Abstract--The connection of distributed generators (DG) to

distribution networks influences the performance of the networks. This paper focuses on the impact of DG on the feeder protection, specifically the impact on the overcurrent (OC) relay performance. The paper presents simulation results to show the extent of deterioration a DG can cause on the OC relay performance. The paper then presents an approach to solve this problem and restore the overcurrent relay performance.

Index Terms—Adaptive Protection, Distributed Generators, Feeder Protection, Over-Current Relay.

I. INTRODUCTION

t is expected that Distributed Generation (DG) is going to be an alternative for supplying power to some consumers.

Instead of producing power using remote and large generator units, power will be generated using a large number of small distributed generators to meet the load demand. These small generators produce power at low voltage level and are connected directly to the distribution network near the load center.

There are two main types of distributed generators in the market. The first type produces dc power and needs an inverter to convert its output dc power into ac power, such as fuel cells, micro-turbine, and photovoltaic cells. The second type directly produces ac power such as the diesel generators and wind turbines. In this study we will focus only on inverter interfaced DGs.

The distribution network topology, control and protection are designed assuming that power is flowing in one direction; from substation to loads. The connection of a distributed generator to the feeders of the distribution network can cause the power flow to be bi-directional instead of uni-directional affecting the network performance and stability in a number of ways [1-6]. In this paper we will focus only on the impact of DG on feeder protection system.1

Distribution feeders are usually radial with the loads tapped off along the line sections. The feeder protection strategy aims at optimizing the service continuity to the maximum number of users. This means applying a combination of circuit

This work was supported by the office of Naval Research ( ONR ) under

award number : N000014-00-1-0475 M. Baran is with the Department of ECE, North Carolina State University,

Raleigh, NC 27695 USA (e-mail: baran@ncsu.edu). I. Elmarkabi is with the Department of ECE, North Carolina State

University, Raleigh, NC 27695 USA (e-mail: imelmar@ncsu.edu).

breakers, automatic reclosers and fuses to clear temporary and permanent faults. The main circuit breaker at the substation end of the feeder and reclosers on the feeder are equipped with over-current (OC) relays to isolate any permanent fault along the feeder. When the current seen by the relay exceeds a certain pickup value the OC relay sends the breaker a trip signal, the speed of tripping the fault is inversely proportional with the fault current magnitude.

The connection of a DG to a feeder affects the feeder protection scheme in three different areas. These areas are the islanding phenomenon, the effect on OC relay performance and finally the reclosure to fuse coordination [3-6]. In this paper we will focus on the impact of DG on the feeder overcurrent protection relay, such impact will be presented in section two. While section three we introduce our proposed approach for solving such problem, with simulations showing the success of such approach.

II. IMPACT OF DG ON OVERCURRENT RELAY

PERFORMANCE

To illustrate the impact of DG on its feeder’s OC relay performance, the prototype feeder in figure 1 is used. The feeder’s load is assumed to be equally distributed along the feeder. The feeder is protected with an over-current relay, and the feeder power is supplied from the substation via a step-down transformer. (Relevant data in Appendix A)

Fig. 1. The 7 bus feeder

A. Reduced reach

Relays are set to protect a certain distance of the feeder; this is sometimes referred to as the “reach”. The reach of the relay is determined by its minimum pickup current. The presence of a DG will reduce the reach of the OC relay [6], thus leaving medium impedance faults at the end of the feeder undetected (figure 2). The reduction in reach is due to the fact that the presence of the DG increases the equivalent impedance of the feeder, thus decreasing the fault current for

Mesut Baran Ismail El-Markabi

North Carolina State University Raleigh, NC

Adaptive Over Current Protection for Distribution Feeders with Distributed Generators

I

DG

Load-2Load-4Load-6 Load-5

SubstationOLTCT15 3 247 6

Load-3 Load-1Load-7

R

OC-Relay

the same fault resistance (Rf).

Fig. 2. Relay reach with and without DG

To understand the effect of DGs on reducing the reach of

the feeder’s OC relay, the system in figure 1 is simulated for different fault locations. The total feeder load is 6 MVA, a 600 kW DG operating at 0.9 pf is connected to the feeder at bus 6. The pickup current of the OC relay protecting this feeder is set to 700 amps (appendix A).

Before connecting any DGs to the feeder, a fault at bus 7 with Rf = 10.4 Ω will create a current at the relay = 710 amps, and the relay will trip.

After connecting the DG, the same fault will create a fault current at the relay of 650 amps, thus the relay will not trip.

