ICSET 2008

Development of Fault Location Algorithm for Distribution Networks

with DG
Zhu Guo-fang, Lu Yu-ping, Senior Member, IEEE

Abstract—This paper proposes a novel fault location signal, is sent to the master system by FTUs. And the master
algorithm of distribution automation master system for urban system implements the functions such as fault location,
distribution system with distributed generation (DG). A isolation and restoration. Conventional power distribution
differential current ratio (DCR) is constructed in the paper system is radial in nature, characterized by a single source
which contains the differential current information to describe
feeding a network of downstream feeders. The distribution
the feature of faulted segment and differential current
information of normal one. The faulted segment would have a automation fault location has traditionally been designed
DCR that the value is less than one while the normal one would assuming the system to be radial. The conventional fault
have a DCR far larger than one. Basing on the distribution location algorithm, given by paper [3, 4], is suitable for fault
feature of fault current, that the largest fault current would be location with single source radial supply line or single fault
sensed by one of FTUs in the faulted segment, an optimized with multi-source open loop operation.
searching process is given. The core of the fault location
With the introduction of DGs, the distribution system in
algorithm is to search the segment with the smallest DCR. This
method uses only currents and thus avoids the need for near future would be regarded as a multisource system. It is a
installation of voltage transformers. Furthermore, the method well-established fact that fault location in a multisource
does not require time synchronized measurement and therefore, system has to be direction sensitive. There are two types of
more practical and economical to implement on existing directional detecting means available: One using the power
distribution system. The PSS/E simulation results indicate the frequency component and the other employing the
method could locate faulted segment correctly in urban
traveling-wave approach [5]. The problem with latter means
distribution network with DG.
is the short duration available for decision making. Further, if
Index Terms-- distributed generation; distribution a fault occurs at a voltage inception angle close to zero degree,
automation; feeder terminal unit; fault location; differential there will be no appearance of a traveling wave. These two
current ratio; PSS/E aspects greatly affect the accuracy and reliability of the
approach and may cause misoperation. It is apparent from the
previous discussions that a reliable directional detector needs
I. INTRODUCTION the power frequency components of the bus voltages and the

D STRIBUTED generation(DG) is referred to the

generation which is of limited size(few kilowatts to few
megawatts) and interconnected at substation,
line currents availed through different sensors. But for a
distribution automation scheme, such sensors are required in
large numbers. An alternative scheme using fewer sensors
distribution feeder or customer load level[1]. DG will be an attractive choice from an investment point of view,
technologies include renewable energy resources like particularly in the emerging DG system.
mini-hydro, wind turbines and photovoltaic as well as cleaner The paper is devoted to hunt a new fault location algorithm
technologies such as micro turbines, gas turbines and internal only using fault current amplitude information for urban
combustion engines and fuel cells [1]. Connection of DG distribution network with DG. A differential current ratio
fundamentally alters distribution network operation and (DCR) is constructed in the paper which contains the
creates a variety of impacts with relay protection, reclosing differential current information to describe the feature of
and distribution automation (DA) based on feeder terminal faulted segment and differential current information of
unit (FTU) [1, 2].
normal one. The faulted segment would have a DCR that the
Currently, the distribution automation solution, based on
value is less than one while the normal one would have a
feeder terminal unit (FTU), has been successfully applied in
DCR far larger than one. This algorithm also bases on an
urban distribution network. In the scheme FTUs, installed on
assumption: there is a single fault at same time. The core of
the disconnect switches, play the role of the over-current fault
the fault location algorithm is to search the segment with the
detector. The fault information, such as over-current fault
smallest DCR. This method uses only currents and thus
avoids the need for installation of voltage transformers.
Manuscript received May 30, 2008. This work was supported by National Furthermore, the method does not require time synchronized
Natural Science Foundation of China (NSFC) (50577006).
Zhu Guo-fang is with School of Electrical Engineering, Southeast measurement and therefore, more practical and economical to
University, Nanjing, Jiangsu (210096), China. (Tel: +86-25-83791817; Fax: implement on existing distribution system. Basing on the
+86-25-83793371; E-mail: [email protected].) distribution feature of fault current, that the largest fault
Lu Yu-ping is with School of Electrical Engineering, Southeast
University, Nanjing, Jiangsu (210096), China. (Tel: +86-25-83791817; Fax:
current would be sensed by one of FTUs in the faulted
+86-25-83793371; e-mail: [email protected] ). segment, the searching process could also be optimized.

