RCM structure can be carried out through multi criteria decision making methods

* Corresponding author.
E-mail address: [email protected] (M. Rastegar).

https://doi.org/10.1016/j.epsr.2021.107647
Received 20 June 2021; Received in revised form 19 October 2021; Accepted 22 October 2021
Available online 30 October 2021
0378-7796/© 2021 Elsevier B.V. All rights reserved.
M. Enjavimadar and M. Rastegar Electric Power Systems Research 203 (2022) 107647

weights. To run FMEA, researchers in [9] employ the fuzzy belief

Nomenclature structure and the gray relational projection method. The authors of [11]
utilize an approach based on a developed Pythagorean fuzzy (PF)
Sets and indices k-means clustering algorithm and PF-VIKOR MCDM method to collect S,
I Index of load points O, and D factors for failure modes. In [12], a framework is presented to
J Index of failure modes minimize linguistic uncertainties and make use of interval fuzzy sets to
T Set of failure modes in maneuver section calculate the criteria weights of failure modes. In [13], a new approach,
S Index of failure in maneuver section which utilizes a fuzzy set to deal with uncertainties during an evalua­
V Index for risk factor elements such as severity, detection tion, is proposed. In this study, the AHP and the entropy technique are
and occurrence used to calculate the weights of failure modes.
N Number of risk factors The optimal maintenance strategy is the last step in the RCM defined
structure. Strategies must simultaneously meet both costs and reli­
Parameters and constants ability. In [14], a decision-making flowchart is utilized to select the
ccj Cost of coldline repair strategy in component j optimal maintenance strategy. Authors in [15] present the component
chj Cost of hotline repair strategy in component j state model which uses Markov chains and particle swarm optimization
Csc
j Safety cost of coldline repair strategy in component j (PSO) to assign the optimal maintenance strategy to the failed compo­
csh Safety cost of hotline repair strategy in component j nents. Reference [16] presents the feasible structure in implementing
j
RCM. In addition, researchers in [17] put forward the application of
rjc Maintenance duration due to coldline repair in
RCM in the transmission system. The overhead line and underground
component j cable in an electrical system are elaborated in [18] and the power
ENSci,j Energy not supplied due to the application of coldline transformer is the focus of attention in [19, 20]. Researchers in [21, 22],
repair strategy of component j in load point i furthermore, present a practical structure of RCM, which is implemented
ENShi,j Energy not supplied due to the application of hotline in three steps. Additionally, [23] and [24] define a method for the
repair strategy of component j in load point i maintenance budgeting devotion and the cost limitation in power dis­
ENSni,j Energy not supplied due to the application of neutral tribution systems, and the financial maintenance planning in electrical
strategy of component j in load point i power systems is elaborated. Reference [14] proposes the implementa­
βj Risk factor of failure mode j tion of RCM in smart grid to protect devices. In [10], a structure is
Tc Maximum time for coldline repair presented to implement RCM in power transmission components based
on the particle swarm optimization. In [25], a probabilistic structure for
Nh Number of hotline repair team
the implementation of reliability-centered-maintenance in power dis­
Variables tribution networks is also proposed.
Rhj Binary variable that indicates the use of hotline repair Although the past studies elaborated in the preceding paragraphs
strategy in component j undertake the implementation of RCM, the issue of the optimization-
Rcj Binary variable that indicates the use of hotline repair based strategy assignment has not come to their attention. FMEA has
also been the focus of many pieces of research, but the results were less
strategy in component j
fruitful to help the researchers select the maintenance strategy.
nj Binary variable that indicates the use of neutral strategy
Furthermore, no applicable maintenance strategy was proposed to be
in component j
utilized in the distribution systems. Therefore, the lack of maintenance
structure in the distribution systems is always felt. Hence, FMEA results
in selecting the maintenance strategy can be exploited to assign the
(MCDM) seeks to identify system interface function and its performance. maintenance strategy, through a mathematical model, to all failure
Some methods such as AHP, TOPSIS, and entropy analysis are often used modes in an electrical distribution system.
in FMEA. Severity (S), Occurrence (O), and Detection (D) are three In this study, an applicable structure is proposed for the imple­
factors that are incorporated in FMEA. In the classic method, the output mentation of RCM. FMEA approach is, initially, employed to obtain risk
of FMEA is obtained by multiplying these three factors together. In [10], factors of all failures in the distribution systems and prioritize them. The
the application of BWM and the gray relational analysis (GRA) to pri­ presented FMEA approach consists of BWM method to obtain the
oritize failure modes is presented. BWM is used to calculate weight and importance weight of criteria and GRA method to prioritize failure
then GRA is utilized to prioritize failures based on their data and modes. Then, to assign the optimal maintenance strategy to all failure

Table 1
Comparing past literature with the present work.
MCDM for criteria weight in MCDM for prioritization in Uncertainty of Optimization of strategy Application
FMEA FMEA information assignment

[14] ⨯ ⨯ ⨯ improved Jaya algorithm (IJA) smart grid distance protection

systems
[15] ⨯ ⨯ ⨯ Particle Swarm Optimization Electrical transmission system
(PSO)
[16] ⨯ ⨯ ⨯ Optimal Power Flow(OPF) Renewable energy sources (RES)
[18] ⨯ ⨯ ✓ ⨯ Electrical distribution system
[19] ⨯ ⨯ ⨯ ⨯ Electrical transmission system
[20] ⨯ ⨯ ⨯ ⨯ Electrical transmission system
[11] PF-VIKOR k-means clustering ✓ ⨯ Gas refinery
[12] IT2F-DEMATEL IT2F-TOPSIS ✓ ⨯ Plant gearbox
[13] Analytical hierarchy process entropy technique ✓ ⨯ Refinery
(AHP)
Present best-worst method (bwm) gray relational analysis (gra) ✓ Linear mathematical model Electrical distribution system
work