Figure 3 represents the maximum fault resistance at each bus that will create fault current that can still be detected by the relay at each bus with the DG connected at bus 6. For example for a fault at bus 7, the maximum fault the relay can detect is a fault with Rf = 9.4 Ω, any higher fault resistance will go undetected. From the figure we can see that for a fault with Rf = 10.4 Ω to be detected by the relay it has to be located at least at bus 6. If the fault is further down stream it will go undetected. For this fault resistance the reach of the relay is reduced from 100% to 83% of the feeder length. Figure 4 shows the effect of the amount of power injected by a DG on the reduction in the relay reach for three different fault resistances. The reduction in reach is increased with the amount of power injected by the DGs. For a specific Rf as the DG power increase the relay reach decrease, until the whole feeder becomes unprotected against that specific fault resistance. For example for DG total power equal 70% of total feeder load, a fault with Rf equal 10.4 cannot be detected along the whole feeder.

Fig. 3. Maximum Rf detected at each bus with DG injecting

10% of feeder load

Reduction in relay reach vs. DG injected powerfor different fault resistances

27

45

80

33

50

66

100

3

45

80

3

98

0

20

40

60

80

100

120

0 10 20 30 40 50 60 70 80 90 100

DG pow er as % of feeder maximum load

% r

educ

tion

in r

elay

rea

ch

Rf=10.4

Rf=9.5

Rf=8

Fig. 4: Reduction in reach vs. DG injected power

III. ADAPTIVE OVER CURRENT RELAY

As mentioned in section II, the presence of a DG will reduce the reach of the over-current relay, thus leaving medium impedance faults [12] at the end of the feeder undetected. To understand why the presence of DGs will cause such problem we simulated the prototype in section I once without any DGs and once with only one DG connected at bus 4; and created a fault at bus 8. We will then plot the fault current as seen by the relay in the two cases against the relay time-current characteristics.

The over current relay has an inverse time-current characteristic where the tripping time gets longer as the fault current decreases, and the tripping time is maximum for a fault current equals to the Ipickup of the relay. The following equation represents the relationship between the fault current and the tripping time [12].

( ) 2

1 TII

Tt

pick

p += α

Where tp is the time for pickup, T1 is a time constant depending on relay design parameters, T2 is a time constant that accounts for saturation in the magnetic circuit, α is changed according to type of relay (inverse, extremely inverse .. etc.), and finally Ipick is the relay set pickup current. We used the values for these constants that represent the ABB inverse time over-current relay type CO-8, where α = 2, T1 = 13, and T1 = 0.59 [9, 12]. Figure 5 shows the simulation of the relay characteristic using Matlab. In the figure we superimposed the relay characteristics over the fault current seen by the relay. The figure shows that for the case where there are no DGs connected to the feeder it will take the relay about 4.9 seconds to trip. This in fact is the longest trip time because the fault conditions where chosen to yield a fault current almost equal to the relay minimum pickup current. The figure also shows that the relay will not trip for the same fault conditions when a DG is present.

Reduced reach

7 6 5 4 3 2 1

DG

Original reach OC

R

Max Detectable R fault for I pickup= 700 amps 600 kVA DG at bus 6

5

7

9

11

13

15

7 6 5 4 3 2 1Feeder bus number

R fa

ult

Detected region

Undetected region

Rf = 10.4

Fig. 5: Relay characteristic and I fault

To solve this problem we are proposing an approach to change the relay pickup current Ipickup as seen in figure 6, such that the Ipickup decreases as the total amount of power injected by all DGs increase. Thus keep protecting the feeder against the same fault conditions it was originally protected against. But how will we get the values of the Ipickup along that curve?

To try to answer this question we need to understand how the DG will react to a fault at the feeder, the effect of that reaction on the fault current seen by the relay, and finally a tool to estimate the fault current at the relay when several DGs are connected to the feeder prior and during a fault. It is important to mention that this study is focused only on studying inverter interfaced DGs.

0 10 20 30 40 50 60 70 80 90 100

DG power as % of feeder maximum load

OC

rela

y I p

icku

p

Constant pickup current

Adaptive pickup current

Fig. 6. Constant and adaptive I-pickup

A. DG response under fault conditions

To understand the DG performance during fault conditions, we simulated the same example mentioned earlier with a DG connected to bus 4. The DG simulated was an inverter interfaced fuel cell controlled with a PI controller. As shown in figure 7; after the short circuit it took the DG only 110 millisecond to restore its P and Q set points, where Pset = 1 MW and Qset = 0 MVAR. Inverter interfaced DGs with more sophisticated control techniques are expected to have even shorter time duration to restore the output active and reactive powers to setpoint values. Therefore if we neglect the DG’s transients during the initial 0.1 seconds, we can conclude that during a fault on the feeder the DG will continue to supply the same amount of active power at the same power factor as it was just before the fault.

