Transmission ground distance relaying with distribution underbuild

[stevenal]

How do you handle the mutual coupling?

1. Ignore it.
2. Carefully model the distribution circuit(s) and mutual(s) in your fault program.
3. Model just the distribution neutral in your fault program.
4. Measure the line impedances including the mutual coupling with test equipment.
5. Use actual fault data to estimate the parameters.
6. Avoid using ground distance. (In favor of?)
7. Something else. (What?)

Thanks for sharing.

 

[System Protection]

I model it all in the line constants program that I use as its not much more effort to add the underbuild.

But I don't input the mutual impedance between transmission conductors and underbuild into my fault sim program as the distribution system is not modeled.

Definitely would recommend modeling the neutral conductor though.

I usually set my ground distance reach quite a bit less than the phase distance reach to account for modeling errors.
Coordination is much nicer with ground distance; I wouldn't recommend going without.

If you have transfer trip or a differential then you could use ground distance as a backup only if you had concern over your model.

Verifying faults with known origin can be used to verify your model; there is an SEL paper on it somewhere.

 

[stevenal]

SP,

How much less?

Yes we are using differential protection, but wish to have step distance protection if/when the differential channels fail.

Davidbeach please chime in.

 

[System Protection]

Typical would be 85% reach for Phase distance and 70% reach for Ground distance for Z1.

 

[davidbeach]

The distribution phase conductors have very little to do with line impedance, but that neutral is just line any other "shield" wire. Include it in your parameter calcs the same way you would an overhead shield. Also, don't blindly assume rho earth is the conventional 100 ohm meters. I find that 10, here, gives better results. Others have found that on essentially solid granite something more like 1000 gives better results. Our "real" Z0 is around 80% of what would be calculated just looking at the three transmission conductors and assuming 100 for rho earth. But it's a line variable, not a line constant. Soil moisture plays a big part in the actual value; Z0 in August and Z0 in March can be very different. Calculate the March value so you don't over reach and accept the shorter reaches in August.

Where the mutual coupling between the underbuilt distribution and the overbuilt transmission gets really fun is when a distribution ground fault induces circulating 3I0 is a transmission loop. Can totally confuse ground directional elements that expect a common source for the I2 and I0 currents since the I0 is just induced and the I2 comes from the low-side of the delta-wye transformer that's feeding the ground fault.

 

[bacon4life]

Historically I ignored it based on advice that ground fault relays were typically less accurate than phase relays.

For a recent fault I analyzed, the distance shown by the ground relay element was much shorter than the actual fault location. Once I recalculated the line impedance to include the distribution underbuild, the updated zero sequence impedance was about 19% lower. The new calculation matched the recorded fault current to within 100 amps.

My region has very few lines with shield wires on 115 kV lines, so we probably have more error than regions where transmission always includes a shield wire.

Adding the data for mutual impedance to the shared regional Aspen oneliner model would be somewhat challenging. Typically each transmission line is several miles long, but the underbuild often changes configurations multiple times within a single mile. I doubt my neighbors would appreciate me adding that much complexity to the model, even if I had a way to pull the data out of GIS.

I like the idea of just adding the neutral conductor. That would be way simpler than adding mutual impedances.

 

[stevenal]

I found the the SEL paper at https://selinc.com/api/download/99422. Not reliable for determining Z0 on mutually coupled lines they say.
Due to a setting error, both relays reported zero distance to fault. The line patrol found nothing. 87L operated at both ends. One end picked up on Zone3 also. No distance elements picked up on the other end.
Using the fault program, I found the best fit location.
Using that location, I was able to reproduce the results by stepping up my original Z0. A multiple of 6.5 worked.
Seems like if I added a second shield wire at the neutral location or reduced rho I would only be reducing Z0, not increasing it.

 

[System Protection]

How are you finding the best fit with your fault program?

Could the fault resistance be the reason your Z0 seems low?

 

[davidbeach]

How do your Z1 and Z0 compare? I've never seen that much discrepancy between values arrived at using the more conventional rules of thumb and those arrived at using more detailed analysis. Any unmodeled ground sources in the neighborhood?

 

[stevenal]

How are you finding the best fit with your fault program?

Could the fault resistance be the reason your Z0 seems low?

Fault program has a routine to find fault location by inputting fault current. It runs lots of faults and spits out the best fit location.

I can get the zone results by putting in fault resistance of 7 ohms. This throws the impedance line way off to the right, though. SynchroWave shows it much more reactive.

