Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

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/TPWRS.2020.2981508, IEEE
Transactions on Power Systems
1

Optimization Model-based Reliability Assessment


for Distribution Networks Considering Detailed
Placement of Circuit Breakers and Switches
Zihao Li, Wenchuan Wu, Senior Member, IEEE, Xue Tai, Boming Zhang, Fellow, IEEE

 LN Set of load nodes.


Abstract—Requirements of economic and reliability criteria  N
Set of system nodes.
are often stressed simultaneously in operating and planning  SS Set of substation nodes.
models for distribution systems. However, conventional reliability Parameters
evaluation algorithms are hard to be integrated to these
optimization models analytically. Recently, several optimization C CB Unit cost of circuit breaker.
model-based reliability assessment methods have been proposed, C SW Unit cost of switch.
which can be embedded into such models. Nevertheless, the
detailed placement and actions of circuit breakers and switches Li Peak demand at node i .
are ignored or oversimplified in these methods, leading to an M A sufficiently large positive number.
inconsistency with the real world situation. Thus, we propose a NCi Number of customers at node i .
new optimization model-based reliability assessment method that
fully considers detailed placement and actions of circuit breakers SijC Maximum transmission power of branch ij .
and switches in distribution networks. Based on fictitious fault
S Cf Capacity of the transformer connected to feeder f .
flows, strategies for tripping circuit breakers to cut fault current,
operating switches to isolate the fault and processing post-fault Active demand at node i .
Pi
network reconfiguration are linearly modeled as constraints.
Therefore, this method can be easily integrated into operating and Qi Reactive demand at node i .
planning models of distribution networks. Case studies on 54-node, s
V Source voltage at the head-end of a feeder.
85-node, 137-node and 417-node distribution networks show the
scalability and efficiency of the proposed model. h Duration of load level h .
biji ,NO Equal to 1 when circuit breaker placed at end i of
Index Terms—Reliability assessment, distribution network, branch ij is closed under normal operating
circuit breaker, switch.
condition.
NOMENCLATURE bijj ,NO Equal to 1 when circuit breaker placed at end j of
branch ij is closed under normal operating
Indices
h Index for the load levels used to represent the condition.
loading condition. piSC The probability of scenario i.
i, j Indices for nodes. Equal to 1 when switch placed at end i of branch
siji ,NO
Sets ij is closed under normal operating condition.
H Set for the load levels used to represent the Equal to 1 when switch placed at end j of branch
loading condition.
sijj ,NO
ij is closed under normal operating condition.
 SC Set for scenarios.
 Set of branches. rij , xij Resistance and inductance of branch ij .
 B
I
Set of branches equipped with circuit breaker at  iCIF CIF requirement of node i .
the left end.
 BJ Set of branches equipped with circuit breaker at  iCID CID requirement of node i .
the right end.  EENS EENS requirement of the distribution network.
 SI Set of branches equipped with switch at the left
end.  SAIDI SAIDI requirement of the distribution network.
 SJ Set of branches equipped with switch at the right  SAIFI SAIFI requirement of the distribution network.
end.
xy Outage rate of the branch xy .
i Set of nodes connected to node i .
h Loading factor of load level h .
F Set of feeders.

This work was supported in part by the National Science Foundation of Power Systems, Department of Electrical Engineering, Tsinghua University,
China under Grant 51725703. Beijing 100084, China. X. Tai is with the State Grid Shanghai Municipal
Z. Li(18811363920@163.com), W. Wu(Corresponding author, e-mail: Electric Power Company, Shanghai, 200233, China.
wuwench@tsinghua.edu.cn and B. Zhang are with the State Key Laboratory of

0885-8950 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on June 02,2020 at 03:49:40 UTC from IEEE Xplore. Restrictions apply.
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/TPWRS.2020.2981508, IEEE
Transactions on Power Systems
2

 xyRP Duration of the repair-and-switching interruptions reconfiguration due to a fault at branch xy (or under
associated with the failure of the branch xy . normal operation condition if xy = NO ).
 xy
SW Duration of the switching-only interruptions that Qijxy Reactive power flow through branch ij (from node i
occur when the branch xy fails. to node j) after post-fault network reconfiguration
Binary Variables due to a fault at branch xy (or under normal
biji , xy Equal to 1 when circuit breaker placed at end i of operation condition if xy = NO ).
branch ij is closed before post-fault network Square voltage at node i after post-fault network
U ixy
reconfiguration due to a fault at branch xy . reconfiguration due to a fault at branch xy (or under
Equal to 1 when circuit breaker placed at end j of normal operation condition if xy = NO ).
j , xy
bij
branch ij is closed before post-fault network Vi xy Voltage at node i after post-fault network
reconfiguration due to a fault at branch xy . reconfiguration due to a fault at branch xy (or under
i , xy
sij Equal to 1 when switch placed at end i of branch normal operation condition if xy = NO ).
ij is closed after post-fault network fi xy ,RA fi xy ,RA  [0,1] and equals to 0 when node i loses
reconfiguration due to a fault at branch xy . power supply right after (RA) the action of circuit
breaker due to a fault at branch xy .
sijj , xyEqual to 1 when switch placed at end j of branch
ij is closed after post-fault network fijxy ,RA Fijxy ,RA  [0,1] and equals to 0 when branch ij loses
reconfiguration due to a fault at branch xy . power supply right after the action of circuit breaker
due to the fault at branch xy .
pixy Equal to 1 when node i is affected by the outage
due to a fault at branch xy . fi xy ,PF fi xy ,PF  [0,1] and equals to 0 when node i is
xy Equal to 1 when the demand of node i is isolated after post-fault (PF) network reconfiguration
qi
completely recovered after post-fault network due to a fault at branch xy .
reconfiguration of the network following a fault at f ijxy ,PF fijxy ,PF  [0,1] and equals to 0 when branch ij is
branch xy .
isolated after post-fault network reconfiguration due
i ,CB
xij Equal to 1 when a circuit breaker is placed at end to a fault at branch xy .
i of branch ij .
j ,CB
xij Equal to 1 when a circuit breaker is placed at end I. INTRODUCTION
j of branch ij .
xiji ,SW Equal to 1 when a switch is placed at end i of
branch ij .
R eliability is defined as the ability of a power system to meet
the power demand of end users continuously, in the light
of both quantity and quality [1]. Typically, the reliability of a
j , SW Equal to 1 when a switch is placed at end j of
xij distribution system can be systematically characterized by
branch ij . several indices including system average interruption duration
NO
lij Equal to 1 when the branch ij is connected under index (SAIDI), system average interruption frequency index
normal operation condition. (SAIFI), and expected energy not supplied (EENS) [2]–[7].
lijxy Equal to 1 when the branch ij is connected after These indices are generally taken into account in operational
and planning problems, and are counted conventionally using
post-fault network reconfiguration due to a fault at
analytical methods [1], [2], [6], [8]–[15] and simulation-based
branch xy .
methods [16]–[20]. However, these methods face some
Continuous Variables unavoidable problems in applications [24], especially for the
ASAI Average service availability index.
integration in the operating and planning models of distribution
CIDi Customer interruption duration at node i . networks.
CIFi Customer interruption frequency at node i . The analytical methods usually derive the fault-impact-
EENS Expected Energy not supplied. ranges (FIRs) by topology analysis [1], [2], [6], [8], or by
SAIDI System average interruption duration index. formulating fault incidence matrix based on the topology
SAIFI System average interruption frequency index. information of the distribution networks [13]–[14]. For
Pi xy Active demand at node i after post-fault network example, in [6], based on the results of topology analysis,
reconfiguration due to a fault at branch xy (or under algebraic equations for reliability indices can be formulated
normal operation condition if xy = NO ). based on the failure rates and the repairing interruptions of
equipment. In [13], grouping buses are firstly merged based on
Pijxy Active power flow through branch ij (from node i to the placement of circuit breakers and switches and three types
node j) after post-fault network reconfiguration due
of fault incidence matrices are constructed considering the
to a fault at branch xy (or under normal operation
influence type of each fault event on the load points. In [14],
condition if xy = NO ). based on the upstream and downstream relationship, two
Qixy Reactive demand at node i after post-fault network matrices are defined to describe the interconnection