2
M. Enjavimadar and M. Rastegar Electric Power Systems Research 203 (2022) 107647

modes in power distribution systems, this paper proposes an optimiza­ [ )]

∑q ( ) 1∑ q (
tion model which considers the practical constraints. The objective ̃ jv = 1
T T kjv ,
k
T jv (3)
function contains the outage and operation costs. Three binary decision q k=1 q k=1
variables are considered in optimization to assign hotline, coldline or
These average interval numbers can be put in matrix Das shown in
neutral strategies to failure modes. Time, budget, and crew restrictions
(4) [12]. The normalization of matrix D is done to put interval number in
in the practical application of maintenance strategies are considered as
the range of [0, 1]. The Normalized matrix is shown by matrix D∗ and its
the optimization constraints. The maintenance strategy cost, the outage
process described in (5) and (6) [11].
cost, the safety cost of different strategies and the energy not supplied
⎡ ⎤
(ENS) of all load points in distribution feeder when failure mode occurs
̃ 11 T
T ̃ 12 ⋯ T ̃ 1n
are employed in the proposed method. This method assigns optimal ⎢
⎢T
⎥
̃ 21 T̃ 22 ⋯ T ̃ 2n ⎥
maintenance strategy to each failure modes while it minimizes the total D=⎢ ⎢ ⋮
⎥ (4)
⋮ ⋱ ⋮ ⎥
maintenance cost. Table 1 represents a comparison between past liter­ ⎣
̃ ̃ ̃
⎦
T m1 T m2 ⋯ T mn
ature and present work in order to highlight the novelty and contribu­
tions of this paper. [ ]
T jv Tjv { }
̃ ∗ jv =
T , max , Tvmax = max Tjv (5)
2. Proposed failure mode analysis T max
v Tv 1≤j≤m

⎡ ⎤
Since RCM planning must be performed in industries to manage cost ̃ ∗ 11
T ̃ ∗ 12
T ⋯ ̃ ∗ 1n
T
and reliability, all steps must be applicable and possible to be performed ̃ ∗ 21 ∗
⎢T T ∗ 22 ⋯ ̃ 2n ⎥
T
D =⎢
∗ ⎥ (6)
in distribution system components. Failures are inevitable in system ⎣ ⋮ ⋮ ⋱ ⋮ ⎦
components; therefore, there must be a prioritization to specify which ̃ ∗ m1
T ̃ ∗ m2
T ⋯
∗
̃ mn
T
failure is more severe or less severe. Thus, it’s necessary to initially
implement FMEA to identify those failure causes which have a higher
risk. MCDM methods are then applied in this step to calculate the risk 2.4. Calculating weighted normalized matrixD∗
factor based on some major factors such as severity, detection, and
occurrence. In order to obtain weighted normalized matrix (̃ J jv ), it’s necessary to
∗

In the light of [10], all failure modes in a system are listed and three ̃ v ) and normalized
calculate the product of risk factor interval weights (w
criteria are defined to evaluate the failure modes. Then, BWM is applied matrix (T
̃ ∗ jv ). ̃
Jjv is described as [26]:
to attain the criteria weights based on experts’ opinion. Finally, GRA is [( ) ( )]
used to prioritize all failure modes. The mentioned FMEA method is ̃
J jv = ̃ ∗ jv ∗ w
T ̃v (7)
summarized as follows [11–13]:
The gray number multiplication in (7) can be implemented using
2.1. Making a list of possible failure modes formula (8) [26]:
[ ( )]
⎡ ⎤
A list of all possible failure modes, which occur in the system, needs min T jv ∗ wv , T jv ∗ wv , T jv ∗ wv , T jv ∗ wv ,
to be prepared. Fj = F1 , F2 , F3 , ......Fm is the set of all possible failure J jv = ⎣ [
̃ ( )] ⎦ (8)
modes. max T jv ∗ wv , T jv ∗ wv , T jv ∗ wv , T jv ∗ wv

2.2. Calculating the weights of risk factor elements [ ]

̃
J jv = J jv , J jv (9)
The goal of this step is to calculate the interval weights of severity,
detection and occurrence based on the BWM method. The mentioned The weighted normalized matrix D∗∗ can be displayed below [26]:
interval weights are the results of q experts’ opinions. This weight can be ⎡ ⎤
̃ ̃ ̃
represented as: w ̃ 1v , w
̃ 2v , w
̃ 3v , .........., w
̃ qv which are a set of interval number J
⎢ 11
⎢̃
J 12 ⋯ J 1n ⎥
J 2n ⎥
J 22 ⋯ ̃
J 21 ̃
andv = 1, 2, 3, ......, n. The average amount of interval weights based on D∗∗ = ⎢
⎢ ⋮
⎥ (10)
⎣ ⋮ ⋱ ⋮ ⎥ ⎦
experts’ opinions can be shown as [13]: ̃ ̃ ̃
J m1 J m2 ⋯ J mn
q ( )
1∑
w
̃v = ̃ kv
w (1) Among all possible failure modes,Fj = F1 ,F2 ,......,Fm , a referential set
q k=1
(Fmax ) can be obtained through (11) [26].
Interval risk factor weights consist of upper and lower bounds; [
⎡ max
] [ ]
⎤
therefore, (1) can be written as [13]: max max max
1≤j≤m J j1 , 1≤j≤m Jj1 , 1≤j≤m J j2 , 1≤j≤m Jj2
⎢ ⎥
[
q ( ) ∑ q ( )] F max = ⎣ [ ] ⎦ (11)
1∑ 1
(2)
max max
w
̃v = k
wv , k
wv , ........ 1≤j≤m J jn , 1≤j≤m Jjn
q k=1 q k=1

