PWM dump controller for grid tied inverter

[stephendv1]

Hi all,

I'm at the planning stages of a wind turbine install for an off-grid system and after considering the various options, I've settled on the idea of using a GTI connected to the turbine to backfeed the SMA Sunny Island inverter through AC coupling.  Main reason being that using this system I can locate the turbine in a very good spot which turns out to be about 200m from the batteries, but I have an existing AC cable about 100m away.

To complicate things, it's a low wind area and I'm on a budget, so I'd like to use a 500W Windmaster programmable GTI with a turbine capable of 1500W max output (A Huaya "1kW" turbine).  My plan is to buy/build a PWM controller that can apply a variable additional load to keep the turbine at the right RPM if it's producing more than the 500W the GTI can pull.  Also, the Sunny Island may kick the GTI offline every now and then if the batteries are already full, in which case the PWM dump must be able to provide all of the load.

Since I'm a software not electronics guy I tend to have a naive view of electronic components as discrete boxes that take input, do magic, then provide output.  Below is an idea of using mostly off-the-shelf components to build a PWM dump controller, would appreciate your opinion and whether there are existing controllers that do this on the market (at a reasonable price!)  Most of the commercial GTI "controllers" I have only provide a mechanical relay as a failsafe in the case of overvoltage.  In my case since the 500W will be over volted quite often, I'd prefer a variable PWM load rather than suddenly slamming in a huge resistor.

The components of the controller will be:
- 3 phase bridge rectifier: http://uk.farnell.com/ixys-semiconductor/vuo62-12no7/bridge-rectifier-63a-1200v-3ph/dp/9359443
- Arduino (which has a 0-5V analog in pin, and a PWM output facility
- Mosfet control board: http://shop.ciseco.co.uk/arduino-high-power-mosfet-shield-motor-lighting-pwm-dc-control/
and replace the mosfet with a 100A, 100V model: http://uk.farnell.com/nxp/psmn8r5-100ps/mosfet-n-ch-100v-100a-to-220ab/dp/2254226
- 7 x 10.8ohm, 300W dump load resistors.

The turbine is the 1kW 48V Huaya model, I don't have the power curve, but I'd expect it'll be between 30V and 90V.
The windmaster GTI has an input voltage range from 30V-150V.

The plan would be to program the GTI to suck all 500W up to about 50ish volts, then start applying the PWM dump gradually as voltage continues to increase.  And could limit voltage to about 60V since the power above this will just be lost anyway. 
Does this sound reasonable?  Anything I've overlooked, or any components missing from my nasty sketch?  Will the simple voltage divider work or do I need a capacitor somewhere there?

[DamonHD]

I'm not understanding how your GTI is connected to the DC side of the rectifier and the (presumably DC) MOSFET dumpload.

Rgds

Damon

[stephendv1]

The GTI and mosfet are connected in parallel.

[stephendv1]

Just had another thought, what if instead of the arduino+mosfet solution, I connected a Tristar in diversion mode and instead of a battery used some capacitors, then configure it to absorb and float at 60V ?

[DamonHD]

I may be completely misunderstanding, but maybe you are also!

A GTI is a fairly sophisticated box of tricks looking to interact with an AC grid, and checking parameters such as voltage, frequency and so on.

You can't just connect it, unless again I'm misreading your diagram, to a DC-based system including a rectifier and expect anything other than smoke and possible injury to happen.

At the very least you should be crystal-clear what bits of the GTI connect where (battery, AC, etc) and exactly what facilities you expect it to provide since there is a wide range of available behaviours under that set of initials.

And capacitors: do you have any idea what cost and size of caps would likely be needed to be relevant?  I wish it were otherwise otherwise my systems would be based on them!

Rgds

Damon

PS. I promise that I'm not mocking you, but the devil is in the detail here.

[stephendv1]

Ok, the battery based off-grid component of this is not necessary to understand what I'm looking for. 
Imagine a turbine connected to a normal wind GTI connected to the grid. 
Except in my case the GTI is only capable of 500W and the turbine is capable of 1500W.  So I need some kind of PWM dumping mechanism that will kick in above the 500W and start dumping gradually until all 1500W are absorbed. 

[mab]

well, provided that the GTI will work with a wind turbine (most are configured to find the max power point of a PV array) then I don't see why that wouldn't work (not sure I understand your problem with the GTI DamonHD - A GTI is designed to connect to a DC power input).

