Some thoughts on load control of a wind turbine, feed forward or feedback ?

[Warpspeed]

I think we are all aware that the combination of wind turbine blade design, plus the electrical generator characteristics together, create a complex system that cannot by itself work at peak efficiency over a very wide range of wind speeds.
 
At some point the blades will begin to stall from mechanical overload, reducing the effective electrical output.

We need some way of controlling either blade pitch or the electrical loading on the machine, so that rpm can be allowed to rise, and the blades recover from stall.

Discounting pitch control as too mechanically complex, we are left with pulse width modulating the electrical output to hopefully find and track changes in wind speed to stay at max torque, just below the onset of blade stalling over as wide a range of wind speeds as possible.

The intuitive approach is usually to propose some kind of maximum power point tracking system as used very successfully with solar applications. 
This works extremely well for solar, because any load change on the panels registers pretty much instantly as a measurable power change.

You can with software, change panel loading by say 1% and measure the small change up or down in actual power output within perhaps a few milliseconds. So the system software can repeat the process perhaps ten times each second, and track passing clouds for example.

Things are rather different with a wind turbine.  If the blades are just beginning to stall, changing the loading by just 1% is probably not going to do anything measurable, and even if it does, its going to take a finite time for the effects of rotor inertia to settle to a new stable rpm and the effects of the load change show up as a definite measurable true change in power.

It might work if the wind velocity is constant, and we go about correcting machine loading very gradually.  If we over correct, its possible to get some instability and even cyclic surging.
In gusty conditions, very slow load correction might actually make things worse.

Feedback systems work very well where there is a very fast response time through what you are controlling, and where external conditions change relatively slowly. 
Where what you are controlling is very slow and sluggish to respond, and external conditions change very rapidly, feedback becomes much less effective, and can actually make things worse.

The concept of feed forward control may be new to some of you, but the idea is fairly simple.
A good example is controlling the ignition advance in a gasoline engine.
Its well known that the ignition timing needs to be advanced and retarded with changes in engine rpm and load.

If we are really clever, we can build into our distributor an advance curve using springs, weights, levers and cams so that as engine rpm changes, the correct ignition advance is generated automatically.  A secondary input could be engine load sensed by a vacuum can and a spring that further modifies the centrifugal advance.
Nothing is fed back !
All the control is from direct inputs of engine rpm and engine load.

Modern electronic ignitions do it all from an ignition map in a read only memory.

A wind machine control system might consist of a small fast responding anemometer as the primary control input, and a secondary control input might be battery voltage. (we don’t want to overcharge our battery).

So we measure our instantaneous wind speed with the anemometer, and we use that to set the required PWM duty cycle and loading so that for every measured wind speed there is a correct load level applied to the turbine to provide optimum output.

Like the ignition advance curve, we can set this up to respond instantly to changes in wind velocity.
Sudden gusts or lulls will track without having to allow time for rotor rpm to stabilize.
It can never become unstable or surge.

The advantages of feedforward correction are extreme speed of response and stability where a sluggish machine can be very quickly corrected for rapid changes in external conditions.

I actually have a working home brew 5Kw sine wave inverter that has no feedback system.
It measures the dc voltage and current from the battery supply, and applies the exact amount of correction to the ac output voltage from those alone.
It works over a 2:1 input voltage range and from zero load to power surges well over 5Kw.  And the output voltage corrects better and faster than any feedback system ever could.
AC voltages are slow to measure because its constantly changing sinusoidaly anyway. You need to rectify, and average over several cycles, then apply feedback slowly as to not over correct.
With feedforward an instant snapshot of incoming dc voltage and current can be taken, and the ac output voltage instantly corrected.  Absolute minimal light flicker when a big load kicks in or out.

But that is all another story….

Feed forward definitely works, and for the adventurous I believe there are some very worthwhile benefits possible for controlling any kind of wind turbine from a fast responding anemometer.

[SparWeb]

To visualize what I think you mean, let me make up a plot of CP with wind speed for a typical wind turbine (OK, it's actually mine).




And you're looking for a way to get the last bits out of the power curve using the control algorithm.
Push up the CP at low wind speeds to capture more energy in light winds, and raise the peak as much as possible.
I don't believe it's wise to have the control algorithm trying to push out the top end of the power curve, because it's the furling tail that's limiting the output there.  If the tail didn't furl, the CP curve would just continue out asymptotically to the Betz limit, for a while, until the machine blows up, that is.




