Don't worry about the battery side I have that more or less covered. I just need to find a way to adjust the voltage in that 15v window @ 40amps.
How do you have the battery side more or less covered when controlling the voltage in that window is half the problem?That comment aside...
It's conceivable that the autotransformer under current sensing or current limiting control will be able to keep the current within acceptable charging limits.
As for the diode shunting. THat's just basically a "digital" linear regulator with discrete voltage output steps. It involves a crapload of switches (and multi-level totem-pole stacked high-side switches at that which involve a crapload of extra drive circuitry of you are using transistors) and has only discrete voltage steps. High-side switches won't matter if you use relays but man, that's a lot of sparking and switching and wearing out. You might as well just make a giant linear regulator that won't have those relays or high side
If that's the case you might as well go with an actual linear regulator. Albight, it would be very massive and produce a lot of heat but no more heat than your initial diode-shunt concept would (Worst case is 15V@40A = 600W, but charge current should be reduced as it charges via the autotransformer which should help things a lot). I'd try an op-amp linear regulator.
Do you know enough to understand how the regulator works? Enough to know what to modify so that the pass transistor's power comes from the generator but the comparator power supply comes from your low voltage DC source? It's simple as cutting the wire between Q1 and RZ and connecting them to their respective power sources.
**broken link removed**
It's also possible to make the reference adjustable. This involves replacing the zener diode with a resistisve divide, however then you need a stable regulated low voltage DC supply (like from a regulated wall adapter as opposed to an unregulated one).
A DC supply is going to be needed for the op-amp and voltage reference to indicate the "fully charged voltage of the cells". This voltage need neither be very large or high current. THe op-amp does have to be the kind with a high enough supply voltage to properly control the pass transistors but not so high that it can damage their base/gate terminals. This means the op-amp should be around 15-20V supply probably for most large transistors. I think a 24VDC wall adapter should do the trick.
With so much heat you'll need both heatsinking of the transistors as well as parallel transistors. I'd consider going with multiple parallel transistors in SOT-227 packages. That way you can use screw terminals to pass the massive amounts of current easily between batteries and generator using wires, whereas transistors with pads or pins need a printed circuit board which is both expensive (very expensive for high currents) and takes time to design. And with SOT-227 and similar packages, you can bolt them to giant blocks of metal for heatsinking which is much easier than dealing with heatsinking on a circuit board. They cost quite a bit, but sounds like it is on the cost scale of your project. 95% of the cost is going to be in these transistors.
**broken link removed**
**broken link removed**
**broken link removed**
There are things to be aware of and understand...first is completely ignore the current rating given. It's acquired under ideal non-real world conditions. aAlculate it for yourself from the heat dissipated and thermal resistance with heatsink. It will be much much lower. Second is P-channel types are less efficient, more expensive, and hard to find but are very straightforward to use (this has to do with how the transistor is on the high-side, not the low-side). N-channel is more efficient, common, and cheaper but not straightforward to use (they are most straightforward to use if they are on the low-side but they are not for this application). For you, it might be worth the effort of just going with the more expensive P-channel and getting more of them to deal with the heat since it greatly simplifies things and the circuit can pretty much be built as in the schematic. Things get...complicated...if you try to use N-channel and that would require more research and learning on your part to figure out what it is you have to do to get it to work. TO learn more read this:
**broken link removed**
http://www.microchip.com/stellent/groups/designcenter_sg/documents/market_communication/en028089.pdf
Note that there are two types of transistors and two types of each. MOSFETs have N-channel and P-channel. BJTs have NPN and PNP. As far as the high-side/low-side issue goes P-channel and PNP are in the same boat. N-channel and NPN are in the same boat. This will help you understand the article better as you translate between the MOSFET and BJT version of the circuit since they talk about BJTs most of the time but you will really only find MOSFETs in the the large, screw terminal, bolt-onto-heatsink packages.