IEEE ELECTRICAL POWER CONFERENCE, OCTOBER 22–23, 2009, MONTREAL, QUEBEC, CANADA 1

Fault Contribution of Grid-Connected Inverters

Dave Turcotte, Member, IEEE, and Farid Katiraei, Senior Member, IEEE

Abstract— The distribution grid is mainly built on a radial cases, forcing power to flow upstream. This change has various
configuration where power is coming from one transformer implications, one of which being that this new generation is
substation to supply clients. Up to recently, in the rare cases contributing to faults occurring on the distribution grid or in
where distributed generation existed, it was almost exclusively
constituted of rotating machines, which have quite a different transformer substation. While the fault current will always be
behaviour under fault than inverter-based sources. Consequently, increased by adding generation, the consequences on the fault
current connection impact assessments rules were built on years clearing elements can be in two opposite directions:
of rotating machines experience and often misrepresent inverter- • If a fault occurs upstream of the fault clearing device
based sources.
This paper presents an overview of the issue of short-circuit (i.e. toward the substation), the clearing device will see
contribution with respect to distributed generation and highlights the current flowing upstream to contribute to the fault.
the distinctions between rotating and inverter-based sources in This is not totally new since even motors will behave as
this regard. A typical inverter and synchronous machine short- generators, except with less capacity.
circuit current model is presented as well as simulation results • If a fault occurs downstream of the fault detection and
for a 7.5 MW implementation on a typical Canadian network.
clearing device, the fault current seen by the clearing
Index Terms— short-circuit, fault, inverter, distributed gener- device will be reduced and may even be shadowed by the
ation, DG, interconnection.
contribution of the local generation and thus remain un-
detected. The fault current will still be increased though.
I. I NTRODUCTION
This paper addresses protection issues of DG interconnec-

I NTERconnection of distributed generation (DG) is reg-

ulated by a comprehensive set of parameters to ensure
proper power quality and stability of the grid. While small
tion in terms of total fault current contribution and changes
in short-circuit capacity requirements of circuit breakers or
under-load disconnecting switches only. In this regard, all
and isolated independent power producers can connect quite equipment have a limit with respect to the amount of energy
easily, medium to large penetration projects on a single they can dissipate. Wires, the building block of distribution
distribution feeder are subjected to more scrutiny. Amongst lines, transformers, rotating machines are mainly limited in
the interconnection barriers is the DG potential to increase terms of the maximum temperature they can reach before
the fault current level of distribution substation or primary isolation get damaged or material melts. Fortunately, the
networks beyond the permissible short-circuit capacity of thermal mass of wires is significant and provides an important
substation/feeder equipment. grace period before an abrupt current rises causes overheating.
Up to recently, the grid was exclusively built with rotating For instance, a linnet conductor (ACSR, 336 kcmil) is rated at
machines that have quite a different behaviour under fault 510 A but can withstand 40,000 A for half a second before
than inverter-based sources. Consequently, current connection reaching its melting point[1].
impact assessments (CIA) often misrepresent inverters due to Semiconductor devices are also limited in terms of temper-
lack of experience or knowledge of these sources. ature at the junction but the thermal mass is very low and
This paper highlights the distinctions between rotating and must rely on their conductance with an external thermal pool.
inverter-based sources with respect to their fault current con- The thermal resistance between the junction and the thermal
tribution. A typical inverter short-circuit current model will pool is the key factor limiting power dissipation, even for
be presented as well as simulation results for a 7.5 MW DG short periods of time. Moreover, the internal connections in
connection study on a typical Canadian network. a semiconductor device consists of tiny wires that limit the
maximum current of the device. In this regard, semiconductor
II. S HORT-C IRCUIT AND D ISTRIBUTED G ENERATION devices are very sensitive to overcurrent but fortunately, they
Historically, power in electrical grid was flowing from can be turned off very quickly to re-route excess current before
centralized power generation plants to distributed load centers. the junction temperature becomes critical.
Power flowing in a single direction allowed for unpolarized For a breaker, two cases must be distinguised. A closed
(non-directional) protection schemes based on this assumption. breaker, that behave like any other conductor, and an opening
DG is shuffling the cards by adding generation points in dis- breaker. A breaker opening on a faulted circuit is exposed to
tribution grids, reducing the visibility of the load and in some the liberation of an enormous amount of energy. Neither before
nor after the fault the breaker operation should be significantly
Funding for this study has been provided by the Government of Canada affected by the fault current or distinguishing arc. The breaker
through the Program on Energy Research and Development (PERD). Dave
Turcotte is with the Canmet ENERGY, Varennes, J3X 1S6 – Canada is there to open the circuit before other components reach a
([email protected]) and Farid Katiraei is with Quanta Technology, damaging temperature level.
Canada ([email protected]). Consequently, to be free of any concern, the fault current
must remain present until the opening of the protection device