Designing a WT for any load control scheme like MPPT works best if the generator is over-sized for the blades.  This way, the control scheme can reduce the effectiveness of the generator to match the energy capture of the blades.

The opposite is needed when using mechanical control like blade pitch, which reduces the effectiveness of the blades so as not to overwhelm a generator working at its maximum efficiency, or preferably, tune the blades so that they work at their maximum efficiency.

You're talking about the first kind of control scheme, of course. 
You will need feedback to make this work.  The latency and randomness in wind is awful, just the worst kind of system to derive feedback from.  I might start with current from a reliable and rapidly sampled ammeter, first.  You need an accurate and quickly responding ammeter.  The response from each step-wise change in the PWM buck-boost circuit will be instantaneous, and this instantaneous response will be too quick to include a response from the WT blades right away.  Which is good because the blades are probably doing something random, as usual.  It will take many cycles later for the WT to change speed, and in the meantime the algorithm can monitor the trend in the current after each adjustment.  The first change in the current after altering the PWM duty will be the obvious proportional change to the average current load.  The trend immediately after that will indicate if the response of the WT is going in the right direction.

If I wanted to try something like this, and I was doing it DIY by myself, I wouldn't want to make the computer too smart.  I'd just want it to make some adjustments and record what the responses were.  Later I could analyze the results myself and determine if each change was an improvement or not.  Then make each improvement the new norm for the PWM control and try it again.  At some point this is going to sound a lot like manually building a MPPT curve for this turbine's characteristics.

The close coupling between the PWM duty cycle and the resulting average current trend seem to make them natural partners for this tracking scheme.  How would you want to do it based on an anemometer alone?
You definitely need the anemometer to be as close to the turbine as possible, without actually being in its wake, or the wake of the tower. 

[Warpspeed]

Designing a WT for any load control scheme like MPPT works best if the generator is over-sized for the blades.  This way, the control scheme can reduce the effectiveness of the generator to match the energy capture of the blades.

Yes indeed, that is definitely the essence of it, you need to start out with more voltage and an oversized generator to keep cut in speed as low as possible.
And further up the scale of wind speed, have something to control and throttle back the loading with pulse width modulation control.

You will need feedback to make this work.

Not if you have already determined your exact load points at every specific increment of measured wind speed from prior testing.

Agree that  the latency and randomness of wind combined, makes any change of electrical loading take some considerable time to settle down at some new rpm and power level.
But once you know for example, that at 10 Kph a specific pwm duty cycle will produce maximum output and optimum blade loading after much prior experimental trial and error, you can program the system to instantly go to that sweet spot any time the anemometer hits exactly 10 Kph.

Likewise at 15 Kph after much testing you determine another optimum loading point and pwm duty cycle.

Once you have both those points settled, its possible to interpolate settings for 11, 12, 13, and 14 Kph.
It should all work out very close, even if its not 100% optimum.

The trick is you work out before hand what the system needs to do at each measured wind speed.  Then apply that correction to a lookup table without having to wait for the whole contraption to finally settle at the new operating point.

With highly variable wind, the loading will be all over the place, but at least it will be very rapidly following the optimum curve, even though it may never reach a true final equilibrium of steady rpm and load.  Gusts will rapidly ramp up the loading without having to wait for the machine to actually change rpm.

It requires no software "smarts" or feedback.  All that is programmed into it during prior testing and setting up.  It just blindly  and instantly follows what has been programmed into it for each increment of measured wind speed.

Yes, it cannot all keep increasing to infinity, at some point it all becomes dangerous, and furling/braking needs to take over.

But the anemometer can also indicate when that danger point has been reached.

[MattM]

Too bad it is so impractical to build some kind of voltage regulation into the tower.  If it was practical then you'd probably want to do it at the mounting point because of the weight.  You'd also probably have to dampen the fluctuations since the changes would happen so often and out of any consistent pattern.  Dump loads at the tower would be necessary.  I guess you could gear down a flywheel, to run ridiculous slow at low wind speeds to mechanically dampen some of it.  As wind picks up momentum the resistance of the flywheel would become resistance against rapid changes.  That kind of dampening would unfortunately become gyroscopic as wind speeds pick up.  If you could use that gyroscopic action to dampen the tower oscilation without preventing furling, it wouldn't be as bad.  Voltage regulation isn't cheap but for larger systems it wouldn't be ridiculous considering the costs already involved.  If all this was possible, what voltage range would be the ideal range to design around becomes a question to tackle.

