Generator voltage output mismatch

[stratford]

Dear all, on site we have a 170MVA generator with 15kV voltage output, Y connection, earthed through transformer and resistor.
On this generator we notice from the first day of operation a consistent difference on the terminal voltage, meaning that on normal operation we measure phase A 14.99kV, phase B 15.15kV, phase C 15.18KV.
This is noticed during full speed no load (when the generator circuit breaker is open) and when the generator is synchronized to the grid via the step-up transformer.
Can you give me some points of searching or testing?
Voltage is measured on the terminal box of the generator (before the star connection) and on the brekaer side.

 

[JBinCA]

The magnitude of error is just over 1%, depending upon your accuracy class, this may not be an issue. Since the indication remains the same after you parallel to the grid, I don't think you have to worry that there is some problem in the generator windings that causes the issue. I think it's just indication inaccuracies.

The PTs ration can be checked using a TTR and the meter can be checked with a relay test set. I think you'll find that errors in one of the two components account for your issue.

As far as I'm concerned your readings seem within the normally expected error band.

Best Regards,

JB

 

[davidbeach]

If you have the opportunity to do so, roll the VT connections so that the present A VT is connected B on the generator, B to C and C to A. If the low voltage stays with the VT, it is the VT. If the low voltage stays with the phase, it is the generator. Probably the VT.

 

[waross]

If this is a high speed machine and you are able to check the air gaps you may find the cause of the discrepancy. I would first "Roll" the PTs as davidbeach suggests.
respectfully

 

[JBinCA]

If the gen did truly produce and unbalanced no-load voltage, wouldn't we expect that voltage imbalance to disappear when connected to the grid and show up as an imbalance in var production?

JB

 

[waross]

Good point JBinCA.
respectfully

 

[ScottyUK]

How well balanced are the line currents? Assuming this generator is connected to its own dYN1 GSU transformer this magnitude of imbalance should show up as an imbalance on the line currents. Your negative phase sequence relay would pick up on a significant imbalance in order to protect the rotor.

Are you measuring at the VT secondaries or reading data from a DCS screen? Measure the VT secondaries directly using the same meter for all measurements. After that the suggestions to 'roll the VTs' is a good one.


----------------------------------

Sometimes I only open my mouth to swap feet...

 

[JBinCA]

Is anyone particularly concerned by the magnitude of the error observed? I know I'm not, so I wanted to find out if I should be...

JB

 

[ScottyUK]

I would be if it is real. That level of voltage imbalance would cause load current imbalance, in turn causing high level negative phase sequence currents at double line frequency to circulate in the rotor forging. NPS heating very bad news for the rotor, especially since there is no direct temperature measurement on the rotor. The NPS relay is critical because it isn't backed up by any other relay which can protect the rotor. Most other functions have some degree of overlap so a single relay failure is covered by other functions to some extent.

So... is it real or just a transducer error?

 

[rbulsara]

I am wiht JBinCA. I do not see a problem.

 

[JBinCA]

ScottyUK,

I must say your answer surprises me. Not the discussion of concerns related to negative sequence currents, but rather the magnitude of voltage imbalance being a concern to you.

I'm used to seeing a negative sequence current relay/function set in the 10 % range. Granted, I usually deal with gens in the 5MVA range. Would you expect a 1% voltage mismatch to cause negative sequence currents around 10% of rated curent, or are the larger gens just protected more tightly? I had a gen manufacturer report that their unit could handle 70% negative sequence current a couple of years ago!

Of course, if 1% is too much error, the next logical question is what is the threshold that is tollerable? Then there's the question of PT accuracy. A pair of .5% accuracy PTs can't be trusted if they show a mismatch of 1% becuase one could be .5% high and the other .5% low at precisely the same voltage.

Looking forward to your response.

