Islanding detection on grids with distributed energy resources

[chao_david]

There are lots of existing methods for detecting islanding conditions when the circuit breaker at the side of the utility unintentionally trips. I've seen passive methods, active, and intelligent systems that detect the islanding and trips the DG breaker within 2 seconds as required by IEEE standards. I think I'm missing the basics here because from the diagram, why not just coordinate the tripping of DG CB2 and utility CB1? Why not just automatically trip both CB1 and CB2 when there is a fault at the CB1?

I've thought about this because active method equipment are expensive to install and would distort the power quality.

 

[davidbeach]

It may be impossible for relays at CB2 to see all faults for which CB1 needs to trip.

I’ll see your silver lining and raise you two black clouds. - Protection Operations

 

[chao_david]

Is there no such thing as tripping CB2 on the basis of the status of CB1? Basically, CB2 won't have a fault sensing ability on the utility side but will just "copy" the status of CB1.

 

[wcaseyharman]

Inverter based resources are pretty limited when it comes to faults:
They are limited in overcurrent capability - 110% of rated output does not seem uncommon. This makes overcurrent protection almost impossible to apply since maximum load current and fault current are practically indistinguishable. For the battery/microgrid I am involved in the company used sequential undervoltage tripping instead.
They may have high negative and zero sequence impedance, mean they will not feed unbalanced faults. In addition, voltages may become highly unbalanced as well, possibly leading to customer overvoltages (may not be “effectively grounded.) in this case it’s even more important to quickly isolate the battery to prevent customer damage and possible system surge arrestor failures.
For all the reasons above, in my utility it’s
required to install fast transfer trip for larger distributed generation particularly for batteries and other inverter based resources to ensure they are isolated during faults.

 

[jghrist]

Why not just automatically trip both CB1 and CB2 when there is a fault at the CB1?

CB1 is at the utility substation and is often remote from CB2. Local methods to detect islanding will eliminate the communication expense needed for direct transfer tripping.

 

[Hoxton123]

This is an extremely complex area.

Firstly your single line diagram does not show distance, the breakers could be many kilometers apart. Intertripping reliably over that distance could be difficult / expensive.

The point at which the system is disconnected may not be CB1, it could be much further away and at a higher voltage level. The disconnect may not be a circuit breaker. In the UK we had a major outage in 2019. A lighnng strike affected a 400kv substation, and 400 kv CBs opened. 2000 MW of generation was disconected and the system low frequencey relays operated to remove load before the whole system shout down. intertripping for that scenario would be difficult and expensive.

The planning is moving towards getting the distributed generation to 'ride through' events where possible to help the system recover.

 

[FreddyNurk]

The fundamental is that the plant may be islanded without CB1 opening, it could be something upstream that causes a blackout and CB1 won't open, rather it'll remain closed and reconnect when the network returns. There's often no obligation on the utility to put a UV element on the CB, and if they did they'd need to be contacted to restore power every time it went off.

Whatever the reason, the plant must have the means to detect when it's islanded at it's point of connection, and a means to automatically disconnect when an islanding condition is detected.

EDMS Australia

 

[stevenal]

s there no such thing as tripping CB2 on the basis of the status of CB1?

Sure, it's known as transfer tripping. As Horxton123 and FreddyNurk indicated, CB1 may not be the only location an island might form. Even if the islanded load exceeds the DG capacity, tripping might not occur within 2s, and it may cause power quality issues.

 

[cranky108]

I have attempted to use frequency as a determent of grid detachment. It is not perfect, but it works most of the time. That said I have a synchronous machine for a generator, not an IBR. There are other ways to determine detachment, but can only go more costly.
With IBR's, so many are not on-line all the time, so real coordination may not be possible. What I hear is being done in other countries, is a requirement that IBR's produce negative sequence.

 

[Hoxton123]

In the UK we have fixed on ROCOF (Rate Of Change Of Frequency) relay to detect grid separation. There will usually be a power flow at the separation point. Interupting this will result in a step change in genset frequency before the governor corrects it. The ROCOF relay should detect this.