164
978-1-4244-1888-6/08/$25.00 
c 2008 IEEE
 Rest of the paper is organized as follows. Section II the complex current is negative component and negative
introduces the proposed scheme in general. Section III gives component is steady fault component, it is also insensitive to
the definition of differential current ratio and Sensitivity load current.
analysis of the algorithm is outlined in section IV. Section V Based on the current complex information from FTUs, the
provides an optimized searching process. Simulation master system calculates the DCR of each segment and
approaches and results are presented in section VI. Section locates the faulted segment by using DCR information. The
VII is a summary and makes conclusions. details about DCR and fault locating procedure are given in
the following sections.
II. PROPOSED SCHEME
The distribution system is divided into a number of III. DEFINITION OF DIFFERENTIAL CURRENT RATIO
network segments as shown in Fig.1 for fault isolation When there is a fault anywhere in the system, the sum of
purposes. Circuit breakers (CB) are installed at the current contributions from all sources (the main source
interconnecting points to facilitate the isolation of a network and all DG) would be equal to the fault current. Once a fault
segment from the rest of the system for a fault inside it. FTUs in the system is sensed, total fault current complex can be
are installed at the busbar interconnecting different network determined using the following simple equation:
segments. Currents transformers (CT) installed on the n
interconnected branches feeds the FTUs. These CTs are I f = ¦ Isi DŽ (1)
i =1
connected to measure the currents leaving the node. A
communication network facilitates data exchange between where I f is the total fault current (phasor) and Isi is the fault
FTUs and the master system. current contribution (phasor) from source i . “ n ” is the total
number of sources (including the main source) in system.
According to Kirchhoff’s current law, the current
injected into a normal segment on one side is extracted at the
other side. If this criterion is not fulfilled, part of the current is
obviously deviating via some unwanted path. To normal
segments, the sum of the current (phasor) would be equal to
zero. This criterion could be described by using the following
equation:
m

¦ I ki =0 (2)
i =1

where Iki is the current (phasor) following through switch i

Fig.1 structure of the proposed scheme bounding the segment. “ m ” is the total number of switches
bounding the segment.
The isolation zones in urban distribution network could be
Considering add-assist current of the multiple sources,
defined as 3 types: single-switch segments, double-switch
the extracted current is the largest one among the segment and
segments and multiple-switch segments. To single-switch the sum of else current among the segment equals to the
segments, usually the end of the feeder (e.g. segment 8~12), largest one. The equation (2) may be rewritten as:
there is a single measuring node using FTU for it. To the
m
double-switch one, such as the part between two ¦ I ki = − Ik max䯸 Ik max = max{Iki } (3)
pole-top-breakers of overhead line (e.g. segment2, 4, 5), there i =1
i ≠ max
are two measuring nodes for it. To the multi-switch one, such
where Iki and m have same meaning as in equation (2).
as the busbar part of ring main units (e.g. segment1, 3, 6, 7),
T-type line, each switch has a FTU measuring node. FTU Ik max is the largest current sensed by FTUs in the segment.
measuring nodes play a role as a fault detector and report the In the same way, to faulted segments, the sum of the
faulted current information to the master system. current (phasor) would be equal to the total fault current
In the scheme, FTUs record faulted current by using (phasor) I f . And the equation is:
current complex relay. The relay records the current
m
I ¦ = I1 + KI 2 ( I1 is positive sequence current. I 2 stands
¦ I ki = I f (4)
negative sequence current, and K is a constant, usually equals i =1