It is important to mention that although DGs will quickly adjust to fault conditions, yet some faults may cause the DG to disconnect from the feeder. If the location of the fault is close enough to the DG, that will cause the DG bus voltage to drop below acceptable limits, this will then trigger the under voltage relay protecting the DG to trip. The speed of the under voltage relay is stated in the IEEE P1547 standard, where the DG must trip within 2 seconds if bus voltage is 0.5 < VDG-bus < 0.88 p.u, and the maximum trip time should not exceed 0.16 seconds if the DG bus voltage is < 0.5 p.u. Therefore the status of the DG during a fault on the feeder will be determined via the under voltage protection relay.

Fig. 7: DG output power during a fault

In fact the DG reaction to fault that we just discussed is

also clearly stated in the IEEE standard 929-2000 [13] as follows:

“Solid-state inverters do not behave like rotating generators. Solid-state inverters have no inertia in their output and can respond immediately to changes in the ac power system. These inverters generally sense a short circuit by an associated voltage drop, rather than by actually sensing short circuit current. Thus, the voltage-trip recommendations are really the equivalent of short circuit protection.”

B. Effect of DG on fault current

Now we will try to understand the effect of the DG’s response to faults on the fault current seen by the relay. We simulated the example mentioned earlier with a 10 ohm fault resistance, and figure 8 shows the fault current at the substation end of the feeder; with and without DGs. The higher curve with fewer transients shows the fault current when there are no DGs connected to the feeder, as seen in the figure the fault current settles to a steady state value of 708 amps after only 20 milliseconds. The lower curve is the fault current when the DG is connected to bus 4, in this case the fault current settles to a steady state value of 660 amps after milliseconds. The figure also shows effect of the DG in the transients of the fault current.

Fig. 8. Relay fault current with/without DG

C. Solution approach

As mentioned earlier our solution approach depends on estimating the fault current at the relay, while taking into consideration the DG location and output current. Therefore we need first a system model. To determine our system model we will make two approximations, the first approximation is to neglect the transients in the DG response, that is neglect the DG response during the first 0.1 seconds of the fault duration. This approximation is acceptable since it is more conservative to account only for the steady state current. This approximation will allow us to model the DG as a current source with constant output power during fault duration.

The other important approximation is considering the fault current at the relay to be constant and equal to the steady state fault current value, as seen in figure 9. This will introduce an insignificant error that could be neglected. With these two approximations our system model can be represented as seen in figure 12.

Fig. 9. fault current approximation

Unfortunately, the standard short circuit analysis cannot

handle or account for a current source; therefore we need to find another technique to solve this circuit. We propose to use a power flow technique, where the current and voltage at every node can be evaluated for a given DG output. This will result in a set of nonlinear equations which can be solved using the Gauss-Newton method. Such approach is tested and its results were compared with the results from simulating the feeder using PSCAD, and the current value at the relay was found to be identical with ± 1 ampere accuracy.

Now that we estimate accurately the value of the fault

current at the relay, we can adjust the minimum pickup current of the relay depending on the feeder configuration. As seen in figure 10, when there are no DGs connected to the feeder in our example, we will set the relay to a minimum pick up current of 700 amps. When a DG is connected we can recalculate the value of the fault current, which for this case is 660 amps, and change the relay minimum fault current setup to 660 amps as seen in the figure. This adaptive technique will ensure that the feeder is always protected from the same range of faults whether DGs are connected to the feeder or not.

Fig. 10. Actual and modified relay characteristic with I fault

One last issue we need to address is the fact that some DGs

might trip during a fault for under voltage reasons. To resolve this issue we adopted a conservative approach, which assumes that all DGs remain online, even if some of them may trip during the fault. This decision is justified due to the fact that fault current seen by the relay will be higher if some of the DGs drop during fault. Therefore the relay will trip faster than the simulated if indeed some DGs trip. According to fig. 11 if we set the relay pickup current to fault current when 3 DGs are considered it will take the relay time “t2” to trip, but if only two DGs remain connected, the actual fault current will be higher and it will take the relay time “t1” to trip.