2.3. The evaluation of failure modes 2.5. Calculating risk factor index

Every failure mode is evaluated in this step with respect to the risk In order to calculate risk factor index, initially, a comparison be­
factor element. The failure mode j which is related to the risk factor tween referential set (Fmax ) and failure modes should be done to compute
element v and evaluated by expert q can be shown by interval number as possibility [27]. This possibility can be shown in (12):
Tjvq where j = 1, 2, 3, ......, m and v = 1, 2, 3, ......, n and q are the number of n [ ( )]
( ) 1∑
experts. Tjvq consists of upper and lower bounds and its final value, (12)
max
P Fj ≤ Fmax = P ̃J jv ≤ ̃
Jv
n v=1
presented here (3), is based on the average of all experts’ opinions
presented [11]. Eq. (12) can also be estimated and described in the form of (13) [28]:

3
M. Enjavimadar and M. Rastegar Electric Power Systems Research 203 (2022) 107647

Table 2 not be imposed during the maintenance on the load points of distribu­
Details of the first case study. tion system. The Reduction of Energy loss, high efficiency, and social
Hotline Coldline Maximum Total βthreshold Load type satisfaction are the advantages of the hotline maintenance strategy. The
safety safety time devoted maintenance high expense of maintenance implementation can be regarded as one of
cost cost to coldline budget the disadvantages of this strategy.
(Toman) (Toman) maintenance (Toman)
(minute)
3.2. Coldline maintenance strategy
30,000 20,000 240 30,000,000 0.254396 residential

In comparison to the hotline maintenance strategy, the coldline

⎡
( ( ))
⎤ strategy causes interruption during the implementation of the mainte­
( ) 1 ∑n max 0, L∗v − max 0, J jv − J max
j nance activity. Furthermore, this strategy suffers from the maintenance
P Fj ≤ F max
= ⎣ ⎦ (13) time limitation which can be accounted as another disadvantage of the
n v=1 L∗v
coldline strategy. Although this strategy incur high expenses, these are
usually lower than the expenses of the hotline maintenance strategy.
Where L∗jv demonstrates the sum of the length of interval number dis­
played in [28]: 3.3. Neutral maintenance strategy
( ) ( )
(14)
max
J jv + L ̃
L∗jv = L ̃ Jv In the maintenance strategy selection, the last alternative is to assign
the neutral maintenance strategy when other strategies are not efficient
Then, the risk factor index can be calculated through (15) [28]. This considering the cost and time. Therefore, the best strategy is to let the
Equation indicates that the higher the risk factor index, the higher the facility fail and repair or substitute it with a new one after a breakdown.
risk will be. This index can be used to prioritize possible failure modes. This strategy will result in an outage and more ENS. Consequently, it is
(
β = 1 − P Fj ≤ Fmax
)
(15) performed when there is no affordable and proper maintenance strategy.

The output of FMEA, the risk factors, is directly utilized in the se­ 3.4. Proposed optimization-based strategy selection
lection of the maintenance strategy.
3.4.1. Cost function
3. Proposed method to assign the optimal maintenance strategy In order to assign the optimal maintenance strategy, it is essential to
propose a method to select a proper strategy based on the defined
The final step in the implementation of RCM is to assign the main­
tenance strategy to all failure modes. Optimality can also be defined Table 4
from both the cost and reliability point of view. It is vitally important to Maintenance strategy planning in scenario 0 of the first case.
clarify what the maintenance strategies are and where they can be Strategy Failure mode
applied. Hence, the definition of maintenance strategies are here
Rhj {F2, F11, F13, F14, F15, F16, F17, F18, F19, F20, F21, F22, F23, F24, F25,
introduced in the first place:
F26, F27, F28, F29, F30, F31, F32, F33, F34, F35, F38, F39, F40, F41, F42,
F43, F44, F45, F47, F48, F49, F50, F51, F52, F53, F54, F55, F56, F57, F59,
F60, F62, F63, F66, F67 F68}
3.1. Hotline maintenance strategy Rcj {F1,F3, F4, F5, F6, F7, F8, F9, F10, F12, F36, F37, F46, F58, F61, F64,
F65, F69, F70}
Today, outage is an important issue in all distribution companies. nj {}
Hotline strategy is implemented without any outage and any ENS will

Table 3
Result of FMEA in the first case.
Failure mode β Priority Failure mode β Priority Failure mode β Priority