c NATURAL RESOURCES CANADA, 2009

2009-163 PP-PPR 411-MULINV, °
IEEE ELECTRICAL POWER CONFERENCE, OCTOBER 22–23, 2009, MONTREAL, QUEBEC, CANADA 2

3 Phase

3 Phase

3 Phase
ABC->G Logic

RMS

RMS

RMS
2 x 47 MVA Transformer
Fault
Timed X=18.5% each
A
End Mid Bus
COUPLED COUPLED COUPLED
B
PI PI PI CBbus Vs r 0.00199 [H]

CBsplit
CBPowerPlant SECTION SECTION SECTION C
Timed
Fault
Logic ABC->G

P+jQ
0.2 [MW] 0.066 [MVAR] 8 [MW] 3.87 [MVAR]
Lump value of 3 adjacent feeders
6 [MW] 2.91 [MVAR]
PC 60MW, 0.9pf

Power
Plant

Fig. 1. PSCAD model of the study system

in charge of fault clearing. Typical minimal operating time modulation (PWM). In all cases, the inverter bridge must be
for a distribution breaker is 3, 5 or 8 cycles. Consequently, a followed by a filter to take out harmonic distortion and allow
DG source capable of tripping within 50 ms would have no a controllable power flow.
effective contribution to the short-circuit capacity of a system. PWM inverters can quickly cease delivering power to the
grid by stopping gating the power devices. With switching fre-
III. M ODELING quencies of a few kilohertz, this provides tens of opportunities
In order to accurately assess the impact of the short circuit for disconnection in a single cycle. When dealing with short
contribution of inverters, a PSCAD model of a typical distribu- circuits, the two decision-making mechanisms for disconnec-
tion grid was developed as well as the model of a large-scale tion are under-voltage and over-current measurements. Table I
inverter-based photovoltaic plant. For the sake of comparison, summarizes the functions related to these two quantities in a
the same work was done with a synchronous generator model typical inverter.
representing a small hydro plant. The following sections An important consideration with inverters is the presence of
explain the various models implementations. an instantaneous over current relay (function 50). Solid-state
devices being acutely sensitive to over current (See section
A. Distribution Grid II), manufacturers must equip their converters with fast over
current sensors to avoid the power bridge to self destruct at
Figure 1 shows the distribution grid used for this study. It
the first abnormal situation.
represents a typical feeder used in Ontario-Canada, for which
Figure 2 shows the PSCAD core model of the photovoltaic
the short circuit contribution would be considered already too
inverters used for this study. It is essentially an average repre-
high. The transmission line is supplied by a substation of two
sentation of an inverter using a controllable voltage source (V)
47 MVA transformers with an impedance of 18.5% each. The
with a small resistance representing the resistance of the source
27.6 kV line spans over 25 kilometres with a total load around
(equivalent to switching and conduction losses). It is followed
15 MW. Adjacent feeders load (about 60 MW) are lumped as
by a breaker (GATES) that is used to simulate the gating status
a single load. The model is built for two fault scenarios i.e.
of the converter. Together the source and the breaker represent
a three-phase to ground fault at the transformer station or a
the converter bridge while not modelling the switching per se.
similar fault type at the line end. The DG plant is located also
The bridge is followed by a LC filter and a tie breaker (BRK)
at the feeder end.
that controls the connection of the inverter to the grid.

