load matching through pulse width modulation

[electrondady1]

i want to be able to hold my vawts at 60% of wind speed.
regardless of what that wind seed is.
i think i will need an anemometer to establish a frequency .
and a comparative circuit(s) that will sense the frequency of the mill(s)
and switch the load on and off to keep the speeds at 60 % of the anemometer's frequency.
i want to avoid using voltage output as a medium of control as the alternators all produce differing voltages.
the glitch is, i have zero knowledge of electronics.

modern automobiles continuously monitor engine functions.
is there a way to scavenge something from one of those systems?

[TomW]

Some used to use magnetic pickups in the ignition system Particularly timing pickup behind the main pulley  on an '89 Ranger but I only know this from having to replace it.  I am no wrench tho so no clue if they can be retasked for maybe a pickup on an anemometer or the turbine itself?

Just spitballing ideas see what sticks to the wall.

Tom

[electrondady1]

yes tom, thanks.
 i was thinking of a mag and a sensor on the free wheeling anemometer to give a base frequency and another one on  the mill to give it a frequency .
some how those two pulses need to be compared and an circuit closed to add load to the mill.

some members of the forum  are using those bike computers/speedomitors
i wonder if there is a way to tie some of those together to get what i want.

i'm not afraid of mosfets , i just don't know what they look like.

[electrondady1]

i know that computers can be programed to act as oscilloscopes.
so i know they can be utilized as a monitoring system.
so what is required is a program that monitors the frequency(or voltage, for that matter) of the anemometer
and calculates 60% of that number,
 and acts to match the windmill to that number   by sending a control signal to a mosfet.

[bob g]

for the sake of discussion, and fleshing out the idea a bit

what if one were to do the following

use an anemometer to measure windspeed, and provide that frequency to a micro controller
input as one data input

use the alternator frequency and feed as another data input to the same microcontroller

develop a lookup table for the microcontroller to compare windspeed/frequency and compare it to alternator rpm/frequency
and provide an output to a mosfet driver or a controlled rectifier or to provide control of a buck converter.

this leaves the problem of how to control the dump load, because as the machine produces more than what is needed the duty cycle
of the controlled rectifier/buck converter will reduce to a low number, so

we use another micro controller that takes an input from the duty cycle of the driver output of the first microcontroller, it then has its own
lookup table which is similar to the one used for the first microcontroller.

so when for instance the duty cycle of the first controller goes down to 25% the remaining off time (75%) would turn over control to the second
controller which would use the lookup table and compare wind speed, alternator speed, and compare against the available duty cycle (in this case 75%) and set a duty cycle appropriate to the driver for the dump load that would keep the alternator speed at the optimum speed for the available windspeed.

the two controllers would have to hand off one to the other control as the duty cycle shift, giving priority of course to the first controller which provides for primary battery charging.

there would also need to be a fail safe controlled by a watchdog timer to take control of the windgen should one or both controllers lock, fail, or lose their minds.

in all likelihood this could all be done on a single micro controller, but i think it might be safer to do it with two microcontrollers, wherein if one fails the
other would provide some level of safety.

the controller part should be fairly straight forward, and relatively inexpensive, the buck converter however might be significantly more expensive unless one is pretty handy at surplus parts.

i am no expert on microcontrollers, but i think i could put together a working prototype using a couple bs2 microcontrollers, and those that are
very good with pic chips could probably put together a working system in an afternoon if they had a clear objective related to them on what is needed.

as for the buck converter, they don't get terribly complicated until they get to running at above 25khz or so, below that they are noisey but
from what i can tell easier to work with?

i would sure like to see some development in this area, and see it done in the diy community rather than some commercial concern where
there is no way to tell how they do it.

bob g

[Janne]

I've been working with a bit similar concept of what bob g is proposing, but i'm not using any anemometer as an input for the lookup table, just only the generator RPM. The diary about the project is located here; http://fieldlines.com/board/index.php/topic,130334.0.html

So far I've nailed the problems regarding the FET driving, and the thing seems to be happy with a constant 500W going through it to the dump load, in my test bench. I plan to install it to the axial flux turbine later on during the summer holiday.

While working on the system, If i'd now start from scratch I'd do some things differently. For example, I think the idea of using a current feedback might be non-necessary, and the system could work with open loop control, as in just setting a defined PW-value for the PWM generator based on the current generator RPM. That approach would take a bit more trialling and erroring, mayby with a potentiometer controlled pwm generator at first, but it would make for a much more simple hardware. Also, For 12V systems I'd go with synchronous rectification, since the freewheeling diodes waste a lot of power on 12V.

