2014 International Conference on Power System Technology (POWERCON 2014) Chengdu, 20-22 Oct.

2014

Reliability Evaluation for Distribution System

Connected with Distributed Generations
Hong-shan ZHAO, Member, IEEE, Hang-yu ZHAO, and Song DENG

 probability model based on the historical load data, but it do not

Abstract—This paper presents a reliability index calculation consider the correspondence of DG output and load size on the
method for distribution system connected with Distributed time. Reference [6], [7] set up the island partition model with
Generations (DGs). First, the sequence model of DGs and load are the objective function of the largest equivalent load according
established on the assumption that the time unit of power changes
is an hour. Then, on the basis of the concept of segmentation, four
to the important degree of each load point in distribution system,
types of data structures are presented to represent the topological but the uncertainty of load is not considered. Reference [8]
structure and parameters of the distribution system. According to establishes the DG model and load model based on historical
the differences of each segment affected by the fault, the statistics data and calculates the reliability indices with
distribution system will be divided into six types of areas, and the analytical method. Reference [9] shows two kinds of fault
outage time of each load point are determined by the result of area impact matrix and calculates the reliability indices with time
division and load optimization in the island. At last, the
synchronous sampling of DG output and load size is realized by
sequential Monte Carlo method.
time sequential Monte Carlo simulation and the reliability indices This paper establishes the sequence model of DGs and the
are calculated. load on the assumption that the basic unit for power changes is
an hour, divides the system into different types of area on the
Index Terms—distribution system; DG; reliability evaluation; conception of segmentation, and realizes the synchronous
time sequential Monte Carlo simulation sampling of DG output and load size by time sequential Monte
Carlo simulation. Thus the proposed method improves the
rationality of reliability evaluation. The data structures
I. INTRODUCTION proposed in this paper can also bring great convenience to the

W ITH the flexible and environment-friendly features,

more and more distributed generations will be embedded
into the power distribution system. The penetration of DGs will
reliability evaluation.

II. SYSTEM MODELLING

bring a series of impacts on the structure and operation of
A. Wind Turbine Generator Model
distribution system, and it will also change the original
reliability evaluation method [1-2]. Under the condition of the standard air density, the curve
At present, scholars have carried out many studies on the between the output power of WTG and wind speed is called the
reliability evaluation method of distribution system connecting standard power curve of the WTG, as shown in Fig.1.
with DGs, and some articles have been published. Reference [3]
calculates the reliability indices of distribution system
connected with DGs based on interval computation, but the
output of DGs and the load size are treated as constant values,
thus the uncertainty of DG and load is not considered.
Reference [4], [5] establish the wind turbine generator (WTG)
model based on the historical wind speed data and the power
characteristic curve of WTG and establish the load state

This work was supported in part by the China SNF under Grant 51077053
and the project-sponsored by SRF for ROCS, SEM, China.
Hong-shan ZHAO is with North China Electric Power University, Baoding,
Hebei 071003 China (e-mail:[email protected]).
Hang-yu ZHAO is with North China Electric Power University, Baoding, Fig.1. Standard power curve of the WTG
Hebei 071003 China (e-mail:[email protected]).
Song DENG is with North China Electric Power University, Baoding, Hebei
071003 China (e-mail:[email protected]). The mathematic expression of the relationship between

