Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
I. INTRODUCTION
Transmission system
Load 5
Load 2
Load 1
DG3
DG1
Load 3
Fig. 1.
DG2
DG4
Load 6
Load 4
BRAHMA AND GIRGIS: DEVELOPMENT OF ADAPTIVE PROTECTION SCHEME FOR DISTRIBUTION SYSTEMS
1
From Source
57
Z1
Z4
B1-2
B4-5
F1
F2
Flt1
Flt 2
S/S B
S-1
Z2
F3
B2-5
B2-3
B3-5
Z3
DG
Z5
B5-6
Z6
(a)
Source
Recloser
DG
(b)
Fig. 2.
Tapped
Line
Load
Current
stream of the recloser has to be disconnected before the first reclose takes place to avoid connection without synchronism.
These solutions are not practical. As mentioned earlier, DG
is getting popular because it can serve the load without adding
to T&D burden. Throwing off all DG from system every time
a temporary fault occurs would make the system very unreliable. This paper offers a comprehensive, system-independent
adaptive protection scheme for distribution systems with high
DG penetration that would not undermine the system reliability
after connecting DG. Sections IIIV describe the scheme and
its implementation in detail.
II. PROPOSED SCHEME
A. Outline of Scheme
The ideal situation for any protection scheme is to isolate only
the faulted section from the system. In this case, this is impossible, since the section-controlling protective device is a fuse
here and as explained in Section I, coordination between fuses
is lost in presence of DG. Moreover, fuses cannot be controlled
by an external signal. Therefore, they will not respond to some
tripping signal generated by say, a relay.
The next best approach is to divide the system into zones as
shown in Fig. 3. A zone should be formed such that it has a
reasonable balance of load and DG, DG capacity being a little
more than the load. In addition to this, at least one DG (usually the biggest in the zone) should have load frequency control capability. As shown in Fig. 3, these zones should be separated by breakers. These breakers should be capable to repeatedly open or close on receiving a signal from a main relay located in the substation. Most manufacturers make breakers and
reclosers with remote communication capability. The breakers
should also be equipped with check-synchronization function.
The main relay would be computer-based, capable of storing
and analyzing large data and able to communicate with other devices like zone breakers and DG relays. The relay would sense
a fault, identify the type of fault and the faulted section (and,
hence, zone) on line, and isolate the faulted zone by tripping
appropriate breakers and DG connected to this zone. This way,
the remaining zones can still function as usual. Reclosing to take
care of temporary faults would be performed by main relay itself as explained later in Section II-F. The following subsections
explain the scheme in detail.
58
B. Measurements
The first requirement for a protection scheme is the actuating
input. This comes through measurement. The following continuous measurements are recommended for this scheme:
synchronized current vectors for all three phases from
every DG in the system and from the main source;
a signal indicating current direction in every zone-forming
breakers.
Synchronized vector measurement using a synchronizing
clock pulse from global positioning system (GPS) receiver and
phase measurement unit (PMU) is described and used by J.
Jiang et al. [15][17]. It is possible to achieve a synchronization
with this method.
accuracy of better than 1
C. Off Line Calculations and Data Storage
This method requires a load flow study and a complete shortcircuit analysis for all types of faults involving different phases.
In addition to fault current for different types of faults at each
bus, this analysis should also find out fault current contribution
from each DG and from the main source. The method also requires the minimum melting (MM) characteristics of all fuses in
the system to be stored in the relay database. From these characteristics and short-circuit analysis, the time before some fuse in
the system gets damaged in case of fault can be determined. Significance of this time will be mentioned in Section II-F. However, as explained later in section III, this requirement can be
avoided altogether. The load flow and the short-circuit analysis
need to be updated after every significant change in load, DG,
or the system configuration. Whereas a change in load and/or
DG would require to merely run the load flow and the short-circuit analysis again, any change in the system configuration (say,
taking off a line) would require to update the bus admittance and
impedance matrices as well. These are routine procedures supported by software used by utilities. With modern computing
power and memory, the off line calculations and data storage
would not be a concern.
D. Sensing Fault and Determining Fault Type On Line
Current phasors from the main source and all DG are continuously available. In normal operating condition, the sum of
all these phasors would be equal to the total load on system. In
case of a fault on any part of system, this sum would exceed the
total load substantially. This is how the relay would sense a fault
in the distribution system. This in a way is similar to a current
differential scheme. The monitored zone here is the distribution system itself. When there is a fault anywhere in the system,
the sum of the current contributions from all sources (the main
source and all DG) would be equal to the fault current. On the
other hand, if there is a fault in a DG, since the DG is outside
the monitored zone, the sum would be zero. This is how a fault
in DG can be distinguished from a fault in the system.