F17 0.42831 1 F43 0.29653 21 F11 0.22104 46

F4 0.42831 1 F42 0.29653 21 F10 0.22104 46
F69 0.39857 3 F41 0.27584 27 F9 0.22104 46
F68 0.39857 3 F40 0.27584 27 F8 0.22104 46
F67 0.38653 5 F39 0.27584 27 F7 0.22104 46
F66 0.38653 5 F37 0.27584 27 F6 0.19563 54
F60 0.38653 5 F36 0.27584 27 F5 0.19563 54
F56 0.38653 5 F35 0.25836 32 F48 0.19563 54
F26 0.37859 9 F34 0.25836 32 F3 0.19563 54
F49 0.37859 9 F33 0.25836 32 F2 0.18703 58
F22 0.37225 11 F32 0.25836 32 F1 0.18703 58
F21 0.37225 11 F29 0.25836 32 F14 0.16667 60
F20 0.37225 11 F28 0.25836 32 F65 0.094973 61
F70 0.37225 11 F27 0.25836 32 F64 0.094973 61
F63 0.37225 11 F50 0.24951 39 F61 0.094973 61
F62 0.32154 16 F25 0.24951 39 F58 0.094973 61
F59 0.32154 16 F24 0.24951 39 F51 0.094973 61
F57 0.32154 16 F23 0.24951 39 F46 0.020361 66
F55 0.32154 16 F19 0.24951 39 F38 0.020361 66
F54 0.32154 16 F18 0.24951 39 F31 0.020361 66
F53 0.29653 21 F47 0.24951 39 F30 0.020361 66
F52 0.29653 21 F16 0.22104 46 F15 0.020361 66
F45 0.29653 21 F13 0.22104 46
F44 0.29653 21 F12 0.22104 46

4
M. Enjavimadar and M. Rastegar Electric Power Systems Research 203 (2022) 107647

Fig. 2. Strategies applied to the feeder studied in scenario 1 of the first case.

Table 6
Fig. 1. Strategies applied to the failures in the business as the usual scenario Optimal maintenance strategy in scenario 2 of the first case.
(Scenario 0) of the first case study.
Strategy Failure mode

Rhj {F27, F28, F29, F31, F32, F33, F34, F35, F36, F37, F38, F39, F40, F41,
Table 5 F42, F43, F44, F45, F46}
Optimal maintenance strategy in scenario 1 of the first case. Rcj {F1, F2, F3, F4, F5, F6, F7, F8, F9, F10, F11, F12, F13, F14, F15, F16, F17,
F18, F19, F20, F21, F22, F23, F24, F25, F26, F30, F47, F48, F49, F51, F52,
Strategy Failure mode F53, F54, F55, F56, F57, F58, F59, F60, F61, F62, F63, F65, F66, F67, F68,
Rhj {F1, F2, F3, F4, F5, F6, F7, F8, F9, F11, F12, F13, F14, F16, F17, F18, F19, F69, F70}
F20, F21, F22, F23, F24, F26, F27, F28, F29, F31, F32, F33, F34, F35, F36, nj {F50, F64}
F37, F38, F39, F40, F41, F42, F43, F44, F45, F46}
Rcj {F15, F30, F47, F48, F49, F51, F52, F53, F54, F55, F56, F57, F58, F59,
F60, F61, F62, F63, F65, F66, F67, F68, F69, F70} ∑∑ [( ) ( ) ( )]
nj {F10, F25, F50, F64} IC = c
VoLLi × ENSi,j × Rcj + ENSi,j
n h
× nj + ENSi,j × Rhj
i j

(18)
criteria. To satisfy this need, a numbers of equations from (16) to (26)
( ) ( )
are presented. (17) describes objective functions that consider the ∑
OC = Rcj × ccj + csc + Rhj × chj + csh (19)
interruption cost and the total operation cost simultaneously. The
j j
j
interruption cost, which is presented in (18), is the multiplication of the
value of lost load at the load point i (VoLLi ) and ENS at this load point. 3.4.2. Constraints
The coldline, hotline and neutral maintenance strategies, respectively, A constraint on the duration time of the coldline maintenance
lead to ENSci,j , ENShi,j and ENSni,j for the year ahead. Based on reference strategy in a distribution system is defined in (20). Risk factor shown in
[29], if one of the proposed strategies lead to the replacement of (21) also presents a constraint which leads to avoid applying the neutral
equipment, the failure mode will be reduced enormously. But if strate­ maintenance strategy to the components of distribution system. βthreshold
gies lead to minor repair, the failure rate will not change considerably. is the threshold of risk factor which is the average of all failure modes of
According to (16), the ENS in the load point irelies on the failure rate, risk factor. In the light of (22), one of the mentioned strategy should be
the outage rate, and the amount of load. Based on the cost of all main­ applied to each failure mode. Based on the hotline and coldline main­
tenance strategies and the safety cost, (19) calculates the total operation tenance costs related to each failure mode, (23) argue the constraints on
cost. the total maintenance budget. Following (24), if the failure mode s in the
maneuver section is assigned a coldline strategy, all failure modes will
ENSi = Li × λi × ri (16)
be assigned a coldline strategy in the maneuver section. Maneuver
Min OF = IC + OC (17) section is a part of the distribution network which is located between
two maneuver switches. We have linearized Eq. (24), and the linear form
of this constraint can be shown in (25). According to Eq. (25), P is a
variable and P1 is a positive variable. B1 and B2 are binary variables.
According to (25-a), if any of the failure modes in maneuver section is

5
M. Enjavimadar and M. Rastegar Electric Power Systems Research 203 (2022) 107647

Fig. 4. Strategies applied to the feeder studied in scenario 3 of the first case.

Fig. 3. Strategies applied to the feeder studied in scenario 2 of the first case.

