Charge controller for wind

[rossw]

Something you might want to consider, use a dpdt contactor to disconnect the DC side at the rectifier and short the DC side of the rectifier, it makes for a softer stop for the turbine.

If we're just trying to stop the thing going over-voltage as would (presumably) occur if the MPPT decided to drop load entirely while it re-profiled the load, or the batteries had reached float and simply couldn't take the extra load - why do we have to short the turbine at all?

Isn't a suitably rated 3-phase dump load easy enough to construct, and provides enough load to at least tame the output while whatever is going to happen, happens? A seperate shutdown (short) would be applied for extended periods, or in preparation for a storm, or after an equipment failure etc, sure.

[joestue]

If we're just trying to stop the thing going over-voltage as would (presumably) occur if the MPPT decided to drop load entirely while it re-profiled the load, or the batteries had reached float and simply couldn't take the extra load - why do we have to short the turbine at all?

Its useless to "remap a wind turbine".
Ideally your mppt would tell you how hot the coils* are by filling a few tens of kilobytes of lookup tables with historical data!
*technically, the difference in impedance of the overall system between what it is yesterday and what it is today.
Actually finding the MPPT profile of a turbine for the first time is rather complex because you don't have enough sustained winds to do a clasical purturb and adjust to find the right tsr.
however if you provide the cpu the rpm of the turbine it gets a lot easier, because you can suck power out or let it free spin** for a few *tenths* of a second and measure the change in rpm, which would give you the wind speed.
take the change in rpm over the change in power and use that to determine which direction you need to go.
** it should never let it free spin, certainly not cut the power to zero for any length of time, for the simple reason that due to the high impedance of micro hydro, turbines etc due to low efficiency, the mppt converter should be able to manage a generator with an unloaded voltage well above what its rated to. (This voltage rise follows the inverse of the efficiency of the system.)


back to a more realistic dump load.
8 igbt's, 8 resistors can get you zero to XXXX.Y watts in 256 steps
if you tell the A/d converter to sample the line at 10,000 times a second, this is plenty slow enough for the IGBT's to avoid any significant switching losses but fast enough that you'll never notice its operating.
all you need is an opamp to compare the dc bus with a referance, filter this with a cap and resistor to slow everything down and feed it right into the a/d converter.
no need for crazy loop compensation, provided you don't expect it to absorb a lightning bolt.
realistically you're only looking at a hand full of parts.
if you keep the switches on the ground it makes it that much more simpler.

[cardamon]

boB, or whomever,

So I assume midnite's clipper does diversion as well is clipping  or does the owner supply his own diversion system? 

[boB]

boB, or whomever,

So I assume midnite's clipper does diversion as well is clipping  or does the owner supply his own diversion system?  

PWM diversion is how the clipper works.  This can be on the 3 phase AC turbine side (preferred method) or can be on the DC side of the 3 phase rectifier.
(or however many phases there are)

MidNite supplies the power resistors as part of the Clipper.  The 3 phase AC Clipper uses 3 resistors.  The DC Clipper uses 2 or more resistors in parallel.

(typically this is the resistor complement)   Larger turbines ( 5 to 10 kW) would require more Clippers (and Classics of course)

Clippers are more complicated than one might think although not THAT bad.  Well, sometimes they are.

boB

[zvizdic]

Hi GHurd

Ohm's Law say Amps multiplied by Amps multiplied by Ohm's = Watts.
For this discussion, Watts is heat in the fet, and I like it below 1W in a TO-220 package (physical size of the fet).

What about TO-247 ?

[ghurd]

Hi Darko,

I am not comfortable posting anything about TO-247 packages because I have seen so many variations and have no experience worth mentioning.  Some people like them better than TO-220s, though personally I do not at this time.  It seems to me like plastic would not conduct heat from a metal tab to the heat sink as fast as metal.  I do not know any details.

Similarly, I do not know much about TO-263 packages either.  It takes a lot of calculations to determine what kind of watts/heat they can deal with, and the calculations depend on the surface area of the PCB, the copper close enough to matter, the effective thermal characteristics of the copper in the region, and the thickness of the copper on the PCB.

Let us assume the difference is two rare $3 fets with $2 each heat sinks, compared to six $1 fets with $0.5 each heat sinks.
Often the complex mathematics involved in attempting to save $0.01 is best avoided.
Often the complex mathematics involved in attempting to save $0.01 results in a system that blows up.