Once a fault in the system is sensed, total fault current in each
phase can be determined using the following simple equation:
(1)
BRAHMA AND GIRGIS: DEVELOPMENT OF ADAPTIVE PROTECTION SCHEME FOR DISTRIBUTION SYSTEMS
59
start
IFMAX
IFMAX
IFMIN
IFMIN
0 Length of a line section i-j (PU)
bus i
1
bus j
at the two buses connected to this section from that source for a
given type of fault. As will be explained by an example system
in Section III, when the number of DG increases in the system
and system loses its radial nature in most parts, this method is
very effective.
Fig. 5.
the current direction signal from that breaker. The next closing
signal would be sent only after this confirmation. Finally, DG
breaker would be closed and system restored to normal. In case
of a permanent fault, the fault would have to be cleared by
maintenance personnel before incorporating that zone back to
the system. In such cases, relay will adapt to the new network
configuration by running a load flow and short-circuit analysis
again. There is a possibility here that some part of system may
be islanded after disconnection of one zone. In Fig. 3, this would
or
are faulted. Since load and generhappen when either
ation are balanced in each zone and at least one DG in each zone
has load frequency control capability, such islands can continue
to function.
G. Sources and Handling of Error
The source of error in this scheme can be fault resistance.
Since the short-circuit analysis is made with zero fault resistance
value, a fault with resistance may produce currents which
would lead to the predicted section being adjacent to (or
even farther than) the actual faulted section. If the adjacent
section still lies in the same zone, relay would still operate
correctly. In case this section lies in a different zone, there
is a chance of mal-operation. Therefore, the relay should get
a signal indicating current direction from each zone-forming
breaker as explained in Section II-B. With the help of this
signal, relay can crosscheck before tripping a zone. Flowcharts
given in Figs. 5 and Fig. 6 sum up the off line and on line
calculations, respectively.
There is a reason to believe that the error due to fault resistance would be mostly negligible. Burke [21] reports a four-year
fault study performed in the U.S. on 50 feeders from 13 utilities at various voltage levels. It is concluded by the closeness
of actual and calculated values that there was essentially no
fault resistance observed in the recorded faults. This reference
also states that the readings could be even closer had there been
60
J=1
start
J=J+1
Select section J
K=1
K=K+1
Select measured contribution (maximum
phase current) from source K. Call it IMK
Is the sum
zero?
Yes
Fault is in
DG.
Let it clear.
No
Is the sum
significantly
greater than
total load?
No
No
Is IMK between
IFRJ and ITOJ?
Yes
Yes
This is fault
Determine type of fault
Is K= Total
number of
sources?
No
Yes
Faulted section is J
Is fault
temporary?
Yes
Open breakers
that isolate
faulted zone
No
Update
system data
Do off-line analysis
No
Is fault
still
sensed?
Open
Yes substation
breaker
Stop
available in market take system data in terms of sequence impedances. This does not fulfill the requirements of an unbalanced and untransposed distribution network. Therefore, shortcircuit analysis software was developed based on three-phase
analysis. This software was tested first on a 22-bus balanced
multisource system. Results with conventional sequence component techniques matched exactly with results using the threephase algorithm.
Then an actual 60-bus distribution feeder from a utility in
the southeast U.S. was taken as a test system. Fig. 7 shows the
single line diagram, locations of reclosers, fuses, and DG, as
well as DG capacities. Load details are given in the Appendix.
DG shown in Fig. 7 are assumed and their capacities are selected such that they feed some load around their location. The
feeder load is 2.5 MVA and the total load fed by all DG is 45%.
BRAHMA AND GIRGIS: DEVELOPMENT OF ADAPTIVE PROTECTION SCHEME FOR DISTRIBUTION SYSTEMS
103
102
TIME (SEC)
61
30A TC
Coordination Range Without DG
Coordination Range With DG
30A MM
101
100
Recloser 'C'
Recloser 'A'
10-1
10-2
101
102
103
CURRENT (A)
104
105
(a)
102
25A MM
TIME (SEC)
DG were modeled as sources behind reactances and the interconnecting transformers were considered D/YG. DG and transformer reactances were chosen using [6] and [22].
