Power factor correction will reduce copper losses in the alternator, and to some extent iron losses as well.
but at 2Khz, it will be difficult to get power factor correction, because you have such a wide range for the input frequency.
i see no reason to use power factor correction. some improvement might be made with a choke after the three phase diode bridge.
to make the system more complicated, you might go with this topology:
**broken link removed**
note that you can add phases as well with this design.
and the mosfets can be paralleled with IGBTs
for 40KW i would still look at at least 4 phases. unless you want to go with CCM for the inductors.
to operate the inductors in discontinuous mode (high frequency) you're going to need more phases because the current ripple on those three caps (the ones connected in delta) are going to be insane.
furthermore, because the alternator puts out up to 2KHz, those capacitors have to be relatively small, otherwise they are going to present a very non negligible reactive load on the alternator.
so you might have to go with more than 4 phases so those caps can be reduced in size.
this isn't a bad thing, because the more phases, the smaller and cheaper the capacitor just after the 3 phase diode bridge needs to be.
do you need constant power over the full range 120vac to 420vac?
or does the power increase with the voltage?
420vac has to be boosted to above 600 volts to maintain discontinuous current in the boost inductors.
600 volts could be stepped down to 140vdc with a fixed ratio transformer running at a fixed 50% duty cycle. the 600 vdc could be bucked down to 140vdc more efficiently, but then you lose voltage isolation. (do you need it?)
unfortunately, shorting three phases together causes the dc output of the rectifier to have a high frequency ripple with relation to the generator.
with multiple phases this would be reduced but it won't be eliminated completely.
to keep this system as simple as possible just run the alternator into a three phase bridge, an inductor and capacitor sufficient to handle the ripple current of the multiphase buck converter and buck it down to 140 volts dc. i would run it in forced synchronous mode to keep the feedback loop simple.