Building a voltage booster

[Flux]

You should be ok to test with resistive load but watch the output volts and don't let it go too high.

A current limited power supply should be safe to test into a battery load.

I am not sure what you can learn about the control, it will at least prove the converter bit.

Possibly a resistor in series with the power supply would be the best way to simulate the alternator, with a battery load.

With about 10v in, the pwm should be near 50% and let you pass a small current into the battery. As you raise power supply volts the converter should phase back and let the input volts rise and pass more current to the battery. With about 10A into the battery the psu volts should be up to about 24 and converter phased right back.

Probably a series resistance of about 0.1 ohm would be in the right order.

Flux

[elt]

Here's the corrected low pass filter -

I've also swapped the pins of the analog in and PWM out signals. The Tiny45 has over 30 port functions that can be assigned to its six pins but the assignment in the original schematic wasn't a valid combination; now it should work.

- Ed.

[elt]

I guess my understanding of MPPT with the booster circuit is stuck - this is all more or less a question:

If the FET is held in the off state, the prop will spin as fast as the wind will turn it. The converter and battery will cause no load on the alternator/prop because the volts are too low for current to flow into the battery.

In what I call zero duty cycle, the FET is turned off and the mill would be free-wheeling below battery cut in voltage.

When the duty cycle is increase up from zero, here's a range where the circuit may be boosting but output voltage is still too low to conduct current to the battery. Except for some inefficiency in the boost circuit, there's still no load on the alternator and the blade doesn't change speed (much) and the input voltage doesn't change (much). (Is this correct or is there something about the battery that will allow it to use even a very small amount of inductor-stored energy? If so, then the alternator is loaded, the blade slows down and we're into the next case...)

Increasing the duty cycle allows the inductor to store more energy and eventually a point is reached where the voltage achieved is just high enough for current to flow to the battery. This point isn't stable because now there is a load on the alternator and it will slow the mill down some and the input voltage will drop and the duty cycle will have to be increase to get the output voltage up a little bit more so that current will continue to flow into the battery.

If we were just sweeping the PWM from zero on up, we'd go from no load on the alternator and no power and eventually get to the point where the blade has slowed to to it's maximum power point (at a slower blade RMP and lower voltage then the free-wheeling, unloaded alternator.)

Here's where I get really confused:

What happens if we increase the duty cycle a little bit more from that point? I realize that if the duty cycle goes to 100%, then the alternator is shorted and mill stalls/stops. But what happens if th duty cycle is just increased a little? The inductor takes more of a charge but it can't develop more volts to the battery. Likewise, it can't deliver more amps to the battery because that would be more power and, by definition, this case is where we increased the duty cycle past the maximum power point. Or does it briefly deliver more power and slow the blade down to the point where it delivers less power... So, where does the the extra inductor energy go? Heat? To slow down the blade. Or, if it can't discharge, does it continue to draw power from the alternator on the next cycle of the PWM clock? (If not, that would imply that there might be a maximum power "plateau" instead of "point.")

Thank you,

- Ed.

[stephent]

From what you are describing happening and trying to make a small bit of power at that point--it would be better to think of the shape as a sawtooth pattern--really small pointy "top" and no plateau and a very steeply dropping down side.

The coming up side will more/less be a "sine" type shape, good for extracting power from without too many problems--but go over that max point for just a few seconds and the rpm of the rotor will fall too much and the controller will have to correct a bunch. It will be a fine line and will have to follow (lag) the alternator in what the alt can actually do at that point. If you are after that very very  peak in power, good luck, wind changes so quick "usually" at low velocities and heights.

There would be just as many wind mppt controllers as solar mppt controllers if it were easy as solar.

But sooner or later--someone has to build a viable one that will work--go for it.

[Flux]

I think you have to look at this as energy transfer.

With no gate pulse, you are right, there is no conduction below main bridge cut in.

If you gate the mosfet you store energy in the inductor. When you turn the fet off the energy has to go somewhere. There will be a large voltage spike and during the interval when this spike is above battery volts current will flow into the battery.

With small fet conduction angles, the energy will be small and only a momentary current spike will flow. With about 50% fet drive pwm the energy stored will be such that with a 2;1 boost the thing behaves more or less as a stiff 2;1 voltage converter and energy will be transfered during the whole of the fet off period.

The thing will still boost some energy from considerably lower than half battery volts but conduction will be for a shorter period than the fet off period.

The prop is also an energy source. If the pwm is too short the prop will have to run fast to deliver its energy to the battery. If the pwm is too long the prop will run slow.

The maximum energy is transfered when the prop runs at a speed that keeps it on the peak of its power curve for that wind speed.

The system is stable at any point, the available energy from the prop will be transfered. The prop speed will settle to a point where its available energy is transfered. It is the same as a current limited circuit, power flow will sort itself out and the supply voltage will adjust to suit.