[Mary B]

So maybe a slight lag in adjustments to match what the blades are doing, lag based on wind speed/gust duration maybe? Say a short 5 second 40mph gust(sustained 20mph winds) hits, it is to short to really ramp the blades up so it is ignored or a very minor load correction is made, next 40mph gust lasts 20 seconds so a sustained adjustment needs to be made after a delay to let the blades start to ramp up... gust lasts past your correction delay so it kicks in... that delay will need to be found to match your turbine/load... doable in software, current limiting can be done up to 50 amps or so with off the shelf parts...

To visualize what I think you mean, let me make up a plot of CP with wind speed for a typical wind turbine (OK, it's actually mine).

(Attachment Link)


And you're looking for a way to get the last bits out of the power curve using the control algorithm.
Push up the CP at low wind speeds to capture more energy in light winds, and raise the peak as much as possible.
I don't believe it's wise to have the control algorithm trying to push out the top end of the power curve, because it's the furling tail that's limiting the output there.  If the tail didn't furl, the CP curve would just continue out asymptotically to the Betz limit, for a while, until the machine blows up, that is.

(Attachment Link)


Designing a WT for any load control scheme like MPPT works best if the generator is over-sized for the blades.  This way, the control scheme can reduce the effectiveness of the generator to match the energy capture of the blades.

The opposite is needed when using mechanical control like blade pitch, which reduces the effectiveness of the blades so as not to overwhelm a generator working at its maximum efficiency, or preferably, tune the blades so that they work at their maximum efficiency.

You're talking about the first kind of control scheme, of course. 
You will need feedback to make this work.  The latency and randomness in wind is awful, just the worst kind of system to derive feedback from.  I might start with current from a reliable and rapidly sampled ammeter, first.  You need an accurate and quickly responding ammeter.  The response from each step-wise change in the PWM buck-boost circuit will be instantaneous, and this instantaneous response will be too quick to include a response from the WT blades right away.  Which is good because the blades are probably doing something random, as usual.  It will take many cycles later for the WT to change speed, and in the meantime the algorithm can monitor the trend in the current after each adjustment.  The first change in the current after altering the PWM duty will be the obvious proportional change to the average current load.  The trend immediately after that will indicate if the response of the WT is going in the right direction.

If I wanted to try something like this, and I was doing it DIY by myself, I wouldn't want to make the computer too smart.  I'd just want it to make some adjustments and record what the responses were.  Later I could analyze the results myself and determine if each change was an improvement or not.  Then make each improvement the new norm for the PWM control and try it again.  At some point this is going to sound a lot like manually building a MPPT curve for this turbine's characteristics.

The close coupling between the PWM duty cycle and the resulting average current trend seem to make them natural partners for this tracking scheme.  How would you want to do it based on an anemometer alone?
You definitely need the anemometer to be as close to the turbine as possible, without actually being in its wake, or the wake of the tower.

[kitestrings]

Are you familiar with Midnite Solar's Classic charge controller?  I'm not trying to get on a soapbox - this is what we used for our system - to say this is the way, or the best way to go, but it sounds a lot like what you've laid out.  Their approach was to have a Wind Mode, and then pre-program in the power curve of the turbine.  They've de-coupled the input voltage from the turbine, as with a solar controller, and then buck it down to the batteries.  The CC does its best to stay on the user defined power curve.  It seems pretty simple and AFAIK they were the first to offer this in a commercial product like this.  They also have some larger units in development (Hawks Bay, Barcelona... I forget which is which), but a Wind Mode would probably come later if at all.  I suspect there is not much demand these days.  Regards, ~ks

[Warpspeed]

Are you familiar with Midnite Solar's Classic charge controller?  I'm not trying to get on a soapbox - this is what we used for our system - to say this is the way, or the best way to go, but it sounds a lot like what you've laid out.  Their approach was to have a Wind Mode, and then pre-program in the power curve of the turbine.  They've de-coupled the input voltage from the turbine, as with a solar controller, and then buck it down to the batteries.  The CC does its best to stay on the user defined power curve.  It seems pretty simple and AFAIK they were the first to offer this in a commercial product like this.  They also have some larger units in development (Hawks Bay, Barcelona... I forget which is which), but a Wind Mode would probably come later if at all.  I suspect there is not much demand these days.  Regards, ~ks

No, I am not familiar with Midnight Classic's charge controller, but it sounds like EXACTLY what I am suggesting.

