2890 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 28, NO.

3, AUGUST 2013

Fault Analysis on Distribution Feeders With High

Penetration of PV Systems
Hossein Hooshyar, Student Member, IEEE, and Mesut E. Baran, Fellow, IEEE

Abstract—Fault current profile on a PV-dominated distribution

feeder is rather different than a conventional feeder. To estimate
the fault current profile on such a feeder, the paper proposes a new
method which extends the capability of conventional short-circuit
analysis method. The paper also shows that this time-varying fault
current profile makes it also more difficult to estimate the time it
Fig. 1. Main components of a residential PV system.
will take for an overcurrent relay/device to interrupt such a fault
current. The paper proposes a method for estimation of this op-
erating time also. Performance of the proposed methods has been
assessed by simulations on a sample distribution feeder. This paper proposes a method which can provide an esti-
mate of fault current variation on a PV-dominated distribution
Index Terms—Fault analysis, overcurrent relay, photovoltaic
(PV), PSCAD/EMTDC. feeder during the first few seconds of the fault occurrence.
This enhanced fault analysis method will enable the protection
engineers to design and coordinate protection schemes on a
I. INTRODUCTION PV-dominated feeder [4]. In Section II, we show that both the
current injection from a PV system, as well as the protection
scheme employed for the PV, need to be considered in order

O NE of the main impacts of a photovoltaic (PV) system

on a distribution feeder is on the system protection, as
the PVs will contribute to the fault current during a fault [1],
to capture the PV system’s response to a fault. In Section III,
the proposed fault analysis method is introduced. Section IV
introduces the method for estimating the operating time for an
[2]. As the penetration of PV systems on a feeder gets higher, overcurrent relay/device which will see such a time varying
the fault current profiles on such a system will be considerably fault current. Conclusions are drawn in Section V.
different than that of the feeder with no PV [3]. These fault cur-
rent variations will in turn impact the operation and coordina- II. PV SYSTEM BEHAVIOR UNDER FAULT
tion of protection devices on such a feeder considerably [4], [5]. A green hub is a distribution feeder on which almost every
Since conventional short-circuit analysis methods provide only customer has a PV system and that each PV system can generate
one snapshot of the fault current [6], they become inadequate up to the maximum load of the residential unit. Hence, a green
to analyze the fault current variations on PV dominated feeders hub has a very high PV penetration. To analyze the effects of
and estimate their impact on protection relays and devices. PVs on fault currents on such a system, we need a model for the
The impact of a distributed generator (DG) on the fault PV system first.
current levels on a feeder has been investigated in literature, Fig. 1 shows a typical grid connected residential PV system. It
by using mainly time domain simulation methods. Wang et is an inverter interfaced system to accommodate 120- or 240-V
al. proposed a fault analysis method based on the load flow ac connections.
method; however this method considers mainly the distributed When a fault occurs on the feeder, the PV system feeds cur-
synchronous generators [7]. In [8], a fault analysis method for rent to the fault. The current to be injected depends on the PV
inverter-interfaced DGs has been proposed. The method aims inverter design. Since the PV systems are designed to push the
at estimating the initial high current that an inverter interfaced maximum power available from PV panels to the system, the PV
DG under voltage control scheme can inject during the first inverter tries to push this power even under low voltage con-
cycle of the fault. ditions which occur during a fault, i.e., it will try to act like a
constant power source. Hence, the current injection from a PV
Manuscript received June 13, 2012; revised August 09, 2012; accepted Oc- inverter to the system can be approximated as ,
tober 30, 2012. Date of publication December 20, 2012; date of current version
where is the power from the PV panels and is the ac
July 18, 2013. This work was supported by ERC Program of the National Sci-
ence Foundation under Award EEC-08212121. Paper no. TPWRS-00655-2012. terminal voltage. However, if this current gets to be higher than
H. Hooshyar was with the North Carolina State University, Raleigh, the maximum current rating of the inverter, inverter limits the
NC 27606 USA. He is now with the Electric Power Engineering Group,
current at its maximum level. Inverters limit their current to one
Luleå University of Technology, 931 87, Skellefteå, Sweden (e-mail:
[email protected]). to two times the rated current [9]. Hence, for fault studies, we
M. E. Baran is with the North Carolina State University, Raleigh, NC 27606 will assume the limiting current to be twice the rated current, for
USA (e-mail: [email protected]).
the worst case scenario. Thus, the PV system will have the –
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. characteristics shown in Fig. 2. As indicated in the figure, the in-
Digital Object Identifier 10.1109/TPWRS.2012.2227842 verter current and voltage are usually in phase for unity power