6~8) when a fault occurs. In the case of unbalanced fault, the where Iki and m have same meaning as in equation (2).
current complex relay has more sensitive than phase current The paper is hunting for a means only using the
relay. And to balanced fault, they have the same sensitivity. amplitude of current phasor. If only using current amplitude
As 90% of the faults are unbalanced faults in the power to describe the characteristic of normal and faulted segments,
system, it could improve sensitivity of the whole scheme to the equation (3) and (4) could be rewritten as:
use current complex relay. Because the major composition of

165
 m m
ΔI normal  ¦I − 2 * I k max ≈ 0䯸 I k max = max{I ki } (5) / I f −1 * I f / I max + μ
i =1
ki
¦I
i =1
ki

DCRn = (10)
ΔI fault 
m
−I f ≈ 0, I f ≈ ¦ I si
n
(6)
μ
¦I
i =1
ki
i =1
The interval, between DCR f and DCRn − min (the
At the base of equation (5) and (6), an index expressing
minimum value of DCRn to I f / I max ), reflects the
differential current ratio (DCR) could be written as:
ΔI fault + μ I max sensitivity of the algorithm. When I f / I max > 0.5 , the
DCR  (7)
minimum DCRn − min equals to 1 if some segments meet the
ΔI normal + μ I max
m
where I max is the largest current of the whole system sensed / I f =1 . When I f / I max ≤ 0.5 , the minimum
condition ¦I ki
by FTUs. μ is constant, such as 0.1. i =1

DCRn − min happens in the segments which sensed the largest

To normal segments, ΔI normal should be close to zero
current in global ( I max = I k max ). Thus
while ΔI fault mostly should be far bigger than ΔI normal .
Therefore DCR of a normal segment is a value more than one. ­ I max
° 1 > 0.5
ΔI normal and ΔI fault of a faulted segments are quite the ° If
DCRn − min =® (11)
opposites. So DCR of a faulted segment is a value less than ° | 2 − I f / I max | + μ I max
≤ 0.5
one. °
¯ μ If
The equation (5) is invalid to single-switch segments.
So DCR of single-switch segments could be defined as:
Ik − I f T
DCRs  (8) 
μ Ik

where I k is the current sensed by the single measuring node

in the segment. I f is the total fault current.

Assuming that there is a single fault, the basic steps of
searching process could be listed as: ķcalculate out the total 

fault current I f and to find out the largest current 

I max sensed by FTUs in global. ĸ calculate out DCR of each

segment using equation (7). Ĺ find out the segment with
smallest DCR, it is just the faulted one. 
DCRn − min

IV. SENSITIVITY ANALYSIS
 DCR f
To the faulted segment, the largest current sensed by
FTU in the segment is also the largest current in global. Thus I max
m
           If
I max = I k max . On the other hand, ΔI fault = ¦I ki −I f ≈ 0 .
i =1
Fig.2 Sensitivity Analysis to I max / I f ˄ μ = 0.1 ˅
The equation (7) could be rewritten as:
μ I max Through Fig.2, it is concluded that the algorithm may
DCR f = m have a fade area if global largest sensed current is just half of
¦I ki − 2 * I k max + μ I max the total fault current ( I max = 0.5 * I f ).
i =1

μ To conventional distribution networks without DG, the

= total fault current would equal to the largest sensed current by
I f / I max − 2 + μ 1+ μ
(9) FTUs ( I max = I f ). Because DCRn = and DCR f ≈ 0 ,
To normal segment, the equation (7) could be rewritten μ
as the algorithm has good sensitivity.
Considering the add-assist current is distributed, the
faulted segment could sense the largest fault current. Then it