RossW and boB are working in a completely different context than I usually work in.
boB's employer  may purchase components 1,000,000 at a time with a greatly reduced cost, with a known housing/PCB to use as a heat sink for the fets, specified amperage limits, semi-known parameters regarding thermal conditions, etc.  That is the reason some companies can sell completed devices for less than the cost of the components.
I do not interpret anything boB or RossW said as being contrary to anything I said, within the context, circumstances and conditions we typically work in.
G-

[joestue]

Some datasheets will tell you something to the effect of "operation in this area limited by package." Typically igbts, since the bond wires explode long before the die does.

The actual silicon inside a TO-247 STW20NK50Z (17 amp, 500v 0.230 ohms) is about 5mm by 3mm. it has two bond wires spaced about 2mm apart in the center. (i blew one up)
I don't remember how big the silicon die is inside an irfz44 but i seem to recall it being a 2mm square. I stopped using those a long time ago.
I'll crack one open tomorrow just to look at it.

I haven't looked into it yet but i don't believe the larger packages are cost effective if you are passively cooling them, reguardless if they are surface mounted to a circuit board.

take a look at these fets: IRFH5250DTR2PBF .99$ if >10 at newark  1.4 milliohms 25v
I don't believe that they have the same chip inside a more user friendly package, unfortunatly.
Its getting to the point where you don't even need heatsinks.
however as the silicon gets better you have to derate them to an extent.
if you compare the 20nk50 fets with the irfp450's you might expect to be able to run the first at much higher current.
but if your design sucked in the first place you're sure to burn the newer fets up because they both have the same avalance rating.

A Clipper will increase the life of the Classic... Or any other wind MPPT controller unless it just has a huge voltage headroom.

Some people can design and build their own clippers just fine...  But some have had trouble. There are some important design and construction issues that have to be considered, such as inductance of the load resistors, mainly.   Some times these problems aren't found out until it's too late.

It would be nice if folks on this forum can eventually come up with a  less expensive and reliable  working option for clippers.  Choosing the proper load resistance and making sure that the load is not going to get too hot and disconnect (like a water heater) is very important.  Resistors with good ratings and good cooling is also very important of course.

Anything that says "increase the life of" scares me.

There's only two things that limit lifetime: electrolytic capacitors and unknown variables.
The rest is predictable.. mtbf stuff.
If you're having issues with a pwm clipper that can't handle inductive resistors then you're doing it wrong.
Even if 10% of the power dumped into the resistor ends up back in the supply 10,000 times a second.. that's still only requires a 3 amp diode for every 30 amp igbt. one whole extra dollar.
Don't forget the 1-2 extra amp of rms current however, so you might need to stick an extra 99 cents of film capacitors on the dc bus.
Some of the diversion loads posted here don't even have caps for the recovery diodes to discharge into, so of course you will run into issues if the diversion load is located 4 feet from the batteries and you used inductive resistors.

[boB]

Anything that says "increase the life of" scares me.

There's only two things that limit lifetime: electrolytic capacitors and unknown variables.
The rest is predictable.. mtbf stuff.

A Clipper will increase the life mainly because it will keep the voltage from exceeding the maximum safe limit
of the controller.  I guess that would go into the unknown variables category.

boB

[zvizdic]

Thank you Glen

I used 6 IRF640N rather slapped together , one burn in high winds .
Locking to make better designed Ghurd super-controller with 2-stage dump and LVD for 10' machine at 48V.
I am thinking,everything to go  through  LVD(4-irfp150n), all the current from   1. sage 2 FET(300W max) 2.stage 4 FET(900W).


Any input or suggestion

[ghurd]

Keep the fets on separate heat sinks, with separate loads.

I do not usually recommend it, but if it is not possible to have separate loads, each resistor should have its own gate resistor.  I use 56 ohm each because I have them.
( http://i701.photobucket.com/albums/ww20/ghurd1/Sketches/ParalelFetsSinglLoad.gif )

2.stage  Consider using 6 fets.  Cheap insurance.

G-

[joestue]

I used 6 IRF640N rather slapped together , one burn in high winds .

how hot did you let them get?

also, i don't see any reason not to stick the fets on the same heatsink
the on resistance has a rather high temp co, so they parallel easily

[zvizdic]

It was a dump switch  only.
I left a multimetar on top of it and block the cooling .