POWERCON 2014 Paper No CP2126 Page 1/7 1467 Session 2

 2014 International Conference on Power System Technology (POWERCON 2014) Chengdu, 20-22 Oct. 2014

output power of the WTG and wind speed can be shown in (1), weekly load in terms of the annual peak, Pd is the percentage
­ 0ˈ 0 d V d Vci
° of daily load in terms of the weekly peak and Ph (t ) is the
°° V  Vci , Vci d V d Vr percentage of hourly load in terms of daily peak.
P ®Vr  Vci (1)
° P, Vr d V d Vco III. FAILURE MODE EFFECTS ANALYSIS OF THE
r
° DISTRIBUTION SYSTEM
°¯ 0, V t Vco
A. Data Structure
where, P and Pr are the output power and the rated power of
The structure of distribution system is complex and its
WTG. V , Vci , Vr and Vco are the wind speed, the cut-in speed, elements are numerous. The connection of DGs not only further
the rated speed and the cut-out speed respectively. increases the complexity of the distribution system, but also
changes the original radial distribution system operation mode.
B. Solar Cell Generator Model
In order to analyze the failure mode effects of the distribution
Fig.2 shows the typical solar cell generator (SCG) power output system, data structures are established based on the concept of
model presenting a relationship between the power output and the segmentation.
the solar radiation [10].
The segmentation concept means that the distribution system
will be modeled in terms of segments, not components. A
segment is a group of components whose entry component is a
switch or a protective device, and each segment will have only
one switch or protective device [12]. When one component in
the segment breaks down, the other components in the segment
will also lose power supply until the fault component has been
restored. Due to the segments have common entry components,
when one component outside the segment breaks down, the
interruption duration of components in the segment is the same.
This means that all the customers in any segment have the same
reliability level and can be treated as one customer. By
definition, the distribution system shown in figure 3 can be
divided into six segments.
Fig.2 Power output model of SCG

Equation (2) is the general mathematical expression for the

power curve of SCG,
­ Psn (Gbi2 / (Gstd Rc )), 0 d Gbi  Rc
°
Pb ® Psn (Gbi / Gstd ), Rc d Gbi  Gstd (2)
° Psn , Gbi t Gstd
¯
where, Pb and Psn are the output power and the rated power of
SCG. Gbi , Rc and Gstd are the solar radiation, a certain
radiation in the standard environment set, and the rated solar
radiation respectively.
C. Load Model
A detailed customer load profile varies with the type of
customer, the location and time of the day, the day of the week
and the week of the year. This paper uses the model presented Fig.3 A sample of dividing the segment
in [11] to determine the load for hour t using the following
In order to analyze the reliability of distribution system
equation:
connected with DGs, four types of data structure are presented,
L(t ) Ly u Pw u Pd u Ph (t ) (3) namely segment data structure, DG data structure, component
where Ly is the annual peak load, Pw is the percentage of data structure and load point data structure, and they are used
to represent the topological structure and parameters of the

POWERCON 2014 Paper No CP2126 Page 2/7 1468 Session 2

 2014 International Conference on Power System Technology (POWERCON 2014) Chengdu, 20-22 Oct. 2014

distribution system, as shown in Fig.4, where FR is the TABLE I

SEGMENT DATA STRUCTURE
abbreviation for “failure rate” and MTTR is the abbreviation
for “mean time to repair”. Number Code DG Number
1 [1,1,2,1,1] 0
2 [1,2,2,1,1,1,1,1] 1
3 [1,2,2,1,1,3,1,2] 0
4 [1,3,2,1,1,1,1,1,1,1,1] 0
5 [1,3,2,1,1,1,1,1,1,1,2] 0
6 [1,4,2,1,1,1,1,1,1,1,1,1,1,1] 2

The DG data structure is shown in Fig.4 (b), and it includes

four fields, number, FR, MTTR and power. The power field
represents the length of 8760 annual instantaneous output of
DG calculated by (1) or (2).
The component data structure is shown in Fig.4 (c), and it
Fig.4 Data structure diagram of the segment, the DG, the element and the load
includes four fields, number, segment number, FR, and MTTR.
Components usually refer to feeder, distribution transformer,
The segment data structure is shown in Fig.4 (a), and it circuit breaker, isolating switch and fuse. The feeder and
includes three fields, number, code, and DG number. The distribution transformer belong to only one segment. The
number field represents the number of the segment and the DG circuit breaker, isolating switch and fuse belong to the two
number field stands for the number of the DG in the segment. If segments directly connected with them.
the segment does not connect with DG, the field of DG number The load point data structure is shown in Fig.4 (d), and it
is ’0’. includes six fields, number, segment number, peak,
The code field reflects the location of the segment in the instantaneous, level and customers. The peak field represents
network topology [13]. Code diagram is shown in Fig.5, the peak load and the instantaneous field stands for the length of
8760 annual instantaneous value of the load calculated by (3).
The customer field represents the number of users in the load
point.
The four types of data structure form a whole by associated
fields. By the four types of data structure, it can be very
convenient to conduct failure mode effects analysis.
B. System Division
Fig.5 Code diagram of the segment
According to the differences of each segment affected by the
fault, the distribution system will be divided into six types of
areas: they are fault area, upstream island area, upstream
where, the interval number field represents which main feeder isolated area, downstream seamless island area, downstream
interval the segment belongs to, the layer number field isolated area and unaffected area respectively.
represents which layer the segment belongs to, the connection Fault area (A area): The area is the segment that the fault
relationship field represents the connection mode with the component belongs to, thus the area cannot get power supply
upper segment, ‘1’ stands for isolating switch, ‘2’ stands for form the main power or DG until the fault component be
circuit breaker and ‘3’ stands for fuse. The branch number field repaired.
refers to the branch number of the segment and it’s used to Downstream seamless island area (B1 area): The area is
distinguish the parallel segment in the same layer. The grade located in the downstream of the fault area. When fault occurs,
number field refers to grade number of the segment and it’s the area immediately runs on island state.
used to distinguish the series segment in the same layer. Downstream isolated area (B2 area): The area is located in
By definition, the segment data structure in Fig.3 can be the downstream of the fault area. When fault occurs, the area
shown in Table I. cannot immediately run on island state.
Upstream island area (C1 area): The area is located in the
upstream of the fault area. When fault occurs, the area
immediately runs on island state.

