Drive Efficiency Measurement

[johnnyw]

Hello all, I am starting to investigate the measurement/calculation of full load efficiency for a ~1000hp medium voltage motor drive. I would like to get the efficiency as accurate as possible (hopefully to a tenth of a percent - with some uncertainty of course). My plan is to calculate the input power by acquiring the 3 line-line (no neutral available) input voltages (4160 VAC) and 3 line currents, calculating the output power by acquiring the 3 line-neutral output voltages and 3 load currents, and then calculating the efficiency from the input and output power. I will also be recording water cooling flow rate and temperature rise for losses to liquid, but I won't have any information for losses to air so this would likely just be a sanity check for the efficiency calculation based on voltages and currents. I am concerned with the quantity (12) and accuracy (i.e., ~ +/-0.5%) of transducers and the potential impact of harmonics. Also, since an input transformer will be included in the drive efficiency, I think the measurement should occur after thermal equilibrium so that the winding resistance is at the design point.

I plan on acquiring time based data for several minuntes and then post processing to determine the efficiency. To reduce transducer effects, I plan on first calibrating the sensors simultaneously from a known source to fine tune their scales and eliminate offsets. Afterwards, I can also look at the frequency spectrums to determine the harmonic content. I haven't yet searched for the transducers or done the math to determine how accurate/uncertain the calculation will be.

Does anyone have experience/advise for such a task? For example, would a two watt-meter method be preffered for calculating power, would a power meter such as a Yokogawa PZ4000 be better than post processing the data, transducer selection. Thanks.

 

[dpc]

Have you ever looked at the waveforms on the input and output of a PWM drive?

You will not be able to get close to 0.1% accuracy. Your CTSs and PTs will have more error than that especially with the high frequency components you are dealing with.

There are some people more knowledgeable about drives who may have a better idea, but I don't see how you will get closer than 5 to 10% using the methods you're describing.

The drive controller itself may be able to calculate power out - it has the most information.

 

[ScottyUK]

Rogowski coils are available with significantly better than 0.1% accuracy and with sufficient bandwidth to effectively eliminate amplitude and phase errors at the frequencies of interest. High bandwidth HV isolation amplifiers are available, although whether you can find an instrument supplier who will qualify it for service on a utility line is slightly dubious. Revenue metering class VTs are a possibility but will almost certainly limit your ability to measure high frequencies accurately. You may find they have sufficient HF response for your needs.

Power meters such as the Yokogawa and Voltech instruments would be a good choice in my opinion. Instinctively I doubt you will get better than 1% overall accuracy.

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

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

 

[johnnyw]

I should have mentioned that PWM effects will be negligible. This is a custom unit with a high PWM frequency and very effective output filters - EMI could be an issue but transducer and data acquistion system location will be chosen to minimize. Also, the input transformer provides a multi-pulse configuration.

Efficiency will be tested one time in a laboratory setting so there is no intention to permanently connect anything to the line.

My instincts also tell me an efficiency measurement to a tenth of a percent efficiency may be impossible, but I would like to have the best approach to minimize the uncertainty.

 

[CJCPE]

For the most part, input harmonic currents are like reactive currents. They contribute to the VARs but not to the watts unless they are caused by source voltage distortion. The classical rotating disk watt-hour meters measure VFD input power pretty accurately by being influenced only by the fundamental current. I have seen drive efficiency measured this way with an electronic wattmeter on the output.

It might be worth while to insulate the cabinet from the ambient air and measure cooling air flow and temperature rise. If the VFD is about 95% efficient and you can determine that the losses are between 4.5% and 5.5% (+/- 10% measurement) you have determined the efficiency to be 95% +/-0.5%. With input/output power measurement, you need the two power measurements to be within +/- 0.25% to know the efficiency is within +/-0.5%.

 

[Skogsgurra]

I have recently done similar measurements on kilns for iron ore pellets. I used what was available; three old, robust current clamps 2000/1000/500/5 A that had been calibrated and were accurate to better than .2 % when short-circuited, a two wattmeter type U,I to W transducer, a Dranetz PowerPlatform 4300, a Yokogawa 300 recorder and also signals from the drive.

The purpose was to compare VFD drives and hydraulic drives. We had a deep and lengthy discussion about accuracy and we arrived at something between 2 and 3 % uncertainty. We hope to get this verified by using several different techniques and comparing results. I am still waiting for pressure and flow numbers from the hydraulics guys, so I haven't started the report yet. I will get back to this when I know more about the accuracy we (probably) got.

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[jraef]

Even when you attained a measurement, I don't see how you could possibly verify the 0.1% accuracy figure anyway. What is the reason for wanting that much accuracy anyway? That is much higher than revenue metering requirements.

I kind of like CJCPE's idea. After all, the meaningful losses in a VFD are switching losses, which will be converted to heat.

JRaef.com
"Engineers like to solve problems. If there are no problems handily available, they will create their own problems." Scott Adams
For the best use of Eng-Tips, please click here -> faq731-376

 

[ozmosis]

johnnyw
What sort of efficiency are you expecting? if >95%, then you will need some very accurate instruments.
For example (I'll use some easier examples), a drive system rated 1000kW input power and 950kW output power. The losses are therefore 50kW. If you use gunnars recently calibrated CT's at +/-0.2%, then for a 1000kW measurement, there is a +/-2kW tolerance, both at the input as well as at the output. When referred to the determined drives losses of 50kW, this signifies a tolerance of 2 x +/- 2kW representing a measurement tolerance of 12.5%(+or -). It means you can still be correct, even if you measure incorrectly by 12.5%.
Measuring the output voltages from your VFD will also present, imho, certain challenges to accuracy due to the PWM voltage rises.
The accuracy of carrying out the calorimetric (measuring the heat energy losses) method also suggested by CJCPE would be quite impractical as well if high accuracy is required. It is a useful guide' but certainly not accurate. It would mean packaging the complete drive inside a sealed enclosure. Not easy on a 1000hp system.
I think I've done my sums right!

 

[DickDV]

Skogsgurra, if the information you develop from your comparison testing is not proprietary, if would be greatly appreciated if you would post the results.

I have never trusted the published claims for efficiency comparing these two systems and having some actual test results would increase my confidence level considerably.

 

[Skogsgurra]

I am afraid they are proprietary. I get paid by the hydraulics company to do the measurements. But I can surely post what uncertainty we seem to have.

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...