Viewing magnitudes from COMTRADE files 2

[bacon4life]

Is there a good tool to view the RMS magnitude of a channel in a COMTRADE file? In the image below from a DFR (USI M9000 @6000 Hz), the magnitude calculation displayed by SYNCHROWAVE appears to have numeric noise. I suspect the noise may be due to COMTRADE timestamps being limited to whole integer microsecond timestamps whereas the SYNCROWAVE calculation of magnitudes assumes more exact timestamps.

I see a few COMTRADE viewers options listed at

https://en.wikipedia.org/wiki/Comtrade,

and am hoping to avoid going through the lengthy process for getting IT approval just to find out if each option can easily graph magnitudes.

 

[davidbeach]

SynchroWAVe calculates magnitude by using the current sample and the sample (closest to) 90 degrees earlier. On an event record filtered to just the power system frequency and with some multiple of 4 samples per cycle you get a great magnitude calculation.

Sample at something that isn’t a multiple of 4 times the system frequency and you won’t get good results. Have a raw event, even sampled at a multiple of 4 time the system frequency and you won’t get good results.

For the raw event (what I’d expect from a DFR) even if it’s sampled at say 16 or 32 samples per cycle, that value from a “quarter cycle” is only a quarter cycle at the fundamental. At the third harmonic it’s 3/4 of a cycle earlier, etc.

If you really want to smooth it out, I’d suggest dumping it into excel for some post processing, or open the files in MathCad and do the post processing there.

But the first step, once you move away from a filtered, fundamental, event record is to define what you mean by magnitude.

 

[bacon4life]

Thanks David. Our old Qualitrol DFR's came with software that nicely calculated all sorts of things, including RMS magnitude, fundamental magnitude, or magnitude of each harmonic. If I had realized we needed to roll in the cost of Mathcad license for every person who views the DFR records into the price of the DFRs, we certainly would have selected a product besides USI. Not to mention the ongoing cost of engineering labor to post process every file.

 

[davidbeach]

A good reason to move away from DFRs. I've found that money spent on good relays provides a better return than buying DFRs. Long event file from SEL-4xx relays can provide pretty much everything a DFR could. Automatic event retrieval and remote access certainly help in that regard. One of the things we've done is that the 11A relay only triggers on a trip; we go with a 1 second event with 45 cycles of pre-trigger data. That grabs an entire zone 2 event. The 11B relay triggers on various things, and they have given us a virtual DFR, broadly diversely located.

One thing to watch out for with long events is that if you're using traveling wave fault location on the 411L (probably applies the the T400 and T401) you want every fault initiation to be in its own event. There's only one TW event file generated per event; so if you have a long event file that has pre-fault - fault - trip - open interval - reclose - restrike - trip all in the same file you won't get a TW event for the second strike of the fault.

 

[bacon4life]

Do you have fully automated event retrieval? I have been dreaming of this for more than a decade and it stills seems several years away.

Do you cross trigger all relays at a substation to trigger on an event? In the case of the event pictured above, the fault was down stream of distribution substation, so the magnitude of the disturbance was much lower than the pickup levels for any line relays. Typically the DFR is triggered for this kind of event based on 230/115 kV transformer neutral current. In this case, the transmission level DFR was the only piece of utility equipment to capture a fault cleared by a customer fuse. Without records for all line terminals, it would have been much harder to figure out where the fault was located.

Can sychrowave automatically combine records for 20+ relays at a station? With the new DFR and syncrowave, I have to manually open each COMTRADE event file, then open the template corresponding to that substation. I doubt I would review records if I had to manually combine files from multiple relays. With our old DFR, there was a single place that stored a 2 second record that covered all line terminals. The Ben32 software let me set up a custom display that showed the most commonly triggered channels all on one page. This made it really easy to differentiate kinds of events at a glance. Once every couple months I would review all the DFR records, and it usually took less than 2 seconds per record to differentiate between normal events (i.e. cap bank switching) and unusual events. Moving from one event to the next was just one click.

 

[davidbeach]

Fully automated event retrieval for a significant portion of the system. We don't have a hard-wired cross trigger but use voltage triggers so we have multiple relays all self triggering for the same cause.

I've had well over 20 records open in the window. If you want to open a bunch of event files and don't care how they get numbered, you can zip them up and then open the zip file with SynchroWAVe. If you just want a bunch of files open, you can highlight them, right click, and select open. That works for up to about 15 events.