JB

 

[ScottyUK]

Hello JB,

On the subject of VT accuracy, I would expect the VTs to be a set of identical transformers. In my experience wound transformers of the same design, and very probably from the same batch, are usually very well matched even if their absolute accuracy might be Class 0.5. The chances of getting one at the top of the tolerance and another at the bottom is very small. Of course if the VTs are of different designs then it is quite likely that the matching will be poorer. I've not seen a generator supplied with different designs of VT from original build but I guess it could happen somewhere. Assuming that the VTs are of identical design, a significant mismatch in secondary voltage would suggest that the primary voltage has a similar mismatch.

The NPS current relay is usually set around 10% or so with a definite minimum time function. The UK Grid Code has a requirement continuous operation with a defined level of negative sequence voltage / current present. I'll have to check what our is (at home right now). We keep a close watch on anything unusual which could lead to generator damage simply because a big generator is such an expensive asset. Voltage imbalance is an anomaly which would get looked at because we do not typically see an imbalance greater than a few tens of volts.

We normally see a very well balanced set of voltages on the UK grid, at least where I've worked and where the network is robust. If I was seeing an imbalance where previously none had existed then it would set alarm bells ringing that something had changed on that unit. One probability with a lot of scope for causing huge amounts of trouble for maintenance would be high current joint on a bushing or at the neutral bar going 'high' resistance. With perhaps 10kA passing through the LV joints it does not take much increase before things become troublesome.

The OP said the imbalance had existed since Day 1, so it may not be a symptom of a progressive failure but equally could be a symptom of a problem from construction. Do the revenue metering VTs on the HV side show a similar imbalance? If not then it points to a problem local to the generator. Is there a sister unit in the plant of similar construction which can be used for comparison? Does it display an imbalance?

This discussion is proving interesting.


----------------------------------

Sometimes I only open my mouth to swap feet...

 

[waross]

How about the burdens on the VTs.
Are the VTs equally loaded?
respectfully

 

[stratford]

Tahnk you all for the answers and the help. On site there is another generator , connected to another separate winding of the 3-winding step-up transformer. This unit is from GE, 324MVA, but there the voltages are 15.73-17.78-15.79kV either at full speed no-load or synchronized to the grid. From the other hand the grid itself has an assymetry 403-407-403 kV on the high voltage of step-up transformer.
About the VT's used their are from the same batch at each generator. Fir the ST generator, 15kV raing, the same difference in measurements apply for both the VT's located at the start point and those located at line side of the generator circuit breaker. Finally i want to mention that the ST generator has assymetry at the currents as well 4238A-4072A-4106A.

 

[ScottyUK]

Assuming that 17.78 is a typo for 15.78, those values from the big machine look much more typical of what I would expect to see. Have you measured the secondary voltages of the suspect set of VTs or are you looking at a display on the turbine controls?

I've not seen a three winding transformer used as shared GSU for two large generators before. I assume it is a ddYN* design with the two generators each having a dedicated high resistance earthing arrangement? Is the 324MVA set the STG you refer to or is it the 170MVA set? Are you saying that one machine has a load current imbalance while the other machine on a separate winding of the same transformer does not have (significant) imbalance?


----------------------------------

Sometimes I only open my mouth to swap feet...

 

[stratford]

Scotty UK, this is a 3-winding Step-Up transformer, YNdd1.
The 2 secondaries are connected to each generator output, via generator circuit breakers. The 17.78 is a tyoe fault, stand for 15,78kV. Also both generators are earthed via trasnformers and recistance in order to limit stator earth fault. The generator with the voltage unbalance is from ST generator 15kV nominal rating 170MVA nominal.

 

[JBinCA]

Have to agree with ScottyUK that this post is getting interesting.

Regarding the issue at hand:

1. The voltages listed from the second generator have about a .4% error / mismatch.

2. The voltages listed from the grid have about a 1% error / mismatch.

3. The currents listed for the generator in question have about a 4.2% error / mismatch.

I can believe that an actual error (not a measurement error) of 1% in the no load voltage could result in a 4% error in phase currents as VARs are drawn into the machine from the grid due to the low excitation in the phase that exhibited the lower voltage. The only thing that doesn't add up for me is that the OP states that the voltage error is observed when gen is isolated and when it is synched to the grid. I would have expected the voltages to balance and the VARs (and current) to show the imbalance after synched to the grid. There cannot be an actual voltage mismatch between the gen and grid while in parallel without a current flowing throught the bus, breaker, and XFMR impedances trying to balance the two voltages. Of course, now we know the grid has a 1% voltage imbalance at the point of common coupling. Any thoughts here?