[JeffD]

I also went with a synchronous buck converter since I use a 12v battery bank.  Building the buck converter and controller was a steep learning curve but the results were well worth the struggle.  I didn't use the buck converter on a VAWT but on one of my small HAWTs.  I used 200v n-mosfets on the high and low side and a floating high voltage half-bridge n-mosfet driver chip.  The buck converter is running at 31 Khz and the duty cycle is controlled by a Freeduino board using an Atmel AVR atmega328 microcontroller.

I've got four control algorithm's programmed in that can be selected through a menu on the LCD display.  Still playing with each one since none of them are perfect, yet .

  The simplest algorithm which is not mppt but does a pretty good job at maintaining a constant TSR is the RPM/PWM duty cycle table lookup method.  All the computer does is get the turbine rpm and look up the pwm duty cycle in the lookup table using linear interpolation and then adjusts the duty cycle a little bit based on battery voltage.  Up to 20 points can be put in a table but 10 seem to be plenty so far.  The microcontroller can generate a new table based on turbine diameter, blade Cp,  blade TSR, Alt volt dc/rpm, Alt phase resistance, transmission resistance, and nominal battery system voltage.  The table can be modified manually through the UI ( a pain if you are doing a whole table) or the system can be put in Train/Optimise mode after auto table generation.  The Train/Opt mode will adjust PWM duty cycle to try and maximise power output when RPM is fairly steady (turbine not speeding up or slowing down).  This mode uses rpm, current feedback, battery voltage, and wind speed to determine if its on track.  If a new PWM duty cycle is determined (by changing duty cycle up and down up to 20%) that improves output by more than 5% then the rpm/pwm% table is adjusted.  Depending on how the wind is behaving this can take a long time to optimize the whole table.  Multiple tables can be saved on EEPROM and selected to be used at any time. 

I do have one very small savonius VAWT that generates up to about 10 watts at 13m/s but I am not using a buck converter on it.  I used a boost converter which is controlled by a KA7500 PWM chip.  Someday I may rewind the stator for high voltage and use a buck converter since it is more efficient.

[electrondady1]

oh boy!
 thanks so much .
you guys will get a laugh out of my lack of knowledge.
back in my school, we spent a great deal of time on ,
why does something look good and something else doesn't  
and , how much does that cost and how can we make it less expensive .

i had to look up, "lookup table"
and found out it eliminates the need for the devise to calculate, .
and wikepedia helped me understand micro controllers as opposed to micro processors
understanding mosfets will take a bit longer.
electronics is a subject that requires you to start reading at page 1 in order to understand the lingo and concepts.
difficult to  jump ahead to the next chapter.

maybe hawts are different, (i haven't been paying attention)
but for a vertical mill i believe 60% of wind speed provides maximum output.

i may be missing something here ,but with an mppt devise ,and a purpose built alternator ,  isn't a buck converter redundant?
and as flux alluded to in Janne's link , furling itself could possibly  be made redundant .
(that ought to raise some eyebrows)

as far as the charge controller and dump load, i guess they could be incorporated but why mess up a good thing ?
those are separate devises and functions and would still work with a mill that is load controlled by pulse width modulation upstream from them.

i need to study janne's method to understand it better

it just seems that by controlling the mill by  comparing frequency, you don't need to get involved with the high amperage output of the alternator,other than turning it off and on.


ok, getting smarter all the time .
a buck converter is not another name for a boost converter!

[dyslexicbloke]

Interesting post, I have some ideas that may help.

Add a magnetic pickup to your rotors and some small magnets to drive them.
You will get tiny pulses with a somewhat variable shape.
Feed these to a compactor to get a square wave with fixed duty cycle and variable frequency that is proportional to your RPM.
Count and compare the pulses ….:-
You could do this with a micro or PLC but you can also do it with simple logic chips


Use desecrate counter chips
Mill feeding chip 1, anemometer feeding chip 2
Reset both if either carry.
Add some load on chip 1 carry.
Shed some load on chip 2 carry.

It also occurs to me that you could perhaps build a PWM circuit that accepted a digital input to set its duty cycle.

Add a third up/down counter or shift register, driven from the carry pins of the first two, to modify your load.
PWM could be derived with a multiplexer chip tied to the output of counter/register 3.
A system like that would always find its natural equilibrium and you could easily scale either input. (more counters).
Hope this is of some use.
Al

[JeffD]

The beauty of the Atmel AVR microcontrollers is that they have built in hardware support for doing PWM, timing events (counters), and analog to digital conversion.  An Atmega328 costs about $5 and has 6 PWM channels, 6 channel 10-bit ADC inputs,  USART, 32KBytes for code, 2KB ram, 1KB EEPROM. 