0885-8950 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on June 02,2020 at 03:49:40 UTC from IEEE Xplore. Restrictions apply.
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/TPWRS.2020.2981508, IEEE
Transactions on Power Systems
3

relationships between components and circuit breakers and the [26], it is also assumed that only one circuit breaker is placed
relationships inside the components. The reliability indices then for each feeder and two switches are equipped at both ends of
can be linearly expressed by these matrices. In [15], based on each branch. Actually, in real distribution networks, it is hard
the topology of the distribution network, linear algebraic to guarantee that each branch is equipped with switches because
equations are built and the reliability indices can be quickly of budget limitation. And obviously more than one circuit
calculated. Obviously, these topology-depend methods need the breakers can be placed at large feeders to improve the reliability.
network’s information as a priori, which are evidently not All these cases lead to an inconsistency between the models and
suitable for the applications of distribution network planning. the actual situation.
Besides, for the methods proposed in [6], [13] and [14], the A practical reliability assessment method need fully consider
formulas or the matrices may be frequently changed and the placement and actions of circuit breakers and switches in
reformulated as the network slightly adjusts. Moreover, in [13]– distribution networks [27]–[33]. For example, circuit breakers
[15], circuit breakers and switches are modeled without are usually installed in the middle of a feeder to divide it into 2-
considering their contribution to load restoration through 4 sections in order to improve reliability of distribution
optimally operating the switchable equipment, which is an networks [34]–[36]. On this basis, accurate relay protection can
important method to improve reliability in real-world practice. locate fault branch and trip the appropriate circuit breaker to
Thus, a network reconfiguration optimization model needs to minimize the affected loads [27]. Besides, considering the cost
be formulated, which cannot be simply expressed by algebraic of investment, the placement of switches is earnestly designed
formulation. [37]–[40]. Existing model-based reliability assessment methods
On the other side, simulation-based methods still face cannot properly handle these situations.
computing problems, though the placement and actions of This work proposes a new optimization model-based
circuit breakers and switches can be taken into consideration in reliability assessment method for distribution networks
detail [19]. In these methods, numerous Monte Carlo samples considering the detailed placement of circuit breakers and
are generated to capture all outage scenarios (typically one switches. The strategies for tripping circuit breakers, operating
sample one outage scenario) [16]–[17]. During each outage switches and processing post-fault network reconfiguration are
scenario, the fault power flow is firstly calculated to recognize expressed as linear constraints. Fictitious fault flows stemmed
affected demands in the distribution network, and appropriate from faulty branches are appropriately modeled to be tripped by
actions of circuit breakers and switches are generated, then circuit breaker and isolated by switches. Under this
post-fault network reconfiguration is processed [18]–[20]. circumstance, the placement of circuit breakers and switches
These methods always work but are time-consuming (refer to are explicitly described as an input of the model. The model is
Section IV). Besides, simulation-based reliability assessment then cast as an instance of mixed integer linear programming
methods are unable to be directly integrated into the operating and can be effectively solved by off-the-shelf software.
and planning models. For example, in order to satisfy the The main contributions of this paper are listed as follow:
reliability requirements of the planned distribution networks, 1) The proposed model firstly uses fictitious fault flows for
planners often adopt a posterior simulation step to validate the both situations where a fault just happens and the post-
reliability for alternative solutions [21] or utilize metaheuristics fault network reconfiguration processes, and the detailed
[22]–[23]. placement and actions of circuit breakers and switches
Motivated by the lack of exact methods for reliability- can be considered. It is a new model (not a slight revision)
constrained distribution optimization problems, recently, compared to [24]–[26] since the existing models cannot
several optimization model-based reliability assessment be extended to describe the detailed placement of the
methods have been proposed to improve the equipment switchable equipment. The proposed model is more
flexibility and the computational efficiency [24]–[26]. A linear- practical and needs much less topology assumptions.
programming-based approach to reliability assessment is 2) The proposed model takes a linear programming form and
presented in [24] where fictitious power flows were defined to can be conveniently embedded into operating and
describe the nodal affiliation of each feeder. Based on the planning optimization models of distribution networks.
results of optimal fictitious power flows, reliability indices are Especially, besides applications in the planning of new
linearly expressed and calculated. A fast application is studied branches as [25], this model also provides the ability for
in [25], in which the reliability assessment model proposed in planning new circuit breakers and switches on the
[24] is used to capture the reliability indices of a distribution branches since the detailed placement of these equipment
network planning problem. In work [26], load restoration are included in this model.
through tie-lines are further fully considered in the model-based The remainder of this paper is organized as follows. Section
reliability assessment method. However, in [24]–[26], in order II presents the illustration of reliability assessment. Section III
to properly model the actions of equipment after a fault happens, describes the reliability assessment model. In Section IV,
some impractical assumptions are made for simplicity. For numerical tests are reported and analyzed. Finally, relevant
example, in [24] and [25], it is assumed that only one circuit conclusions are drawn in Section V.
breaker is placed at the output of the substation for each feeder
and all branches are equipped with a switch that enables the II. ILLUSTRATION OF RELIABILITY ASSESSMENT
isolation of the part of the feeder downstream of the fault. In Distribution system reliability is assessed on a yearly basis

0885-8950 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on June 02,2020 at 03:49:40 UTC from IEEE Xplore. Restrictions apply.
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/TPWRS.2020.2981508, IEEE
Transactions on Power Systems
4

through well-known reliability indices, such as CIF (Customer  45RP and  56RP are equal to 2.0, 2.5, 3.0 and 4.0 hours per failure.
Interruption Frequency), CID (Customer Interruption Duration), Finally, switching-only interruption durations for these
SAIFI, SAIDI, ASAI, and EENS. CIF and CID represent the branches,  12SW ,  23SW ,  45SW and  56SW are equal to 0.15, 0.20, 0.50
frequency and the duration of interruptions at each load node
respectively. SAIFI indicates how many sustained interruptions and 0.60 hours per failure, respectively.
an average customer will experience. SAIDI is an index of how Under the assumptions mentioned above, nodal reliability
many interruption hours an average customer will suffer. ASAI indices for these four protection configurations are calculated
is the proportion of hours in a year that the service is available and listed in Table I. Specifically, an outage scenario is
to an average customer. EENS is calculated as the total energy discussed in detail when a fault happens on the branch 2-3. For
that is expected to be curtailed. Further, the analytical reliability configuration N1, the fault trips the circuit breaker B1, so the
assessment considering the effect on reliability of different loads connected to nodes 2, 3 are curtailed. Next, switch S1 is
placement of circuit breakers and switches is demonstrated. opened to isolate and repair the fault. Under this circumstance,
demands at nodes 2 and 3 are out of service during whole
A. Assumptions switching-and-repairing interruption duration  23RP . As for
In this paper, the following practical assumptions are configuration N2, the fault similarly trips the circuit breaker B1,
customarily adopted for the sake of tractability. but next switch S2 is opened to isolate the fault. Thus demand
1) The contingency set only contains single branch outages at node 2 only suffers the switching-only interruption duration
for simplicity, which are characterized by failure rates and  23SW . When it comes to configuration N3, the fault on branch 2–
interruption durations. The contingency set is scalable if needed.
3 trips the circuit breaker B2, and after  23SW , switch S2 is
Based on the methods proposed in [2]-[4] and the IEEE
Standard [5], failure rates and interruption durations of opened to isolate the fault. In this situation, the demand at node
equipment in distribution networks are predefined as the input 2 is totally not influenced by the fault. For configuration N4,
data of the proposed model. firstly the fault trips the circuit breaker B2, and then switches
2) A radially-operated but mesh-constructed distribution S3 and S4 are opened to isolate the fault. At the same time,
system is considered wherein the root of each feeder is equipped switches S5 and S6 are closed in order to restore the demand at
with a circuit breaker. Besides, one or more extra circuit node 3. Therefore, the demand at node 3 could only suffer the
breakers also may be equipped inside each feeder. When a switching-only interruption duration  23SW . It can be seen that
sustained fault happens, the relay protection system would customers reliability indices are significantly influnced by the
recognize the faulty section/branch of this feeder and trip the placement of circuit breakers and switches. Later in Section III,
appropriate breaker to minimize affected loads. the proposed reliability assessment method is applied to the
3) Once a sustained fault occurred, the nearest circuit breaker configuration N3 in Fig. 1 as an illustrative example.
upstream the fault trips, thereby curtailing all downstream load
B1S1 S2 B2
demands. After a switching-only interruption duration (from N1 1 2 3 4 5 6

the fault happens to the network reconfiguration implements),


B1S1 S2 S3 S4 S5 S6 B2
corresponding switches upstream and downstream of the fault N2 1 2 3 4 5 6

are opened to isolate the fault, then the system topology is


B1S1 B2S2 S3 S4 S5 S6 S7 B3
reconfigured by operating switches and reclosing circuit N3 1 2 3 4 5 6

breaker to reduce the non-supplied demand. Finally, once the


B1S1 S2 B2S3 S4 S5 S6 S7 S8 B3 S9 S10B4
isolated fault is cleared, after a switching-and-repairing N4 1 2 3 4 5 6

interruption duration (from the fault happens to the damage is


Substation node Circuit breaker (closed) Switch (closed)
repaired), the system topology is reinstated.
Load node Circuit breaker (open) Switch (open)
4) Linearized power flow equations based on the model
Figure. 1. Illustrative example.
proposed in [41] is adopted, and linearized capacity constraints
based on the piecewise relaxation method proposed in [42] is TABLE I. RELIABILITY INDICES OF SIX-BUS NETWORK
utilized. Different loading factors are taken into consideration Load
Customer Interruption Customer Interruption
when counting reliability indices as different scenarios. Frequency (1/year) Duration (hour/year)
node
2 3 4 5 2 3 4 5
B. Illustrative Example N1 0.8 0.8 0.6 0.6 1.75 1.75 2.2 2.2
N2 0.8 0.8 0.6 0.6 1.06 0.825 0.84 0.84
To intuitively illustrate how the placement of circuit breakers N3 0.5 0.8 0.6 0.6 1 0.825 0.34 1.7
and switches affect system’s reliability, a six-bus network is N4 0.5 0.8 0.6 0.4 0.075 0.135 0.34 0.24
designed as depicted in Fig. 1. This network comprises two
substation nodes, node 1 and node 6, and four load nodes, nodes III. PROPOSED RELIABILITY ASSESSMENT MODEL BASED ON
2-5. Four configurations are designed with different placement FICTITIOUS FAULT FLOWS
of circuit breakers, switches and tie-line, labeled as N1~N4. In this section, we describe the proposed reliability
Branch failure rates, 12 , 23 , 45 and 56 are equal to 0.5, assessment model based on fictitious fault flows for distribution
0.3, 0.2 and 0.4 failures per year, respectively. Repairing-and- networks. Firstly, a brief explanation on fictitious fault flows is
switching interruption durations for these branches,  12RP ,  23RP , presented. Secondly, the objective function and constraints are
given for this model. The utilization of big-M method in this