Stratford, do you have any additional indications you can provide? Per phase kW, VARs, KVA, and pf would be nice. Also, positive, negative, and zero sequence currents and voltages. I would think these values would be available on a system the size of yours. By having this discussion on the basis of voltages and currents alone, we are only able to guess what changes in VARs and sequence voltages / currents are driving the indications.

On to the side issues:

1. I Fully agree with ScottyUK that any change in an indication that has significant history must be explained. Since the OP stated this was a problem from day one, this maxim is not applicable.

2. I also agree that VT (usually referet to as potential XFMRs - PTs in US) arrays are almost always comprised of identical VTs. Exceptions to this rule are very few and are usually due to in service failures (rare) or cheap contractors making bad decisions to save a few pennies (more common).

3. I think my experience with VT accuracies tells a different tale than ScottyUKs, and I'm interested in finding out why... I expect it's because my experience is with the cheap end of the market.

I've ratio tested hundreds of VTs in service in industrial plants. Typical applications are gens under 5MVA, and services under 50 MVA. It would probably be worth me reviewing the last year's data sheets, but off the top of my head I'd say errors in the 1% range are pretty common.

Regards,

JB

 

[scottf]

What are the rated accuracy class and burdens of the VTs?

In general, VTs loaded with the same ratings (regardless of manufacturer) that are loaded with the same burdens would not exhibit that much of a difference between units.

For example, if a VT is rated 0.3% accuracy up to a burden of 75 VA, then most VT manufacturers will place the 0 burden point as low (positive error) as they can in the accuracy class and place the 75 VA point as high (negative error) as they can.

As was said above, it is not likely to have VTs that have the same burden being reading at opposite ends of the accuracy range...in fact it's pretty near impossible.

If the VTs have a metering accuracy rating, are loaded with the same burdens, and are wired correctly, I would not expect to see a 1% difference between 2 of the outputs due strictly to VT errors.

 

[davidbeach]

I'd go back to my first post in this thread. Something is seriously wrong, but what isn't known is whether it is the VT or the generator.

<aside>
While PT is common usage in the US, the IEEE standardized on the use of VT sometime in the late 70's. The use of PT was/remains so deeply ingrained that PT is still more commonly used than VT.
</aside>

Rolling the VT connections will easily allow determination if the problem is the VT or the generator. At no load (no current, therefore no voltage drop across the internal impedance of the generator) you will only get different voltages per phase if you have a different number of turns per phase or some geometry problem in the stator.

Differing voltages while current is flowing could point to differing impedances of the three windings, but with no current you have the internal voltage at the terminals if internal leakage currents are neglected.

Maybe, before rolling the phases, a TTR test of the VTs might provide an explanation. The possibly bad VT shows the possibility of too many primary windings or too few secondary windings. Fortunately that wouldn't suggest shorted turns on the primary, but could suggest shorted turns on the secondary.

But in any case, it is imperative that the problem be isolated and determined to be either a generator problem or a VT problem. Until then we are just speculating.

 

[JBinCA]

Scottf,

Your answer (assuming I understand you correctly) implies that each individual VT is tested for accuracy at the two burden points, then modified as necessary to adjust the error in the desired direction prior to leaving the factory. I'll freely admit that I don't know, but I would be surprsed to learn that such a process takes place.

I figured that some significant effort went into developing the manufacturing process for a given model of VT, then the process is baselined to ensure that it produces XFMRs that vary radomly around the desired error / accuracy target within the acceptable range. Sure, Testig of ratio and insulation, etc take place before a unit is shipped out, but I would be surprised to learn that a manufacturer reworks each VT to get ratio accuracy closer to 0 error if the unit is within the accuracy band. We're talking about relatively cheap commodities here - especially at the low end of the market.

Anyone else have thoughts on this point?

Thanks, JB