Even if the voltage converter had a precisely defined ratio it would still be stable because of the series characteristic of the prop's power capability.

You must limit the gate drive, it must never go to 100% ( limit it to 60% should do)

At 50% the alternator - converter combination will be too stiff and the prop will stall just beyond cut in.

The analog control works by feeding back a current signal that phases back the pwm.

A high gain loop would result in constant current output and the prop would run away.

As you lower the gain the current rises, the lower the gain the higher it rises until at zero gain you are back at fixed pulse width and stall.

By choosing a suitable gain you let the prop speed rise at a rate close to that needed to track maximum power.

I have no idea how you do this by digital means, you may be able to implement a variable gain or you may have to attack things from a completely different direction but the end result must be the same. You need a pwm for each wind speed, chosen to keep the prop output a maximum at that wind speed.

Flux

[TAH]

Wouldn't you be able to use a double curve for wind speed and power that had about five cal points to match it to any given generator? If you can calculate the rpm the particular calibration should be able to always load the mill to a set current at that speed. At any speed of wind the output power available will always be pretty much the same, it may not be perfect but probably close enough.

[altosack]

Flux,

I keep hearing you say that you have no idea how this would be done digitally, but you are a smart guy, and I don't believe it for a minute. Just in case you have a stumbling block, let me try to give a really short comparison of analog and digital. I am by no means an expert, so it'll be for my benefit too (if someone comes along and corrects me).

Realize that anything you feedback with an analog signal could be converted from analog to digital, processed in the microcontroller (the same way your analog part would), converted back to analog, and you would have the same feedback signal. For example, instead of adjusting the gain of your feedback with a potentiometer, you can modify the multiplication factor in the program. Sorry if this is really simplistic, I don't mean to talk down to anyone.

If all microprocessors could do was this, they would still have advantages in some cases because you can replace a lot of parts with some software (and without desoldering). However, they can also skip steps like converting RPM to an analog voltage with a tacho chip. A disadvantage to the microprocessor is that analog feedback devices usually respond faster (if they're doing exactly the same algorithm, which doesn't take advantage of the microprocesser), and with nearly infinite precision.

To me, the big advantage to microprocessors is that they can take big equations that relate disparate things in a way that is difficult or impossible with discrete parts. For example, in this case, everyone keeps assuming that we need to feedback on the power by directly reading the power (or current) to track the maximum power point. What if we start with two equations for power, set them equal, cancel out the power, and only use RPM and battery volts (along with a few constants) to calculate the duty cycle ?

This way, the feeback signal we are using (RPM; battery volts is only slowly changing and is a constant at any given moment) already has the rotor response built into it and we have no problems with feedforward (and gain rate instabilities) at all like we do when we feedback on current.

I'll present my (digital) solution to this problem in another thread another day; I started it here, but it got too messy, I deleted it, and I'm tired. By the way, once the equation is solved (it's iterative), it can be approximated by something much simpler that could even be implemented with analog feedback. The simpler equation works particularly well over a small range like what we need for low-wind-only boost. Of course, it can also be approximated by trial and error, without the big hairy derivation.

Best Regards,

Dave

[altosack]

Yes, this is exactly what I'm talking about. I don't know the number of calibration points necessary, but the lookup table will of course be specific to each given generator. You'll have to measure the RPMs/Volt and approximate the expected Rotor Cp and optimum TSR as best you can to generate the table.

"Close enough" is really the key here, because the TSR plateau lets our close enough be pretty good (certainly a great improvement over not using a converter), even if other assumed parameters aren't perfect.

Dave

[Flux]

Thanks for your comments Dave.

I basically follow your idea of how to implement gain by multiplying by a factor.

My problem is mainly that I was brought up with amplidynes and thyratrons. I have never fully got to grips with basic logic and I am absolutely hopeless at maths, so anything that needs that type of thought is difficult.

You are right that this can be done from speed, that is the first way I did it, and I also suggested to Elt that it was a way. I did not find it as easy as current, but I have simplified many things from that first attempt.

I assume that a processor can calculate speed from the interval between voltage crossings. That would seem more accurate and less bothered by noise than using input volts, which is related to speed, but a bit indirectly.

To you it is probably easier to mess with software than change a few wires, but not the case for me. Just a way in which modern ideas move on.

The rest of the world seems to be hung up on how to control this thing, my problem has been in adapting basic converter ideas developed for a few watt supplies to something that can handle serious power.

I only proposed this boost converter as a simple solution and one that I have used for years. The Buck converter option is a far more elegant solution that reduces the troubles of volt drop in diodes and cables. It is a far more challenging problem when all the machine output has to pass through the converter rather than a few hundred watts in low wind.

I am sure the solution comes closer every day and when it is done reliably we shall not have to worry about fried stators and we shall see outputs several times what is safely possible with resistance matching.

Flux