0885-8950 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on June 02,2020 at 03:49:40 UTC from IEEE Xplore. Restrictions apply.
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/TPWRS.2020.2981508, IEEE
Transactions on Power Systems
5

model is described in detail. An illustrative application based  NC CID i i


on the configuration N3 in Fig. 1 is given. Then, a simple SAIDI = i LN

example of integrating the proposed model with a planning  NC i


i LN
model for circuit breakers and switches is discussed. (1)
 
A. Explanation of Fictitious Fault Flows  NCi   xy xySW pixy +  xy ( xyRP −  xySW )(1 − qixy ) 
=
i LN  xy xy 
Fictitious fault flows are modeled as variables which are
stemmed from the faulty branch, transmitted through nodes and  NCi
i LN
branches and isolated by circuit breakers and switches in order where NCi is the number of customers at node i and CIDi is
to imitate the real actions taking place after a sustained fault
occurs. Based on the assumptions demonstrated in Section II.A, the customer interruption duration for node i . Such objective
two types of fictitious fault flows are assumed in our model function actually takes no practical meaning, because the
during the three stages as shown in Fig. 2. These two types of SAIDI for a given distribution network takes a constant value
fictitious fault flows are labeled by “RA” and “PF” respectively which does not need to be optimized. However, it aims to force
corresponding to different stages. Take the configuration N3 in circuit breakers and switches to adopt appropriate actions as the
Fig. 1 as an example, as shown in Fig. 3, right after a sustained assumptions mentioned in Section II with the constraints in the
fault occurs in branch 2-3, the fictitious fault flow labeled by model. One can also replace the objective function (1) by
“RA” is stemmed from branch 2-3 and is isolated by tripped maximizing the total demand in all outage scenarios, or
circuit breaker B2 and normally-open switch S3. Thus, during minimizing EENS (this has been verified by case studies).
the entire switching-only interruption duration, the fictitious C. Constraints in the Problem
fault flow labeled by “RA” passing through node 3 indicate that The constraints are divided into three blocks by time periods.
the demand at node 3 loses power supply. As shown in Fig. 4, The first block describes how the fault current is cut by breaker
after post-fault network reconfiguration, the fictitious fault flow just after a sustained fault happens. The second block indicates
labeled by “PF” is stemmed from the faulty branch 2-3 and how faulty branch is isolated and network is reconfigured by
isolated by switches S2 and S3. Thus, the demand of node 3 opening/closing switches to restore as much outage loads as
keep out of service from stage 2 to stage 3. possible. The third block calculates reliability indices for the
Fictitious fault flows labeled
distribution network.
Stage 1:
The fault just occurs.
by RA , can only be The first block of constraints describes how the fault flow is
tripped by circuit breakers.
Switching-only cut by circuit breaker after a sustained fault happens.
interruption duration
Stage 2: Fault isolation and
load restoration by network
Fictitious fault flows labeled
by PF , can only be
f xyxy ,RA = 0 (2)
reconfiguration. isolated by switches.
Switching-and-repairing −(1 − siji,NO )M + fi xy ,RA  fijxy,RA  (1 − siji,NO )M + fi xy,RA , ij SI , ij  BI (3)
interruption duration
−(1 − sijj ,NO )M + f jxy,RA  fijxy,RA  (1 − sijj ,NO )M + f jxy,RA , ij  SJ , ij  BJ (4)
Stage 3:
The fault is cleared.

Figure. 2. Explanation of the fictitious fault flows. −(1 − biji , xy ) M + fi xy ,RA  fijxy ,RA  (1 − biji , xy ) M + fi xy ,RA , ij   BI (5)
Source of fictitious fault −(1 − biji , xy ) M + f jxy ,RA  fijxy ,RA  (1 − biji , xy ) M + f jxy ,RA , ij   BJ (6)
flow labeled by RA .
fijxy ,RA = fi xy ,RA , ij  SI , ij   BI (7)
B1S1 B2S2 S3 S4 S5 S6 S7 B3
N3 1 2 3 4 5 6
The fictitious fault flow spreads through the
f ij
xy ,RA
=f j
xy ,RA
, ij   , ij  
S
J
B
J (8)
network and is isolated by tripped circuit
breaker and normally-open switches. b i ,NO
ij + b j ,NO
ij −1 = b i , xy
ij + b j , xy
ij
(9)
ijiB ij Bj ijiB ij Bj
Figure. 3. Example of the fictitious fault flows labeled “RA”.
0  fi xy ,RA  1, i   LN (10)
Source of fictitious fault
flow labeled by PF . 0 f ij
xy ,RA
 1, ij   (11)

1
B1S1
2
B2S2
3
S3 S4
4
S5 S6
5
S7 B3
6 fi xy ,RA
= 1, i   SS
(12)
N3
The fictitious fault flow spreads p = 1 − fi , i  
i
xy xy ,RA
(13) LN

through the network and is isolated


by switches with open status. xy   for (2)-(13)
Figure. 4. Example of the fictitious fault flows labeled “PF”. Here, we use superscript xy to describe different scenarios. For
B. Objective Function xy   , it represents the outage scenario due to a fault at
The objective function for the reliability assessment model is branch xy . For the case of normal operation scenario,
formulated as superscript NO is used. For existing distribution networks, the
status of switches and breakers under normal condition,
s 
i ,NO
ij ijiS
j ,NO
, sij  ij Sj
, biji ,NO ijB , bij
i

j ,NO
 ij Bj
are known.
xy ,RA
We use fi xy ,RA and fij to indicate the fictitious fault flow

0885-8950 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on June 02,2020 at 03:49:40 UTC from IEEE Xplore. Restrictions apply.
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/TPWRS.2020.2981508, IEEE
Transactions on Power Systems
6

stemmed from the outage branch right after the fault occurs. If Pfxy  S Cf , f   F (36)
the fictitious fault flow passes through a branch or a bus (which
is equal to zero), it is affected by this sustained fault and the Q  S , f  
xy
f
C
f
F
(37)
corresponding fictitious flow variable is equal to zero. For the Q fxy + Pfxy  2S Cf , f   F (38)
healthy part of the network, the fictitious flow variables are
equal to one. Pi = Pi q , i  
xy
i
xy LN
(39)
Constraint (2) defines the source of the fictitious fault flow Q = Qi q , i  
i
xy
i
xy LN
(40)
labeled “RA” during the outage scenario due to a fault at branch q = fi
xy xy ,PF
, i   LN
(41)
i
xy . Through linking paths, the fault current spreads around, as
1 − p  q , i  
i
xy
i
xy LN
(42)
shown in (3)-(8), and is cut by a tripped circuit breaker, as
xy   for (14)-(42)
shown in (5)-(6), which are explained in Section III.D in detail.
Constraint (9) restricts that only one circuit breaker is going to After a circuit breaker trips, by taking post-fault network
be tripped when a sustained fault happens. Constraints (10)-(11) reconfiguration, the status of some switches needs to change in
limit the lower and upper bounds of the fictitious fault flow order to isolate the faulty branch and restore as much as the
variables. Constraints (12) shows that substation nodes should affected demand. The isolated part of network after network
not be affected by the fault flow. Finally, the load at node i is reconfiguration is labeled by another fictitious flow variables
xy , PF
affected by the outage of branch xy if and only if pixy = 1 , as fi xy , PF and fij . If the fictitious fault flow (which is equal to
shown in constraint (13). zero) passes through a branch or a node, it means that it is
The second block of constraints indicates how faulty branch isolated and the corresponding fictitious flow variable is equal
is isolated and network is reconfigured by opening/closing to zero. For the other part of the network, the fictitious flow
switches to restore as much affected loads as possible. variables are equal to one.
As shown in constraint (14), the fictitious fault flow labeled
f xyxy ,PF = 0 (14)
“PF” is also stemmed from the outage branch. Linkages of
−(1 − siji , xy ) M + fi xy ,PF  fijxy,PF  (1 − siji, xy ) M + fi xy,PF , ij   SI (15) fictitious power flows are expressed as constraints (15)-(18).
Constraints (19)-(20) limit the lower and upper bounds of the
−(1 − sijj , xy ) M + f jxy ,PF  fijxy ,PF  (1 − sijj , xy ) M + f jxy,PF , ij   SJ (16)
fictitious fault flow variables. Constraints (21) restricts that
fijxy ,PF = fi xy ,PF , ij  SI (17) substation nodes should not be affected. Constraints (22)-(25)
comes from linearized power flow equations [41] for each
fijxy ,PF = f jxy ,PF , ij  SJ (18)
outage scenario. Constraint (26) represents nodal voltage
0  fi xy ,PF
 1, i   LN
(19) constraints. Power flow of branch ij is restricted by the status
0 f ij
xy ,PF
 1, ij   (20) of switches on the branch, as shown in constraint (27)-(30).
Linearized capacity constraints of branches based on the
fi xy ,PF = 1, i   SS (21)
piecewise relation method proposed in [42] are presented as
P xy
ki = P
ji
ij
xy
+ Pi , ki  
xy
(22) constraints (31)-(33). Constraints (34) and (35) represent the
power of feeder f supplied by its corresponding transformer
Qkixy = Q
ji
xy
ij + Qixy , ki   (23)
tr f , and constraints (36)-(38) represents capacity limitations of
− M (1 − sijxy ) + 2(rij Pijxy + xijQijxy )  U ixy − U jxy
transformers.
After post-fault network reconfiguration, if and only if the
(24)
 M (1 − sijxy ) + 2(rij Pijxy + xijQijxy ), ij   load at node i is supplied (including unaffected nodes and
2
restored nodes), the situation is labeled as qixy ,t = 1 , as shown in
U ixy = V s , i   SS (25) constraints (39)-(41). For the nodes which are not affected by
U  U  U , i  
i
xy LN
(26) the outage, the loads of these nodes should not be influenced
after post-fault network reconfiguration, which is restricted by
− Ms i , xy
ij  P  Ms , ij  
ij
xy i , xy
ij (27) constraint (42).
− Ms i , xy
ij  Q  Ms , ij  
xy
ij
i , xy
ij (28) The third block of constraints calculates reliability indices for
the distribution network. Based on the indicator variables pixy
− Ms j , xy
ij  P  Ms
ij
xy j , xy
ij , ij   (29)
and qixy mentioned above, common used reliability indices take
− Ms j , xy
ij  Q  Ms xy
ij
j , xy
ij , ij   (30) linear expressions with these variables, as listed in constraints
−SijC  Pijxy  SijC , ij   (31) (43)-(47). The first term on the right side of equation (43),
−S  Q  S , ij  
C
ij
xy
ij
C
ij (32)  xy xySW pixy , indicates that if node i is affected by the outage
xy