0885-8950/$31.00 © 2012 IEEE

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE ENGENHARIA DE ITAJUBA. Downloaded on April 27,2021 at 21:42:28 UTC from IEEE Xplore. Restrictions apply.
 HOOSHYAR AND BARAN: FAULT ANALYSIS ON DISTRIBUTION FEEDERS WITH HIGH PENETRATION OF PV SYSTEMS 2891

Fig. 2. – characteristics of a PV System.

Fig. 3. PV model for fault analysis.

TABLE I
PROTECTION SCHEME FOR A PV SYSTEM

factor operation. This – characteristics of a PV system has

been verified by using a detailed PV system model simulated
on PSCAD, and will be introduced in Section III.
We propose to use the – characteristics on Fig. 2 in the
fault analysis in order to estimate the fault current contribution
from a PV system. This can be achieved by modeling the PV
system as a controlled current source whose magnitude depends
on its terminal voltage and PV panel power. Fig. 3 illustrates
this model. Note that this model is for a typical voltage-sourced-
based inverters used for grid-connected PV systems. If the in-
verter is set to operate in a different control mode—for example,
for constant power factor mode—then the model needs to be re-
vised accordingly.
Another important factor which affects the current contribu-
tion from a PV system during a fault is the protection scheme
employed for the inverter. Table I summarizes the protection Fig. 4. Response of PVs on a feeder during a fault.
scheme recommended by the IEEE standard 929 for such sys-
tems [10]. As the table indicates, the protection module discon-
nects the PV from the grid when it detects grid abnormal voltage scheme is that on a feeder with many PVs connected along its
or frequency. Note that, as the table shows, the time it will take length, the PVs will be disconnected from the feeder at different
for the protection scheme to disconnect the PV depends on how times during a fault, as they will see different voltages. Fig. 4
low the terminal voltage will be during a fault. If the terminal illustrates such a case. As shown in the figure, when a fault oc-
voltage gets below 0.5 pu during the fault, then the PV will be curs (at 0 s) towards the end of the feeder, the PVs closest to
disconnected in less than six cycles, however, it may take up to the fault, PV1–PV3, will see their terminal voltages drop below
120 cycles if the terminal voltage remains higher than 0.5 pu. An 0.5 pu, and hence they will be disconnected in 0.1 s (six cycles).
important implication of such a voltage dependent protection But PV4 and PV5 terminal voltages will be above 0.5 pu, and

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE ENGENHARIA DE ITAJUBA. Downloaded on April 27,2021 at 21:42:28 UTC from IEEE Xplore. Restrictions apply.
 2892 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 28, NO. 3, AUGUST 2013

the PVs identified. The algorithm then updates the node voltages
and branch currents by repeating the calculations done at the
first step. This loop is repeated until the PV protection does not
disconnect any more PVs.
Finally, note that, if the PV system protection is not IEEE 929
compliant, then this particular protection scheme should be used
in the PV protection simulation part of the analysis.

