Register to reply

Reading Error by Two Unsynchronized Power Meter

Share this thread:
ayinaim
#1
Feb19-12, 10:32 PM
P: 5
Hi everyone,

I've been searching and googling for the answer to no avail, so I'm just wondering if anyone at PF could help me out. It's for my work assignment, by the way.

I have two power meters connected at two ends of a transmission line (assume it's 33kVAC). At each power meter, there is a Data Acquisition computer that pull reading once every (500ms to 1s) with a timestamp attached to each data. For this purposes, we would like to get the power measurements (assume in MW) at BOTH meters instantaneously. When we are matching the data from the 2 meters based on its timestamp, the chances are the readings from the 2 meters could be off from each other at 500ms at most.

For example,
-reading at meter 1 at t=0s is 200MW
-reading at meter 2 at t=0.5s is 170MW

My question is, is there a way to figure out the actual value of meter 2 at t=0s in terms of probability around the obtained 170MW at t=0.5s?

In other words, we knew that we'l have the possibility of obtained mismatched data as big as 500ms, is there a way for us to calculate how much error had occured as compared to t=0s reading for meter 2.

I hope it's not too confusing and i really hope someone could guide me out regarding this matter. Thanks a lot in advance.

Regards,
ayi

P.S. meter used : EPM 5500, schneider PM1200, goosen A2000
Phys.Org News Partner Engineering news on Phys.org
Printing the metals of the future
New gadget helps the vision-impaired to read graphs
3D printing helps designers build a better brick
ayinaim
#2
Feb19-12, 10:35 PM
P: 5
Just a follow, the purpose of this question is to figure out the actual power loss taking into account the error associated with unsynchronized meters.

I guess the easy way to ask the question is,

If i read a power measurement at a meter at t=1s to be 170MW, is there a way I could calculate the actual reading at t=0s based on probability, etc?

Thank you.
jim hardy
#3
Feb19-12, 11:19 PM
Sci Advisor
PF Gold
jim hardy's Avatar
P: 3,518
""Just a follow, the purpose of this question is to figure out the actual power loss taking into account the error associated with unsynchronized meters.


I guess the easy way to ask the question is,

If i read a power measurement at a meter at t=1s to be 170MW, is there a way I could calculate the actual reading at t=0s based on probability, etc? ""

I'm assuming these are as you said power(mw) readings not energy(mwh) from the "integrate" feature on those meters you listed.

Am i correct to think since you're trying to measure line loss not detect a fault, there's no hurry?

Then I would do it this way:
Make a calculated input for each line. name it Delta MW, and treat it as if it were just one more analog input. It would be the difference in readings between the two meters of course. Apply software smoothing to the point perhaps a minute running average(? make that time user-tunable ?).

That way the constant little up-down swings in both meters should average out for you.

If you wanted to get fancy you could put line loss versus power data into a lookup table and compare measured to expected, or use that table to estimate loss when one of the megawatt meters is out of service.

ayinaim
#4
Feb20-12, 12:03 AM
P: 5
Reading Error by Two Unsynchronized Power Meter

Hi Jim,

First of all, thank you for your advice. Just to clarify, I did mean power (MW) and yeah, it's meant to measure line loss.

I do get your 'second method' suggestion about comparing the measured and expected line losses.

However, i do not quite get the your 'first method' suggestion. Are you suggesting that I obtain a certain period worth of 'measured losses' and average them out to obtain an expected loss value for future references? The problem that I could see with this method is that we could not guarantee that the initial power (MW at point A) is constant. It might be changing, and thus, causing the power (MW at point B) to change as well.

Secondly, do you know of a way to 'theoretically' measure these errors, rather than measurement-based deduction. Just wondering.

Thank you so much for your help,
Ayi.
jim hardy
#5
Feb20-12, 01:09 PM
Sci Advisor
PF Gold
jim hardy's Avatar
P: 3,518
""However, i do not quite get the your 'first method' suggestion. Are you suggesting that I obtain a certain period worth of 'measured losses' and average them out to obtain an expected loss value for future references?""

No, though that's not a bad idea you have there.