− 2 S  P  Q  2S , ij  
C
ij ij
xy xy
ij
C
ij (33) of branch xy ( pixy = 1 ), the demand of node i would lost power
Pfxy = Ptrxyf , f   F , tr f   (34) service during a switching-only interruption duration  xySW . The
second term of equation (43) indicate that if node i still lost
Q fxy = Qtrxyf , f   F , tr f   (35)

0885-8950 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on June 02,2020 at 03:49:40 UTC from IEEE Xplore. Restrictions apply.
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/TPWRS.2020.2981508, IEEE
Transactions on Power Systems
7

power service after network reconfiguration ( qixy = 0 ), the D. Utilization of the Big-M Method in the Model
demand of node i would lost power service during a whole The big-M method is used in this problem to coordinate
switching-and repairing interruption duration  xyRP . binary variables with real variables and to represent logical
constraints. For example, in constraints (15)-(16), the big-M
CIDi = 
xy
SW
xy xy  pixy +   (
xy
xy
RP
xy −  xySW )(1 − qixy ), i   LN (43)
method is utilized to maintain the consistence of the fictitious
fault flow variables for a branch and for the nodes at its both
CIFi = 
xy
xy pixy , i   LN (44)
ends with respect to the on/off status of the corresponding
 NC CIF
switches. As depicted in Fig. 5, node i and node j are connected
i i
by branch ij through switches Siji and Sijj whose on/off status
SAIFI = i LN
(45)
 NC i are represented by siji , xy and sijj , xy when a sustained fault occurs
i LN
on branch xy. As shown in constraint (15), if siji , xy = 1 which
SAIDI
ASAI = 1 − (46) means that the switch Siji takes closed status after post-fault
8760
 reconfiguration actions, and the constraint (15) is then
EENS =  h  CIDi h Li (47) expressed as
hH 8760 i LN
−(1 − 1) M + fi xy ,PF  fijxy ,PF  (1 − 1)M + fi xy ,PF (48)
After defining objective function and constraints, the total
reliability assessment model can be presented as: which can be rewritten as
minimize (1) fijxy ,PF = fi xy ,PF (49)
subject to (2)-(47), xy   i
It shows that if the switch S was closed, the fictitious fault
ij

The proposed model evidently takes mixed-integer linear flow would pass through this switch, resulting in the
programming (MILP) formulation and can be effectively solved consistence between the post-fault fictitious flow variables of
by off-the-shelf commercial software. Suppose the numbers of node i and branch ij. If siji , xy = 0 , which means that the switch
nodes and branches in the distribution network are nN and nB ,
Siji takes open status after post-fault reconfiguration, the
the total numbers of variables and binary variables of the
proposed model must be less than nB ( 7nB + 5nN + 1) and
constraint (15) is expressed as
−(1 − 0) M + fi xy ,PF  fijxy ,PF  (1 − 0) M + fi xy ,PF (50)
nB ( 4nB + 2nN ) respectively. The number of constraints is also which can be rewritten as
limited by second order terms of nN and nB , and increases − M + fi xy ,PF  fijxy ,PF  M + fi xy ,PF (51)
with the number of circuit breakers and switches. For
convenience, all model variables except reliability indices are xy,PF xy,PF
f i f j
listed in the following Table II with their dimensions. The i,xy xy,PF j,xy
number of circuit breakers and switches are known as nCB and Node
sij f ij sij Node
i Branch ij j
nSW . xy xy
Pij,Qij
TABLE II. DECISION VARIABLES AND THEIR DIMENSIONS Figure. 5. Explanation of the utilization of big-M method.
Decision variable Type Dimension
{b }, ij  , xy  
i , xy
Boolean nCB nB
ij
Since M is a big enough number, the constraint (51) becomes
{b j , xy
ij }, ij  , xy   Boolean nCB nB invalid in the model. Therefore, M should be big enough to
{siji , xy }, ij  , xy   Boolean nSW nB make sure that the corresponding constraints become invalid
when M works in these constraints. In our model, we utilize big-
{s j , xy
ij }, ij  , xy   Boolean nSW nB
M method in constraints (3)-(6), (15)-(16), (24), (27)-(29). A
{ p }, i  , xy  
i
xy N
Boolean nN nB real active number can be selected which is larger than one, the
{q }, i  , xy  
i
xy N
Boolean nN nB sum of the total load demand and the max deference of nodal
{Pi }, i  , xy  
xy N
Float nN nB
voltages, which can be expressed as
 i LN | Pi | 
{Q }, i  , xy  
i
xy N
Float nN nB  
M  max  i
{U }, i   , xy  
xy N
Float nN nB  LN | Qi |
i  (52)
{Pijxy }, ij  , xy   Float nB2 U − U 
 
{Qijxy }, ij  , xy   Float nB2 1 
{ fi xy ,RA }, i   N , xy   Float nN nB E. Illustrative Application on the Six-bus Network
{ fi xy ,PF
}, i  , xy  
N
Float nN nB The proposed method is applied to the configuration N3 in
{ fijxy ,RA }, ij  , xy   Float nB2 Fig. 1. Specifically, we consider the situation when a fault just
happens on the branch 2-3. At this circumstance, the first block
{ fijxy ,PF }, ij  , xy   Float nB2
of constraints is described as:

0885-8950 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on June 02,2020 at 03:49:40 UTC from IEEE Xplore. Restrictions apply.
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/TPWRS.2020.2981508, IEEE
Transactions on Power Systems
8

f 2323,RA = 0 (53) Pi 23 = Pi qi23 ,, i = 2,3, 4,5 (80)


f 23,RA
12 =f 2
23,RA
(54) Q = Qi q ,, i = 2,3, 4,5
i
23 23
i
(81)
f 23,RA
23 =f 3
23,RA
(55) q = fi
23
i
23,PF
, i = 2,3, 4,5 (82)
f 23,RA
45 =f 4
23,RA
=f 5
23,RA
(56) 1 − p  q , i = 2,3, 4,5
23
i
23
i
(83)
f 23,RA
56 =f 5
23,RA
(57) The model is to maximize the sum of fi , i = 2,3, 4,5 . It can 23,PF