B. Test Results
To evaluate the performance of the proposed method on a
green hub, an actual 22-kV distribution system serving mostly
residential loads in a suburb of Raleigh, NC, is considered as
the test case in this study. The feeder has three-phase primary
with several single-phase underground cables tapped off from
the main circuit to feed the customers within the same neigh-
borhood. For this study, this system is simulated using PSCAD
in order to get detailed time-domain simulations and to use them
for comparison with the proposed method. Since it was not prac-
tical to include all of the loads and PV systems individually in
the simulation, the loads and the PV systems on the single-phase
laterals served off the main line sections have been aggregated
Fig. 5. Proposed fault analysis algorithm for a green hub.
as one lumped load and PV system as shown in Fig. 6. Hence,
the simulated system is a three phase circuit with unbalanced
loads and PVs represented on a phase basis. We will refer to
hence their protection schemes will start counting to disconnect this sample feeder as the green hub.
the PVs in 2 s (120 cycles). PV6 on the other hand will ride The PV system is simulated by adopting a typical model,
through the fault as its terminal voltage is higher than 0.88 pu. which uses average models for the converters, and includes
However, after PV1–PV3 are disconnected at 0.1 s, the maximum power point tracking [11]. Fig. 7 shows the –
system voltage will decrease further due to increased fault cur- curve of the adopted PV system model obtained from the
rent from the substation, as shown on the figure; and this in turn simulation results. As shown in the figure, the PV system in-
will cause PV4 to be disconnected faster at 0.2 s and PV6 to be verter operates under two modes of operation: constant power
disconnected at 2.1 s as well. and constant current. Figs. 8 and 9 compare the results from
Finally, note that, as a result of PVs disconnecting at different the proposed method with the results obtained from PSCAD
times on the feeder, the fault current will be varying, as the simulations.
fault current profiles on Fig. 4 clearly shows. Hence, in order The fault currents given are the fault currents through the
to be able to determine such a fault current profile on a feeder, feeder circuit breaker under single-phase and three-phase faults.
we need to extend conventional fault analysis method by incor- As the figures show, the method calculates the current levels and
porating both the – characteristics and protection schemes the time instants at which the current level changes.
for PVs. Section III introduces the proposed scheme for this Fault currents are calculated for faults at four different nodes
purpose. (4, 8, 10, and 13) and with varying fault resistance. These re-
sults show that the proposed method estimates are very close to
III. FAULT ANALYSIS METHOD FOR GREEN HUB the ones from simulations. The maximum difference between
the current magnitudes calculated by the proposed method and
A. Algorithm for Fault Analysis obtained from the simulations is 6.16%.
Fig. 5 shows the proposed algorithm for the fault analysis
method for a green hub. As the figure shows, the fault is assumed IV. ESTIMATING OVERCURRENT RELAY TRIP TIME
to occur at , and then the algorithm first calculates the As illustrated in Section III, the fault current varies in the
node voltages ( ) and branch current during the first period green hub, and therefore it becomes more challenging to esti-
after the fault where all PVs respond to the fault and contribute mate the time taken for an overcurrent relay, which will see this
to the fault. Since PV current injection depends on its terminal fault, to trip. We developed a method for this purpose.
voltage, an iterative procedure is needed to obtain the solution, Since electromechanical relays and digital relays respond to
as the figure illustrates. The PV model indicated in the figure is the varying fault currents differently, we need to consider both
the controlled current model developed in the previous section. cases.
In the next step, response of the PV protection is emulated in
order to determine the PVs that will be disconnected by their A. Electromechanical Relays
protection system and how long it will take for the protection Fig. 10 depicts the contacts movement of an electromechan-
system to disconnect them. Once this time is identified, the ical for the case that the current changes from to at sec-
algorithm increments the simulation time by and disconnects onds, and and are the angular distances that have been

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE ENGENHARIA DE ITAJUBA. Downloaded on April 27,2021 at 21:42:28 UTC from IEEE Xplore. Restrictions apply.
 HOOSHYAR AND BARAN: FAULT ANALYSIS ON DISTRIBUTION FEEDERS WITH HIGH PENETRATION OF PV SYSTEMS 2893

Fig. 6. Test feeder—a sample green hub (circles indicate the nodes where the faults have been simulated).

Following the same logic, the following equation gets the

relay trip time if the fault current changes times from to
at to :

Fig. 7. – curve of the simulated PV system.

