Transcript
Introduction to High Order Harmonics
Good afternoon, everyone, and welcome to today’s white paper webinar. Today, I have Andrew with me, and this is an interesting paper on high order harmonics and how you see these in PQ Canvass and also in ProVision.
I’m gonna start by introducing what we mean when we talk about high order harmonics or super harmonics or high frequency noise. First, talk about where you might see this, and then we’ll show you how to see this in a quality report.
What Are High Order Harmonics?
The concept here is that we’re looking at a waveform distortion in the frequency domain that is past the typical 50th harmonic, or 51st harmonic. If we’re going into the kilohertz range and past 3 kilohertz, this might be four, five, six, seven kilohertz. This is traditionally not a source of voltage distortion.
Traditional synchronous loads would generate third harmonic, fifth and seventh harmonics, where the distortion is synchronized at 60 hertz. When we’re getting past the third harmonic, we’re kind of losing the concept of synchronous distortion, and these generally are sources that are inverter based. You have switching transients, you have pulse width modulation, inverter schemes, CVFDs that are switching. These are switching in the kilohertz range, often three to maybe up to seven or eight, 10 kilohertz, and their switching frequency is not necessarily related harmonically to 60 hertz.
By definition, if it’s a harmonic frequency, it’s a frequency that’s an exact multiple of 60, and as you go up in frequency, that starts to lose its meaning, especially when the noise sources themselves are not directly tied to 60 hertz, like these inverter switching things. They’re indirectly tied to 60 hertz, but not directly.
At these upper frequencies, we’re looking at 60 hertz bins, but not really because the distortion is a multiple of 60, but rather we kind of transition to a 60 hertz wide bin width for generic spectral display. So when we’re looking at a spectrum like we see in some of the figures here, this X axis is harmonic number, but that really should be interpreted, especially in the upper region past the 50th harmonic, as a generic kind of frequency axis. We’re just looking at a spectral display that happens to have 60 hertz bin spacing. That’s our resolution in this spectral graph. You can think of it more like a spectrum analyzer when we’re looking at these upper frequency ranges.
Terminology for High Frequency Content
Now there are many terms for this kind of content. High order harmonic analysis, super harmonics. When we talk about super harmonics, the idea is that these aren’t harmonics anymore because they’re not technically multiples of 60. They’re above the harmonic range, so that’s super, meaning above or past the harmonics. Some folks like that term, others don’t, but that’s what that’s supposed to mean.
High frequency noise is another term for broadband or non-synchronous noise that’s above the third harmonic, or above 3 kilohertz. But whatever you call it, it’s a problem these days.
Sources of High Frequency Noise
There’s many more switching devices than there used to be, and even residential loads are often variable frequency drive based or inverter based, or any sort of AC to DC conversion, where we’re using active converters to rectify the AC voltage are gonna introduce switching transients that have the potential for introducing high frequency noise.
These are often small components, but they can be amplified by resonances. Traditionally, resonances that were in the distribution network are usually low frequency, just given the size of power factor caps and substation transformers. A typical distribution resonance would be at, say, 500 hertz, under a kilohertz generally, and any sort of high frequency resonances that were present, especially on the secondary side of a transformer, never really came into play because there was no energy there to excite that resonance. There was no high frequency steady state distortion that would energize a resonance that was at, say, 3600 hertz.
That’s not the case anymore when we have more and more switching noise and inverter noise that is small but will be amplified by local resonances that in the past were still there, but not necessarily a problem because you didn’t have any sort of load that was energizing this.
Sensitive Loads and Real-World Impact
So it’s become more important to see this content past the 3 kilohertz, and in parallel with that, we have loads that are more sensitive to that. We have LED bulbs that are sensitive to high frequency noise. They can be fooled into thinking there’s a dimmer. We have things like GFCIs or arc fault breakers that can get confused by those waveform shapes. GFCIs in particular are often far more sensitive with current in the two to three kilohertz range than at 60 hertz.
It’s not uncommon to have, say, a solar site on a weak feeder that’s introducing a few kilohertz of high frequency noise that causes random breaker trips in a residential neighborhood. So there’s an increasing need to see this content above the 51st harmonic.
Measurement Standards and Waveform Capture
Now there’s no standard yet for how you’re supposed to measure this, or certainly no standards for limits on this yet. That’s still gonna be a ways away, but in the meantime, we can certainly still look at the data in a raw manner.