−(1 − b 1,23
12 )M + f 1
23,RA
 f 23,RA
12  (1 − b 1,23
12 )M + f 1
23,RA
(58) 1,23
be seen that the optimal solution is s12 , s45 4,23 5,23
, s45 6,23
, s56 =1 ,
−(1 − b 2,23
23 )M + f 2
23,RA
 f 23,RA
23  (1 − b 2,23
23 )M + f 2
23,RA
(59) 2,23
s23 3,23
, s34 4,23
, s34 = 0 and the fictitious flow variables for load
−(1 − b566,23 ) M + f623,RA  f5623,RA  (1 − b566,23 ) M + f623,RA (60) buses are f 223,PF = f 423,PF = f523,PF = 1 and f 323,PF = 0 . Thus we
2=b 1,23
12 +b 2,23
23 +b 6,23
56
(61) have q223 = 1 , q323 = 0 , q423 = 1 and q523 = 1 by equation (83),
0  fi 23,RA
 1, i = 2,3, 4,5 (62) as shown in Fig. 7. The results show that the switch S2 is
0 f 23,RA
 1, ij = 12, 23,34, 45,56 (63) opened to isolate the fault, which is consistent with the previous
ij
analysis in Section II.
fi 23,RA = 1, i = 1,6 (64)
p = 1 − fi , i = 2,3, 4,5
23
i
(65)23,RA 23
q2 =1
23
q2 =0 (not restored)
23
q4 =1
23
q4 =1
23,PF 23,PF 23,PF 23,PF 23,PF 23,PF
f 1 =1 f 2 =1 f 2 =0 f 4 =1 f 4 =1 f 5 =1
Because the objective function is to minimize the weighted sum
B1S1 B2S2 S3 S4 S5 S6 S7 B3
of CIDs, which takes positive relations with pi23 , i = 2, 3, 4,5 as N3 1
23,PF
2 3 4 5 6
23,PF 23,PF 23,PF 23,PF
f 12 =1 f 23 =0 f 34 =1 f 45 =1 f 45 =1
shown in constraints (43). Thus, the model would maximize the
Figure. 7. Distribution of fictitious fault flow labeled “PF” when a fault occurs
sum of fi 23,RA , i = 2,3, 4,5 . It can be easily seen that the optimal at branch 2-3 based on configuration N3.
1,23
solution is b12 = b56
6,23
= 1 , b23
2,23
= 0 and the fictitious flow
Referring to the equation (43), when a fault happens on the
variables for load buses are f 223,RA = f 423,RA = f523,RA = 1 and
branch 2-3, the interruption durations tiID for load nodes 2-5
f 323,RA = 0 . Thus we have p223 = 0 , p323 = 1 , p423 = 0 and
can be expressed as
p523 = 0 by equation (65), as shown in Fig. 6. The results show t2ID =  23SW p223 + ( 23RP −  23SW )(1 − q223 ) = 0 (84)
that the circuit breaker B2 trips and only the demand at node 3
is curtailed during the switching-only interruption duration, t3ID =  23SW p323 + ( 23RP −  23SW )(1 − q323 ) =  23RP (85)
t4ID =  23SW p423 + ( 23RP −  23SW )(1 − q423 ) = 0
which is consistent with the previous analysis in Section II.
(86)
t5ID =  23SW p523 + ( 23RP −  23SW )(1 − q523 ) = 0
23 23 23 23
p2 =0 p2 =1(affected) p4 =0 p4 =0
23,RA 23,RA 23,RA 23,RA 23,RA 23,RA
(87)
f 1 =1 f 2 =1 f 2 =0 f 4 =1 f 4 =1 f 5 =1
B1S1 B2S2 S3 S4 S5 S6 S7 B3
N3 1 2 3 4 5 6 F. Planning Model for Circuit Breakers and Switches with
23,RA 23,RA 23,RA 23,RA 23,RA
f 12 =1 f 23 =0 f 34 =1 f 45 =1 f 45 =1 Reliability Constraints
Figure. 6. Distribution of fictitious fault flow labeled “RA” when a fault occurs The proposed reliability assessment model is integrated with
at branch 2-3 based on configuration N3.
a planning model for circuit breakers and switches for
distribution networks. In this model, only the placement of
Besides, the second block of constraints is described as
circuit breakers and switches are optimized with reliability
below (power flow constraints (22)-(38) are not listed here due
constraints, and the objective function is to minimize the
to space limitations):
investment cost of these equipment. The model is presented as:
f 2323,PF = 0
minimize C CB  ( xiji ,CB + xijj ,CB ) + C SW  ( xiji ,SW + xijj ,SW ) (88)
(66)
−(1 − s12
1,23
) M + f123,PF  f1223,PF  (1 − s12
1,23
) M + f123,PF (67) ij ij

−(1 − s 2,23
23 )M + f 2
23,PF
 f 23,PF
23  (1 − s 2,23
23 )M + f 2
23,PF
(68) subject to (2), (9)-(13), (14), (19)-(21), (39)-(42), xy  
−(1 − s 3,23
)M + f 23,PF
 f 23,PF
 (1 − s 3,23
)M + f 23,PF
(69) −(2 − biji , xy − siji ,NO ) M + Fi xy ,RA  Fijxy ,RA  (2 − biji , xy − siji ,NO ) M + Fi xy ,RA , ij   (89)
34 3 34 34 3

−(1 − s34
4,23
) M + f 423,PF  f3423,PF  (1 − s34
4,23
) M + f 423,PF (70) −(2 − bijj , xy − sijj ,NO ) M + Fjxy ,RA  Fijxy ,RA  (2 − bijj , xy − sijj ,NO )M + Fjxy ,RA , ij   (90)
−(1 − s45
4,23
) M + f 423,PF  f 4523,PF  (1 − s45
4,23
) M + f 423,PF (71) −(1 − siji , xy ) M + fi xy ,PF  fijxy ,PF  (1 − siji , xy ) M + fi xy ,PF , ij   (91)
−(1 − s45
5,23
) M + f523,PF  f 4523,PF  (1 − s45
5,23
) M + f523,PF (72) −(1 − sijj , xy )M + f jxy,PF  fijxy,PF  (1 − sijj , xy )M + f jxy,PF , ij   (92)
−(1 − s 6,23
)M + f 23,PF
 f 23,PF
 (1 − s 6,23
)M + f 23,PF
(73)
56 6 56 56 6
siji , xy + sijj , xy − 1  lijxy  ( siji , xy + sijj , xy ) / 2, ij   (93)
f 23,PF
=f 23,PF
(74)
q = l
12 2
xy xy
(94)
f 2323,PF = f323,PF (75) i LN
i
ij
ij

f5623,PF = f523,PF (76) xy   for (89)-(94)


0  fi 23,PF
 1, i = 2,3, 4,5 (77) (22)-(38), xy    xy = NO
0  fij23,PF  1, ij = 12, 23,34, 45,56 (78)
biji , xy  1 − xiji ,CB , ij   (95)
fi 23,PF
= 1, i = 1,6 (79)
b j , xy
ij 1− x j ,CB
ij , ij   (96)

0885-8950 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on June 02,2020 at 03:49:40 UTC from IEEE Xplore. Restrictions apply.
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/TPWRS.2020.2981508, IEEE
Transactions on Power Systems
9

siji , xy  1 − xiji ,SW , ij   (97) computationally acceptable sub-models and solve them in a
parallel way, and then count the reliability indices of the
sijj , xy  1 − xijj ,SW , ij   (98) distribution network, as shown in [26].
xy    xy = NO for (95)-(98) On the other side, for the planning model of circuit breakers
and switches, due to the correlation of binary variables for
xiji , SW  xiji ,CB , ij   (99)
whether to place switchable devices, the model cannot be
x i , SW
ij x i ,CB
ij , ij   (100) separated as the reliability assessment model. Our repeating
Pi NO
= Pi , i   LN
(101) tests has proved that for a distribution network whose node
number is less than 300 and branch number is less than 500, the
Q i
NO
= Qi , i   LN
(102)
proposed planning model can be effectively solved by off-the-
i ,NO
s
ij +sj ,NO
ij −1  l NO
ij  (s i ,NO
ij +s j ,NO
ij ) / 2, ij   (103) shelf software like CPLEX. Obviously, in real applications, we
can manually separate the large network to some smaller
n LN
= l NO
ij
(104)
ij
networks based on geographical distribution and place the
(43)-(47), switchable devices individually to reduce computation burden.
CIFi   iCIF , i   LN (105) H. Discussions about the Uncertainties of Failure Rates and
CIDi   CID
, i   LN
(106) Interruption Durations for Outages
i

SAIFI   SAIFI (107) Actually, practical failure rates and repair times of outages in
distribution networks should be statistically counted based on
SAIDI   SAIDI (108) real data, which could vary from year to year and take
EENS   EENS (109) uncertainties. By solving the proposed reliability assessment
The C CB
and C SW
are the investment cost of a circuit breaker model, after obtain the output variables pixy and qixy for all
i ,CB j ,CB i , SW j , SW load nodes and xy   , uncertainties of failure rates and
and a switch respectively. The xij , xij , xij and xij
interruptation durations can be taken into consideration in the
are binary decision variables indicating whether to equip a
reliability indices as
circuit breaker at side i and side j on branch ij , and whether
CIDi =  pixy E SW RP  xy SW +
 xy , xy , xy ~  xy  
xy
to equip a switch at side i and side j . Since the placement of xy 
(110)
( )
these switchable devices is not known as a priori, constraints (3)
( )
 i  xy , xy , xy ~xy   xy  RP , i   LN
SW
-(8) are rewritten as (89)-(90), and constraints (15)-(16) are 1 − q xy
Ε − 

SW RP xy xy
xy
transferred to (91)-(92). Equations (93)-(94) are the radial
operation constraints for outage scenarios. Logical relations where E   represents the expectation operation. Stochastic
between the open/closed status of circuit breakers (or switches) SW RP

and the equipped/unequipped status are shown as (95)-(96) (or variables  xy ,  xy and  xy obey probability distribution
(97)-(98)). Circuit breakers are always collocated with switches, function  xy . The expectation (110) can be counted by
which is constrained by (99)-(100). Load demand condition and numerical methods like Buffon's needle sampling.
radial constraints for normal operation scenario are restricted
by (101)-(102) and (103)-(104). Reliability of load nodes and I. Discussions about the Uncertainties of Power Injections
the distribution system are guaranteed by constraints (105)- The reliability assessment model is scalable and the demand
(109) with predefined reliability requirement  . It can be seen variations, uncertainty of distributed generations (DGs) and
that not only systematic reliability indices but also nodal protection failures can be considered. The uncertainties of load
reliability indices can be taken into consideration in the and DGs can be simulated by multi-scenario techniques [14],
reliability-constrained planning model for the circuit breakers [26], [46]. Through scenario generation and scenario reduction
and switches. This model can also give the best open/close based on the method proposed in [46], required multiple
status of switches under normal operation condition as siji ,NO scenarios are obtained with probability piSC , i  SC and load
and sijj ,NO . and generation data. Then the reliability indices for each
scenario are calculated using the proposed method and labeled
G. Discussions about Computation Efficiency of the Proposed as SAIDI SC −i , SAIFI SC −i and EENS SC −i . Therefore, the final
Models annual reliability indices for distribution network can be
Binary variables have to be implicated in this model because expressed as
the open/closed status for circuit breakers and switches only can SAIDI =  piSC SAIDI SC −i (111)
be described as Boolean type. Furthermore, for the planning i SC

model of circuit breakers and switches, whether to place a SAIFI =  piSC SAIFI SC −i (112)
switchable device is also a binary variable. For a MILP problem, i SC

the number of binary variables directly decides the computation EENS =  piSC EENS SC −i (113)
efficiency of the model. i SC