As for the connection to the Arduino - I would include a zenner diode to limit the voltage on the input pin of the ucontroller - just in case of something going wrong, and I probably would add a small cap just to smooth the ripple from the DC input and to absorb any spikes/ interference that may find their way there.

otherwise I think it might just work.

don't forget a big heatsink for the rectifier and probably a bigger one than supplied for the FET(s).

I assume you are supplying the arduino from an isolating ac line voltage power supply.

[mab]

as for the Tristar version - I don't know much about them - but you would need to know what frequency it would switch at so you could size your cap bank to limit the ripple voltage accordingly - if it operates a Khz it should be OK - if it operates at Hz then you would need a very big cap bank. And bear in mind the caps may have a limited ripple current and stay within that.

Actually, thinking about that, the Arduino version should probably have a cap on the DC bus too (although there may be one in the GTI) , though again it's size would depend on your PWM frequency and it needs to be able to handle the ripple current.

[joestue]

don't use pwm, use your a/d converter to output 6, 7 or 8 I/O pins to turn on or off a bunch of resistors.
size the resistors such that they pull 16, 32, 64, 128, 256, 512 watts or some multiple of this, in this way you can for example dump anything from 16 to 1008 watts in 16 watt steps, with just 6 resistors and 6 mosfets from 6 io pins and almost no code at all.

in this manner, you won't have any current ripple to deal with and there won't be a "pwm frequency" because the transitions from one power level to another are smooth and simultaneous.

you can use pwm the problem is the current ripple will be pulled out of your GTI's input capacitors, and if they are already marginal, then that extra heat will shorten their life.

[domi]

Heat water instead of dumping such a lot of energy.

[boB]

How much more $$ is a larger GTI than the 500 ?   Something like a 50 Hz  Aurora or larger Mastervolt ?

Is the 500 isolated ?

[stephendv1]

As for the connection to the Arduino - I would include a zenner diode to limit the voltage on the input pin of the ucontroller - just in case of something going wrong, and I probably would add a small cap just to smooth the ripple from the DC input and to absorb any spikes/ interference that may find their way there.

Ok, thanks.  I'll need to read up on caps to understand how to find the right one.

I assume you are supplying the arduino from an isolating ac line voltage power supply.

Yeah, maybe add a small battery backup for the arduino just in case. 

[stephendv1]

don't use pwm, use your a/d converter to output 6, 7 or 8 I/O pins to turn on or off a bunch of resistors.
size the resistors such that they pull 16, 32, 64, 128, 256, 512 watts or some multiple of this, in this way you can for example dump anything from 16 to 1008 watts in 16 watt steps, with just 6 resistors and 6 mosfets from 6 io pins and almost no code at all.

Interesting idea.  Since I don't care about the power above 500W, it could be dumped any which way- the PWM idea is basically to protect the turbine so that it doesn't stall suddenly when a huge additional load is suddenly connected. 
So using steps of loads will also work, but with my limited electronic knowledge I'd need a prebuilt circuit that can control the mosfets, like this: http://arduino-info.wikispaces.com/Brick-4ChannelPowerFetSwitch
(will have to replace the fets with higher voltage versions and mount on heatsink)

I don't have a power/V/RPM curve from the turbine manufacturer yet, but using a different 1kW 48V turbine's data.  It looks like I'll have to program the windmaster inverter to draw 500W at 350RPM/36V.  Since everything above that is lost, the dump loads should be programmed to just maintain that speed.  Using 4 x 300W loads they should come on at:
1. 420RPM/46V
2. 480RPM/48V
3. 490RPM/49V
4. Just above that as this resistor is only used when the GTI is offline.

All values to be adjusted in practice since the charts are based on a 2.5ohm constant load, will be quite different with a variable load applied.

In the case of the GTI being offline, it would mean that the turbine would be free spinning till it hits 420RPM, when the first 300W load comes online.  Will this be dangerous or detrimental to the turbine health?
I noticed that the GTI has a 5 second wait period between detecting adequate voltage and actually starting to inverter.

Bringing the Sunny Island into the equation, it's designed to work with SMA GTIs which can vary their output based on the grid frequency.  So as the batteries require less power, the Sunny Island increases grid frequency from 50Hz gradually up to 52Hz and it expects the connected GTI to reduce its power proportionally.  Since the windmaster 500 inverter doesn't support this, it will simply disconnect entirely when the "grid" frequency goes out of range... which I think is at 51Hz. 