POWERCON 2014 Paper No CP2126 Page 3/7 1469 Session 2

 2014 International Conference on Power System Technology (POWERCON 2014) Chengdu, 20-22 Oct. 2014

Upstream isolated area (C2 area): The area is located in the system by operating circuit breaker. DG provides power for the
upstream of the fault area. When fault occurs, the area cannot load points in the island continuously. Because of the
immediately run on island state. uncertainty of the DG output power and load power, not all the
Unaffected area (D area): The area is the segments that do loads in the island can get power supply. During the island, in
not affected by the fault. order to reduce the switching frequency, the output of DG is
The B1 and B2 area are collectively known as B area, and the treated as a constant and the value is the minimum output power
C1 and C2 area are collectively known as C area. Adopt the during the island. The load power is also treated as a constant
protection logic in [14]: “the reclosers are coordinated in such a and the value is the maximum load power during the island.
way that only the minimal number of neighboring reclosers The goal of load optimization is largest the sum of the product
operate and isolate the fault.” When a component in the of load size and load weight coefficient, as shown below:
distribution system breaks down, according to the segment data max L max ¦ Zi u La (i ) (4)
structure and protection logic, the procedure to divide the iD
distribution system into the six types of areas is as follows:
--Step 1: Find out the segment that the fault component
s.t. ¦ La(i) d P
iD
DG (5)

belongs to. Use m to represents the code of the segment. Then, where, La (i ) is the power of load point i , Zi is the weight
find out the interval of the segment, use set S to represent all
segments in the fault interval, other segments not in the fault coefficient of load point i , PDG is the total DG output power in
interval belong to D area. the island, D is the greatest possible connected area in the
--Step 2: Traverse each segment in set S. island.
--Step 3: If the code of segment is the same as m, the
D. OUTAGE TIME
segment belongs to A area; If the layer of the segment code is
one more than m and the one more layer is connected through For the load point not in the island, the outage time directly
fuse, the segment belongs to A area. depends on the type of area. For the load point in the island, the
--Step 4: If the layer of the segment code is more than the outage time depends on the type of area and island load
layer of m and the code inherits m completely, the segment optimization. The details are as follows:
belongs to B area. If the layer of the segment code is the same A area: the outage time is TTR.
as the layer of m, and all the code is the same as m except that D area: the outage time is 0.
the last bit is greater than the last bit of m, the segment belongs C2 area: the outage time is fault isolation time (FIT).
to B area. C1 area: If load point is selected, the outage time is 0.
--Step 5: Find out the circuit breaker or fuse that located Otherwise, the outage time is FIT.
upstream and is closest to the fault component, use k to B1 area: If load point is selected, the outage time is 0.
represents the layer number. If the code of the segment is the Otherwise, delimit the segment into B2 area to determine the
same as m from the 3rd bit to the 3k+1 bit, the segment belongs outage time.
to C area. B2 area: Optimize the load points in B2 area and the load
--Step 6: The rest of the segments do not meet the points not selected in B1 area. If load point is selected, the
condition in step 3, 4 and 5 in set S belong to D area. outage time is FIT. Otherwise, the outage time is TTR.
--Step 7: For all the segments in area B, group them
according to the connectivity. When the entry component of the IV. DISTRIBUTION SYSTEM RELIABILITY
group is circuit breaker, all the segments in the group belong to EVALUATION METHOD
B1 area. Otherwise, all the segments in the group belong to B2 A. Reliability Indices
area. Reliability indices consist of load point indices and system
--Step 8: For all the segments in area C, group them indices. Load point indices include failure rate ( O ), outage
according to the connectivity. All the segments in the group time (r), annual unavailability (U) and energy not supplied
which layer of code is the smallest belong to C2 area. For the (ENS). System indices include system average interruption
left groups, when the entry component is circuit breaker, all the frequency index (SAIFI), system average interruption duration
segments in the group belong to C1 area. Otherwise, all the index (SAIDI), energy not supplied (ENS) and average service
segments in the group belong to C2 area. availability index (ASAI), etc.
C. Island Load Optimization Model B. Component State Sampling
When the distribution system works in the normal state, DG In this paper, time sequential Monte Carlo simulation
and substation provide power for loads together. When fault method was adopted to realize the reliability assessment.
occurs, isolate the subnet containing DG with main distribution Component state sampling is based on the two state model.