Table 8
Optimal maintenance strategy in scenario 4 of the first case.
Table 7
Optimal maintenance strategy in scenario 3 of the first case. Strategy Failure mode

Strategy Failure mode Rhj {F1, F2, F3, F4, F6, F7, F8, F9, F11, F12, F13, F14, F17, F19, F21, F22,
F23, F24, F27, F28, F29, F32, F33, F34, F35, F36, F37, F38, F39, F41, F42,
Rhj {F1, F2, F3, F4, F5, F6, F7, F8, F9, F11, F13, F14, F16, F17, F18, F19, F20, F44, F45, F46}
F21} Rcj {F15, F18, F20, F30, F43, F47, F48, F49, F51, F52, F54, F56, F57, F58,
Rcj {F12, F15, F22, F23, F24, F26, F27, F28, F29, F30, F31, F32, F33, F34, F59, F61, F62, F63, F65, F66, F67, F68, F69, F70}
F35, F36, F37, F38, F39, F40, F41, F42, F43, F44, F45, F46, F47, F48, F49, nj {F5, F10, F16, F25, F26, F31, F40, F50, F53, F55, F60, F64}
F51, F52, F53, F54, F55, F56, F57, F58, F59, F60, F61, F62, F63, F65, F66,
F67, F68, F69, F70}
nj {F10, F25, F50, F64}
{
1 if Rcs = 1 ∀ s ∈ T
RcT = (24)
0 if Rcs = 0 ∀ s ∈ T
assigned the coldline maintenance strategy, for example 3 failure modes,
the variable P has a positive amount. Hence, based on (25-b), P1 has ⎧ ( )
∑
positive non-zero amount and consequently based on (25-c) B2 should ⎪
⎪ P = R c
− 1 (25 − a)
⎪
⎪
be zero. Eq. (25-d) depicts that only one of binary variable must have
s
⎪
⎪ s
⎪
⎪
value, so B2 is zero and B1 is one. As a result, Eq. (25-e) illuminates that ⎪
⎨P − B = P (25 − b)
(25)
1 2
all failure modes should be assigned the coldline strategy. In contrast, ⎪
⎪ P1 ≤ M × B1 (25 − c)
based on (25-a), if no failure mode are assigned the coldline strategy, P ⎪
⎪
⎪
⎪
will have a negative variable. Consequently, according to (25-b), B2 has ⎪
⎪
⎪ B + B2 = 1
⎩ 1
(25 − d)
non-zero amount. So, based on Eq. (25-d), B2 is one and B1 is zero. RcT = B1 (25 − e)
Finally, according to (25-e), since B1 is zero, no failure modes in ma­
∑
neuver section are assigned the coldline maintenance strategy. (26) Rhj ≤ N h (26)
proposes a constraint on the number of hotline repair team in the dis­ j

tribution network.
A mathematical mixed integer programming is used in order to solve
∑
rjc × Rcj ≤ T c (20) the optimization problem and obtain the optimal maintenance strategy.
j The whole optimization is implemented in GAMS, and Cplex solver is
used to solve the optimization problem.
nj = 0 ∀ βj ≥ βthreshold (21)
4. Numerical study
Rcj + Rhj + nj = 1 (22)
∑ 4.1. The first case study
Rcj × ccj + Rhj × chj ≤ M (23)
j In order to test the proposed method, this study is implemented in a

6
M. Enjavimadar and M. Rastegar Electric Power Systems Research 203 (2022) 107647

Red, blue, and yellow indicate hotline, coldline, and neutral mainte­
nance strategies, respectively. As shown in Table 4, 51 and 19 failure
modes are, respectively, assigned the hotline and coldline maintenance
strategies in scenario 0.
In addition, four scenarios are designed in this study to prove the
efficiency of the proposed optimization-based method in comparison to
the business as the usual operation of a distribution system and to
analysis various factors on the strategy assignments, as follows:

• Scenario 1: All constraint except for (24) are considered in this

scenario.

Table 11
Maintenance strategy planning in scenario 1 in the second case study.
Strategy Failure mode

Rhj {F1, F2, F3, F4, F5, F6, F7, F8, F9, F11, F12, F13, F14, F15, F16, F17, F18,
F19, F20, F21, F22, F23, F24, F26, F27, F28, F29, F30, F31, F32, F33, F34,
F35, F36, F37, F38, F39, F40, F41, F42, F43, F44, F45, F46, F47, F48, F49,
F51, F52, F53, F54, F55, F56, F57, F58, F59, F60, F61, F62, F63, F65, F66,
F67, F68, F69, F70, F71, F72, F73, F74, F75, F76, F77, F78, F79, F80, F81,
F82, F83, F84, F85, F86, F87, F88, F89, F90, F91, F92, F93, F94, F95, F96,
F97, F98, F99, F100, F101, F102, F103, F104, F105}
Rcj {F106, F107, F108, F109, F110, F111, F112, F113, F114, F115, F116,
F117, F118, F119, F120, F121, F122, F123, F124, F125, F126, F127,
F128, F129}
nj {F10, F25, F50, F64, F130, F131, F132, F133, F134, F135, F136, F137,
F138, F139, F140, F141, F142, F143, F144, F145, F146, F147, F148,
F149, F150, F151, F152, F153, F154, F155, F156, F157, F158, F159,
Fig. 5. Strategies applied to the feeder studied in scenario 4 of the first case. F160, F161, F162, F163, F164, F165, F166, F167, F168, F169}

Table 9 Table 12
Costs in different scenarios of the first case study. Maintenance strategy planning in scenario 0 in the second case study.

