First-order low-pass filter

[domwild]

Hi,

Windmill MPPT is a complex problem unlike solar or hydro. A Brazilian Master Thesis is suggesting the reading of wild AC coming off the mill to drive a SEPIC buck/boost converter via PWM and to charge batteries that way.

The sampling of V and Amp is done within the sine curve with its high-frequency harmonics to get at the instantaneous DC power component. This is one signal.

The other signal is the RPM figure, which is put into a look-up table to get at the power to be produced. These two signals are subtracted and the error value drives the PWM.

Googling for "first-order low-pass filter" Wikipedia enlightened me how the digital simulation of this filter can be done:

y(i) are the output power samples,
x(i) are the input power samples,
dt or deltat is the time interval between samples,
RC is the time constant of the filter, and
n is the number of samples.

alpha = dt / (RC + dt)

for i = 1 to n
y(i) = y(i - 1) + alpha * (x(i) - y(i - 1))

Or: "The change in filter output is proportional to the difference between the last output and the current input"

Questions:

1. Assuming a mill AC frequency of 100 Hz, how many n samples should be considered?
2. Time constant of RC filter?
3. Sample interval called dt?

Anyone out there who can solve this? Is there a connection to Nyquist's sampling theorem i have come across once?
Thanks.

 

[BeanieBots]

I'm no expert on wind power generation or MPPT for such devices but I do have extensive knowledge of SM power supplies and MPPT with solar.
Firstly, I would strongly sugget avoiding a SEPIC (single ended primary inductance converter). These are very nice in theory and have a small componet count, but in reality, they are very inefficient and the most common design couples with a capacitor sending high currents through that capacitor. For anything other than a few watts, that would require finding a capacitor of several hundred uF that is non polarised and can handle many amps You'd find life much easier with a flyback or quad switcher.
see link for quad approach.http://www.linear.com/pc/downloadDocument.do?navId=H0,C1,C1003,C1042,C1116,D8738

Also, I would suggest monitoring the converter output rather than the input. Afterall, it is the amount of useful energy that is important, not how much the turbine can produce. This could then be coupled in with turbine speed to produce the final MPPT algorythm.

As for nyquist, yes, you should sample at a rate of greater than 10 times the rate at which your measured variable can change by the amount of your lowest acceptable resolution.

 

[manuka]

I tend to go along with BBs A1 response to this,but ask -

A Brazilian Master Thesis

Can you give a reference to this please?

wild AC coming off the mill to drive a SEPIC buck/boost converter via PWM and to charge batteries

I assume only modest power levels are being considered here?

MPPT (Maximum Power Point Tracking) charge controllers were discussed in some depth on the old forum back in late 2006 => www.picaxeforum.co.uk/archive/index.php/t-5769.html

 

[domwild]

Manuka,

You first! The Master thesis, which i do not have is called:

Fabiano Daher Adegas, 2006, "Analysis, Simulation and Experimental Implementation of a Low-Power System with MPPT and PFC for Small Wind Turbines Generators in Stand-Alone Applications". Master Thesis for Electrical Engineering Degree, Pontifical Catholic University of Rio Grande do Sul, Brazil.

Have emailed Senhor Adegas at the given email address and am awaiting his reply how i can possibly get my grubby fingers on his thesis.

The article you can read and where this reference appears is:

http://www.ewec2006proceedings.info/allfiles2/846_Ewec2006fullpaper.pdf

Keep in touch!

 

[domwild]

Manuka (again),

Yes, the power levels are small, only 400W.

Beany: Thanks for your suggestions. As i do not have a black belt in electrickery i cannot argue for or against this academic paper. However, if his master thesis gives me all the maths to calculate the components and the maths are not too complex, i can at least try this out as a stimulating mental exercise.

SEPIC converters, according to one academic, are not too well understood. If wind MPPT were an easy problem to solve, then we would see cheaper units around for us wind enthusiasts. This said, the site www.thebackshed.com lists Picaxe code for V, Amp, RPM and dump load connection for a low power mill and the Picaxe supports PWM, so why not see if the MPPT holy grail can be cracked. The paper gives PFC (power factor correction) in addition to buck/boost, which is what one wants ideally.