I'll be controlling loads on the AC side of things so as not to waste this energy, but if that fails then it's entirely possible that the dump load and GTI will be in an on-off switching loop, where the GTI goes offline for a bit while grid frequency is too high, then the turbine spins to 420RPM until dump loads kick in, then GTI comes online, turbine slows, dumps go off and the process repeats.

That's where I think it would be neater to have a PWM system so that everything can be controlled gradually to reduce wear on the turbine.

To avoid this ripple current on the GTI's input, what if I used 2 different 3 phase bridge rectifiers, 1 for the mosfets and dumps and the other for the GTI?



     

[stephendv1]

How much more $$ is a larger GTI than the 500 ?   Something like a 50 Hz  Aurora or larger Mastervolt ?

Is the 500 isolated ?

The larger inverters are about 4 x the price =~ 1000 euros.  Cheapest option would be a 1.5kW Ginlong which supports the low voltages needed for a 48v turbine.  But even if I had a bigger inverter there's a good chance my mini "grid" will be kicking it offline periodically.  The windmasters can also be paralleled, so if I realise that I'm losing a lot of valuable energy I can just get another.

The 500 is a "galvanically isolated Class II HF transformer".

[stephendv1]

BTW, if I removed the GTI from this circuit it would effectively be an MPPT turbine heating controller.  Surely these already exist?

[joestue]

with regard to the pwm current ripple, adding different diode blocks won't change a thing.

regarding my proposed idea of r-2r resistors, those individual resistors are all in parallel so with just 6 io pins, you can get for example, 0 to 1008 watts in 16 watt steps.
this is equivalent to a 6 bits of pwm because you turn them on the same way you express 0 to 128 in binary and yeah, if you want to you can even pwm the smallest 16 watt resistor to get the equivalent of 8-10 bits of power draw.
only problem being that the resistors won't be matched perfectly so you will need to program in some hysteresis.

you could even use just 4 resistors with that fet board you linked to, using pulse width modulation on the smallest and still get nearly infinite power control without significant current ripple.
yeah of course you can pwm one large resistor but the problem is regardless what the pwm percentage is you are  pulling the current out of the wind turbine in 20 amp pulses (1000 watts at 50 volts) and this current ripple is what causes everyone problems, from overheating the capacitors in the inverters which are near by to blowing up the mosfet with avalanch failures at turn off, etc, etc. and if there is no capacitor, then you essentially have made a mosfet smoker.

now regarding the control loop issues you are going to have a number of them.

In the case of the GTI being offline, it would mean that the turbine would be free spinning till it hits 420RPM, when the first 300W load comes online.  Will this be dangerous or detrimental to the turbine health?

no, the problem will be insufficient hysteresis in the load control logic, and the effect that has on the loads.

[Mary B]

Those Chinese made non name brand GTI's are a fire waiting to happen. No way would I put one in my home

[boB]

Do you have a power curve for your turbine to put into the 500W inverter for tracking ?

Power vs. turbine voltage ?

That will most likely matter for your resistor values.  I have always thought a binary weighted load could
work well.  It can also be PWM'd to make it even better if ever necessary.

[boB]

When the grid goes away for any reason, you might want to think about making your controller semi-slowly
drag the turbine down to the fullest load so that it basically stops running when there is no place for
the power to go.  This would be in addition to just limiting the unloading of the turbine and/or
over-voltaging of the inverter.

What's the max input V of that inverter ? 600 ?

[stephendv1]

Joestue,  thanks for the ripple clarification, so if the static resistors solve the ripple problem and it's not too detrimental to the turbine if loads come on in 300W increments then that seems a good route to go.  Less electronics to worry about    Regarding hysteresis on the loads, I can control that in the arduino code and since we're in a low wind area hopefully I'll have a window where I can test it all in relatively safe winds.

MaryAlana, it's not a chinese inverter it's a Dutch made Mastervolt Windsaver 500W, http://www.mastervoltsolar.com/solar/products/windmaster/windmaster-500/

boB, Huaya the manufacturer have sent me a "power curve" unfortunately it looks like it's the curve of a battery connected system as the voltage steadily increases and then stays rock steady at 56V.  "Binary weighted load" ?  would it in effect be a stepped load in finite increments similar to what Joestue suggests?
The factory programmed curve for the inverter is attached, although there is software available to change this to an arbitrary curve.  Depending on how well my load management on the AC side of our mini-grid works, it might be that the sunny island is continually knocking the wind turbine offline every few minutes.  So wouldn't it be better to keep the turbine spinning on load at a low speed, rather than letting it spin up and then force it to a stop again?  Also, I don't know how I would get the arduino to sense when the GTI disconnects because of grid over-frequency, if I measure AC voltage at the grid connection everything will be fine.  Perhaps an LDR strapped to the error LED of the inverter!    Would be simpler if the dump loads could work independently of the GTI being connected or not.