costscenario maintenance outage cost Safety cost Total cost Strategy Failure mode
cost (Toman/ (Toman/ (Toman/ (Toman/ {F1, F2, F3, F4, F5, F6, F7, F8, F9, F10, F11, F12, F13, F14, F15, F16, F17,
Rhj
year) year) year) year) F18, F19, F20, F21, F22, F23, F24, F25, F26, F27, F28, F29, F30, F31, F32,
0 21,700,665 7,510,650 2,780,000 31,991,315 F33, F34, F35, F36, F37, F38, F39, F40, F41, F42, F43, F44, F45, F46, F47,
1 19,642,585 7,712,700 2,440,000 29,795,285 F48, F49, F50, F51, F52, F53, F54, F55, F56, F57, F58, F59, F60, F61, F62,
2 15,844,375 7,611,600 2,190,000 25,645,975 F63, F64, F65, F66, F67, F68, F69, F70, F71, F72, F73, F74, F75, F76, F77,
3 16,393,130 7,712,700 2,140,000 26,245,830 F78, F79, F80, F81, F82, F83, F84, F85, F86, F87, F88, F89, F90, F91, F92,
4 15,575,505 8,116,500 2,050,000 25,742,005 F93, F94, F95, F96, F97, F98, F99, F100, F101, F102, F103, F104, F105,
F106, F107, F108, F109, F110, F111, F112, F113, F114, F115, F116,
F117, F118, F119, F120, F121, F122, F123, F124, F125, F126, F127,
F128, F129, F130, F131, F132, F133, F134, F135, F136, F137, F138,
real electrical distribution feeder containing 62 load points, 463 over­
F139, F140, F141, F142, F143, F144, F145, F146, F147, F148, F149,
head lines, 19 switches and 41 cables, as the first case study. Table 2 F150, F151, F152, F153, F154, F155, F156, F157, F158, F159, F160,
displays a summary of the details of the first case study. The value of the F161, F162, F163, F164, F165, F166, F167, F168, F169}
lost load in residential load points is considered 0.1 $/MWh, and it is 1.5 Rcj {}
$/MWh in others [30]. 70 failure modes such as the insulator failure nj {}
mode, the jumper failure mode, the cutout fuse failure mode, the switch
failure mode, the transformer failure mode and the cable failure mode
are considered in this study. Failure modes are excluded from the in­ Table 13
spection of a sample feeder in the real distribution system. All failure Costs in different scenarios in the second case study.
modes have distinct latitude and longitude. The risk factor of these 70
CostScenario maintenance outage cost Safety cost Total cost
failure modes are obtained as shown in Table 3. Based on this result, F4 cost (Toman/ (Toman/year) (Toman/ (Toman/
and F17 have the highest priority from the point of risk, and βthreshold is year) year) year)
0.254396.
0 173,636,485 204,726,600 5,070,000 383,433,085
The business as the usual scenario is called Scenario 0, and strategies 1 102,703,610 229,305,000 3,510,000 335,518,610
applied to failures are concluded in Table 4 and are displayed in Fig. 1.

Table 10
Details of the second case study.
Hotline safety cost Coldline safety cost Maximum time devoted to coldline maintenance Total maintenance budget βthreshold Load type
(Toman) (Toman) (minute) (Toman)