I am aware of the MPPT discussions on Fieldlines (Otherpower) and thebackshed, one of them i restarted!

Thanks.

 

[domwild]

Manuka (again),

The puter cut off my link, so you better Google:

Maximum Power Point Tracker for Small Wind Turbines Adegas

to get the full link.

 

[domwild]

Manuka,

Forget the "cutting off", the ellipsis (...) does not mean a cutting off of the web link.

Beanie: Thanks for the link to Linear Technology. Have found references to their chips, which can handle up to 100V, which is what I need because my Fischer&Paykel motor generates a very high voltage and I have a 36V (!) UPS, a gift. And the application notes also tell me how to calculate the components, which is something i need.

Question: The Amps stated as the maximum for those buck/boost converters, are they the max. Amps for driving the gates of the MOSFETs (paralleled, no doubt for charging batteries) or what?

As these chips from LT can be slaved with "one's own" PWM generator, like a Picaxe, where is the wind MPPT complexity then? Please enlighten me!

 

[Dippy]

I'd like to second the thanks to BB for the Linear Link.
That looks a useful chip and quite easy (on the face of it) to use.

Why do we talk about MPPT as if it was some magical work of genius? It's just a jolly sensible and obvious idea. Its just a way of getting the best out of whats coming in. I think most people call it 'efficiency' and the best specific method will vary depending on the app. and Switched Mode /Pulse/PWM techniques will probably be at the heart of all these things. Apologies if I'm not overawed.

 

[domwild]

Beanie,

Thanks. The problem I face is as follows: Have seen one buck converter from an Indian project of "Engineers without Borders" in Gujarat (I believe it was called that) with only 65% efficiency. A bit disappointing.

Thanks for the link to Linear Technology; am using a Fisher & Pykel washing machine motor and can expect a high voltage coming off it. Have a 36 V (!) UPS, a gift to be used in my experiments. When I enter Vmax as 100V input, Vout as 42V for charging a 36V battery bank and 5 Amps, then the LT calculator tells me they do not build a DC-DC converter for such a high voltage.

If I find something on the net with formulae for calculating those components, then I can build it.

The link to www.thebackshed.com gives me the Picaxe code and components for measuring RPM, V, Amps and also dump load switching. The Picaxe also gives me PWM capability, which could be used to drive a buck or buck/boost converter. So I am half-way there already.

I have clicked on the black "POST REPLY" icon and the puter has lost my replies from yesterday to Manuka and yourself. What have I done wrong?

Thanks.

 

[domwild]

My previous messages have reappeared now that I used "quick reply". Magic!

 

[domwild]

Dippy,

Thanks. From all my reading and hunting for solutions all I can ever find is engineers NOT using those ready-made chips, they are all building their own idea of battery-charging DC-DC converters for those wind mill MPPT converters. What might be the reason for this? Is it the high voltage spread/high Amps requirement?

 

[Dippy]

Well, I guess you've answered your own question.
Believe it or not I haven't looked at every driver-on-a-chip but the few I have looked at would confirm your idea about voltage range and, ultimately, the current handling.
And, of course, running a P chan from a higher voltage and driving the Gate to 0V may be uncomfortable.

Obviously the spec depends on your input voltage and your current capacity and a good MOSFET driver design requires a good deal of experience, test equipment, time and being able to get your hands on the most suitable components.
Your driving transistors need to be fast with predicatable switching thresholds so that one is ON when the other is OFF. If they are both ON at the changeover you get an unwanted current surge going straight through the complimentary pair - this is called shoot-through.
Bunging any old pair of transistors into a basic 'Boys Book of Electronics' complimentary pair arrangement may work but not necessarily very well. They may not get their ON-OFFs timed too well, so you could have shoot-through or a deadtime on the gate. No driver is 'perfect' but the ready made IC based ones have been designed, tested and built by people with brains and cash.But they are limited....

Your drivers need to be able to handle amps easily and very, very quickly. Switching times of <<100nS really are ideal. You can do the calcs yourself. get the Data Sheet for a big fat MOSFET and it will show the capacitances and charges associated with the Gate. And you'll know from schoolboy theory about the relationship of current and coulombs and time.

