Harmonic Filter

Your approach to the harmonic filter is a bit puzzling. But, it is much better to have questions, than getting surprises.

1,. Generators are designed to put out fairly clean fundamental frequency. Harmonic filtering need arise on the load side. Arc lamps, switching power supplies as an example are highly nonlinear, and sources of high harmonics, and imbalances in 3 phase systems. There is a need to correct both on the load side, to minimize the effects entering the grid.

2,. A well designed filter may be capacitive near DC, but near resistive at 50 / 60Hz. It may even look near resistive for the first (few) harmonics. Its purpose - beyond the obvious - is to reduce the possibility of high harmonic circulating current in the generator. That is heating you do not need, and the controller may act strangely.

3,. Can you run a generator with no load? That question is equivalent = should a generator blow up, if it suddenly loses its load? Obviously, well, from 0 - 100% load and anywhere inbetween. There is one caveat: no excitation does NOT produce zero output. Remanent magnetisation of the core produces some low voltage output, the voltage regulator may not like. Study the manual for the controller on this and ramp-up issues, that you may not get surprised.

4,. The generic answer to leading / lagging load currents is that a rough correction right at the load(s) is the correct approach. There exist old gensets with old controller, that is misbehaving at excessively capacitive loads. Modern controllers ought to handle any kind of power factor. But, to be sure of the facts, study the manual.

5,. If you do not have the book(s), get them from the manufacturer. You need them anyhow.

Thank you for your response.

perhaps I was too general in my question.

I understand the purpose of a harmonic filter is negated if there is no inductive load on the system to induce transient frequencies in the first place. As I said in my initial post, "this is not the normal operation of this apparatus". During normal operation there is nearly 500MW of primarily inductive load on the system, which is supplemented by a further 10 identical synchronous power generation turbines. This particular 19MVAR filter is just one of many and tends to harmonics produced by ancillary motors that service some of the largest ore mills in the world.

The fact that the generator is running without load is of much less consequence to my inquiry than the possibility of increased voltage as a result of a capacitive load. It has been suggested to me by some of the site engineers that they would expect to see up to 20% higher voltages during this testing, however they provided no reasoning for thier assertion. I was hoping someone on this forum could provide some clarification??

Obviously, I am aware that a generator has the ability to shed all of it's load almost instantaneously without ill effect. My question is can it be run indefinitely over a long period of time under these conditions?

Our generator and bus bar protection logic is such that the circuit breaker cannot be closed onto a dead bus, until it registers a minimum 25% excitation voltage at the AVR. Residual field is not substantial enough to even register on the 33kV protection relay which governs this circuit breaker. So in order to achieve this "dead bus close" the AVR is defeated and we manually change the excitation current to allow us to ramp the voltage as required. With protections and AVR circumvented, we need to be very sure of how the system is going to behave. For instance we could have a flashover if voltage exceeds our SF6 Gas insulated bus bar voltage rating.

In regards to consulting the manuals. This would be my first line of approach in most cases. However my Cantonese is extremely limited.

I have to take your thoughtful note in pieces.

1,. My and my son's cantonese is about on par with yours. But, in my opinion, it is insane to run an expensive equipment without good manuals. My son happened to work for a translation service. If you care to send me a private email (as CR4's management frowns upon anything, that might be an advertisement), you might get an offer, or info on who can do the translation for you.

2,. Since the voltage regulation have to work, or kick off the generator, I do not understand the 20% voltage increase at all. Consult the manual, again.

3,. I understand the rest of your concerns, even if I know not enough to comment intelligibly. An SF6 switch switching uncontrollably, anywhere but at voltage 0 is detrimental for its life expectancy.

4,. A generator with zero load produces a low, baseline heating of the coils. No problems at all.

You will get a higher voltage from the generator on capacitive load than on inductive load for the same excitation current (this is a vector thing...). Just use a lower excitation current and you will get the same result. The AVR might be upset about the capacitive load because it is tuned for an inductive / resistive load so try this in manual control by raising the field current gently and watching the generated voltage.

As an aside, this looks like a risky way to test your filter. Why not simply measure the capacitance of each branch after powering off and discharging them properly? This way, you will not put the whole system at risk in an unorthodox maneuver.

I generally agree with your response but your item #2 is questionable. Are you certain of this?

RLC (in series) filters are capacitive up to their resonance frequency, resistive at resonance, and inductive above. Depending on the ratio of the series resistance and the inductance, the transition band will be more or less defined in width.

I have never heard of somebody tuning the filter around 50/60Hz. We usually tune it between the 4-5 order to minimize the size and do the most work where it counts at the 5th and 7th harmonics. It also decreases the chance of hitting a parallel resonance at low frequency where it is difficult to mitigate.

But I do most on my work on industrial low voltage. HV guys may have other practices. If it is the case, please enlighten me.

Regards,

Marco

Low voltage and / or electronic filter and high power filters are two different animals entirely.

The first employs relatively high Q components for good selectivity etc.

Q = X_L / R the reactance divided by the resistance assigned to the component

The fact is, that wherever a parallel resonance occur, the circulating currents in that part of the filter magnified by Q. At series resonance the components voltage is magnified by Q times.

High power filters cannot tolerate such 10x - 100x increased stresses without blowing up. Hence the topology, components, installation all aims at broadband, low Q operation. Two basic forms exist: the T and the PI.

T; series L, parallelC,........series L

PI: parallel C, series L....... parallel C

Both are lowpass filters. They are set up to pass the fundamental frequency with minimal, precisely computable loss, and gradually block / divert / absorb harmonics. Which of the three, that depends on the internal construction details.

As a general rule, the higher the harmonics, the higher its current, the higher is the concern about it. SCR's (thyristors) are quite capable producing harmonics approaching the Megaherz range, when the filtering at the source is substandard.