B. Inverter
There are various ways to implement large inverters. HVDC Iat A BRK

inverters are usually built around 6- or 12-pulse line-frequency L A

C

Ibt B
C

converters. On the other hand, modern low voltage invert- L B

C

Ict
ers are usually voltage source converters using pulse-width C
L C
GATES

V
C

TABLE I
T YPICAL T RIP L EVELS
R

R
V

Fn [2] Description Setting Time (ms) Vra Vrb Vrc

27 Undervoltage 88% 500

27HS Undervoltage 50% 30
50 Inst. Over Current 125% 5
51 Over Current 115% 100 Fig. 2. PSCAD Utility Interconnected Inverter Model
IEEE ELECTRICAL POWER CONFERENCE, OCTOBER 22–23, 2009, MONTREAL, QUEBEC, CANADA 3

Output Current (fault at t=2.5s)

Ivalves Iout
5.0
4.0
Vref
3.0
Hydro : Controls Exciter_(AC1A) 2.0
Excitation Speed VT

y (kA)
1.0
1.5 1.2 Ef0 IT 3
Ef If 0.0

3 -1.0
-2.0
Ef0 A
Ef If -3.0
V t (s) ...
2.480 2.500 2.520 2.540 2.560 2.580 2.600 2.620 2.640
...
S ...
1 1 Te
1.13254 1.055 DIST Fig. 4. Fault at the Substation – PV Output Current
Tm
w Tm
Trip signals (fault at t=2.5s)
Tm 0 27 50 51
1.20 27
Tm0
1.00
w w Tm 0.80
z z
51

Trip signal
0.60
Hydro Gov 1 Hydro Tur 1
A 0.40 50
1.002 Ctrl = 1 z0 zi
Wref Wref 0.20

B 0.00

Ctrl -0.20
t (s) 2.480 2.500 2.520 2.540 2.560 2.580 2.600 2.620 2.640 ...
...
DIST ...

Fig. 3. Synchronous Generator Fig. 5. Fault at the Substation – PV Protection Tripping Signals

C. Rotating Machine bus breaker decreases by interruption of the supply side. The
For the sake of comparison, a simulation model of a rotating system load is also represented by constant impedance models
machine based DG unit is also developed for the study. The that have no contribution to the fault. Although the fault
model represents a small 7.5 MW hydro DG based on a current reaches about 16.8 kA at 50 ms subsequent to the fault,
synchronous generator. The resulting PSCAD model is shown the fault current flowing through the bus breaker is only about
in Figure 3. 0.45 kA.
The model uses standard components from the PSCAD Figures 7 and 8 show similar graphs for the cases of
library. The synchronous generator is excited by the IEEE type connecting a photovoltaic (PV) or a hydro generation plant to
AC1A exciter [3]. The governor is based on a mechanical- the feeder that delivers 7.5 MW of active power at unity power
hydraulic transfer function. Multi-mass representation and/or factor. The hydro plant exhibits a 0.64 kA of fault current at
shaft torsional dynamics are neglected in this model since 50 ms and opens about 280 ms after the fault. On the other
the generator is small and sub-synchronous resonances are hand, the PV generation is already disconnected at 50 ms.
not of interest for this study. A tie-breaker is installed at As shown in Fig. 8 the bus current is slightly increased
connection point of DG and used to disconnects the unit. due to the presence of the hydro generation plant, 0.69 versus
Protections of the tie-breaker operate based on an inverse-time 0.45 kA. However, the fault current follows a trend similar
over current relay (function 51) for a fault at the PCC or along to that of the first case (no DG), 17.2 versus 16.8 kA. From
the feeder. The relay is set at 250A and uses the extremely these results, it can be concluded that presence of a 7.5 MW
inverse characteristic of IEEE C37.112-1996 [4]. generation unit at the end of a line has marginal impact on
a fault at the substation and almost no impact if it is a PV
IV. FAULT S TUDY generation facility.
The simulation results also shows that the first signal that
For the purpose of this paper, a fault is simulated at the initiated tripping of the PV plant was the instantaneous over
transformer station or at the line end. Current is measured at current protection of the PV inverter. Figure 4 shows a close-
three locations: at the fault (IF ault ), at the generation plant up of the output current with all the safety features disabled.
(IGen ) and at the bus breaker (IBus ). The measurements are Ivalves is measured just before the capacitor bank while Iout
done with no distributed generation, with a 7.5 MW photo- is measure just after. The peak current observed subsequent
voltaic plant and with a 7.5 MW hydro plant. In all cases, the to the fault occurrence is essentially due to discharging of the
plant is located at the line end i.e. 25 km from the substation. inverter output capacitors (the filters). Figure 4 also shows that
even if the inverter would continue to feed the fault, it would
A. Fault at the Substation be at a level around 1.4 times rated current of the unit. This
level is manufacturers specific and may varies in the range of
The first fault scenario investigated is a fault at the trans-
1.1 to 1.5 times rated current of the unit.
former station, upstream of the main feeder breaker (the
bus breaker in the model). Figure 6 shows the short circuit
current without any generation. Obviously the current in the
IEEE ELECTRICAL POWER CONFERENCE, OCTOBER 22–23, 2009, MONTREAL, QUEBEC, CANADA 4

