Island Mode Generators - Load Acceptance

[Rodmcm]

A facililty has several old gensets plus a couple of new ones. They run them in an 'n-1' config for load.They have had several instances where if a genset trips then the others do not easily pick up the load. The fairly long delayed automatic load shedding comes into play and they loose big sectiond of the distribution system.
I have suggested to them that maybe the old governers/AVRs have gone out of spec for response times. Can anyone the suggest any other thoughts? Also, except for off line drop/accept load tests are there any other test methods to determine response. Gensets range from 700kW to 2000kW at 11kV

 

[dpc]

The engine and governor probably have more to do with this than the AVR, unless the load shedding is voltage-based.

It would help to know how the governors are configured to share load and the droop settings.

Are these engines or steam turbines?



David Castor

http://www.cvoes.com

 

[Rodmcm]

Load shedding is frequency based. Governors are a mix of hydraulic and electrohydraulic. Load sharing frequency control is by a PLC operating directly on speeder motors for the old ones. Droop about 4 or 5% I think ( will check)

 

[waross]

Did you mean N+1? N-1 will have the generators overloaded even before one drops off.
Are these sets old and tired? If they carry and share the load normally, but can't hold frequency with one missing it may be a sign that what you think is N+1 is actually N+0.5 or less.
If you can't hold frequency then you are quite likely below N. One more set to come up to N and one more for N+1.
As well as being old, light diesel fuel may result in the sets being unable to deliver nameplate kW. Rather than depending on nameplate values of capacity you may consider actually testing the maximum output of the sets.
The test is quite simple and does not require a shutdown.
When three or more sets are running, advance the throttle of one set to the maximum. You may do this by increasing the frequency or speed setting until there is no further increase in kW output. Read the kw meter and return the throttle to the normal position. Repeat with the other sets. Caution, do not do this test when the load is so light that there is a possibility that the set under test may develop more than 100% of the connected load.
This is the maximum output of the set and should be 110% of the nameplate rating for a prime rated set. If the maximum out is less than 110% of the rating, then the prime mover is no longer delivering its rated power. Either repair/rebuild or relabel the nameplate temporarily (Duct tape and a felt pen will be good) to 100/110% (91%) of the maximum output. Use the new figure to determine the number of sets required for N+1 loading.

Bill
--------------------
"Why not the best?"
Jimmy Carter

 

[rbulsara]

Yes, it better be N+1, at least and not N-1. There may be more than one issues; the governor and related settings, old age of gens and most importantly the logic of the load management or "spinning reserve".

You cannot have a simple N+1 logic with different sizes of generators. Dropping off a 2000kW unit is equivalent to dropping (3)700 kW units and a simple N+1 algorithm will not work. You need to verify that it is written to take care of dropping out the largest unit running at any time and the system is capable of supporting dropping out a 2000kW unit. Failing that what the system is doing (shedding and adding load) is the expected response. At least that part appears to be working.

As for picking up the load, each generator needs to be individually tested for "step--load" (or block load) response and go from there. Upgrading the governor to electronic ones may be on the cards and overhaul of the old generators and its fuel pumps as well.



Rafiq Bulsara

http://www.srengineersct.com

 

[dpc]

Load sharing frequency control is by a PLC operating directly on speeder motors for the old ones

This sounds like trouble. PLCs can be slower than you think.

There are other here who know a lot more about governor systems than I do, but obviously, governors were sharing load long before PLCs were around.

David Castor

http://www.cvoes.com

 

[Rodmcm]

Gentlemen... The common definition as used in this part of the world, is n-1 means that the total load will be covered with the loss of the largest machine... However, we are all talking about the same concept


Your thoughts on governor response same as mine. Will try out maximum load idea... Thanks Waross

 

[DANSES]

Dear Rodmcm,

I am not agree. Engineering document call this "N+1".
Anyway, It's difficult to set fast acting load sheding for Generator using "mechanical governor" instead of "electronic governor". As far as I know to parameter that you shall be consider are "voltage" and "frequency".
For keep your continuity you can use load bank during addjustment.
Also, at some not essential load you add "undervoltage relay".
It will be trip some load during recovery by other generator.
If the generators running parallel have became stabil, you could go to load panel which have been tripped by under voltage and you reset to "ON" position.
It is good for your system that may be have a less spinning reverse.