166
could narrow the range to search by using the largest fault rules: (1) the most serious fault current could be sensed by
current. Considering the possible measure error, the largest FTUs in the fault segment; (2) the fault zone has the smallest
current is not a value but a value range. So the current should DCR. Based on the first rule, the master could reduce the
be regarded as the ‘suspected largest current’ if it is lager than range to be searched. And based on the second rule, the
K P times the total largest current ( I k > K P I max ,0 < K P < 1 ). On master could locate the faulted. The main steps of algorithm
the basis of the ‘suspected largest current’, ‘suspected faulted are: the master should find out all nodes flowing through the
segments’ could be find out according to the connecting suspected largest fault current at first. According to these
relation between nodes flowing through ‘suspected largest nodes, the suspected fault segments could be found. Then the
current’ and segments. If only the DCRs of ‘suspected faulted faulted segment could be located by comparing the DCRs of
segments’ are compared, the sensitivity of algorithm could be the suspected fault segments. Flow chart for proposed fault
improved. location algorithm for the master system is shown in Fig.4
The flowing part is sensitivity analysis of the optimized
method. The equation (11) could be rewritten as:
Start
­ | 2 K p − I f / I max | + μ I max 1
° ≥
° μ If 2K p
°° I 1 ,QSXWQHWZRUNSDUDPHWHUDQGIDXOW
DCRn − min =® 1 0.5 < max <
I 2 K FXUUHQWGDWD
° f p
° | 2 − I f / I max | + μ I max
° ≤ 0.5
°̄ μ If ,GHQWLILFDWLRQRIQRGHVVHWIORZLQJ
WKURXJKþVXVSHFWHGODUJHVWFXUUHQWÿ
(12)
I
The new curves of DCR f and DCRn − min to max are
If )LQGRXWþVXVSHFWHGIDXOWHG
shown in Fig.3. VHJPHQWVÿVHW

&DOFXODWHWKH'&5VRIþVXVSHFWHG
IDXOWHGVHJPHQWVÿVHW

'HWHUPLQHIDXOWHGVHJPHQWWKDWKDV
VPDOOHVW'&5

DCRn − min
End

Fig.4 Flow chart of fault location algorithm for the master

DCR f
VI. TEST CASE
As a test system, we consider the 69-segment, 8-lateral
distribution feeder, based on the paper [6], and distributed
Fig.3 Sensitivity analysis after improvement ˄ μ = 0.1 ˈ generators are placed at the end of each lateral, at buses 26, 34,
K P = 0.9 ˅ 38, 54, 58 and 90. The feeder layout is shown in Figure 5. We
assume the capacity of DG and connecting transformer are
1MVA, the capacity of generator and transformer connecting
V. FAULT LOCATION ALGORITHM bus 0 are 10MVA. Other parameters please refer to paper [7].
As previously described, this algorithm based on one Using PSS/E 30.1 simulation software, the paper simulates
assumption: there is a single fault. And the algorithm has two 3-phase short circuit fault in the test system at the operation
model that the tie line, between 48th bus and 88th bus, is open

