How best to measure frequency dip in response to load change

[DLClark]

This question has to do with power quality best practices. If you have a large engine-generator, one of the tests should be to measure its response to change in load such as, for test purposes, 0% to 50% load step. In my case the test method is not specified so I developed a test using the power suite measurements which run on an FPGA device that is controlled by a PXI computer.

I observe that the EPS frequency measurement is filtered to agree with an international standard so my measurement of frequency dip during the load step is, say, 97% of the 50 hz line frequency.

I also observe that if I use rpm as a measurement of the dip, it will dip to a lower percentage of the standard rpm. Or I can convert the rpm to frequency (engine shaft is locked to generator shaft) . The calculated frequency can then be 96% which may fail the test.

My question is: does anyone on the forum perform this kind of test? Since the customer only specifies frequency and power quality as measurements of pass and fail criteria I designed the test this way and did not use rpm in the calculation. The loose end here is that we could use frequency from the IEC certified equipment or I could use the "actual" frequency developed from shaft rpm. The customer doesn't specify. So, what is the standard practice for this type of measurement?

Thanks for reading this too-long post.

Dave

[James_McN]

Hi David,

Are you talking about the FPGA resample (that gives n samples per cycle?)

If you are I would warn away from that. It looks to lock onto a 50Hz signal by filtering around that and using a PLL. This means it doesn't deal well with step changes in frequency and amplitude as it takes time for the PLL to react.

There are of course other ways to derive the frequency electrically and I've not come across anything hard about whether to measure mechanically or electrically but certainly the resample isn't the bet way.

My immediate thought though is electrically makes the most sense seeing as that is the main concern and so gives the best future proofing. If the change a design to add some battery storage for example then electrically still works but mechanical doesn't.  (Obviously a huge number of assumptions in this paragraph!)

[DLClark]

Thanks for the reply, James. Your comment about the frequency measurement using the PLL makes sense. Does this only happen during resampling? We do resample using the standard EPS functions but I thought that all of this still meets the IEC 61000 standard and if so is it still ok to use? I ask mostly because the customer will likely repeat the test during commissioning and my assumption is that he will also use IEC compliant measurements.

[James_McN · Accepted solution]

This will always happen when using the IEC resampling block. The toolkit is specifically designed for IEC 61000-4-30 power quality standard where the error caused by this isn't severe enough to cause issues once the output is averaged over 10 seconds which is the measurement interval in 61000-4-30. I'm not sure how fast the changes you are interested in are.

However we discovered this when looking at higher speed data from it - in our case, there is a noticeable (1%) error in the half-cycle RMS as this issue combines with a similar problem with the RT side of the processing and it loses track of the zero crossing points for a couple of cycles.

Unfortunately my experience has been grid measurements where frequency is highly stable so I don't have the same experience from a generator point of view. My guess would be if they are concerned with power quality the measurements should probably be taken from the electrical signal for higher confidence but I would avoid the resampler if you are looking for fast changing effects.

Possibly the easiest method might be to use the zero crossing function in Power Measurement > Advanced and you can derive the frequency on a cycle by cycle basis from this. There is a frequency function but this uses an FFT so again, it depends on how dynamic your change is since this will give a more average speed over the FFT window.