Like many engineering concepts, its hardly surprising that other engineers reach similar conclusions and come up with similar solutions to a problem.

The difficult part is trying to explain the concept of programming a simple pwm controller to optimally load the turbine at every small increment of directly measured wind speed. Many people just cannot escape the logic of some kind of feedback being absolutely necessary to do this.

I can understand that people just want a commercial off the shelf controller that they just wire in and forget about, like a solar mppt controller. 
But a wind turbine controller needs individual tuning and the user needs to get fairly deeply involved in that.  Its more of a hands on project for the enthusiast rather than some kind of magic black box that would be a commercial success.

It could be something really simple and low cost to build such as a row of potentiometers.
Each potentiometer covers a certain range of wind speeds.
As the wind speed goes up and down, each pot comes into play controlling the pwm.
A led next to each pot tells you which one is active.
Each can be adjusted for best battery charging current at each range of wind speed.

No need for any software.  Tweak each potentiometer for best results and then just let it run by itself.
Its a bit crude, but should be quite effective as a first attempt.

[joestue]

Boost mode power factor control chips using boundary mode for the inductor are simple and easy to put together. They already have an onboard multiplier that takes the incoming voltage and multiplies it by a value to get the peak current X, at which the boost switch turns off.

I actually figured out how to interleave N number of transition mode boost converters by mixing the signals from each current sensor together.

so anyhow you take the input to the multiplier and run both inputs to the voltage on the output of the turbine after the rectifier, and i think  it will mostly work.

some non linear elements to decrease the squared function may be needed.

[Warpspeed]

Current mode boost topology is certainly one possible option for control, and interleaving several, an excellent way to increase the power potential. 
There are quite a few possible options for adjusting the loading on the turbine.

so anyhow you take the input to the multiplier and run both inputs to the voltage on the output of the turbine after the rectifier, and i think  it will mostly work.

If you do that, the multiplier will produce a square law output, for input voltage change.
Square law voltage would produce cubed law power output.
Interesting.....

[Adriaan Kragten]

Stalling of wind turbine blades only happens if the load keeps the rotational speed almost constant. This is the case if an asynchronous motor is used as generator for a grid connected wind turbine. However, if a PM-generator is used for battery charging, the generated voltage is almost constant and more power is only supplied by increase of the current. Increase of the current requires increase of the rotational speed. So the Pmech-n curve is much less steep as for an asynchronous generator. If the correct matching procedure is followed, so if the generator has the correct size and the correct number of turns per coil, the matching can be rather good for a large wind speed interval like for wind speeds in between 4 m/s and 9 m/s. At higher wind speeds, the power has to be limited by the safety system of the wind turbine. Only at wind speeds below 4 m/s, the matching is poor and below a certain wind speed no power is produced at all as the open DC voltage of the generator is lower than the battery voltage. But the energy content at very low wind speeds is very low and I don't see much use of making an electronic device which increases the matching at wind speeds below 4 m/s except for regions with very low wind speeds.

Wind turbines which are grid connected by an inverter accept a much larger input voltage variation than battery charging windmills. Some inverters can even be adjusted such that the optimum cubic line of the rotor is followed. So with these inverters, it is possible to get a perfect matching from wind speeds above about 2.5 m/s.

[Warpspeed]

Here is a quick and dirty block diagram of the general concept.

 wind turbine control.pdf (287.74 kB - downloaded 50 times.)

Dc output voltage from a cup anemometer goes into an analog to digital converter. The binary output  of the A to D splits the wind speed into as few or as many separate wind speed ranges as you wish to have.

A decoder chip drives a series of LEDs to indicate the current active wind speed range.
Associated with each LED is a potentiometer that can be used to tune the amount of loading on the alternator for each wind speed range.

It should be just a case of finding a peak in the dc output current in a steady breeze appropriate to each potentiometer.

The correct potentiometer to tweak is indicated by the LED, and the correct potentiometer dc output is selected by an analog multiplexer.  The dc voltage sets the duty cycle of a pulse width control chip, which in turn adjusts the loading on the output of the rectifier via a buck or boost switching power supply.