If the relay does not operate, then other factors will detect the separation, voltage etc. Sychronous machines usually operate with the generator excitation in power factor control, separation from the grid will result in voltage to drift out of limits.

Frequency alone is not a reliable indicator of separation. It is possible to have a small generator operating isolated from the grid, but supplying a small local load. Unless the load exceeds the generator capacity shutting the generator down on overload, then the generator may run on. Uttilities do not loke ths for safety and other reasons

 

[wcaseyharman]

At my utility I haven’t heard of too many distribution connected generators island, mostly I think because of the transfer trip requirement. We’ve had at least four transmission level unintentional islands. Voltage and frequency have not worked in all the transmission island causes due to the governor frequency response and fast acting exciters.

 

[cranky108]

Locally, the generation we see most by customers is solar panels, so the interface is with an IBR, which falls under IEEE requirements.
The only generation that is sizable on the distribution system is our own hydro units. The larger units are synchronous, the small is induction. So if the units stay on, it's not much of a detail, as all the synchronous units should be able to pickup some load and stay on. That said, they will likely need to drop load to connect back to the grid.
The largest unit like this is a 28 MVA, and will pick up load and keep running (unforeseen tested), which is amazing as it normally runs at very low load, and mostly as spinning reserve.

The induction units will drop, unless the load, and frequency, and voltage, and PF are just right. Which is not very likely.

 

[bacon4life]

Our distribution grid is designed so that any customer can be fed by at least one alternate substation by reconfiguring the 15 kV utility grid. Thus rather than a single breaker as you have shown as CB1, we actually have a primary breaker plus 1-3 alternate breakers that could feed the customer site. One solution is that the DG has to stay offline if the customer is fed from one of the alternate breakers.

We had one case of unintentional transmission level islanding where UFLS relays repeatedly shed and reconnected enough load to keep the hydro generators within their frequency limits.

 

[cranky108]

Interesting that UFLS would reconnect load, as parts of the US the plan will not allow that.
I am in one of those areas that does require a reconnect of some of the load automatically. But as the plan does not specify any time limits for that, I have inserted a limit timer for a reconnect. No need to surprise an operator.

 

[bacon4life]

Within the Northwest Power Pool we have three restoration blocks with defined timing:
load/Setpoint/time delay
1.1% 60.5 Hz 30 sec
1.7% 60.7 Hz 5 sec
2.3% 60.9 Hz 0.25 sec

 

[crshears]

"UFLS - Under Frequency Load Shedding" - I have never encountered nor even heard of UFLS incorporating load restoration; bloody dangerous in my view.

CR

"As iron sharpens iron, so one person sharpens another." [Proverbs 27:17, NIV]

 

[davidbeach]

Well, you’re in the Eastern Interconnect. The Western Interconnect makes provisions to deal with overshooting the target. A major event that could result in load shedding is probably going to result in a lot of RAS action. If connections between load and generation become severed the brake may be activated and in the short term the frequency may be all over the place. As things stabilize it may be that the slower blocks are subsequently shed or it may be that a certain amount of auto restoration is needed.

The comparative differences between the two interconnects means that each does things that would not work on the other. Overshoot response may be one of those things.

I’ll see your silver lining and raise you two black clouds. - Protection Operations

 

[cranky108]

Besides the West plan was designed by a blue ribbon presidential committee. Who designed the plans in the East?

 

[bacon4life]

The load shedding has to occur very fast, so it can be hard to trigger just the perfect amount of load to be shed. If too much load is shed, automatic restoration is in place to address frequency overshoot. The total armed load to shed is about 30% of armed load and the automatic restoration is only 5%.

The automatic restoration circuits must be designated as part of the UFLS plan. All other restoration must be manual.

The screenshot below from the WECC 2019 UFLS report shows a case where large amounts of load had to be shed to keep the frequency nadir above 57.0 Hz, but the momentary overshoot reached 60.8 Hz and the final frequency ended up at about 60.3 Hz.

 

[stevenal]

Some of the customer load shed by UFLS may need to be restarted manually on the customer side. Automatic restoration of 5% base on the amount shed may miss that mark.