POWERCON 2014 Paper No CP2126 Page 4/7 1470 Session 2

 2014 International Conference on Power System Technology (POWERCON 2014) Chengdu, 20-22 Oct. 2014

Assuming the component failure rate and repair time all submit percentage of weekly load in terms of the annual peak, the
to exponential distribution, the time to failure (TTF) and time to percentage of daily load in terms of the weekly peak and the
repair (TTR) can be calculated by (6) and (7), percentage of hourly load in terms of daily peak see [13].
1 SDG parameters: rated power is 1 MW, Rc is 0.15 kW/m2,
TTF  ln E1 (6)
O Gstd is 1 kW/m2. The solar radiation data is generated by the
TTR r ln E2 (7) HOMER software [17].
where O and r are the average failure rate and mean time to
repair respectively, E1 and E 2 are random numbers of uniform
distribution on the interval [0, 1].
C. Time sequential Monte Carlo Simulation Procedure
For a distribution system with n elements, the specific steps
of utilizing time sequential Monte Carlo simulation method to
calculate reliability indices are as follows:
--Step 1: Input the time need to simulate, initialize the
system simulation time.
--Step 2: Generate a random number on the interval [0, 1]
for each component in the system and convert it into
TTFi (i 1,2,, n) corresponding to (6).
--Step 3: Determine component i with the minimum TTF,
and it represents component i breaks down first.
--Step 4: Generate another random number on the interval
[0, 1] for component i and convert this number into time to
repair (TTR) according to (7).
--Step 5: According to the failure mode and effect analysis,
optimize the load points in the island. Count the outage time,
times of outage, energy not supplied in each load points when
component i breaks down.
--Step 6: System simulation time move
forward TTFi  TTR .
--Step 7: Determine whether simulating time reach the
specified simulating time. If it reaches, calculate each load
Fig.6 Connection diagram of distribution system
point and system reliability indices. Otherwise, return to step 2
to continue simulating.
TABLE II
V. CASE STUDIES RELIABILITY DATA OF THE COMPONENTS
Case studies have been conducted using a small reliability Components Failure Rate MTTR(h)
test system based on the IEEE RBTS BUS6. The test system is Feeder 0.05 f/km·yr 5
composed of 30 feeders, 23 distribution transformer and 23
Distribution Transformer 0.015 f/yr 20
load points. Connect a WTG to the system at feeder 18, a SDG
to the system at feeder 30, as shown in Fig.6.
TABLE III
Assuming the circuit breaker action absolutely reliable, the FEEDER LINE DATA
fault isolation time (FIT) is 1h. Reliability data of the
Length(km) Feeder Number Length(km) Feeder Number
components is shown in Table II, feeder line data is shown in
Table III, load point data is shown in Table IV, the weight 0.6 7,13 2.5 2,6,18,23,26
coefficient of load level 1, 2, and 3 are 0.5, 0.3, and 0.2 0.75 27 2.8 1,12,16,22,25,30
respectively. 0.8 9,21 3.2 11,17,19,24,29
WTG parameters: rated power is 1.1 MW, cut-in speed is
0.9 4,10 3.5 14
2.5m/s, rated speed is 12.5m/s, cut-out speed is 25m/s, the wind
speed can be generated by the ARMA model in [16]. The 1.6 3,5,8,15,20,28 — —