For the reliability assessment model, for larger systems,


owing to the constraints of different outage scenarios share no IV. NUMERICAL TESTS
common variable except the ones under normal operation In this section, the proposed reliability assessment model is
condition, we could separate the proposed model into several

0885-8950 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on June 02,2020 at 03:49:40 UTC from IEEE Xplore. Restrictions apply.
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/TPWRS.2020.2981508, IEEE
Transactions on Power Systems
10

20 20

demonstrated on 4 tested systems. We use a PC computer with 17

1
51
3
17

1
51
3
19 9 19 9

an Intel Core i7-7700 processor at 3.60 GHz and 32 GB of 18


2
7 4 18
2
7 4
5 5

RAM. Involved MILP problems are solved by CPLEX 12.5 21


22 23 24
25
8
26
6 21
22 23 24
25
8
26
6

with default settings. 54

10
32
27

28
54

10
32
27

28
33 33
31 34 35 31 34 35
30 36 30 36

A. 54-node Test System [25] 29 37 38 39

40
53
29 37 38 39

40
53

41 41
43 43
This system comprises 50 load nodes, 4 substation nodes, 8 13 45
44 15 16
13 45
44 15 16

42 42

feeders and 61 branches, as shown in Fig. 8. For reproducibility 12 12


11 14 46 47 11 14 46 47

purpose, system data can be accessed from [44]. Failure rate for 48 48
50 50

feeders is set as 0.1 per year and kilometer [45], and interruption 52
49
52
49

(a) Case A (b) Case B


durations for repairing-and-switching and switching-only are 20
17
20
17
51 51
3 3
set as 3 hours and 0.5 hour respectively. The loading condition 19 9
1

2
19 9
1

2
18 7 4 18 7 4

is modeled by three load levels with loading factors equal to 70% 22 23 24


25 26
5
22 23 24
25 26
5

21 8 6 21 8 6
(2,000 h/year), 83% (5,760 h/year) and 100% (1,000 h/year) of 54
27
54
27
32 32

the corresponding peak demand, respectively [26]. We carry 31


10
33
34 35
28

31
10
33
34 35
28

30 36 30 36

out six cases where the different placement of circuit breakers 29 37 38 39

40
53
29 37 38 39

40
53

41 41
and switches is designed (which are described in Table III and 43

45
44 15 16
43

45
44 15 16
13 13

depicted in Fig. 9) to test the proposed method, and the existing 12


42

12
42

status of tie-lines is also discussed. The results are verified 11 14 46 47

48
11 14 46 47

48

compared by Monte Carlo simulation-based method proposed 52


50
49
52
50
49

in [19], and the reliability indices of the tested system are also (c) Case C (d) Case D
20 20
17 17
51 51
evaluated by the method proposed in other model-based 19 9
1
3
19 9
1
3

2 2
methods, e.g., the models presented in [24] and [26]. 18

22 23 24
7 4
5
18

22 23 24
7 4
5

25 26 25 26
21 8 6 21 8 6

20 27 27
54 54
17 32 32
51 10 28 10 28
3
1 33 33
19 9 31 34 31 34
30 35 36 30 35 36
2 29 37 38 39 29 37 38 39
18 7 4 53 53
5 40 40
23 24 41 41
22 26 43 43
25
21 8 6 44 15 16 44 15 16
13 45 13 45
42 42
27
54
32
12 12
10 28 14 46 47 14 46 47
11 11
33
31 34 35
30 36 48 48
29 37 38 39 50 50
53 52 52
49 49
40
43
41
(e) Case E (f) Case F
44 15 16
13 45 Circuit breaker Closed switch Open switch
42

12 Figure 9. Different configurations for five tested cases.


11 14 46 47

52
50
48
TABLE IV. RESULTS OF TESTED CASES
49
Case A Case B Case C Case D Case E Case F
Load node Normally connected branch
Substation node Normally disconnected branch Proposed 0.4864 0.6669 2.0765 1.8961 1.1338 1.9141
Figure 8. One-line diagram of 54-node system. SAIDI [19] 0.4864 0.6669 2.0765 1.8961 1.1338 1.9141
(h/year) [24] 2.0765 2.0765 2.0765 2.0765 2.0765 2.0765
TABLE III. DESCRIPTIONS OF TESTED CASES ON 54-NODE SYSTEM [26] 0.6669 0.6669 2.0765 2.0765 0.6669 0.6669
Descriptions of the placement. Tie- Proposed 0.9271 1.2881 1.2881 0.9271 0.9271 0.9271
Circuit breaker Switch line SAIFI [19] 0.9271 1.2881 1.2881 0.9271 0.9271 0.9271
Sectional circuit breakers are All branches are (1/year) [24] 1.2881 1.2881 1.2881 1.2881 1.2881 1.2881
Case A place at branches 4-5, 8-27, equipped with two √ [26] 1.2881 1.2881 1.2881 1.2881 1.2881 1.2881
9-17, 10-23, 12-45, 14-50. switches at both ends. Proposed 5.7464 7.8785 24.532 22.401 13.394 22.612
All branches are EENS
Case B [19] 5.7464 7.8785 24.532 22.401 13.394 22.612
No sectional circuit breaker. equipped with two √ (MWh/
([26]) [24] 24.532 24.532 24.532 24.532 24.532 24.532
switches at both ends. year)
[26] 7.8785 7.8785 24.532 24.532 7.8785 7.8785
All branches are Solving time of the
Case C
([24])
No sectional circuit breaker. equipped with one switch – proposed method (s)
2.19 2.29 0.47 0.56 2.15 2.53
at the upstream side. Time of simulation-
Sectional circuit breakers are All branches are 213.5 199.4 88.1 92.5 225.6 94.2
based method [19] (s)
Case D place at branches 4-5, 8-27, equipped with two –
9-17, 10-23, 12-45, 14-50. switches at both ends.
Table III describes the characteristics of the tested cases.
Sectional circuit breakers are Selected branches are
Case E place at branches 4-5, 8-27, equipped with switches √ Table IV summarizes the reliability indices of six cases, which
9-17, 10-23, 12-45, 14-50. at one end or both ends. are assessed based on the proposed method, simulation-based
Sectional circuit breakers are Switches are only placed method [19], and model-based methods proposed in [24] and
Case F place at branches 4-5, 8-27, with circuit breakers and √ [26]. From Table IV, firstly, we can see the proposed method
9-17, 10-23, 12-45, 14-50. at tie-lines.
generates the same results with the simulation-based method in
[19], but uses much less computation time. Secondly, the
reliability indices assessed by the methods proposed in [24] and
[26] in Table IV are marked by light orange when the indices

0885-8950 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on June 02,2020 at 03:49:40 UTC from IEEE Xplore. Restrictions apply.
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/TPWRS.2020.2981508, IEEE
Transactions on Power Systems
11

underestimate the reliability of the distribution network, and are From Table VI, we can see that the proposed method costs
marked by light green when the indices overestimate the much less time than the simulation-based method and the
reliability. Only for case C when the assumptions in [24] and solving time does not increase too much with the growth of bus
[26] are simultaneously satisfied, the reliability indices are number. For larger systems, the proposed method is
assessed with one accord by different methods. Because the decomposable (for different outage scenarios) and parallel
models proposed in [24] and [26] cannot recognize the detailed computing technique can be used [26].
placement of circuit breakers and switches, erroneous reliability
indices are computed, as shown in Table IV. Last but not the 20
17
3
20
17
3
1 1
least, as shown in Table IV, we can see that the placement of 19

18
9

2
7 4
19

18
9

2 7 4

circuit breakers and switches poses great impacts on the 21


22 23 24
25
8
26
5

6
22 23 24
25 26
5

21 8 6

reliability indices. 32
27
32
27

In real-life (i.e. utilities), considering the wear and tear when 30


31
10
33
34 35 36
28

30
31
10
33
34 35 36
28

operating the switches, the number of switching operations is 29 37 38 39

40
29 37 38 39

40

often the most important parameter for load restoration. Based 43


44 15 16
41
43

45
44 15 16
41

13 45 13

on the configuration of Case A which comprises 11 tie-lines, 42 42

12 12

the numbers of switching operations of switches under all 11 14 46 47 11 14 46 47

48 48

outage scenarios of 61 branches are listed as Table V. The 50


49
50
49

average number of switching operations takes 3.1475 per fault. (a)  SAIDI = 3 hour/year (b)  SAIDI = 1 hour/year
As shown in Table V, when the number of switching operations 20
17
20
17
3 3

is equal to zero, it means the fault occurs at a tie-line, which 19 9


1

2
19 9
1

2
7 4 18 7 4

does not affect the normal operation of the distribution network. 18

22 23 24
25 26
5
22 23 24
25 26
5

21 8 6 21 8 6

27 27
TABLE V. NUMBERS OF SWITCHING OPERATIONS OF SWITCHES UNDER ALL 10
32
28 10
32
28

OUTAGE SCENARIOS OF 61 BRANCHES 30


31 33
34 35 36 30
31
33
34 35 36
29 37 38 39 29 37 38 39
Outage Num. of Outage Num. of Outage Num. of 40 40
41 41
branch operations branch operations branch operations 43
44 15 16
43
44 15 16
45 13 45
1 1 22 4 43 4 13
42 42