Each potentiometer can be individually set for optimum blade loading for each range of wind speed.
If slider potentiometers are used, the physical position of the individual slider should end up being in a nice even curve of recognizable shape, rather like an audio graphic equalizer. 

Something much more sophisticated could be built with a microcontroller with EEPROM memory, but a fairly crude basic system as suggested above, should make an interesting experiment and not be too difficult or costly to put together experimentally.

[Astro]

See now you guys are starting to think about what I am trying to build. There are obvious reasons that trying to advance the power curve to harness more from less wind is not going to yield large gains, so to me the idea then becomes to keep it in the upper part of the power curve for as long as possible. I know you all are probably tired of hearing it, but I really think that a big part of that is going to require viewing the mag plates as flywheels. Because what we are trying to do from a mechanical as well as from an electrical standpoint is to get it spinning and keep it spinning at a certain speed in a constantly changing environment. To ignore the small fluctuations. I think from an electrical standpoint, that is going to mean slow sampling. What we are trying to do is keep a spinning object between a high rpm set point and a low one (staying on top of the power curve). Applying resistance of whatever kind you desire as it approaches the peak of the curve is all we really need to worry about, because if it is spinning to slowly, we do not care. (assuming we have a reasonable cut in speed and have designed and built it so that part is ok). Also if we view the mag plates as flywheels, that will (or should) help carry the mill through a lull or when the wind lets up for brief moments. The "flywheel" and it's inertia will help keep the mill in the upper part of the power curve when the wind goes from 18mph to 13mph and back to 18mph all in a minute.   Smooth out the fluctuations and you gain in the machines efficiency.

[Warpspeed]

That sounds about right, keep the great enormous thing turning at a constant speed during highly variable or gusty conditions.

Then change the instantaneous electrical loading to track instantaneous wind velocity.
During gusts, suck out the suddenly available extra power during the duration of the gust, over maybe a very few seconds.
During brief lulls, reduce the electrical loading right down to minimize any slowing down.

The load change response needs to be very fast indeed, and a cup anemometer can very easily track these fast changes.

The main advantage will be in turbulent air.  It will work much better than very slow feedback.
Truly constant steady wind, is about the only condition where a slow feedback system can keep up with the game.
Feedback absolutely must be gradual, or you can get instability and surging.

Feed forward can provide almost instantaneous correction for rapidly changing input with zero chance of instability.

[JW]

Then change the instantaneous electrical loading to track instantaneous wind velocity.
During gusts, suck out the suddenly available extra power during the duration of the gust, over maybe a very few seconds.
During brief lulls, reduce the electrical loading right down to minimize any slowing down.

I'm wondering about rectifier load here. Most battery's tend to "float" right?  considering state of charge of the batt bank.

[Warpspeed]

Quite right.
Whenever the battery is fully charged, there may be little or no load on the wind turbine, and the charge controller needs to back off the output current.
Battery voltage control must be a second control parameter for any type of control system.

Exactly as with a solar MPPT system.
The MPPT part only works during bulk charging, during absorb or float the controller backs off the charging current.
A wind turbine should work in the exact same way.
Max turbine performance is only required during bulk charging.

[joestue]

Advantage to a boost topology is the converter only needs to be large enough to handle the low wind conditions.

You preserve the hard limit of the battery voltage governing max rpm.

[JW]

One of my favorite references

https://www.amazon.com/Engineers-mini-notebook-formulas-circuits-Siliconcepts/dp/B00072TU8Y

[SparWeb]

Whenever the battery is fully charged, there may be little or no load on the wind turbine, and the charge controller needs to back off the output current.

Whoa.  The turbine needs to dump out the current regardless of the battery state.  The voltage may rise a bit but the turbine's load parameters stay nearly the same and it will happily cook the battery until it melts if you let it.  What you need is a diversion for the battery when it is full.  Use the excess for something else, if you can.

The charge controller must NOT back off the output current.  That's a solar CC, not wind.  Backing off the current from a wind turbine will unload the blades and they'll turn faster and faster until they become projectiles.

[JW]

Couldn't this be explained by Ohms law selected of the three configurations possible.

[JW]

Like the difference between Power and amps

[Warpspeed]

Whenever the battery is fully charged, there may be little or no load on the wind turbine, and the charge controller needs to back off the output current.

Whoa.  The turbine needs to dump out the current regardless of the battery state.  The voltage may rise a bit but the turbine's load parameters stay nearly the same and it will happily cook the battery until it melts if you let it.  What you need is a diversion for the battery when it is full.  Use the excess for something else, if you can.