There is nothing complicated about the control algorithm code.  For each iteration in the control loop the rpm is looked up in the table to get the duty cycle for the PWM.  The duty cycle value is then put in the PWM duty cycle register and that is it for calculations.  Its about 6 lines of code to do the lookup, interpolation and final output.  If The control loop only runs every 100ms (ten times a second) and takes less than 1ms to execute (atmega328 running at 16Mhz 5v).  One of the digital outputs of the atmega328 is configured for hardware PWM so when the appropriate PWM duty cycle register is written too, the hardware on the micro-controller does the rest.  The PWM digital output pin is connected to the input of a half-bridge driver chip (a little 8 pin chip) which takes care of driving the two mosfets.  Have a look at Tim Nolan's solar mppt pagehttp://www.timnolan.com/index.php?page=arduino-ppt-solar-charger.  He has a good write up (and a synchronous buck converter schematic) on how to use the popular open source Arduino micro controller board to control a synchronous buck converter.  He also provides the source code that you can adapt for doing wind turbine mppt.

I use Gizmo's rpm sensor circuit (found here http://www.thebackshed.com/Windmill/PicLog.asp) (4 diodes, resistor, opto isolator) which uses the wild ac from the wind turbine alt to generate a digitial signal that feeds into one of the digital inputs of the atmega328.  I did add a step down transform (5:1) since the turbine alt voltage gets up to about 80v just before furling (13m/s).  Each time a change in the digital signal occurs (High to low, low to high) the micro controller gets an interrupt which runs a short little routine that reads one of the timer registers to determine the time since the last change.  Knowing the period of the pulses and number of magnet pole pairs on the rotor its then fairly easy to calculate the rpm.

An Arduino board fully loaded (the brains) runs about $30 Canadian which can be connected to a computer through USB to allow programming and data logging. All of the development software is free so no extra costs there.  The buck converter parts + pcb for two units were bought through http://mouser.com and cost me $18 Canadian not including the shipping.  I purchased a number of other parts for other projects so that the $20 mailing charge didn't hurt as much.

My little Sovanious VAWT uses Benesh type blades and seems to run well between a TSR of 0.6 to 0.8.  I haven't done a lot of data collection on this unit so there could be a lot of room for power output improvement.  I lost interest in it after building a HAWT that had the same wind area and put out almost 3 times the amount of power of the VAWT at the same wind speed and that was before I started working on matching the alt to the blades of the HAWT.  With the buck converter the HAWT now puts out about 5.6 times the VAWT power output at 13 m/s.

[electrondady1]

jeff ,thanks a lot .
 from your description it almost sounds like i could build one.
i will need to read these reply s quite a few more times before it sinks in.
and maybe a lot more on the Internet.
the part that intimidates me most is ensuring all the voltages, amperages , signal wave amplitudes
and  shapes  of the various components are compatible with each other.
i hate the smell of that magic smoke.


perhaps i wont need to totally understand how a radio or t.v. works in order to build it.
   maybe i will be able to fix that old  ten band stereo graphic equalizer when I'm finished !
or make my old Traynor amp. sound like am Ampeg s.v.t. by modifying the tone stack.

dislexicbloke, i can't find a reference to any sort of electronic  "compactor"
 with all due respect , in light of your "handle" and all,
could you mean a capacitor ?

[dyslexicbloke]

oops ....
I wanted to say comparator .... must have misread the spellcheck.
That a problem with spellchecks .... you get a list of 5 words youy cant read 
There are loads available, most with high impedance inputs so design is simplified.
EG .... http://www.maplin.co.uk/Module.aspx?ModuleNo=2949
That one is a quad, 4 comparators on the one chip, and costs less than £1.00 - even at Maplin.

Sorry about the spelling ... I do that sometimes.
Al

[Janne]

Hi,

I think the magnetic picup and extra magnets are unnecessary. It's just easier to use an optocoupler to directly read the frequency of the generator.. At least it works perfectly for me.
I'm sure it's also possibly to do it all with analogue electronics / logic IC's, but it's just much easier with a microcontroller. Much less hardware needed, and it's much easier to change the parameters of it later on.