POWERCON 2014 Paper No CP2126 Page 5/7 1471 Session 2

 2014 International Conference on Power System Technology (POWERCON 2014) Chengdu, 20-22 Oct. 2014

TABLE IV Case2: Connect a WTG to the system at feeder 18, a SDG to

LOAD POINT DATA
the system at feeder 30, as in Fig.6. The output of DGs is
Number of Customer calculated by the model and parameters as mentioned above.
Number Total Peak(kW)
Customers Level Case3: Connect 2 DGs at feeder 18 and 30 with unlimited
2 1 322.9 126 2 capacity.
1,6 2 296.4 147 2 The simulation is performed for a period of 10000 years.
5 1 369.8 132 1 Load point average failure rate index is shown in Fig.7. The
8,11,14,19 4 277.6 79 3 reliability indices of some load points for all cases are shown in
10,12,16,22 4 283.1 76 2 Table V. Table VI shows the system indices.
15,20 2 502.5 1 2
3,13,17 3 651.7 1 3
4,18 2 686.0 1 3
7,23 2 796.5 1 1
9,21 2 737.5 1 1
Total 23 4924.0 565

In order to analyze the impact that DGs will bring on the

reliability of distribution system, the test system is evaluated
for three cases.
Case1: The system is not connected with DGs.
Fig.7 Load point average failure rate index

TABLE V
RELIABILITY INDICES OF SOME LOAD POINTS
Case1 Case2 Case3
Load Point
O (f/yr) r (hr) U(hr/yr) ENS(kW/yr) O (f/yr) r (hr) U(hr/yr) ENS(kW/yr) O (f/yr) r (hr) U(hr/yr) ENS(kW/yr)
LP1 1.5182 3.3923 5.1503 668.67 1.5325 3.4120 5.2289 682.00 1.5160 3.3588 5.0920 664.68
LP8 1.5182 5.0150 7.6138 923.28 1.5325 4.8724 7.4670 920.87 1.5160 2.6216 3.9744 485.06
LP11 1.9924 5.0150 9.9933 1213.82 0.9583 6.4526 6.1835 817.86 0.4667 4.9388 2.3049 279.85
LP14 2.2840 3.9173 9.9933 1088.74 2.3009 3.9688 9.1318 1115.63 2.3058 3.9091 9.0137 1102.97
LP19 2.2586 4.9975 11.2875 1370.54 1.7437 5.2464 9.1671 1158.96 0.7442 5.0432 3.7531 457.21

improve each system reliability index, the greater the DG

TABLE VI
RELIABILITY INDICES OF THE SYSTEM
outputs, the greater the system reliability improved.

Case SAIFI SAIDI ENS(MW) ASAI VI. CONCLUSION

Case1 1.7819 7.4406 39.573 0.9992 This paper presents four types of data structures to represent
Case2 1.5218 6.2041 32.362 0.9993 the topological structure and parameters of the distribution
Case3 1.3800 4.9221 24.976 0.9994 system, determines the power outage time of each load point by
the result of area division and load optimization, and calculates
From Fig.7, it can be seen that the connection of DG can the reliability indices of a small reliability test system. By
reduce the failure rate of load point in the segment where DG is comparing the reliability indices before and after connecting
accessed; the failure rate of load point in other segments DGs to the test system, the role that DGs can play in the
affected by the DG is not obvious. improvement of distribution system reliability is verified. The
By comparing the load point indices under each case in Table conditions in this paper such us protection logic and load
V, it can be seen that connection of DG can reduce the failure optimization depends on the realization of the distribution
rate, outage time, annual unavailability and energy not supplied system automation. Thus, the connection of DG also challenges
of the load point in the segment where DG is accessed; for the smartness of distribution system.
load point in other segments, though the failure rate is not
reduced, the outage time, annual unavailability and energy not
supplied is reduced obviously.
From Table VI, it can be seen that the connection of DG can