(5)

swept by and , respectively. The angular distance that B. Digital Relays

the relay will travel under a current of can be approximated as
Fig. 11 shows the principles of a digital relay operation [13].
[12]
As shown, for a given fault current, the digital relay obtains the
value of from the time–current curve. Then, it construct a
function with the value of . This function is integrated
where (1) until the output of the integral becomes 1. This is the time when
the relay sends the trip signal to the interrupting device.
Following the same procedure, if the current changes from
(2) to at , the relay constructs the function shown in Fig. 12.
By getting integral from the function shown in Fig. 12 and
On the other hand, if we assume that the whole has been equating it to 1, we can obtain , as follows:
swept by constant , we can obtain the following equation:
(6)
(3)

where can be obtained from the relay time–current curve. Following the same logic, the following equation gives the
Now, by equating (1) with (3), we can get an estimate of the relay trip time if the fault current changes times from to
trip time as at to :

(4) (7)

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE ENGENHARIA DE ITAJUBA. Downloaded on April 27,2021 at 21:42:28 UTC from IEEE Xplore. Restrictions apply.
 2894 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 28, NO. 3, AUGUST 2013

Fig. 8. Breaker current for single phase faults on the sample distribution feeder (green hub).

Fig. 9. Breaker current for three phase faults on the sample distribution feeder (green hub).

Fig. 10. Contact travel of an electromechanical overcurrent relay.

C. Algorithm for Trip-Time Estimation

Assuming that the fault current changes times from to
at to , we cannot simply use (5) or (7), because we
do not know in advance at which current level relay will trip.
For example, the relay might trip during the second level of
the fault current, so the rest of the information about the fault
current levels or the separation time instants shall not be used.
Fig. 13 shows the proposed algorithm. As shown in the figure,
the algorithm starts with the value of the first level of the current.
In each loop, the algorithm checks whether the calculated trip
time is less than the closest separation time instant or not. If
yes, it breaks the loop as this means that the relay trips before Fig. 11. Principles of a digital overcurrent relay operation.
the current level changes at the next separation time instant.
Otherwise, the algorithm repeats the loop by using the value of
the next level of current together with the values of the previous Figs. 14 and 15 compare the results obtained from the pro-
levels. posed method with the results from PSCAD simulation for the

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE ENGENHARIA DE ITAJUBA. Downloaded on April 27,2021 at 21:42:28 UTC from IEEE Xplore. Restrictions apply.
 HOOSHYAR AND BARAN: FAULT ANALYSIS ON DISTRIBUTION FEEDERS WITH HIGH PENETRATION OF PV SYSTEMS 2895

Fig. 15. Breaker relay trip times for three-phase faults on the sample distribu-
tion feeder (green hub).

Finally, note that these two methods are needed for investi-
gating protection issues on a PV dominated system. The pro-
Fig. 12. Digital overcurrent relay operation when the fault current changes. posed methods will show: 1) if the protection devices on the
feeder operate properly, or there are coordination issues, and
2) if the feeder protection devices coordinate well with the PV
protection systems. For example, the coordination between the
breaker and the recloser in the sample feeder can be evaluated
by checking the breaker and the recloser trip times for a fault
occurring right after the recloser. If there are issues identified
through these methods, they can then be addressed by revising
protection schemes, usually on the feeder side. An example of
such a study is given in [4]. Note that, for the case considered,
the single-phase faults are cleared faster than the three-phase
faults, due mainly to more sensitive pickup settings of ground
relays.

V. CONCLUSION
This paper shows that the fault current profile on a PV-dom-
inated distribution feeder is considerably different than that of
a conventional feeder with only loads. Both the fault current
contribution from PV systems and the voltage-based protection
Fig. 13. Method to calculate the relay trip time of an overcurrent relay.
schemes they employ contribute to the variation of fault current.
The paper proposes new method that extends the conventional
short-circuit analysis methods and provides an estimate of the
fault current profiles. The simulation-based test results indicate
that the method provides quite accurate estimate.
The paper also proposes a method which takes this varying
fault current a protection relay will see and estimates the time it
will take for the relay to operate. Test results for this procedure
show that the method reproduces the relay operating time quite
accurately.
Given that PV penetration on distribution level will increase
in near future, the proposed methods are much needed tools for
a practicing engineer in assessing the impact of PV penetration
on system protection and developing mitigating strategies, such
as revising relay settings for relay coordination.
Fig. 14. Breaker relay trip times for single-phase faults on the sample distribu-
tion feeder (green hub).