The way you do that is with waveform capture. If you’re sampling the waveform at 256 samples per cycle, in theory that allows you to see up to the 128th harmonic, or another way of thinking of it is you’re sampling roughly 15 kHz, and so in theory you have the opportunity to see components up to 7 kHz, the Nyquist limit. Any existing PMI recorder that you have that’s sampling 256 samples per cycle will have the ability to capture that waveform at that sampling rate, and then in the software, in ProVision or PQ Canvas, you can do a frequency analysis and look all the way up to basically 7 kHz or the 128th harmonic, however you want to think about that.
Using Periodic Capture
Generally, the best way to capture this is with periodic capture. The components are often too small to trigger on, and so a periodic capture where you’re capturing, say, every four hours, every two hours, or on a periodic basis throughout the day and night, gives you kind of baseline snapshots, and that gives you the raw waveform for this high-frequency analysis either in ProVision or in PQ Canvass once the data is available.
Of course, if you’re using a streaming recorder, there’s really no memory limit, so you can set it to, say, every hour, 24 hours a day, and you’re never gonna run out of memory because it’s streaming up to PQ Canvass. But even with a local recording, you can still see a lot of waveforms that way.
Viewing High Order Harmonics in PQ Canvass
So I’ll turn this over to Andrew, who will walk you through how to get to this data in PQ Canvass.
So let’s see how we can get to this data in PQ Canvass. If we go ahead and click on our recording here, now we’re in the Recording tab. If we go down to where it says Events right here, and we click on the number next to the Waveform Capture, that’ll pull up all the waveform captures.
Then if we click on any one of these, it’ll bring up these charts here, and you can click this bar graph icon to see the harmonics graph. Then you’ll see the first harmonic is much higher in magnitude, so you can press the F button to normalize the magnitudes, and then you can click down here, this 51 plus button to show past the 50th harmonic.
Interpreting the Harmonic Spectrum
So here, we see something that’s pretty common. We see here the 56th harmonic, 57, or 58, and then we have 62nd and 64th, and kind of a null here at the 60th harmonic. Now, traditionally, with synchronous loads, you don’t have even harmonics. But with this sort of high-frequency noise and the fact that it’s inverter-based switching noise, that’s not the case anymore.
What you have is a 3600 Hertz pulse-width modulation frequency and sidebands. This is more similar to radio carriers. That’s why you have a suppressed carrier. There’s a null at the switching frequency, and you have sidebands where you’ve got even multiples of 60 on either side of the carrier. So here, this is a source with a 3600 Hertz inverter frequency, and we have these sidebands.
You wouldn’t really say you’ve got a 56th and 58th harmonic problem, per se. What you really have is a 3600 Hertz inverter source, and a resonance anywhere in here is gonna amplify this.
Time Domain View
We can go back in the time domain, and you can see kind of the fuzz here on the waveform. The fact that it’s primarily on the peaks of the waveform, or at various points in the waveform, those are the switching points. We can go to a delta view and get a little bit different view of this, where we’re looking at line to line.
You can see kind of where it’s switching on the phases as each phase-to-phase voltage rises above the bus voltage in this three-phase converter, different pairs of phases will have the noise on it, and that translates in the frequency domain to this sort of pattern that we saw.
THD and Equipment Sensitivity
From a THD standpoint, this may not really be too terrible. Here, it’s still at kind of 1.8 volt. So the magnitude of this high-frequency noise often isn’t, on the surface, that bad. But different types of equipment can show different sensitivities, even with a small amount of high-frequency noise. Again, things like arc fault breakers are very sensitive to noise in the two to three kilohertz range.
This might be a waveform where you’ll have random GFCI trips for various customers. If you’re trying to correlate this back to a load, you’d want to look to see when is it stopping and starting. Is it correlated to if it’s only during the day? Is there solar PV on that circuit that you can relate this to? If you switch it on and off, that can be helpful if you’re able to do that experiment. If it goes away when the inverter switches off, then that’s kind of a red flag.
Resonance Amplification Example
Another thing you might see is a resonance making this worse. If you load another example that we have, here is one where this was utility-scale solar. I think it’s around 10 megawatts on the feeder, and there’s unfortunate resonance in this circuit.