As shown in Fig. 7, this system has two reclosers on main
is 25 A) and laterals are protected
feeder ( is 140 A and
by fuses. In addition to fuses shown in this diagram, load on
each bus is tapped off with a fuse. The section fuses coordinate
coordinates with section fuses downwith these load fuses.
and upstream of .
coordinates with section
stream of
coordinates with
for
fuses downstream of . In addition,
faults downstream of . After connecting DG to this system as
shown, the fuse coordination between fuses protecting sections
5659 and 5960 is clearly lost as explained in Section I. Similar
is the case with sections 3536 and 3637. Fig. 8(a) shows cowith 30-A fuse on section 816 and Fig. 8(b)
ordination of
shows coordination of
with 25-A fuses on sections 3536
and 3542. Curve A is the fast curve and C the slow curve
of reclosers. MM stands for minimum melting and TC for
total clearing characteristics of fuse. Clearly, the coordination
with the fast curve is lost after connecting DG. It is possible to
get this coordination back using microprocessor-based recloser
looking only in downstream direction [14]. But this would mean
that for every temporary fault downstream of , DG1 and DG5
would be cut off and for every temporary fault downstream of
and upstream of , all DG is cut off. This could not be an
acceptable solution as explained in Section I.
Success of implementation of the scheme described in Section II essentially depends on the ability of relay to accurately
identify the faulted section and, hence, the faulted zone. As
mentioned in Section II-E, the proposed method is very effective when the number of sources increases. This can be justified through a simple example on the test system. Thevnin
impedance to main source for fault points on section 78 and
724 can be same. This means that the main source contribution to a given type of fault in these two sections could be
identical. Thus, with only main source in this system, the proposed method would confuse between these sections. However,
Thevnin impedances for such fault points to any of the five DG
cannot be the same. Thus, these sections would be distinguished
by checking contribution from any of the DG. Natural extension
103
25A TC
101
Recloser 'C'
100
10-1
Recloser 'A'
10-2
101
102
103
CURRENT (A)
104
105
(b)
Fig. 8.
62
TABLE I
DEPENDENCE OF ACCURACY OF PROPOSED METHOD ON NUMBER OF DG
CONNECTED IN THE SYSTEM
DG
connected
in system
none
one
two
three
four
five
Accuracy
L-L-L
%
Accuracy
L-L
%
Accuracy
L-L-G
%
Accuracy
L-G
%
30.5
61.8
77.8
94.4
95.1
97.9
26.6
52.7
65.5
83.0
86.1
93.9
26.6
52.7
65.5
83.0
86.1
93.9
26.6
52.7
65.5
83.0
86.1
93.9
was observed only for one data sample whereas for the other
pairs, all three data samples gave wrong results. This confusion is not serious from the point of view of the relay scheme
since such confused sections, being radial, cannot be parts of
two different zones. Therefore, the relay would still trip the
right zone. Moreover, such confusions can be predicted beforehand from network topology, thus making the maintenance task
easier. Table I shows how accuracy of this method increased
with more DG being added to the system.
Accuracy in Table I is the successful identifications as percentage of total data samples. Essentially, the confused sections
remain same as mentioned before, but since there is some confusion in two-phase sections where LLL fault data are not
applicable, accuracy in case of LLL fault is higher.
The algorithm to locate the faulted section was also tested
on two more distribution systems with 21 and 47 buses, respectively. The results perfectly conformed to the observations made
in the previous paragraphs. The data of these systems can be
made available upon request.
From the DG capacities given in Fig. 7 and loads given in
the Appendix, this test system can be divided into four zones,
each zone having one/more sources and adjoining buses (The
guidelines followed here are described in Section II-A.) Zone 1
would include DG3, zone2 would have DG2, zone3 would have
DG1 and DG5, and zone 4 would have main source and DG4.
In this case, islands would be formed only when there is a fault
in zone 4.
From short-circuit analysis with DG and the minimum
melting characteristic of fuses in the system, it was found that
in the worst fault case, the time before a fuse can be damaged
in the system was about 55 ms. Phasor identification in PMU
can take between quarter of a cycle [23] to one cycle [17]. Time
taken by the relay to sense a fault using measurements coming
from PMU, identify the faulted section (it just involves referring
to a lookup table) and send trip signal is insignificantly small.
Hence, the zone breaker and DG breakers in the faulted zone
can be tripped well before this time. This applies to opening
the breaker after each reclosing too.
In the above analysis, care is taken to make sure none of the
fuses gets damaged during the operation of scheme. In fact,
since fuses are no longer coordinated in an environment with
DG, it is even possible (and recommended) to remove all fuses
from the area covered by this scheme. This will no longer require the time constraint explained in the previous section and
last step in the flowchart in Fig. 5.