 

[erdep]

Bacon4life,

Can you send the files (.dat and .cfg) of your case?

I want to try it with gtpplot.

Thank you

 

[bacon4life]

erdep- Thanks. I have uploaded an example at:

https://files.engineering.com/getfi...0-4672-8982-71f54fb0efab&file=DFR_example.dat

https://files.engineering.com/getfi...2-4c50-9bab-c0393e1a24ce&file=DFR_example.cfg

USI apparently offers a viewer that shows just a small amount of distortion, though I am still working through the process to get it installed on my computer.

In Synchrowave, I also realized I could save a user template for each unique DFR. If I then load the correct user template and save it as the default template, I can open files like David suggested. Although a significant improvement, it is still bit clunky to load/save the templates when switching substations.

 

[erdep]

Bacon4life,

I am developing Gtpplot as an EMTP-ATP viewer. The program can read COMTRADE files, but I have not access to true COMTRADE files, I am retired since 1996.

Any example of such file are welcome, I can correct the reader using it as a reference.

The signal has harmonics that are not an integer multiple of fundamental.

Any integer harmonic contributes to the RMS as a constant, but non integer harmonic produces the oscillations as your original graphic and similarly in Gtpplot

The file contains 100 points by cycle, corresponding to 6000 samples per second a Frequency=60 Hz

The Fourier components calculated by Gtpplot (using 100 points per cycle DFT, the speed is not a problem for off-line process)

Using a true sinusoidal source in EMTP-ATP, the curves are the same a naked eye, but the ATP RMS is a horizontal line.
By adding sources of 360 Hz and 420 Hz, the RMS show similar oscillations.

Thanks to your example I am doing corrections to Gtpplot.

 

[bacon4life]

Are the noninteger harmonics actually present in the original signal? Or are they a mathematical artifact of the sampling speed and/or COMTRADE file format limitations? Would a different sampling frequency eliminate any of the harmonics?

The COMTRADE file format appears to offer two ways to specify times. Which option does GTPPLOT use?
A)timestamp to the nearest microsecond for each sample
B)Start time, end time, and sampling frequency

If additional COMTRADE files would be useful for improving GTPPLOT, send me an email via hotmail.com.

 

[erdep]

Gtpplot reads the .cfg file and uses timemul

I your example, timemul is

0.477225214242935

It is in micro seconds

The .dat contains:

1,0,267801390,1676682854,-1947471288,84259708,482838889,-574376845
2,349,401917008,1593459685,-1998414489,111587178,464445025,-587236013
3,698,530874354,1505109808,-2041665140,138914648,446051160,-600095212
4,1048,661035311,1415050998,-2084060886,170796696,423058847,-604381591

--------


With this, the time for first point is zero

The second point time is 0.477225214242935*349= 166.5515998 us

It is not a integer submultiple of the period = 1/60 = 16666.6666666 us

The third point time is 0.477225214242935*698= 333.1031995 us

It is not a integer submultiple of two periods = 2/60 = 33333.33333 us

and so.

The RMS are calculated by any procedure by summing the V**2 and at
the end of each period the sqrt(sum) is divided by the period.

The true period is 1/60 = 16666.6666... s, but the numerical (data)
period is too short or too large.

6000 samples/second at 60 Hz are equivalent to 100 samples per cycle

At 100 samples per cycle

98,33876,-148644860,1899181215,-1758741514,11386454,519626618,-535799278
99,34226,-9886764,1829970919,-1826779984,38713924,510429685,-557231266
100,34575,128527446,1757684526,-1889860854,61486816,492035821,-561517645
101,34924,264362521,1678733644,-1946274669,93368864,478240440,-570090435

The true zero is not at the sample #100, T= 16500 us
#101, T= 16666.6 us

It is a problem due to 60 Hz, I guess. 100 samples/s at 50 Hz have better behavior.

It is seen as non multiples harmonics, and distorted RMS.

May be a floating point COMTRADE file helps, the true time and
variables value are in 32 bits (REAL*4) Fortran numbers.

The integer times are too coarse, I guess.

I am not expert in COMTRADE, my interest is in to open a COMTRADE
file to plot as any EMTP-ATP graphics files.