The controller doesn't make furling redundant, in fact furling is much more important with it than ever, since the task of the controller is to keep the turbine at it's peak power rpm, so there is no stalling to help limit the power. Dump load will have to be installed as before, since the turbine must not run unloaded. When the batteries are full, and if the heat from the dumpload is not utilized, then it's possible to change the controller to maxium loading, driving the turbine into stall, reducing output lessening the burden on the dump load and on the turbine itself.

[BrianSmith]

I am trying to exactly understand what you wanting to do because I think I have something that you can use for this problem. 

Are you trying to create a 60% duty cycle pulse synchronized to the output of the generator?  I built an RPM meter with a display and alarm output a while back.  It is powered by 12V and connects to one of the inputs of the bridge rectifiers.  I don't have the output you are looking for, but the alarm output could be made to perform this function and give you a nice 12V output to drive your FETS.  Here's a picture of it.  Let me know if you would like to try it and how it needs to work and maybe I could tweak the code and mail it to you.

" alt="" class="bbc_img" />

[Fused]

Brian, do you have a 24v model?

Just curious.


Fused

[BrianSmith]

It can take the phase input for upto a 48V generator as is.  With some part changes it could go higher. It can run off of 24V as well as it sits behind a regulator.  The input is protected and made to take the higher voltages. 

[electrondady1]

Brian, thanks for your response.
i think it's true for hawts as well as vawts that for a given wind , a mill can support a specific load .
but as wind speed and its inherent energy vary so does the load it can support.
all i want to do is control the amount of load the mill sees in a given wind .

if i understand the devise you have built, an alarm is triggered (a circuit is closed) when alternator rpm exceeds a designated speed.

[BrianSmith]

Yes that's right.  I thought you could hook up an alarm or a big shorting relay upto it to help with keeping the generator going thru the windy season.  Seems like a lot of the stators go up in smoke if you can't get them shut down in time when their furling doesn't work exactly right.

So for what your describing, a simple plan might be to vary the duty cycle from  100% at low RPMs to a lower and lower duty cycle as it goes faster and has more power to deliver to keep the stator from overheating.  My concern with that is if you unload the mill when its really windy by 50%, its going to want to spin even faster.  Seems like when its turned off, you might want to switch it to a reasonable sized dump load that won't allow it to burn up, but will help slow it down, but I'm not really sure. 

So anyway, I can make it switch the alarm output at some duty cycle relative to the RPM easily enough.  It actually has two outputs so maybe when one is off, the other could turn on to do the dump loading?

[electrondady1]

more like the reverse,
so that in the highest wind ,the duty cycle is 100%
and in lower winds the duty cycle becomes increasingly intermittent.

[BrianSmith]

I am confused about something.  If the wind is high, the output power is high, so you want to reduce the duty cycle (100% is always on, 10% duty cycle is 10% on) so that it can't provide as much current, right?

[electrondady1]

ok
 i think you are going to use your devise as a safety back up to the furling system.

on a vertical mill there is no furling .
in high winds i want all that juice.
i can get a charging voltage @30rpm
but in low winds i don't have enough swept area to constantly push the amps into the battery.
if i can load the mill intermittently i can keep my rpms  up.

[BrianSmith]

I get it.  If you want to give it a try, give me a table with maybe 5 RPM / duty cycle points and I will load one up for you to try.

Something like this:

0-150 RPM - 0 duty cycle
160-200 - 20%
210-260 - 40%
270-330 - 60%
340-400 - 80%
410+ - 100%

Lets start with just something easy like this maybe...  Just put in the numbers you want to try and I will update the code and send it to you to try.  The meter only reads in 10 RPM increments.  The other question is what frequency clock do you want for the duty cycle clock.  I would suggest something slow, maybe 200 HZ or less to keep your FETS from getting too hot from the switching.

[bob g]

lets try this one on for size

if you know the operating parameters that your windgen operates under over a range of windspeeds
then you can measure the wind via an anemometer and log that into a table or develop code to reference that
value

the battery can also be measured to determine state of charge and code can be developed to set the pwm to
optimize the charge rate and keep the wind gen operating at its peak rpm for the specific wind speed measured.

now the resultant off cycle of the pwm has to be dealt with
because in higher winds as the batteries top off the pwm will be very low and the mill will be running unloaded, which
no one wants, so

a second processor can sample the first processors "off cycle"  and use it to drive the dump load via another pwm
algorithm.

the priority always goes to the first processor, to make sure that the mill is always running at peak design speed for the specific
windspeed at the time, and the state of charge of the batteries,

if the windgen begins to overspeed (which it will when the batteries are either nearly charged or there is higher windspeeds) then the second
processor knows to start taking advantage of the "off cycle" of the first processor to drive the dump load enough to keep the windgen speed
down to an acceptable rpm, unless...