This is primary side voltage, and we see it’s absolutely clean. When they switch the inverter on, you have extremely large amount of high-frequency noise. This is a textbook bad case that you never want to see on primary side voltage. You don’t really need the harmonic analysis to understand this is gonna be bad, but it’s pretty clear, even with the 120-volt fundamental on, you can still see this pretty clearly.
Here, it’s even more clear. Again, this is another 3600 Hertz inverter PWM frequency. That’s a very common frequency. And this situation happens to be a 3600 Hertz resonance at distribution that was amplifying this.
Mitigation Techniques
One technique that the customer was able to use was to take pairs of inverters from the solar site and gang them together so that their pulse-width modulation pulses were 180 degrees out of phase, and that helped cancel them out in pairs to reduce the noise level, so even with the resonance in place, it was still acceptable.
It wasn’t really possible to change the resonance because that was formed by underground cable capacitance and the system inductance. That underground cable capacitance was not really something easily changed. If it’s a resonance formed by a cap bank, then you have the opportunity to move that resonant frequency around by changing that capacitor bank size or leaving it out altogether.
Here, we can take a look at the THD, and it’s still not terrible just looking at it that way.
How to Enable High Order Harmonic Analysis
Any captured waveform in ProVision or in PQ Canvas can be analyzed with this method. You just turn on this 51 plus button in PQ Canvas, and it changes the scale to show you past the 100th harmonic, up to the 100th harmonic, or in ProVision, it’s a system setting to enable that for all harmonic analysis tools.
Questions and Answers
If you have questions about the white paper or about doing this or about high-frequency harmonics in general, feel free to type them here in the questions box. Or if you have a question later, give us a call anytime at 1-800-296-4120 or send an email to support@powermonitors.com. We’d be happy to chat.
Recorder Compatibility
The question about what meters can you use: the specs for, say, the Volt or the Revolution talk about the 50th harmonic. Because that’s the highest they can record as individual strip charts, where a harmonic calculation is done in the recorder itself. They’re still sampling at a higher sampling rate though, 256 samples per cycle. So that gives ProVision or PQ Canvass the raw data to go higher in frequency. Even though the recorder can’t compute the harmonic at a higher frequency, the data it’s sampling supports that.
CT and PT Considerations
Another good point, some of the CT accessories are limited to 10 kilohertz. That’s true. In general, CTs roll off fairly quickly. RCTs should support up to the seven kilohertz range that the meter’s sampling at. But if you’re using PTs and CTs, like metering CTs, for a primary metered customer, be watchful for that, because some metering CTs are specifically 60 hertz only. They’re not really concerned with high frequency harmonics, or even regular harmonics.
Especially CTs and remit for relays, for protection, they don’t really have a good frequency content. Even PTs or CCVTs, if you’re using capacitive voltage dividers, they will often have resonances at other frequencies. If you’re looking for accurate measurements above the 51st harmonic, or even up to the 51st harmonic, the CTs and PTs that you’re connecting to can play into that.
Now, with high frequency noise, you’re often just looking at relative values or just the presence or absence of it. Exact measurements aren’t that critical in most cases, because you’re just trying to correlate is it there or not with some sort of load or thing you think is aggravating it. So in that situation, you’re not really trying to make precise measurements because there aren’t really standards yet for what those measurements are limited to in the first place. You’re really looking at qualitative measurements for things going over the 50th harmonic at this point.
Frequency Ranges for Different Sources
A good comment here on the UPRI guidelines for different frequency ranges and what you might see in those. Someone’s posted here, PLCs are in the, say, 9 to 95 kilohertz, photovoltaic converters, 4 to 20 kilohertz, power electronic converters, generally AC to DC converters, 9 to 150 kilohertz. That 9 to 150 kilohertz range is something that’s being worked on in Europe in IEC standards.
There is an advisory method for measuring this, but there’s no limits on it and nothing’s been adopted yet. There’s still a challenge to doing this and also a challenge getting accurate measurements, especially at 150 kilohertz. Those are more often seen on the secondaries of transformers. As you go up in frequency, it becomes more and more difficult for loads to kind of push those high frequencies through a transformer. At some point as the frequency goes up, you end up with capacitive coupling through the transformer rather than true transformer coupling.
Closing
Well, that’s all the questions we have now. Again, if you have a question later, give us a call anytime at 1-800-296-4120 or send an email to support@powermonitors.com. Thanks for attending everyone. Everyone have a great afternoon.