3L-G, No generation, Substation 3L-G, No generation, End of line

IFault IrmsFault IFault IrmsFault
40 2.0
30 1.0
20
y (kA)

y (kA)
0.0
10
0 -1.0
-10 -2.0
IBus IrmBus IBus IrmBus
4.0 4.0
2.0 2.0
y (kA)

y (kA)
0.0 0.0
-2.0 -2.0
-4.0 -4.0
t (s) 49.90 50.00 50.10 50.20 50.30 50.40 ... t (s) 49.90 50.00 50.10 50.20 50.30 50.40 ...
... ...
... ...

Fig. 6. Fault at the Substation – No generation Fig. 9. Fault at the End of Line – No Generation

3L-G, 7.5 MW PV, Substation 3L-G, 7.5 MW PV, End of line

IFault IrmsFault IFault IrmsFault
40 2.0
30 1.0
20
y (kA)

y (kA)
0.0
10
0 -1.0
-10 -2.0
IGen IrmsGen IGen IrmsGen
0.40 0.40
0.20 0.20
y (kA)

y (kA)

0.00 0.00
-0.20 -0.20
-0.40 -0.40
IBus IrmsBus IBus IrmsBus
4.0 4.0
2.0 2.0
y (kA)

y (kA)

0.0 0.0
-2.0 -2.0
-4.0 -4.0
t (s) 2.40 2.50 2.60 2.70 2.80 2.90 ... t (s) 2.40 2.50 2.60 2.70 2.80 2.90 ...
... ...
... ...

Fig. 7. Fault at the Substation – PV Generation Fig. 10. Fault at the End of Line – PV Generation

3L-G, 7.5 MW SG, Substation 3L-G, 7.5 MW SG, End of line

IFault IrmsFault IFault IrmsFault
40 6.0
30 4.0
20 2.0
y (kA)

y (kA)

10 0.0
0 -2.0
-10 -4.0
IGen IrmsGen IGen IrmsGen
2.0 4.0
1.0 2.0
y (kA)

y (kA)

0.0 0.0
-1.0 -2.0
-2.0 -4.0
IBus IrmBus IBus IrmBus
4.0 4.0
2.0 2.0
y (kA)

y (kA)

0.0 0.0
-2.0 -2.0
-4.0 -4.0
t (s) 49.90 50.00 50.10 50.20 50.30 50.40 ... t (s) 49.90 50.00 50.10 50.20 50.30 50.40 ...
... ...
... ...

Fig. 8. Fault at the Substation – Hydro generation Fig. 11. Fault at the End of Line – Hydro Generation
IEEE ELECTRICAL POWER CONFERENCE, OCTOBER 22–23, 2009, MONTREAL, QUEBEC, CANADA 5

Trip signals (fault at t=2.5s)

1.20
27 50 51 at t + 103 ms. Low voltage detection is significantly faster
1.00 since the fault occurs close to the plant and thus causes a
0.80
27
significantly larger voltage drop. Instantaneous overcurrent is
51
still the first parameter to cause tripping. Figure 13 presents the
Trip signal

0.60

0.40 50
current evolution of the PV plant with safety features disabled.
0.20

0.00
V. C ONCLUSION
-0.20
t (s) 2.480 2.500 2.520 2.540 2.560 2.580 2.600 2.620 2.640 ...
...
This paper quantifies the fault contribution of a distributed
...
generation plant installed at the end of a long rural feeder. The
Fig. 12. Fault at the End of Line – PV Trip Signals
primary conclusion of the presented studies is that inverters are
capable of stopping delivery of power within the first cycle or
Output Current (fault at t=2.5s) few cycles subsequent to a fault. The fast disconnection of
Ivalves Iout
5.0 inverter-based DG units is achieved on the basis of utilizing
4.0
3.0
very sensitive and highly precise instantaneous over-current
2.0 protection schemes (function 50) supported by under-voltage
y (kA)