the anemometer reading is such that one clearly knows that he does not want his machine running in, then the processors can default to
control of a linear actuator to pull the tail and furl the system.

the microprocessors and their related code can have any amount of hysteresis built in to keep the system from surging or cycling uncontrollably and to account for days where the wind comes in gusts.

basically the first processor is the master, and it takes priority controls the pwm charge rate and also provides a contol handle of furling, and controls the on time of the second processor (slave) which controls the dump load primarily and a second control handle of the furling.

if either processor fails, the other can take control of the windgen and initiate furling and/or braking

i think it could work, and work well

bob g

[BrianSmith]

I basically agree with what you are saying.  Here's the way I see it (actual results of this plan may vary  )

1.  Get a performance curve of the wind gen that determines the power vs RPM characteristic.  Determine what the maximum power you want to draw from the generator to keep it from smoking.

2.  Setup the 1st PWM to start loading at some RPM a little past optimum and allow it to load it up (increase duty cycle) to keep it at the optimum RPM.

3.  If the batteries are full or the wind is too strong, even when the PWM is full on, it won't be sufficient to slow the generator at which point you begin loading the dump load in PWM fashion.

4.  So the second PWM would be enabled when the 1st PWM is full on and the output RPMs are higher than the optimum setting.  If you had a linear voltage output proportional to the RPMs generator, it would be pretty straight forward to do this with a couple comparator gates (one is configured to oscillate, the other comparing the oscillator signal (ideally a triangle waveform) and the RPM voltage to make a duty cycle that varies with RPM voltage.

5.  Somewhere between the two PWM loops, you still need a furling mechanism that triggers based on too much speed or power.

How does that sound? 

[bob g]

sounds to me like, if one were to have the hardware in place (anemometer, master controller, slave, and brake/furl)
then getting it all to play nice together would boil down to proper program coding.

once it comes down to that it would seem anything is possible.

or so it seems to me

bob g

[electrondady1]

i was at an electronic supply store last week picking up some rectifiers.
i mentioned the Arduino board and the owner had heard of them and felt he could get one for me.

is it realistic of me to think i could put something together with out any real knowledge of electronics?

[Bruce S]

is it realistic of me to think i could put something together with out any real knowledge of electronics?

Edaddy1
There's tons of stuff you can do with these little puppies
go up to the .instructables. website and do a search for arduino projects and have a blast reading how versatile it can be. My current interest in the fully automated growing unit.

Bruce S

[JeffD]

http://www.arduino.cc/ has lots of tutorials and reference material about using the arduino ie interfacing to the real world and programming. 

[JeffD]

bob g: I agree with you on having a modular system.  My system was designed to use 3 Arduino boards:
 
Arduino board 1: controls voltage clamp on the input to the wind turbine buck converter.  It monitors the voltage input to the buck converter ie just after the rectifiers and If the the voltage starts to go above 70 vdc then the arduino uses one of its PWM channels to control a dump load that is also fed off the input to the buck converter.

Arduino board 2: monitors the wind turbine performance and controls the buck converter.

Arduino board 3: monitors the battery bank state of charge, performs 3 stage battery charge control, monitors Solar panels, and controls the dump load that is fed off the battery.

I built the 3 different I/O boards but had them all connected to one arduino board while testing, trouble shooting and code modifications.  It was easier and quicker to upload code changes to one board than trying to manage 3 boards while trouble shooting and making improvements.

  Once the system was stable I had intended to install the other two arduino boards and use I2C for communication between the systems but after 9 months of the system being in full operation there is still only one arduino board controlling and monitoring everything.  The other two boards are still in their boxes.  With one micro controller doing everything there is a much higher potential for failure but there is a back-up system installed so I haven't been to worried.

  I have two home made Ghurd type dump load controllers installed as back up.  One controller (modified) is for the voltage clamp backup on the input to the buck converter and activates at 75VDC.  If for what ever reason the buck converter or arduino fail then the voltage clamp at 75VDC should keep the turbine loaded in high winds.  I've tested it twice (summer wind rarely gets high enough) and so far its worked well.  The other Ghurd type controller (unmodified) is for the battery dump load and gets activated when battery voltage gets to 15 VDC if the arduino 3 stage charge controller fails to do its job.

[WindriderNM]

It should be fairly easy to do this using a few op-amps, a 555 timer and a FET driver ic.