REFERENCES
[1] J. Keller, B. Kroposki, R. Bravo, and S. Robles, “Fault current contribu-
breaker relay for single-phase and three-phase faults. The max- tion from single-phase PV inverters,” in Proc. 37th IEEE Photovoltaic
imum difference is 5.97%. Specialists Conf., Jun. 19–24, 2011, pp. 1822–1826.

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE ENGENHARIA DE ITAJUBA. Downloaded on April 27,2021 at 21:42:28 UTC from IEEE Xplore. Restrictions apply.
 2896 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 28, NO. 3, AUGUST 2013

[2] T. S. Sidhu and D. Bejmert, “Short-circuit current contribution from [13] S. E. Zocholl, “Testing dynamic characteristics of overcurrent relays,”
large scale PV power plant in the context of distribution power system in Proc. 20th Annu. Western Protective Relay Conf. , Spokane, WA,
protection performance,” in Proc. IET Conf. Renewable Power Gener- Oct. 1993, pp. 350–355.
ation RPG, Sep. 6–8, 2011, pp. 1–6.
[3] J. Gao, W. Hu, Z. Shi, and B. Tang, “Research of effect on distribution
network with penetration of photovoltaic system,” in Proc. 45th Int.
Universities Power Eng. Conf., 2010, pp. 1–4.
[4] M. E. Baran, H. Hooshyar, Z. Shen, and A. Huang, “Accommodating
high PV penetration on distribution feeders,” IEEE Trans. Smart Grid, Hossein Hooshyar (S’12) received the Ph.D. degree
vol. 3, no. 2, pp. 1039–1046, Jun. 2012. from the North Carolina State University, Raleigh, in
[5] R. A. Walling, R. Saint, R. C. Dugan, J. Burke, and L. A. Kojovic, 2011.
“Summary of distributed resources impact on power delivery systems,” Since then, he has been with the Electric Power
IEEE Trans. Power Del., vol. 23, no. 3, pp. 1636–1644, Jul. 2008. Engineering Group, Luleå University of Technology,
[6] F. Soudi, D. Shirmohammadi, and C. S. Cheng, “A distribution short Skellefteå, Sweden, as a Postdoctoral Fellow. His
circuit analysis approach using hybrid compensation method,” IEEE research interest includes digital simulation of
Trans. Power Syst., vol. 10, no. 4, pp. 2053–2059, Nov. 1995. power systems and integration of renewable energy
[7] D. Y. Wang, Z. Zabar, and D. Czarkowski, “Load flow based distri- resources.
bution system short-circuit algorithm incorporating distributed syn-
chronous generators,” in Proc. IEEE Int. Symp. Circuits Syst., May
21–24, 2006, pp. 1–4.
[8] M. E. Baran and I. El-Markaby, “Fault analysis on distribution feeders
with distributed generators,” IEEE Trans. Power Syst., vol. 20, no. 4,
pp. 1757–1764, Nov. 2005. Mesut E. Baran (S’85–M’88–SM’05–F’11) re-
[9] Draft Guide to Conducting Distribution Impact Studies for Distributed ceived the Ph.D. degree from the University of
Resource Interconnection, IEEE.7/D4.0, . California, Berkeley, in 1988.
[10] IEEE Recommended Practice for Utility Interface of Photovoltaic (PV) He is currently a Professor with North Carolina
Systems, IEEE 929, 2000. State University, Raleigh. His current research in-
[11] Model Available on the Website of ECEE Department at the University terests include distribution and transmission system
of Colorado at Boulder [Online]. Available: http://ecee.colorado.edu/ analysis and control, integration of renewable
~ecen2060/matlab.html energy resources, and utility applications of power
[12] S. H. Horowitz and A. G. Phadke, Power System Relaying, 3rd ed. electronics-based devices.
New York: Wiley, 2008.

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE ENGENHARIA DE ITAJUBA. Downloaded on April 27,2021 at 21:42:28 UTC from IEEE Xplore. Restrictions apply.