POWERCON 2014 Paper No CP2126 Page 6/7 1472 Session 2

 2014 International Conference on Power System Technology (POWERCON 2014) Chengdu, 20-22 Oct. 2014

REFERENCES [10] Jeongje Park, Wu Liang, Jaeseok Choi, et al, “A Probabilistic Reliability
Evaluation of a Power System Including Solar/Photovoltaic Cell
[1] BIE Zhaohong, LI Gengfeng, and WANG Xifan, “Review on reliability Generator”, Proceedings of IEEE Power & Energy Society General
evaluation of new distribution system with micro-grid,” Electric Power Meeting, pp.26-30, July, 2009.
Automation Equipment, vol.31, no.1, pp.1-6, Jan. 2011. [11] Peng Wang, and Roy Billinton, “Time Sequential Distribution System
[2] TIAN Hongxun, YUAN Rong, and ZHAO Yuan, “Computation of Reliability Worth Analysis Considering Time Varying Load and Cost
Probability Distribution for Reliability Indices of Distribution System Models,” IEEE Transactions on Power Delivery, vol.14, no.3,
Containing Distributed Generation,” Power System Technology, vol.37, pp.1046-1051, July. 1999.
no.6, pp.1562-1569, Apr. 2013. [12] Y.M.Atwa, E.F. EI-Saadany, and Anne Claire Guise, “Supply Adequacy
[3] QIAN Ke-jun, YUAN Yue, and ZHOU Cheng-ke “Study on Impact of Assessment of Distribution System Including Wind-based DG During
Distributed Generation on Distribution System Reliability,” Power Different Modes of Operation,” IEEE Transaction on Power Systems,
System Technology, vol.32, no.11, pp.74-78, Nov. 2008. vol.35, no.1, pp.78-86, Feb. 2010.
[4] Atwa Y M, and El-Saadany E F, “Reliability evaluation for distribution [13] WANG Haoming, “Reliability Evaluation of Distribution System
system with renewable distributed generation during islanded mode of Including Distributed Generations,” D.E. dissertation, Tianjin University,
operation,” IEEE Transcations on Power Systems, vol.24, no.2, Tianjin, 2012.
pp.572-581, May. 2009. [14] Aleksandar Pregelj, Miroslav Begovic, and Ajeet Rohatgi, “Recloser
[5] Xu Yuqin, and WU Yingchao, “Reliability Evaluation for Distribution Allocation for Improved Reliability of DG-enhanced Distribution
System Connected with Wind-Turbine Generators,” Power System Network,” IEEE Transaction on Power Systems, vol.21, no.3,
Technology, vol.35, no.4, pp.154-158, Apr. 2011. pp.1442-1449, Aug. 2006.
[6] LIU Chuanquan, and ZHANG Yan, “Distribution network reliability [15] R. Billinton, and S. Johnnavithula, “A Test System for Teaching Overall
considering distribution generation,” Automation of Electric Power Power System Reliability Assessment,” IEEE Transactions on Power
Systems, vol.31, no.22, pp.46-49, Nov. 2007. Systems, vol.11, no.4, pp.1670-1676, Nov. 1996.
[7] LIU Chuanquan, “The distribution network reliability assessment [16] P. Wang, and R. Billinton, “Time-sequential simulation technique for
considering distribution generation,” Master of Engineering dissertation, rural distribution system reliability cost/worth evaluation including wind
Shanghai Jiao Tong University, Shanghai, 2008. generation as alternative supply,” IEE Proceedings: Generation,
[8] In-Su Bae, and Jin-O Kim, “Reliability evaluation of distributed Transmission and Distribution, vol.148, no.4, pp.355-360, 2001.
generation based on operation mode,” IEEE Transactions on Power [17] Lambert T, Gilman P, Lilienthal P, “Micropower system modeling with
Systems, vol.22, no.2, pp.785-790, May. 2007. HOMER,” Wiley-IEEE Press, pp.279-418, New York, 2006.
[9] SU Ao-xue, FAN Ming-tian, LI Zhong-lai, and LIU Wei, “Reliability
evaluation of distribution system considering wind power effect,” Power
System Protection and Control, vol.41, no.1, pp.90-95, Jan. 2013.

POWERCON 2014 Paper No CP2126 Page 7/7 1473 Session 2