Now, such a filter can be measured very well to reassure a nervous engineer. And I understand being twitchy. Rent a variable audio generator. Feed it into an audio amplifier of a few 100 Watts. Terminate it into 4 Ohm + 0,1 Ohm of a proper wattage. The 0,1 Ohm will feed the filter. Terminate the filter with 0,1 Ohm. Use a 2 channel oscilloscope to compare the input to the output, both in voltage and phase. It will tell you both the damping and the amount of inductive / capacitive behaviour of the filter at all frequencies of interest. I admit to a degree of arbitrariness in selecting 0,1 Ohm for both the input and output termination. Close enough for government work, so to speak. The manual may provide closer values.

I would be willing to bet dollars to donoughts, that the filter will turn out to be a nonproblem. Even, if something is shed, its input or output loading impedance stays still connected to it, keeping its operation in the low Q regime.

The ABB paper and I look at the very same coin from different sides.

To round out the filter talk for now, let's mention some filter basics. The filtering capacity is normally measured in dB (deciBel). that is a relative logarithmic scale between input / output voltage and power.

A few numbers are worth remembering:

3dB 1/2 power 0,707 voltage ratio

6dB 1/4 1/2

12dB 1/8 1/4

20dB 10^-2 0,1

30dB 10^-3

60dB 10^-6 and so on

Now, let's view your reality. When the SCR's chop the incoming feed, they turn at least a few % into harmonic energy. That, in your case means 1 MVA or higher to be handled by the filter. Such filters are normally reflective. For conversation's sake I assume a respectable 20dB. It means, only 1% passes thru to the generator, to be turned into heat, and trying to get unwanted movement in the winding. For reference: the original question mentioned a 10MVAR harmonic filter, a massive construct, capable handling stresses beyond that ought to occur. Oversizing never hurt, while undersizing regularly leads to "interesting" results.

The harmonic sampling circuit ought to be decoupled by 30dB. Much higher, and unwanted signal leakage might become a problem. In this case Kilowatt(s) are available to be filtered and metered for the particular harmonics. Such circuits were calibrated originally with the method described in my previous note.

As the SCR's produce square waves, expect even numbered harmonic to predominate. While in a general case higher order harmonics are lower in amplitude, there is no guarantee for such orderly behaviour in your particular case. On the top of it, as the SCR controller vary firings, your harmonic contents will vary too.

An interesting scenario indeed, but I have a further question: How is this setup supposed to test your harmonic filter's functionality if the devices that cause/generate these harmonics; i.e., the normal loads on the system, are not connected? You didn't say what type of harmonic filters you are testing or how they selected/tuned, nor the frequencies you are trying to minimize, but here's my quick take on your situation.

The simple answer to the no-load question is no, there is no problem running a generator at zero load and leading power factor, that's how a synchronous condenser works. However, if the reason the power factor is leading is because the load is highly capacitive with minimal real load then you must be very careful not to have a resonant condition present. The other danger is that of a runaway condition if your power factor correction/harmonic filter is dynamic.

Under normal operating conditions when additional inductive load is added to a generator the voltage drops incrementally, the AVR sees this drop and incrementally raises the excitation level to compensate for the drop.

When the generator is operating in the under excited region of the generator capability curve and the load becomes more capacitive, the AVR sees the voltage rise due to the additional capacitance and incrementally lowers the voltage by reducing the excitation level.

If your PFC/HFC is dynamic it sees the voltage being lowered and adds more capacitance to raise the voltage, which in turn causes the AVR to back off the excitation even more, until there is insufficient field strength to maintain synchronization. Now at zero real load you shouldn't slip poles, but at anything above that your governor may not be able to hold the desired frequency.

So you're faced with two potential problems, overvoltage if a resonant condition exists, and undervoltage if a race condition exists. I assume that is why you want to ramp the voltage slowly, and the best way to do that is to have the AVR out of the circuit and use the manual excitation control to slowly ramp the voltage to check for any instability, then put the AVR in service (it should have a bump-less transfer of control) and watch for any sharp rises caused by any resonances.

You are wise to be cautious in your approach, and you might find some additional information in this, especially section 8.3.2:

http://www04.abb.com/global/seitp/seitp202.nsf/0/18aa8879b8cc0186c125761f005035b7/$file/Vol.8.pdf

My apologies. I worded my initial question very poorly. The test we are conducting is not intended to test the effectiveness of the harmonic filter in reducing harmonics, (that will come later), but rather to test the auxillary components of the system, such as voltage meters, protection relays etc. Basically, ensuring the whole system does not become a 19MVA molten copper firework. This test is being performed at the request of ABB, (the vendor for this particular harmonic filter), and forms part of their comisioning activities. Hence the removal of all non-essential load from the system, so as to avoid any collateral damage if our engineers have had a calculator malfunction.

I should also have mentioned that this harmonic filter is monitoring the 4th, 8th and 12th harmonic ranges of 50Hz. This is due to the enormity of the ore mills being run. The mills operate at very low frequencies, in the order of 5-10Hz, and as such are supplied by a cyclic converter. The cyclic converter is essentially a 12 pulse VSD. The gating of high current SCRs is the offending component which necessitates the harmonic filter.

The filter is most definitely fixed as opposed to dynamic. It can be tuned to accommodate other frequencies, but this must be done offline by manually changing inductor tap positions.

I should also mention that I, as a lowly technician posed this question to CR4 users in the hope of furthering my own understanding. Fortunately for this very expensive machinery, this test is being orchestrated by far greater minds than my own. All of you who have contributed to this thread, have been extremely helpful and I greatly appreciate you all volunteering your collective knowledge so freely.

The test will be carried out this afternoon and I will keep you posted on the results! :-)