In fact if you don't get it right then you will pathetic efficiencies like 65%
Or 0% and smoking FETs.
I'd love to have a bash at a big switched mode but playing doesn't pay the bills.

 

[domwild]

Thanks, Dippy. Yes, unless I can get my hands on something like this Master Thesis from Brazil and am shown how to do the calculations, I am going to be "stuffed".

 

[hax]

Hi Domwild,

So you say that the two inputs to the calculation are the RPM of the turbine and the instantaneous power being used. The output of the calculation will then control the PWM of the batteries to be charged. Right? So the aim is to regulate the charging of the batteries while keeping the power output of the turbine at its maximum?


OK here is another way I would tackle it:

Strap your wind turbine to your car and on a still day, drive your car at 10km/h, 20km/h, 30km/h .....up to 100km/h

You will need a variable load on the turbine.

For each constant speed multiple, vary the variable load and note the point at which maximum power is produced. Then write down the RPM of the turbine at that point.

Now you know what RPM the turbine will have to be running at to produce power at maximum efficiency for any given wind speed.


Next, when the turbine is mounted on a pole, all you have to do is have a separate wind measuring instrument mounted somewhere on the same pole.


To generate at maximum efficiency, the problem is then very simple.

main:

check wind speed
check turbine RPM
lookup table
is turbine RPM lower than best efficiency? If yes turn off charging (this will increase RPM) if no, turn on charging (the load will decrease RPM)

Goto main


The faster this program runs, the faster the PWM. But even a dead slow 10Hz PWM would be fine. When the load is off, energy is being built up in the turbine in the form of higher RPM.


Now for the charging circuit. You mention you are using a Fisher and Paykel washing machine motor:

If you don't re-wire it, it will give you a relatively high voltage, somewhere around 100 to 350V DC. This is great for a reasonably long run of cable back to the shed.

Most people like to fiddle with this and re-wire to 12 or 24V. I wouldn't bother.


You just need to find a way to drop this to a useable voltage once at the batteries. Well it just so happens that a computer power supply is very efficient, has a SMPS built in, is cheap, and can work on a DC input of around 90 to 350V. (check the lable) The first thing a SMPS does is run 240V AC through a bridge rectifier and then to a capacitor. So DC won't hurt it.

Now if you need more than 12V then you can stack multiple power supplies together. But there is a small modification you need to do to do this safely. (I will elaborate if you like)

If you have a 36V pack, you will need 4 computer power supplies. Wire the 4th power supply between the 5V output and 12V output (7V). This will give you a total of 43V which is about 14.3V per 12V battery. Close enough.

Then select your favourite mosfet to pulse the current when required.


Is this not doing the same thing, without complex maths?

OK so I didn't take into consideration the effect that the temperature of the air will dictate its density and perhaps alter the maximum efficiency at a particular RPM. This might make the lookup table only work at one particular air temperature. But I dont know the maths to work out whether this is a negligible effect or not.

 

[domwild]

Thanks, Haxby et. al. for the suggestions. Will have to read through them and try to digest them. Read the same question at www.fieldlines.com and at www.thebackshed.com. Your idea and my similar idea of just upping/downing the pulsewidth via Picaxe depending on, in my case Watts produced to get at the "sweet spot" was not accepted completely by the members of one forum.

Anyway, will read the suggestions more thoroughly.

 

[Dippy]

Don't worry about it Dom - everyone is the World's expert

How many Forum contributors does it take to change a light bulb?

Answer: - it would never get changed. There would be safety issues, arguments about who is qualified to do the job, there would be an argument about the selection of the bulb, the production of a spreadsheet to calculate the environmenal impact, an argument about the price, the delivery delay from China, whether or not to use anti-static precautions, should we test it on a breadboard first , a three week pointless & circular discussion on the Forum and finally whether or not to connect a resistor in series. Then report back to the Forum with each party ensuring they get the last word in on the Forum.

It's tough. Try and absorb what everyone has contributed. And then apply the BS filter and ask someone qualified.

 

[BeanieBots]

And don't forget to post your results.