The 500 inverter has an operating voltage range of 35-150 V DC and MPPT range of  65-125 V DC


 

[stephendv1]

Huaya has responded with some more RPM/V/power data, not sure what the load was when getting this data.

[stephendv1]

Given that curve above and the fact that the GTI is 500W, could I get away with just using 2 mosfets and 2 resistors a 500W and a 1000W, and only switching on or off, no PWM?
E.g.:
-  At 124V turn on 500W resistor
-  At 130V turn off 500W and turn on 1000W
-  At 140V turn on both.
I expect those voltage will be lower in practice as they're free spinning no load values.  The GTI will overvolt at 150V.


Going back to the PWM idea and the ripple current problem.  What if I used an extremely slow PWM, in the order of 1Hz?

[OperaHouse]

I like that four channel brick.  That opto design is basically the same as I use my PWM power point water heater and charge controller.   If you have never tried this before, you can develop control loop hunting issues between the sample rate of the A/D and software delays.  And many of those wall wart supplies operate at 50V for isolated operation.

[boB]

Unfortunately, it doesn't look like Huaya can supply a power vs. voltage curve for use with an MPPT controller or inverter like you are using.  So, you will just have to play with the curve in the inverter for a while as you optimize it.
If you have a good amount of wind, it shouldn't matter too much as the inverter won't be able to use much of
that power anyway.  At 50 volts output, if I can believe their curve, the power output will be twice as much as
the inverter can take anyway (> 500W)

The reason that the weighted resistive clipper load is nice is because it doesn't have to pulse on and off fast in order to emulate a certain fixed resistance (as Joestue elluded to.)   One thing to be careful of for certain inverters and controllers when loading the DC input of the devices' is that they may be bi-directional and while the electronics is busy tracking and moving its input voltage around, power may be supplied from the grid (if inverter) or battery (if controller) TO the R load momentarily.

If it needs the load, then that power should be coming from the turbine and shouldn't really be a problem but
just beware of that possibility.  If the inverter has backwards flow protection, it may turn off and on once in
a while with widely varying wind speeds.

This is one reason why I like to do wind turbine clipping on the AC side of the rectifier if possible.

There are other good reasons to clip on the AC side as well and in my experience, is easier.

[stephendv1]

This is one reason why I like to do wind turbine clipping on the AC side of the rectifier if possible.

There are other good reasons to clip on the AC side as well and in my experience, is easier.

Dammit boB, just when I thought I had the design semi-clear you throw this curve ball 
So I'd need a 3 phase SSR capable of being switched by the max 5V, 40mA from the arduino pins and can handle about 30A AC AND can switch at the high frequency of the AC out of the turbine.  The one mentioned on the midnite forums by keyturbogears was this one: http://uk.farnell.com/crydom/d53tp50d/ssr-50a-4-32vdc/dp/1200208  but the spec says max frequency is 65Hz.  Will that still work with the wild AC from the turbine?

[boB]

NO, DC will work OK.  The frequency for DC may also want to be fairly low frequency because the self inductance of wire wound power resistors can be fairly high.  If you are heating water with low L heating elements then it might not be such a problem.  PWMing at high frequencies (hundreds or thousands of Hz) "CAN" cause the DC voltage to actually rise instead of drop in some cases.

AC side clipping is done with at least 3 power resistors and triac based SSRs and always turn OFF at zero current and so there is no inductive voltage spikes.

boB

[joestue]

I hope Stephen you understand that you can't pull energy out of a thing if that energy isn't stored somewhere.

Shorting out the ac (or dc) side with a switch (no resistor at all) is actually a great way to use the transmission line and the internal inductance of the ac generator to form a boost converter.

If you want to set up a pwm dump load using a single resistor, supply at least 100uf per amp of dump load current, and have this capacitance setup correctly.
Here is a photo of a very cheap and simple low inductance topology. http://johansense.com/bulk/50_buck2.jpg. you see that the capacitors are soldered to adjacent sides of the board and the switching fets drop through a hole in the board to reach though the back side.
point to point wiring has two orders of magnitude higher inductance and will cause you problems.