167
loop. The algorithm could be verified by using the simulation PSS/E illustrate DCR work well in urban distribution
data. automation system. The new algorithm is suitable for
If the 3-phase short circuit fault locates at middle-point of conventional distribution network and the distribution
the 30th zone, the fault current (in pu) distribution also shows network with DG.
in Figure 5. The total fault current equals to 1.697 pu and the Once the faulted segment is located, it is could be
largest current in global is 1.097 pu. If K P = 0.9 ˈ μ = 0.1 , the removed. Then islands may form involving groups of DGs. A
suspected faulted segments set is (bus11, bus11-12, bus12-13, load shedding scheme is necessary to match the loads
bus13-14, bus14-15, bus15-16, bus16-17, bus17-18, generating capability of the island. It is also required a
bus18-19, bus19-20, bus20-21, bus21-22, bus22-23, mechanism to reconnect the removed segments after clearing
bus23-24, bus24-25). The DCRs of suspected faulted the faults. Further research is underway to add these
segments are (5.93, 5.53, 5.53, 5.53, 5.53, 5.53, 5.53, 5.53, additional functions to the scheme.
5.53, 5.53, 5.53, 5.53, 5.53, 5.53, 0.24). The bus24-25
segment is the faulted one as its DCR (0.24) is the smallest. REFERENCES
[1] Dugan, R. C. Price, S. K. “Issues for distributed generation in the US,”
in Power Engineering Society Winter Meeting, 2002. IEEE,
G
pp.121-126.
0.667
0.667
0 [2] Zhang Chao, Ji Jian-ren, et al. “Effect of distributed generation on relay
0.667
1
2
protection and automation of distribution network,” East China
Electric Power, vol.34, pp.23-26, Sep.2006.
[3] Liu Jian, Ni Jianli, Du Yu. “A unified matrix algorithm for fault section
0.798
detection and isolation in distribution system,” Automation of Electric
3
0.868
Power System, vol. 23, pp.31-33, Jan.1999.
0.070
0.868
4 0.066
27 0.068
[4] Zhu Fa-guo, Sun De-sheng et al. “Optimized matrix arithmetic of line
35 5 0.066
28 0.068
27e
fault location based on field terminal unit,” Automation of Electric
0.070 0.868
36 6 0.066
29
28e Power System, vol.15, pp. 42-44, Aug.2000.
0.070 0.868 0.068
37 7 0.066
30
65 [5] J. Roberts and A. Guzman, Directional Element Design and
0.868 0.068
0.070 66 Evaluation [Online]. Available: www.selinc.com/techpprs/6009.pdf.
38 8 40 0.066
31 0.068
0.957
9 41 0.066 67 [6] Pregelj A., Begovic et al. “On optimization of reliability of distributed
DG 32
0.957
10
0.090
42 0.066
0.068
68 generation-enhanced feeders,” in Proceedings of the 36th Hawaii
33
55
0.957
11
0.090
43 0.066
0.068
69
International Conferences on System Sciences, IEEE Computer
1.097
12
0.090
44
34
0.068 Society, Hawaii(USA),pp.1-6
56 70
0.184
1.097 0.090
45 DG 0.068 [7] Kojovic Lj. A., Willoughby R. D. “Integration of distributed
57 13 88
0.184 1.097 0.090 0.068 generation in a typical USA distribution system ,” in Proceedings of the
58 14 46 89
1.097 0.090 0.068 16th CIRED, Amsterdam ,pp.1-5.
15 47 90
DG 1.097 0.090
16 48
1.097
17
0.090
49
DG VIII. BIOGRAPHIES
1.097 0.090
18 50
1.097 0.090
19 51
1.097 0.090
20 52
Legend:
1.097 0.090
53
Zhu Guo-fang is currently pursuing a Ph.D. degree
21 Load
1.097 0.090 in electrical engineering from Southeast University
22 54
1.097 G Generator of China. He received the M.S. degree from
23
1.097 DG Southeast University in 2006. His research interests
24
0.552
DG
Distributed
Generator
are protection and control technique to distribution
0.552
25 system with DG.
26

DG

Fig.5 The test system with six DG located at lateral ends and
Current distribution chart of the 30th zone fault Lu Yu-ping has received the Ph.D. degree in
electrical engineering from the City University, UK
in 2003. He is working as a professor in Southeast
University of China. His research interests are power
VII. CONCLUSIONS system protection, especially digital relaying of
A fault location algorithm for distribution automation generator-transformer unit, and protection and
control technique in distribution system with DG.
master system was proposed. This method uses only currents
and thus avoids the need for installation of voltage
transformers. Furthermore, the method does not require time
synchronized measurement. In the new scheme, FTUs only
implement the functions including over-current judgment
using current complex relay and fault-report sending same as
in the conventional DA scheme. The performance of the
algorithm was investigated through PSS/E simulations of a
typical distribution system with DG. The simulation results of

168