My thought was this -

in a half second the power flow down a line should not change by a lot unless there's a system upset underway.
But watching any real life measurement one sees small fluctuations about an average value, and indeed power around a system is a flow with small short term fluctuations.

and my thought was an average value taken over an interval several times longer than your half-second should be statistically pretty well smoothed out.
So by doing that at each end one could subtract the averages

Or one could average those almost simultaneous differences between ends and achieve same smoothing. I like this because difference shopuld be near zero at low power and that'll tip you off when one of the meters needs adjustment.

If the fluctuations are random you'll be okay,

If you find them to have a definite period the grid operator needs to know about that.
Sustained low frequency "sloshing" of power around a system should be looked at because it might indicate a stability issue. We found 2/3 hz in a power system but dispatchers learned how to damp it out......


""do you know of a way to 'theoretically' measure these errors,""

no i dont. Your client's system people know what is resistance of the lines and I^2*R would be first approximation for calculated loss to compare to measured, and you could adjust R for local temperature.
What is accuracy of PT's and CT's you are using for primary sensors? That dictates how much refinement is waranted.

"When you want to make a machine do something exceedingly well, watch the skilled workingman who does it best of them all and make your machine do the same thiing"... my mentor
old jim
jim hardy
#6
Feb20-12, 01:13 PM
Sci Advisor
PF Gold
jim hardy's Avatar
P: 3,518
That GE Scada terminal you mentioned looks pretty intelligent.
Can you synchronize its scans to best clock available and place these measurements at start of list?
We kept our computer clock synched to WWV.

jus wondering.
ayinaim
#7
Feb20-12, 09:42 PM
P: 5
It is actually our custom-made software running on a Moxa's embedded computer that do the data pulling. I dont believe these computers at substation-level are synchronized, but rather run on its internal clock since previous purposes were not extremely time-sensitive.

Our client actually wanted us to develop an adaptive line overload automation scheme based on the existing meters and and sub-level embedded computer, without installing pricey PMUs, and they are worried about the mismatched timestamps for data pulled from all the participating substations.

I guess I'll put forward a proposal on tackling the error issue based on what we have discussed here. I really can't thank you enough for your advices on the matter. Thanks.

Regards,
Ayi
jim hardy
#8
Feb21-12, 01:24 AM
Sci Advisor
PF Gold
jim hardy's Avatar
P: 3,518
Thanks for your kind words....

If your client feels this synchronization of readings is important and it becomes "much ado ",, here's a thought that might defuse a "situation". We all know what can happen in meetings.....


I dont believe these computers at substation-level are synchronized, but rather run on its internal clock since previous purposes were not extremely time-sensitive.
Are time stamps affixed by those substation-level computers?
Does that mean we really dont know whether the readings from opposite ends of the line were in fact separated by the delta-time inferred by their time stamps? Seems that'll make your job a lot harder.

Anyhow reason i got back on this was to mention another thought i had today.

IF you were sure of accurate time stamps at both ends of line, you'd know how far they are offset.
Could you not for one end do linear interpolation between previous and current reading, and estimate a reading for same time as other end's reading? I doubt it'd be any more accurate than time-averaging mentioned earlier , but it may have appeal.

I am skittish about extrapolating forward but interpolating backward seems rather inherently stable.
Sounds too easy, doesnt it? Poke at the idea for a flaw. Often these things look great at first but dont work out, especially late at night (like now).

old jim
ayinaim
#9
Feb22-12, 10:01 PM
P: 5
As far as i know, they do not have a standardized time sync method between substations as of now.

Regardless, I'll take your advices into consideration. This new scheme is still very much in the exploratory phase and I'm looking forward to propose the ideas we have discussed on here to our client.

Thank you so much again for your help :)

Regards,
Ayi


Register to reply

Related Discussions
Fortran77 end-of-file error reading formatted input Programming & Computer Science 5
Electric company meter reading General Engineering 3
Runtime file reading error Programming & Computer Science 7
AC Power Meter & Remote Switch Electrical Engineering 6
Reading an analog meter Electrical Engineering 2