0 1 2 3 4 5

Time for OC relay

OC

rel

ay c

urre

nt

fault current with 3 DGs

fault current with 2 DGs

t1

Relay current-time relation

t2

Fig. 11. Relay minimum pickup current

IV. CONCLUSION

This paper illustrates the impact of DGs on distribution feeders’ protection, especially on the impact of DGs on the relay reach. Simulation results are presented to show that DGs can have a considerable impact on the reduction of the overcurrent relay reach. An adaptive overcurrent pickup

AC

Zfeeder

Loa

d-1

Loa

d-2

DG Loa

d-4

Loa

d-5

Loa

d-6

I1

I2

I3

I4

I5 I6

Zfeeder Zfeeder Zfeeder Zfeeder Zfeeder

I9

I8 I12I11I10I7

V1

Rfault

XS.C

Fig. 12: Feeder model for fault analysis

scheme is presented. The technique updates the OC relay minimum pickup current based on fault analysis of the system. And finally it is shown that the approach gives conservative results when some DGs get disconnected during the fault.

V. REFERENCES

[1] Nick Jenkins, Ron Allan, Embedded Generation, Published by the Institution of Electrical Engineers, London, United Kingdom, 2000, pp.50 – 93.

[2] G. Joos, B.T. Ooi, F.D. Galiana, “The potential of distributed generation to provide ancillary services” IEEE Power Engineering Society Summer Meeting, vol. 3, 2000, pp 1762 – 1767.

[3] P. Barker, “Determining the impact of DG on power systems, radial distribution”, IEEE Power Engineering Society Summer Meeting, vol. 3, 2000, pp 1645 -1656.

[4] M Ropp, M. Begovic, “Analysis and performance assessment of AFD method islanding detection”, IEEE Transactions on Energy Conversion, vol. 14, 1999, pp 810 -816

[5] M. Ropp, “Determining the effectiveness of islanding detection methods using NDZ”, Energy Conversion, IEEE Transactions, vol. 15, 2000, pp 290 -296.

[6] Roger Dugan, “Distributed Generation”, IEEE Industry Applications Magazine 2002

[7] K. Pandiaraj, B. Fox, “Novel voltage control for embedded generators in rural distribution networks,” Proceedings of the power conference on power system technology 2000, vol. 1.

[8] M. Suter, “Active filter for a microturbine,” IEEE, 23rd Telecommunications Energy Conference, INTELEC 2001, pp. 162 – 165.

[9] ABB website, Product technical guide 1ZSE 5492-104. [10] Arthur R. Bergen, Power system analysis, Published by

Prentice Hall Inc., New Jersey, USA, 1986, pp. 151 – 173.

[11] Manitoba HVDC Research Center, “PSCAD/EMTDC V2 User’s Manual”, Copyright Manitoba HVDC Research Center.

[12] P.M. Anderson, Power system protection, published by McGraw Hill, Inc., New Jersey, USA, pages 60-64, 214.

[13] IEEE Recommended Practice for Utility Interface of Photovoltaic (PV) Systems, IEEE std. 929-2000, published by IEEE, 2000.

VI. BIOGRAPHIES Mesut E Baran (S’ 87 – M’ 88) is currently as associate professor at North Carolina State University in Raleigh, NC. He received his Ph.D. from the University of California, Berkeley in 1988. His research interest include distribution and transmission system design.

Ismail M. El-Markabi is a graduate student in Dept. of ECE, North Carolina State University. Currently he is studying towards his Ph.D. degree, he received his M.S from NC state in 2002. He graduated from Cairo University, Egypt in 1997. His research interest includes distributed generation, power electronics control and PSCAD.

VII. APPENDIX A

Simulated System

Feeder: 11 kV, 7 bus feeder. Load: All bus loads are 0.75 MW@0.9 pf. Total feeder load is 6 MVA DG size: 600 kW fuel cell connected at bus 6 and operating at constant 0.9 pf. (DG is about 10% of feeder load) Utility bus is represented by a large synchronous generator Relay current: Without DG Ir = 325 A, with 1 DG Ir = 292 A, with 2 DG’s Ir = 260 A For each feeder segment R= 0.48, X= 0.2859. ∆V between bus 1&7=10% (max allowed) Relay pickup current: The minimum short circuit current for the system is 2.15 kA, and the maximum load current is 325 A. the relay pickup current is set according to the following relationship

322 min

max −<< −

−−sc

uppickload

III

The pick up current of the over-current relay must be greater than twice the maximum load current (650 A), and Isc min /3 = 720 A. So we will set the relay Ipick up = 700 A