2 4 23 4 44 1 12
11 14 46 47
12
11 14 46 47

3 4 24 4 45 4 50
48
50
48

4 6 25 4 46 0 49 49

5 6 26 4 47 4 (c)  SAIDI = 0.8 hour/year (d)  SAIDI = 0.6 hour/year


6 6 27 4 48 0 20
17
20
17

7 0 28 4 49 4 19 9
1
3
19 9
1
3

8 6 29 0 50 0 18
2
7 4
5
18
2 7 4
5
23
9 4 30 4 51 4 21
22 24
25
8
26
6 21
22 23 24
25
8
26
6

10 4 31 1 52 4 32
27
32
27

11 4 32 0 53 4 31
10
33
34 35
28

31
10
33
34 35
28

30 36 30 36
12 4 33 1 54 4 29 37 38 39 29 37 38 39

40 40
13 4 34 4 55 4 43
41
43
44 15 16
41

44 15 16

14 4 35 4 56 4 13 45
42
13 45
42

15 0 36 0 57 4 12
14 46 47
12
11 14 46 47
11

16 4 37 4 58 4 48 48
50 50
17 4 38 1 59 0 49 49

18 4 39 4 60 0 (e)  SAIDI = 0.4 hour/year (f)  SAIDI = 0.1977 hour/year


19 0 40 1 61 4
20 4 41 4 Circuit breaker Closed switch Open switch
21 4 42 6 Branch (Connected) Tie line (Disconnected)
Figure 10. Different placement results of circuit breakers and switches under
TABLE VI. RESULTS OF 85-NODE, 137-NODE AND 417-NODE SYSTEMS different reliability requirement.
System 85-node 137-node 417-node
SAIDI (h/year) 0.5216 0.9396 0.4815 C. Case Studies for the Planning Model for Circuit Breakers
SAIFI (1/year) 0.7633 1.1581 0.8759 and Switches with Reliability Constraints
ASAI (%) 99.9940 99.9893 99.9945
EENS (MWh/year) 12.761 30.646 44.404 In order to test the reliability-constrained planning model for
Number of binary variables 1597 3091 34512 circuit breakers and switches, based on the 54-node system
Number of constraints 3690 6745 73347 presented in Section IV.A, different placement schemes are
Solving time (s) 7.65 20.62 96.59 generated with different reliability requirement. Without loss of
Time of simulation-based
method [19] (s)
448.8 813.7 2854.1 generality, SAIDI (  SAIDI ) is used to represent the requirement
of systematical reliability, and the SAIDI of the planned
B. 85-node, 137-node and 417-node Test Systems [24]
distribution network should be less than the requirement  SAIDI .
The scalability of the proposed method is further
demonstrated by 85-node, 137-node and 417-node systems. The unit prices of a circuit breaker and a switch are set as
Detailed system data can be accessed from [44]. The results are 6400$ and 2320$ respectively. Average computation time for
verified by simulation-based method [19]. Table VI lists the solving the planning model is about 5 minutes. Results are
reliability assessment results. CIDs and CIFs are listed in [44]. shown as Fig. 10 and Table VII.

0885-8950 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on June 02,2020 at 03:49:40 UTC from IEEE Xplore. Restrictions apply.
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/TPWRS.2020.2981508, IEEE
Transactions on Power Systems
12

From Fig. 10 and Table VII, with the increase of reliability Practical Methods and Applications. Hoboken, NJ, USA: Wiley, 2009.
[3] R. Billinton and J. Billinton, “Distribution system reliability indices,”
requirement, the investment cost of circuit breakers and IEEE Trans. Power Deliv., vol. 4, no. 1, pp. 561–568, Jan. 1989.
switches increases and the SAIDIs of the planned distribution [4] R. Billinton and R. N. Allan, Reliability Evaluation of Power Systems.
network are closely less than the  SAIDI . When the  SAIDI New York, NY, USA: Plenum Press, 1996.
[5] IEEE Guide for Electric Power Distribution Reliability Indices, IEEE
decreases to relatively small value, the cost rapidly rise up to Standard 1366–2003, May 2004.
guarantee the reliability requirement, as shown in Fig. 11. With [6] R. E. Brown, Electric Power Distribution Reliability, 2nd ed. Boca Raton,
FL, USA: CRC Press, 2008.
the proposed reliability-constrained planning model of circuit [7] R. C. Lotero and J. Contreras, “Distribution system planning with
breakers and switches, planners can economically and precisely reliability,” IEEE Trans. Power Deliv., vol. 26, no. 4, pp. 2552–2562, Oct.
place these switchable devices with given reliability 2011.
requirement. Besides, constraints (105)-(106) can also be [8] R. E. Brown and A. P. Hanson, "Impact of two-stage service restoration
on distribution reliability," in IEEE Transactions on Power Systems, vol.
applied to ensure the reliability of some important demand. 16, no. 4, pp. 624-629, Nov. 2001.
[9] Sanabria, L.A., Dillon, T.S. Power system reliability assessment suitable
TABLE VII. PLACEMENT RESULTS OF CIRCUIT BREAKERS AND SWITCHES for a deregulated system via the method of cumulants, International
BASED ON 54-NODE SYSTEM Journal of Electrical Power and Energy Systems, vol. 20, no. 3, Mar. 1998,
SAIDI EENS pp. 203-211
requirement Cost CB SW SAIDI SAIFI (MWh/ [10] Meliopoulos, A.P., Bakirtzis, A.G. Power system reliability evaluation
4
(10 $) Num. Num. (h/year) (1/year) using stochastic load flows, IEEE Transactions on Power Apparatus and
 SAIDI (h/year) year)
Systems, vol. PAS-103, no. 5, May 1984, pp. 1084-1091.
3 9.528 8 19 2.9609 0.9870 34.980 [11] Ruiz-Rodriguez, F.J., Gomez-Gonzalez, M., Jurado, F. Reliability
1 14.96 8 42 0.9992 0.9047 11.805 optimization of an electric power system by biomass fuelled gas engine,
0.8 17.18 8 52 0.7879 0.8030 9.3082 in International Journal of Electrical Power and Energy Systems, vol. 61,
0.6 21.07 9 66 0.5991 0.7363 7.0782 Oct. 2014, pp. 81-89.
0.4 28.68 14 85 0.3994 0.5905 4.7190 [12] Ruiz-Rodriguez, F.J., Gómez-González, M., Jurado, F. A method for
0.1977 106.38 122 122 0.1977 0.3497 2.3358 reliability optimization of distributed generation using meta-heuristic and
probabilistic techniques, in Electric Power Components and Systems, vol.
120 43, no. 1, 2 Jan. 2015, pp. 32-43.
[13] C. Wang, T. Zhang, F. Luo, P. Li and L. Yao, "Fault Incidence Matrix
Based Reliability Evaluation Method for Complex Distribution System,"
100
in IEEE Transactions on Power Systems, vol. 33, no. 6, pp. 6736-6745,
Investment Cost (10$)

Nov. 2018.
4

80 [14] C. Chen, W. Wu, B. Zhang and C. Singh, "An Analytical Adequacy


Evaluation Method for Distribution Networks Considering Protection
60 Strategies and Distributed Generators," in IEEE Transactions on Power
Delivery, vol. 30, no. 3, pp. 1392-1400, June 2015.
40 [15] A. Tabares, G. Munoz-Delgado, J. F. Franco, J. M. Arroyo and J.
Contreras, "An Enhanced Algebraic Approach for the Analytical
Reliability Assessment of Distribution Systems," in IEEE Transactions
20
on Power Systems. Early Access, 2019.
[16] N. S. Rau and C. Necsulescu, "An Investigation of Two Methods for the
0 Probabilistic Energy Production Simulation," IEEE Power Eng. Rev., vol.
0 0.5 1 1.5 2 2.5 3
SAIDI requirement (hour/year)
PER-3, no. 8, pp. 35-35, Aug. 1983.
[17] P. Jorgensen, "A new method for performing probabilistic production
Figure 11. Investment cost of circuit breakers and switches under different
simulations by means of moments and Legendre series," IEEE Trans. on
reliability requirement.
Power Syst., vol. 6, no. 2, pp. 567-575, May 1991.
[18] C. Sondergren and H. F. Ravn, "A method to perform probabilistic
V. CONCLUSION production simulation involving combined heat and power units," IEEE
Trans. on Power Syst., vol. 11, no. 2, pp. 1031-1036, May 1996.
This work proposes an optimization model-based reliability [19] Y. Tang, “Power distribution system planning with reliability modeling
assessment method for distribution networks considering and optimization,” IEEE Trans. Power Syst., vol. 11, no. 1, pp. 181–189,
detailed placement and actions of circuit breakers and switches. Feb. 1996.
[20] H. Li, Z. Lu, Y. Qiao and N. Wang, "A Non-Sequential Probabilistic
It is specifically applied to a planning model of circuit breakers Production Simulation Method for Wind Energy Curtailment Evaluation
and switches of distribution networks. As verified in case Considering the Seasonal Heat Supply Constraints," IEEE Transactions
studies, compared to the simulation-based methods, our method on Sustainable Energy, vol. 9, no. 1, pp. 462-473, Jan. 2018.
[21] H. L. Willis and J. E. D. Northcote-Green, "Spatial electric load
uses much less computation time but obtains same results. forecasting: A tutorial review," Proceedings of the IEEE, vol. 71, no. 2,
Numerical tests show that the configuration of circuit breakers, pp. 232-253, Feb. 1983.
switches and tie-lines significantly affect system’s reliability. [22] E. Diaz-Dorado, J. Cidras and E. Miguez, "Application of evolutionary
algorithms for the planning of urban distribution networks of medium
Case studies for the planning model show that, with the voltage," IEEE Trans. on Power Syst., vol. 17, no. 3, pp. 879-884, Aug.
improvement of reliability requirement, the investment cost of 2002.
circuit breakers and switches rapidly increases. It’s worth [23] I. Ziari, G. Ledwich, A. Ghosh, and G. Platt, “Optimal distribution
network reinforcement considering load growth, line loss, and reliability,”
studying the data-driven reliability assessment method based on
IEEE Trans. Power Syst., vol. 28, no. 2, pp. 587–597, May 2013.
historical outage data for future work. [24] G. Muñoz-Delgado, J. Contreras and J. M. Arroyo, "Reliability
Assessment for Distribution Optimization Models: A Non-Simulation-
VI. REFERENCES Based Linear Programming Approach," in IEEE Transactions on Smart
Grid, vol. 9, no. 4, pp. 3048-3059, July 2018.
[1] H. L. Willis, Power Distribution Planning Reference Book, 2nd ed. New [25] G. Muñoz-Delgado, J. Contreras and J. M. Arroyo, "Distribution Network
York, NY, USA: Marcel Dekker, 2004. Expansion Planning With an Explicit Formulation for Reliability
[2] A. A. Chowdhury and D. O. Koval, Power Distribution System Reliability:

0885-8950 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on June 02,2020 at 03:49:40 UTC from IEEE Xplore. Restrictions apply.
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/TPWRS.2020.2981508, IEEE
Transactions on Power Systems
13

Assessment," IEEE Trans. on Power Syst., vol. 33, no. 3, pp. 2583-2596, [46] N. M. M. Razali and A. H. Hashim, “Backward reduction application for
May 2018. minimizing wind power scenarios in stochastic programming,” in Proc.
[26] Z. Li, W. Wu, B. Zhang and X. Tai, "Analytical Reliability Assessment PEOCO, Shah Alam, Malaysia, 2010, pp. 430–434.
Method for Complex Distribution Networks Considering Post-fault
Network Reconfiguration," in IEEE Transactions on Power Systems.
Early Access, 2019.
[27] S. Heidari, M. Fotuhi-Firuzabad and M. Lehtonen, "Planning to Equip the Zihao Li (S’18) received the B.S. degree in electrical
Power Distribution Networks with Automation System," IEEE Trans. on engineering from Tsinghua University, Beijing,
Power Syst., vol. 32, no. 5, pp. 3451-3460, Sept. 2017. China, in 2016. He is currently working toward the
[28] Y. Tian, M. Benidris, S. Sulaeman, S. Elsaiah and J. Mitra, "Optimal Ph.D. degree in the Department of Electrical
feeder reconfiguration and distributed generation placement for reliability Engineering, Tsinghua University. His research
improvement," 2016 International Conference on Probabilistic Methods interests focus on the reliability assessment and
Applied to Power Systems (PMAPS), Beijing, 2016, pp. 1-7. planning of distribution system.
[29] J. C. López, M. Lavorato, and M. J. Rider, “Optimal reconfiguration of
electrical distribution systems considering reliability indices
improvement,” Int. J. Electr. Power Energy Syst., vol. 78, pp. 837–845,
Jun. 2016.
[30] S. Elsaiah, M. Benidris, Y. Tian and J. Mitra, "A Comprehensive Analysis Wenchuan Wu (SM’14) received the B.S., M.S.,
of Reliability-oriented Optimal Distribution System and Ph.D. degrees from the Electrical Engineering
Reconfiguration," 2018 IEEE Industry Applications Society Annual Department, Tsinghua University, Beijing, China.
Meeting (IAS), Portland, OR, 2018, pp. 1-9. He is currently a Professor and director of electric
[31] A. Nader, C. Ruben, S. C. Dhulipala, A. S. Bretas and R. A. da Silva, power systems research institute in the Department
"MILP Model for Reliability Optimization In Active Distribution of Electrical Engineering of Tsinghua University and
Networks," 2018 North American Power Symposium (NAPS), Fargo, ND, Deputy Director of the State Key Laboratory of
2018, pp. 1-6. Power Systems. His research interests include
[32] S. M. Hashemi, A. Fereidunian, H. Mirsaeedi and H. Lesani, "Optimal Energy Management System, active distribution
placement of normally open switches for distribution automation in smart system operation and control, machine learning and
grid," 2017 Smart Grid Conference (SGC), Tehran, 2017, pp. 1-6. its application in energy system.
[33] I. Lim et al., "Design and Implementation of Multiagent-Based Prof. Wu was a recipient of the National Science Fund of China
Distributed Restoration System in DAS," in IEEE Transactions on Power Distinguished Young Scholar Award in 2017. He has published 400 peer
Delivery, vol. 28, no. 2, pp. 585-593, April 2013. reviewed papers including 60 IEEE journal papers. He is an IET fellow and also
[34] A. Pregelj, M. Begovic and A. Rohatgi, "Recloser allocation for improved serves as an Associate Editor of the IET Generation, Transmission &
reliability of DG-enhanced distribution networks," in IEEE Transactions Distribution and IET Energy Systems Integration.
on Power Systems, vol. 21, no. 3, pp. 1442-1449, Aug. 2006.
[35] S. Kazemi, M. Fotuhi-Firuzabad and R. Billinton, "Reliability assessment Xue Tai (S’18) received the B.S. and M.S. degrees
of an automated distribution system," in IET Generation, Transmission & from the Electrical Engineering Department,
Distribution, vol. 1, no. 2, pp. 223-233, March 2007. Tsinghua University, Beijing, China, in 2016 and
[36] A. Abiri-Jahromi, M. Fotuhi-Firuzabad, M. Parvania and M. Mosleh, 2018. She is currently an engineer in State Grid
"Optimized Sectionalizing Switch Placement Strategy in Distribution Shanghai Municipal Electric Power Company,
Systems," in IEEE Transactions on Power Delivery, vol. 27, no. 1, pp. Shanghai. Her research interests include applications
362-370, Jan. 2012. of block chain in power system and reliability analysis
[37] Y. Xu, C. Liu, K. P. Schneider and D. T. Ton, "Placement of Remote- of distribution system.
Controlled Switches to Enhance Distribution System Restoration
Capability," in IEEE Transactions on Power Systems, vol. 31, no. 2, pp.
1139-1150, March 2016.
[38] L. S. de Assis, J. F. V. González, F. L. Usberti, C. Lyra and F. Von Zuben, Boming Zhang (SM’95–F’10) received the Ph.D. degree in
"Optimal allocation of remote controlled switches in radial distribution electrical engineering from Tsinghua University, Beijing,
systems," 2012 IEEE Power and Energy Society General Meeting, San China, in 1985. Since 1985, he has been with the Electrical
Diego, CA, 2012, pp. 1-8. Engineering Department, Tsinghua University, promoted to
[39] N. N. Mansor and V. Levi, "Integrated Planning of Distribution Networks a Professor in 1993. His interest is in power system analysis
Considering Utility Planning Concepts," IEEE Trans. on Power Syst., vol. and control, especially in the EMS advanced applications in
32, no. 6, pp. 4656-4672, Nov. 2017. the Electric Power Control Center (EPCC).
[40] I. Hernando-Gil, I. Ilie and S. Z. Djokic, "Reliability planning of active
distribution systems incorporating regulator requirements and network-
reliability equivalents," in IET Generation, Transmission & Distribution,
vol. 10, no. 1, pp. 93-106, 7 1 2016.
[41] L. Gan, N. Li, U. Topcu and S. H. Low, "Exact Convex Relaxation of
Optimal Power Flow in Radial Networks," in IEEE Transactions on
Automatic Control, vol. 60, no. 1, pp. 72-87, Jan. 2015.
[42] X. Chen, W. Wu, B. Zhang and C. Lin, "Data-Driven DG Capacity
Assessment Method for Active Distribution Networks," in IEEE
Transactions on Power Systems, vol. 32, no. 5, pp. 3946-3957, Sept. 2017.
[43] T. Ding, Y. Lin, G. Li and Z. Bie, "A New Model for Resilient
Distribution Systems by Microgrids Formation," in IEEE Transactions on
Power Systems, vol. 32, no. 5, pp. 4145-4147, Sept. 2017.
[44] Z. Li, W. Wu, B. Zhang, X. Tai. Test bench of Optimization Model-based
Reliability Assessment for Distribution Networks Considering Detailed
Placement of Circuit Breakers and Switches. Accessed on 4, Sept., 2019.
[Online]. Available:
https://drive.google.com/open?id=1HlsK9BThu5UIIMCV06VeVHK_zk
etBE-u
[45] R. C. Lotero and J. Contreras, “Distribution system planning with
reliability,” IEEE Trans. Power Deliv., vol. 26, no. 4, pp. 2552–2562, Oct.
2011.

0885-8950 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on June 02,2020 at 03:49:40 UTC from IEEE Xplore. Restrictions apply.

You might also like