30,000 20,000 240 200,000,000 0.35541 -residential

-agricultural
-public

7
M. Enjavimadar and M. Rastegar Electric Power Systems Research 203 (2022) 107647

• Scenario 2: It considers the practical constraint in (24) in addition to the total cost is justified.
scenario 1. While the outage cost constitute 68.3% of the total maintenance cost
• Scenario 3: In this scenario, the duration time of the coldline main­ for the second case study, this is 25.8% in the first case study. The reason
tenance strategy is considered twice of that in scenario 1. is that, in the second feeder, there is a variety of consumer types con­
• Scenario 4: The hotline safety cost is considered twice of that in taining residential, agricultural and public consumers, which have the
scenario 1. Hence, the safety of operators is highly important, and higher value of the lost load. In addition, the neutral strategies in feeder
any possible operators’ injuries can lead to more costs. 2, in the optimal method, constitutes 26% of the total strategies, while
only 6% of the failure modes were assigned neutral strategy in the first
The results of the optimization in scenario 1 are displayed in Table 5. feeder. Feeder 2 has a higher number of failure modes compared to
In this scenario, the hotline and coldline maintenance strategies are feeder 1, and this leads to a higher total and safety costs in feeder 2 in
assigned to 54 and 12 failure modes, respectively. The neutral strategy is comparison with feeder 1.
assigned to 4 failure modes. Fig. 2 shows the strategies applied to feeder
failures in scenario 1. 5. Conclusion
Scenario 2 considers all constraints including (24). Based on this
constraint, all failure modes which occur in the maneuver section will be In this paper, a comprehensive and practical structure for the
assigned the coldline maintenance strategy, if one of the strategy in this reliability-centered maintenance is implemented. An optimization
section is coldline. 26 failures are identified in the maneuver area. problem is proposed to select a maintenance strategy based on the ENS,
Hotline, coldline, and neutral strategies are, respectively, assigned to 37, the maintenance cost, the outage cost, the safety cost, and the risk factor
29 and 4 failure modes, which are illustrated in Table 6. Fig. 3 shows the of failure modes of different components in the distribution network.
strategies applied to feeder failures in scenario 2. The GRA method is used to prioritize failure modes in the distribution
The results of scenario 3 shown in Table 7, illustrate that 18 failure network; furthermore, the risk factor of these failure modes has also
modes are assigned the hotline strategy and 48 failure modes are been calculated. Optimal maintenance strategies for each failure mode
assigned the coldline strategy. 4 failure modes are also assigned the are obtained separately. The maintenance cost, the outage cost and the
neutral strategy. Fig. 4 shows the strategies applied to feeder failures in safety cost of different scenarios are presented and compared. Two case
scenario 3. studies are done to implement the proposed method. In the first case
As presented in Table 8, in scenario 4, 34, 26 and 12 failure modes study, four scenarios are considered to compare the results of applying
are, respectively, assigned hotline, coldline and neutral maintenance the method in the current study with those of the business as the usual
strategies. Fig. 5 shows the strategies applied to feeder failures in sce­ strategy. The results calculated in different scenarios show about a 7%
nario 4. reduction in the total annual cost of the maintenance if the proposed
Table 9 concludes the cost of different scenarios in the electrical method is applied. When the maneuver section in the proposed method
distribution feeder. First row displays the result of scenario 0. Based on is taken into account, the results show the lowest total annual mainte­
the results in scenario1, the total maintenance cost was reduced 7% nance cost; however, the business as the usual strategy has the highest
compared to that in scenario 0. This occurred since failures 10, 25, 50 total annual maintenance cost. In the second case study, the proposed
and 64 were assigned neutral strategy while failures 10 and 25 and method is compared to the business as the usual scenario. Results show
failures 50 and 64 were, respectively, assigned the coldline and hotline the reduction of annual maintenance cost and the safety cost. However,
strategies in scenario 0. Furthermore, the number of hotline strategy was there is an increase in the outage cost due to the increase of neutral
reduced in this scenario. Compared with that in scenario 0, the total maintenance strategy. To sum up, this paper suggests that the proposed
maintenance cost in scenario 2 also experienced 20% reduction for two optimization-based RCM can decrease the maintenance and outage costs
reasons: first, failures 50 and 64 were assigned neutral strategy and, in comparison with those of the business as the usual situation. More­
second, the number of coldline strategies were increased more than over, failure mode analysis can give a good insight to operators to
scenario 0 because a large number of failures were located in the ma­ choose proper strategies. However, the application of this method re­
neuver area. In comparison with scenario 0, there was a reduction of quires technical and operational details of a distribution system, which
18% and 20% in the total maintenance cost for scenarios 3 and 4, may dynamically change. Hence, the proposed method and procedure in
respectively. this paper should be embedded in a maintenance software for proper
applications in reality in the future. In addition, implementation of RCM
4.2. The second case study in electrical transmission system and consideration of protective con­
straints in the proposed method can be considered as future works.
Another case study is considered here to show the effectiveness of the
proposed structure for the maintenance strategy selection in various CRediT authorship contribution statement
feeders. Table 10 displays a summary of the details of the second case
study. This feeder has 169 failures in different latitudes and longitudes, MohammadHosein Enjavimadar: Conceptualization, Software,
and 37 distribution transformers. This feeder contains residential, Formal analysis, Writing – original draft. Mohammad Rastegar: Su­
agricultural and public loads. Similar to the first case study, scenario pervision, Methodology, Validation, Writing – review & editing.
0 represents the maintenance strategy in accordance with the business as
the usual strategy in the distribution company, and scenario 1 represents Declaration of Competing Interest
the output of applying the proposed optimization model. In the second
case study, the results of the implementation of the proposed mainte­ The authors declare that they have no known competing financial
nance structure for feeder 2 in scenarios 1 and 0 are given in Table 11 interests or personal relationships that could have appeared to influence
and Table 12, respectively. The total cost of repairs allocated to this the work reported in this paper.
feeder is 173 million Toman. The results of studies in feeder 2 show a
12.5% reduction in the cost of scenario 1 compared to that in scenario 0. Reference
Table 13 illustrate the cost of scenarios 0 and 1 in the second case study.
According to the results in Tables 11, the neutral and coldline stra­ [1] G. Crognier, P. Tournebise, M. Ruiz, P. Panciatici, Grid operation-based outage
tegies have been applied to 26% and 14.2% of the failure modes, maintenance planning, Electr. Power Syst. Res. 190 (2021), 106682.
[2] Y.S. Duarte, J. Szpytko, A.M. del Castillo Serpa, Monte Carlo simulation model to
respectively. However, in scenario 0, a hotline strategy has been applied coordinate the preventive maintenance scheduling of generating units in isolated
to all the failure modes of this feeder; therefor, a reduction of 12.5% in distributed Power Systems, Electr. Power Syst. Res. 182 (2020), 106237.