The charge controller must NOT back off the output current.  That's a solar CC, not wind.  Backing off the current from a wind turbine will unload the blades and they'll turn faster and faster until they become projectiles.

Are you absolutely sure about that ?

Unloaded blades will certainly rotate a bit faster, but the speed cannot just keep accelerating to infinity.

Aircraft and ship propellers generate thrust when the blades try to go faster than the the passing air/water.  That takes power to drive the propeller.
If the blades go slower than the passing air/water we have a turbine that can generate some torque.

When the blade speed MATCHES the passing air/water speed it cannot rotate any faster or slower without adding  positive or negative torque. It just free wheels.

Speed cannot ever just keep on rising higher and higher.

[clockmanFRA]

Hi Warpspeed,

I am talking from a real practical point of view in a moisture climate.

My No 1 3.7mdiameter turbine got up to 7 times the speed of air flow.

I tested this once with a moderate wind and let my No 1 turbine go.

Now i thought that i had it well balanced, but at that speed it was shaking and wobbling and started to get into a vibration wobble. Very scary. And my No 1 had light cedar wood blades on.

Wouldn’t even try it with my heavier fiberglass bladed Turbines. They scare me enough in storm conditions as they are.

Its weather here that gets water every bloody where. Just a little bit and the Turbine blades are out of balance.

My turbines are clamped to the 48v 1300ah battery bank. I use Morningstar Tristar diversion controllers to monitor the Battery charging status.  I have 4off Tristars, and since 2008 they have worked flawlessly and even overloaded they just pass on to the next Tristar and carry on working and so on.  Real good well-made bit of gear, and not costly either. Mine dump up to to 8KW of excess.


Photo shows Turbine control box. Note cobwebs the crawly things still get in.

[Warpspeed]

Yes, I realize that a yacht can go faster then the wind speed, when close hauled, but I find it difficult to believe a wind turbine just keeps accelerating to infinity.

[MattM]

Yes, I realize that a yacht can go faster then the wind speed, when close hauled, but I find it difficult to believe a wind turbine just keeps accelerating to infinity.

It cannot, which is why I suggested a gear reduction into a flywheel.  The flywheel resistance will scale with the gear reduction.  The momentum of the flywheel should dampen wind velocity fluctuations.  Sure it will delay start ups, but is that a bad thing?  Why have your windmill spinning in useless wind velocities to begin with?

[PaulJ]

 Many years ago my 2.7m diameter wind turbine had a connection fail open circuit after the rectifier in high but not extreme winds.
The results were terrifying, I still can't quite believe the mill survived. Helicopter noises, tower flexing and rotor RPM at least triple what I'd ever seen before. The furling wasn't working properly either, presumably because the thrust from the unloaded rotor was significantly reduced.

 The mill would normally have been making about 750W in those conditions. Lose that 750W load and where does the power go? Bearing friction isn't really significant, so the blade speed increases massively until parasitic aerodynamic drag eventually limits it. Eventually.

 I like the feed forward idea and the speed and relative simplicity of a lookup table, but failure modes need to be considered. A mill designed for high winds with the electronics boosting the low wind output could be designed so that if the circuitry fries you're left with the mill directly connected to the battery. As SparWeb said you need a separate diversion load (preferably more than one for redundancy) to control battery charging.

[Mary B]

Whenever the battery is fully charged, there may be little or no load on the wind turbine, and the charge controller needs to back off the output current.

Whoa.  The turbine needs to dump out the current regardless of the battery state.  The voltage may rise a bit but the turbine's load parameters stay nearly the same and it will happily cook the battery until it melts if you let it.  What you need is a diversion for the battery when it is full.  Use the excess for something else, if you can.

The charge controller must NOT back off the output current.  That's a solar CC, not wind.  Backing off the current from a wind turbine will unload the blades and they'll turn faster and faster until they become projectiles.

Bingo! Think of delivering max power to the "load" that can be battery or dump load.