 

[waross]

We had an electronic governor paired with a hydraulic governor for starting some large motors that could not be started on either generator alone, even with soft starts.
On a cold start the hydraulic unit could not match the electronic governor but after about 20 to 30 minutes warm-up time the sets worked and played well together. This was quite severe block loading.
While electronic governors may be fast they may not be as fast as you think and hydraulic governors use hydraulic force acting over a small distance and may be faster than you think.


Bill
--------------------
"Why not the best?"
Jimmy Carter

 

[desrod]

A simple contingency design (N-1 in Canada) might not be enough for all generation scenarios.
We used to plan for the biggest unit loss but for small generation pool, the margin have to be increased as the potential loss to total load ratio becomes more significant e.g.: loosing a 3 MW gen at 30 MW load is not the same as loosing it at 10 MW.

It is a function of both generation inertia, change in generation, load level and even load management (shedding) schemes.

We usually perform stability studies to make sure the entire system perform correctly.
Here's an example of varying spinning reserve and how it was estimated:

http://www.carilec.org/Assets/.../Spinning Reserve Presentation.ppt

Daniel D.

 

[catserveng]

Bill,

How true that statement is about the electronic vs hydraulic governors. Recently dealt with an issue where an older unit was upgraded from a UG8 to a 2301D and a UG actuator. When tested end user couldn't figure out why response had not improved. Had to explain there was still the mechanical response to deal with. In the end by making some adjustments we did get some slight improvement in dynamic response, and improved cold start overshoot and stability.

To the original post, did it work well at one time and does not perform as expected? Could be a matter of needing to go back thru each unit one at a time, and do a good tune, evaluate it's actual response, and make dynamic adjustments to controls to make the units respond as similar as possible. In some cases with a mixed system it is possible the "smart" governors may need to be dummied a bit, and the oldest units may need a closer review to assure they can keep up with group.

When I do projects like this I usually take one unit at a time and do some block loads on a load bank, preferably resistive/reactive to also get the best possible adjustment on the voltage droop as well. Use a chart recorder and record the voltage and frequency variation during the load changes. Try to come up with a common response characteristic. The min and max values will likely differ between the units, but you may find if they dip, then slightly overshoot, then settle in about the same time span, your overall system response will be about as good as it can get.

As for PLC based overall control systems, there are lots out there, Enercon, CAT/ISO and others have been doing them for the size range engine driven systems as described by the OP. Main issue I've seen with them is the PLC programmer may not always understand why the engine doesn't respond as fast as the PLC sends the signal, then they usually keep trying to speed things up and get into trouble.

One other note, depending on the governor models used, the speeder motor input linkage may be wearing, many original designs were not intended to be constantly adjusted, and they do eventually wear, even on models like the Woodward UG and PSG governors. Ends up getting a windup and further delays proper response.

Hope that helps,

Mike L.

 

[Rodmcm]

Thank you
Desrod - Cannot open link

Catserveng - We unfortunatley do not know whether it originally worked or not. As to testing, agree with your methodology. the problem lies that some of the gensets are 11kV so presents a problem to do on a remote island.

 

[racookpe1978]

Do you have the actually logged/plotted frequency responses (of the network) as the electric load increases AND the speed losses (time delays ?) of the prime movers as the frequency drops, then attempts to recover?

That is, are you sure you are trouble-shooting/anal;yzing/theorizing the right problem?

 

[Rodmcm]

Could you please amplfy what you mean, The network cannot be divorced from the generation on an island system would have thought.

 

[desrod]

I uploaded the presentation.
Daniel D

 

[Rodmcm]

Thanks