8
M. Enjavimadar and M. Rastegar Electric Power Systems Research 203 (2022) 107647

[3] X. Ge, Q. Chen, Y. Fu, C. Chung, Y. Mi, Optimization of maintenance scheduling for [17] H. Alayo, E. Paucar, A MILP model for maintenance scheduling in transmission
offshore wind turbines considering the wake effect of arbitrary wind direction, systems and an example application to the peruvian system, IEEE Lat. Am. Trans.
Electr. Power Syst. Res. 184 (2020), 106298. 16 (4) (2018) 1099–1104.
[4] M. Catelani, L. Ciani, D. Galar, G. Patrizi, Optimizing maintenance policies for a [18] M. Khodaei Tehrani, A. Fereidunian, H. Lesani, Financial planning for the
yaw system using reliability-centered maintenance and data-driven condition preventive maintenance of power distribution systems via fuzzy AHP, Complexity
monitoring, IEEE Trans. Instrum. Meas. 69 (9) (2020) 6241–6249. 21 (3) (2016) 36–46.
[5] R. Ahmadi, S. Wu, A. Sobhani, Reliability modeling and maintenance optimization [19] A.E. Abu-Elanien, M. Salama, Asset management techniques for transformers,
for a repairable system with heterogeneity population, IEEE Trans. Reliab. (2021). Electr. Power Syst. Res. 80 (4) (2010) 456–464.
[6] R. Ghorani, M. Fotuhi-Firuzabad, P. Dehghanian, W. Li, Identifying critical [20] H.A. Koksal, A. Ozdemir, RCAM model for Turkish national power transmission
components for reliability centred maintenance management of deregulated power system: SF6 circuit breakers, transmission lines, transformer centers and protection
systems, IET Gener. Transm. Distrib. 9 (9) (2015) 828–837. relays. 2014 International Conference on Probabilistic Methods Applied to Power
[7] P. Dehghanian, "A reliability-oriented outlook on the critical components of power Systems (PMAPS), IEEE, 2014, pp. 1–6.
distribution systems," 2012. [21] P. Dehghanian, M. Fotuhi-Firuzabad, F. Aminifar, R. Billinton, A comprehensive
[8] F. Pourahmadi, M. Fotuhi-Firuzabad, P. Dehghanian, Identification of critical scheme for reliability centered maintenance in power distribution systems—Part I:
generating units for maintenance: a game theory approach, IET Gener. Transm. methodology, IEEE Trans. Power Delivery 28 (2) (2013) 761–770.
Distrib. 10 (12) (2016) 2942–2952. [22] P. Dehghanian, M. Fotuhi-Firuzabad, F. Aminifar, R. Billinton, A comprehensive
[9] S. Bahrami, M. Rastegar, P. Dehghanian, An fbwm-topsis approach to identify scheme for reliability-centered maintenance in power distribution systems—Part
critical feeders for reliability centered maintenance in power distribution systems, II: numerical analysis, IEEE Trans. Power Delivery 28 (2) (2013) 771–778.
IEEE Syst. J. (2020). [23] M. Mirhosseini, F. Keynia, Asset management and maintenance programming for
[10] H.-.W. Lo, J.J. Liou, A novel multiple-criteria decision-making-based FMEA model power distribution systems: a review, IET Gener. Transm. Distrib. (2021).
for risk assessment, Appl. Soft Comput. 73 (2018) 684–696. [24] V. Aravinthan, W. Jewell, Optimized maintenance scheduling for budget-
[11] M.M. Shahri, A.E. Jahromi, M. Houshmand, Failure mode and effect analysis using constrained distribution utility, IEEE Trans. Smart Grid 4 (4) (2013) 2328–2338.
an integrated approach of clustering and MCDM under pythagorean fuzzy [25] H.A. Ravaghi Ardabili, M.R. Haghifam, S.M. Abedi, A probabilistic reliability-
environment, J. Loss Prev. Process Ind. 72 (2021), 104591. centred maintenance approach for electrical distribution networks, IET Gener.
[12] S. Boral, S.K. Chaturvedi, I. Howard, V. Naikan, K. McKee, An integrated interval Transm. Distrib. 15 (6) (2021) 1070–1080.
type-2 fuzzy sets and multiplicative half quadratic programming-based MCDM [26] R. Rajesh, V. Ravi, Supplier selection in resilient supply chains: a grey relational
framework for calculating aggregated risk ranking results of failure modes in analysis approach, J. Clean. Prod. 86 (2015) 343–359.
FMECA, Process Saf. Environ. Prot. 150 (2021) 194–222. [27] A. Mohammed, B. Naghshineh, V. Spiegler, H. Carvalho, Conceptualising a supply
[13] M. Yazdi, Improving failure mode and effect analysis (FMEA) with consideration of and demand resilience methodology: a hybrid DEMATEL-TOPSIS-possibilistic
uncertainty handling as an interactive approach, Int. J. Interact. 13 (2) (2019) multi-objective optimization approach, Comput. Ind. Eng. (2021), 107589.
441–458. [28] A. Jääskeläinen, The relational outcomes of performance management in buyer-
[14] M. Rafiei, M.-.H. Khooban, M.A. Igder, J. Boudjadar, A novel approach to supplier relationships, Int. J. Prod. Econ. 232 (2021), 107933.
overcome the limitations of reliability centered maintenance implementation on [29] S. Peng, W. Jiang, W. Zhao, A preventive maintenance policy with usage-
the smart grid distance protection system, IEEE Trans. Circuits Syst. Express Briefs dependent failure rate thresholds under two-dimensional warranties, IISE
67 (2) (2019) 320–324. Transactions (2020) 1–13.
[15] J. Heo, M. Kim, J. Lyu, Implementation of reliability-centered maintenance for [30] F.H. Moghimi, T. Barforoushi, S.M. Mousavi, Optimal scheduling of resources for a
transmission components using particle swarm optimization, Int. J. Electr. Power price-maker distribution company in electricity markets considering network
Energy Syst. 55 (2014) 238–245. component failures, Int. J. Electr. Power Energy Syst. 122 (2020), 106200.
[16] E. Shayesteh, J. Yu, P. Hilber, Maintenance optimization of power systems with
renewable energy sources integrated, Energy 149 (2018) 577–586.