[Astro]

You all are on the trail now.  I can't decide yet if a slow reacting or fast reacting or somewhere in the middle electrical control would be better. I can't make that leap until I get to that point. Each way has advantages and dis advantages.   First thing is we must establish limits. Be that from vibrations or from amperage cooking a battery. So each build (unless you have a factory that can pump out identical mills over and over) is going to be different. As is each battery bank. So to start with we have to have equipment that is adjustable within a range that serves us well. I mean a 48v inverter is not going to work well with a 12v bank. So the equipment must work together with a certain range so as to be able to set the limits, be it voltage limits, amperage limits and tip speed limits.     We will get there. I just have to get this generator built first. Maybe in the next couple of days I will finish up the mag plates and order a few other parts as I can afford.

[Warpspeed]

There are quite a few related issues to consider, but my main purpose of starting this thread was to suggest feed forward control as a solution for optimizing operation in bulk charging mode only, especially in highly variable wind conditions and over as wide a range of wind speeds as are usable.

Every home brew turbine will usually be a unique combination of parts, and every wind site different.  So we need something that is not too difficult to put together, low cost, simple to adjust, and understandable in how it works.

For many people feedback systems are a much more intuitive approach to tackling any type of control system problem, but its not always the best choice in some applications, particularly those that must respond very rapidly in wildly changing conditions.

There will need to be several additional very necessary features and functions, which include protection for both the battery and the turbine.
Those can be approached as entirely separate and different problems to be solved.

Only really interested in discussing best power tracking during bulk battery charging (a form of MPPT if you like) but doing it in a way that does not introduce all the known problems and limitations that PID feedback systems have.

The basic idea is to adjust loading on the machine as fast as possible to match known steady state conditions.

Even though the wind speed is rapidly changing in an unpredictable manner, the loading is also constantly changing, and the machine is always trying to catch up in the right direction.

With feedback, the machine is always headed towards how conditions WERE, it always one step behind what is happening at any particular instant.  Its entirely possible that it can be correcting in the wrong direction, which is never going to be helpful in a rapidly fluctuating wind.

[bigrockcandymountain]

From hours of watching mine behave in pretty severe gusty conditions, the reaction time will need to be very fast.  Around 1 second i would think. 

I also agree that the midnite classic is very similar to the original idea of this thread.  It uses incoming voltage to look up an output amperage.  It is even simpler, since it doesn't need an anemometer.  It reacts very fast, approx 1 second i would guess, but in the gustiest conditions, it even gets a bit behind and the voltage runs high sometimes. 

So, to be clear, the classic is a mppt controller in solar mode, but NOT a mppt controller in wind mode.  It is very similar to what warpspeed originally suggested in this thread and it works very well. 

Yes, you need a load on turine blades at all times. They are scary enough with a load on, let alone freewheeling.  I have only let mine freewheel in the lightest breeze, and it doubles or triples in speed almost instantly.

[Warpspeed]

Incoming voltage must come more or less directly from turbine rpm then I assume ?

A small cup anemometer will respond very fast indeed due to having absolute minimal inertia.

The anemometer does not need to read to extremely high absolute accuracy, only to always be repeatable and read the same value at the same wind speed. 
So if its mounted behind the turbine in some turbulent air, that should not really matter. The turbulence should always be about the same at the same wind speed, or very close.

[adobejoe]

The anemometer feedback/control seems a good approach.  The changes are so quick.  Would an extension rod  say 3 m or more up wind  (in front ) to mount the control device help?

[Warpspeed]

I  doubt in practice if it will make very much difference, as long as its close enough to pretty much duplicate whatever wind speed the turbine experiences.

Its all still just an idea, but its very encouraging that Midnight Classic have already made something similar that is known to work quite well. 

I think between us, we can probably put together our own experimental systems and  come up with a low cost do it yourself version that might work even better....

[joestue]

nothing low cost about reliable diy power electronics, unfortunately.

the lowest cost boost converter i could sell anyone here, would be to simply use the inductance of the transmission line and the turbine as the inductor and simply use 3, mosfets wired back to back (100 volt fets for a 48v system) and turn all of them on and off through a optical isolator. each source of the mosfet is connected together, each drain goes to the 3 phases. you turn them all on at the same time and turn them all off at the same time. regarding power factor? it won't meet the 5% THD requirement for industrial equipment but it is better than a simple rectifier.

you stick with the existing bridge rectifier and battery load. an additional capacitor would need to be added to buffer the inductance of the battery, and a small rectifier and energy recovery snubber could make it work pretty well.

the emi will be pretty incredible, since the entire distance from the rectifier to the turbine is an antenna running at say, 5 to 50Khz with rise times on the order of 1uS or less at around 100 volts peak to peak.