1.0
detection scheme (function 27). Not only they are capable
0.0
-1.0 of doing it but also they are constructed to do so in order
-2.0 to survive disruptive switching events and fault disturbances
-3.0
t (s) 2.480 2.500 2.520 2.540 2.560 2.580 2.600 2.620 2.640 ... occurring on the grid. As a result, the short-circuit contribution
...
... of inverter-based DG units are insignificant.
A secondary conclusion of the presented studies is that
Fig. 13. Fault at the End of Line – PV Output Current inverters, even with disabled protective functions, will feed a
current in the range of 1.1 to 1.5 times their nominal currents
which is significantly lower than the 4 to 10 times fault to
Figure 5 on the other hand shows the sequence of tripping nominal current ratio typically caused by rotating machines.
signals generated by the protection devices with their actions For a worst case scenario, the contribution of an inverter will
disabled. An action is triggered on the rising edge. The signal not exceed 1.5 p.u.
sequence is function 50 at t + 7.5 ms, function 27 at t + 24 ms The third conclusion of the paper is that while medium size
and function 51 at t + 103 ms. It is worth noticing that rotating machines are more prone to feed a fault, their impact
even if the instantaneous overcurrent detection fails, the low on the feeder breaker fault duty is limited or non-existent when
voltage detection is still within 50 ms and thus would provide they are located at the end of a line.
protection to the breaker.
R EFERENCES
B. Fault at the End of Line [1] Southwire Overhead Conductor Manual, 2nd ed., 2007.
[2] IEEE Standard for Electrical Power System Device Function Numbers,
The second fault scenario investigates the presence of a Acronyms and Contact Designation, IEEE Std. C37.2, 2008.
fault at the end of line. Figure 9 shows the short circuit [3] IEEE Recommended Practice for Excitation System Models for Power
current without any generation. Contrary to the first scenario, System Stablity Studies, IEEE Std. 421.5, 2005.
[4] IEEE Standard Inverse-Time Characteristic Equations for Overcurrent
an increase in bus current is definitely observed. Fault current Relays, IEEE Std. C37.112, 1996.
at the bus is of 2.44 kA at 50 ms while the current at the fault [5] J. Das, “Reducing interupting duties of medium-voltage circuit breakers
reaches 1.13 kA. by increasing contact parting time,” in Pulp and Paper Industry Technical
Conference, June 2007, pp. 257–264.
Figure 10 and 11 show similar graphs but considering
the presence of photovoltaic (PV) or hydro generation plant
delivering 7.5 MW and zero reactive power. The hydro plant
delivers 1.66 kA of fault current at 50 ms and opens about Dave Turcotte (S’92 – M’97) received his B.A.Sc. degree in Electrical
25 ms later. Assuming contribution to the substation equip- Engineering in December 1996 from Université de Sherbrooke, Quebec
ment, a slower response of the bus breaker to this fault would Canada. He joined the CanmetENERGY laboratory in Varennes in January
1997 where he has been working on various projects related to photovoltaics
actually be beneficial. This approach is proposed in [5] as a and power conversion. His current responsibilities include planning and
general measure to reduce breakers duty. conducting R&D to investigate the impact of utility-interconnected inverters
As for the PV plant, again it is already disconnected at on the electrical grid in order to ensure the adequacy of current and future
standards for distributed generation.
50 ms, confirming no contribution to the fault. It is worth
noticing that in all cases the fault current flow through the main
feeder breaker (IBus ) remains the same at about 2.44 kA. From
these results, it can be concluded that presence of a 7.5 MW Farid Katiraei (M’01 – SM’09) received the B.Sc. and M.Sc. degrees
generation unit at the end of a line has no impact on the fault in electrical engineering from Isfahan University of Technology (Iran) in
duty of a bus (feeder) breaker. 1995 and 1998 respectively. He received his Ph.D. degree also in electrical
engineering from the University of Toronto (Toronto-Canada) in 2005. He is
Again, the tripping sequence was simulated with trip actions currently manager for market area system testing at Quanta Technology. His
disabled (see Figure 13). The obtained results are: function research includes power electronic applications in power systems, distributed
50 at t + 7.4 ms, function 27 at t + 9.7 ms and function 51 energy generation systems, and microgrids.