IV. OBSERVATIONS
There is always a chance of malfunction in any scheme. In
this scheme, there is a chance that one of the breakers does
not open after getting a command from the relay to isolate the
faulted zone. In this case, since the fault is not isolated, the relay
will continue to sense the fault. In such a case, the relay should
disconnect the whole feeder by opening the breaker at the substation. The flowchart in Fig. 6 includes this possibility. In case
the substation breaker fails to operate, the breaker on the high
voltage side of the distribution transformer will open.
The scheme covers all types of faults normally occurring in
a distribution system. In case of two simultaneous faults in the
same zone (very rare), the relay will still trip the zone using the
current direction signal from breakers, but will fail to detect the
faulted section. In case the faults occur simultaneously in different zones, after clearing one zone, the relay will still sense a
fault. It will open the substation breaker in such a case. In case of
open conductor fault, if the unbalance created is substantial, the
relay can sense it (though it cannot sense the faulted zone) and
isolate the feeder. Otherwise, such a fault would be undetected,
as is the case in conventional protection schemes for distribution systems.
Hart et al. [24] conclude that considering the operating cost
and the connection speed, wire- or fiber-based systems are viable alternatives (telephone lines are too slow) for communication of data in PMU-based schemes. If such a connection already
exists between DG and substation, part of the bandwidth can be
used for relaying. The installation cost incurred by this scheme
would certainly be higher because it introduces many new components in the system, but once installed, the maintenance cost
would be quite reasonable. The installation of measuring and
processing equipment can certainly be managed by utility engineers after training.
The accuracy of fault sensing and section detection depends
mainly on the accuracy of PMU and on the relay algorithm. The
accuracy of PMU has been practically tested [17]. The relay
algorithm, instead of being based on some heuristic method
(e.g., neural network), is derived from simple and proven circuit fundamentals. Therefore, there is no apparent reason why
the scheme cannot be successfully implemented. It should be
noted here that many utilities model their distribution network
as a single line network (assuming it balanced) for load flow
and short-circuit analysis. This will increase the error in section
identification but will not render the scheme useless.
V. CONCLUSION
Current trend and literature show that distributed generation
is going to increase significantly in the coming years. Coordination between fuses and between fuses and reclosers in a distribution system can be disrupted with substantial penetration of distributed generation. The methods proposed in literature to solve
the problem are not satisfactory from operational point of view.
The adaptive scheme proposed here offers a practically acceptable solution to this problem that is independent of size, number,
and placement of DG in the distribution system. The proposed
scheme is adaptive to temporary as well as permanent changes
BRAHMA AND GIRGIS: DEVELOPMENT OF ADAPTIVE PROTECTION SCHEME FOR DISTRIBUTION SYSTEMS
TABLE II
LOAD DATA FOR THE TEST SYSTEM USED
63
REFERENCES
[1] Overcurrent Protection for Distribution Systems, Application manual
GET-1751A, General Electric Company, 1962.
[2] IEEE Power System Relaying Committee report, Distribution line protection practices: Industry survey analysis, IEEE Trans. Power App.
Syst., vol. PAS-102, pp. 32793287, 1983.
[3] M. A. Anthony, Electric Power System Protection and Coordination. New York: McGraw-Hill, 1995, pp. 109121, 342-346.
[4] J. L. Blackburn, Protective Relaying Principles and Applications. New
York: Marcel Dekker, 1998, pp. 383408.
[5] P. M. Anderson, Power System Protection. New York: IEEE Press,
1999, pp. 201240, 249-257.
[6] P. Barker and R. W. de Mello, Determining the impact of distributed
generation on power systems: Part 1Radial power systems, in Proc.
IEEE Power Eng. Soc. Summer Power Meeting, 2000, pp. 16451658.
[7] U.S. Department of Energys distributed power program homepage, Online Available:, http://www.eren.doe.gov/der/basics.html.
[8] Portfolio of DG Technologies. [Online]http://www.distributed-generation.com/technologies. html
[9] H. L. Willis and W. G. Scott, Distributed Power Generation Planning
and Evaluation. New York: Marcel Dekker, 2000.
[10] Role of distributed generation in competitive energy markets, Gas
Res. Inst., Rep. GRI-99/0054, Chicago, IL, http://griweb.gastechnology.org/pub/solutions/dg/distgen.pdf.
[11] N. Hadjsaid, J. Canard, and F. Dumas, Dispersed generation impact on
distribution networks, IEEE Comput. Appl. Power, vol. 12, pp. 2228,
Apr. 1999.