 

[stevenal]

How do your Z1 and Z0 compare? I've never seen that much discrepancy between values arrived at using the more conventional rules of thumb and those arrived at using more detailed analysis. Any unmodeled ground sources in the neighborhood?

Z1= 0.00617 + 0.01717j pu
Zo = 0.01426 + 0.06091j pu

No unmodeled ground sources.

 

[System Protection]

Fault program has a routine to find fault location by inputting fault current. It runs lots of faults and spits out the best fit location.

I can get the zone results by putting in fault resistance of 7 ohms. This throws the impedance line way off to the right, though. SynchroWave shows it much more reactive.

Another fault location method is to compare the ratio of 3I0 current from each end of the line.
That should help account for fault resistance.

You would use Synchrowave to get the 3I0 current from each end. Divide one end by the other.

Then run intermittent faults in your fault sim program until you get a ratio that matches.

I'd be curious if it lines up with your fault program location. (Aspen?)

 

[stevenal]

Another fault location method is to compare the ratio of 3I0 current from each end of the line.
That should help account for fault resistance.

You would use Synchrowave to get the 3I0 current from each end. Divide one end by the other.

Then run intermittent faults in your fault sim program until you get a ratio that matches.

I'd be curious if it lines up with your fault program location. (Aspen?)

The I0ratio during the fault is about .53. I cannot get that ratio in Oneliner with a fault inside the differential zone.

 

[bacon4life]

As I experimented with various configurations, it seemed like X0 and R0 sometimes go in opposite directions for a change in circuit configuration. I don't recall seeing anything change by a factor of 6.5 though.

Perhaps there could be current flows on the distribution circuit cased by the unbalanced source voltage? I haven't quite wrapped my mind around whether there would be a current induced downstream of a delta-wye ground distribution transformer for various kinds of transmission voltage changes.

 

[davidbeach]

I'd be interesting in taking a look at it, you know how to reach me...

Is the line in question in the regional model? If not, enough of a model that I can connect it in would be helpful.

 

[stevenal]

Suppose an evolving fault could throw off the distance elements? I see in the "pre-fault" area of the plot the phase currents are rather unbalanced looking, and 3I0 is not as close to zero as I would expect on transmission serving delta windings. About 17 A primary before shooting to 3000. Normal 3I0 today looks to be 0.1 A. Just wondering.

 

[System Protection]

Any chance this started as a distribution fault? Could explain the prefault.

I remember seeing some strange prefault waveforms and it was due to the underbuilt distribution faulting first.

Or perhaps evolving into a transmission to distribution fault?

 

[stevenal]

I'm seeing nothing simultaneous, but a 17 A line to ground contact would not be picked up as a fault on distribution.

 

[davidbeach]

Ok, the pre-fault unbalance is just a real line in the non-textbook world. Your transmission provider doesn't do much transposition and I'm going to guess that you don't either. Nobody does the "continuous transposition" thing and so balanced 3-phase systems are a useful fiction but not found in the wild. Load current distribution in loops can be drastically impacted by minor differences in per-phase impedances.

Based on "other information" the line in question is part of a three line loop. The loop has a single source at one station, call it "F". From there you have lines to "G" and "H". There's also a line from "G" to "H". Those three lines spend a fair amount of time as double circuit lines, same poles, opposite sides, and a certain amount of time on opposite sides of the road from each other. Lots and lots of mutual coupling.

If those mutual couplings aren't calculated and included in the model you're not going to get good information from the model about where faults happen, nor what the apparent impedance is from any terminal to any fault location. I'm pretty sure you didn't have a 7 ohm fault resistance, but rather 7 ohms mimics the effect of the mutual impedance, for that specific fault location. If no damage was found, your location with a 7 ohm fault may be wildly off the mark.

If the fault was on G-H and there's lots of coupling between F-G and G-H, and much less between either of those and F-H you'll get all sorts of weird current distributions for ground faults that could be significantly different than the current distribution for 3LG or LL faults. LLG faults can also play havoc; phase distance elements are selected by the relays but the mutual coupling plays its weird tricks.

Mutual coupling to the underbuilt distribution probably had nothing to do with the issue, but mutuals to other lines definitely have to be modeled.

 

[stevenal]

I can see there'd be some mutual coupling between F-G and F-H as they run down opposite sides of the same road as they exit F for several spans. G-H does not share a common path with the others, although the distribution underbuild may link them. I'll see what I can do to model the F-G and F-H mutual.
Thanks.