Firstly it is crazy to try and start on load so the first step is to let the blades run up to cut in speed with no load and then connect load ( I see no point in connecting below 6 mph and for heat you may wish to raise that to 8mph).

Your method suggests that you choose the load resistor to match the machine full load rating at near optimum tsr. You will stall at all wind speeds much below full load and will never get started without a converter. The converter will reduce pulse width at all wind speeds below full load, with the pulse width increasing continuously from cut in up to full load.

The alternative and probably simpler method is to switch in a resistor at cut in that matches the prop load at that point. The volts should track wind speed, so the power into that resistor will rise as wind speed squared. Your prop follows wind speed cubed so the load will become inadequate to hold the speed down to constant tsr.  As the wind speed rises you will need to add parallel resistors in stages. The more stages you choose the better the match to the cube law curve, but one or 2 steps should be near enough in real life, the prop will stand a bit of mismatch above or below ideal tsr.

If you want to use the pwm route then one resistor chosen for full load is what you need. If you go the step load change route then one resistor chosen at cut in and perhaps one or two steps of added load should do it. If you attempt to start directly into the first step resistor it will have to be far too high , to let you start, and you will need more steps.

The step load approach can be triggered from prop speed and as Joestue said you will need comparators with hysteresis but again it's not critical, the prop is very tolerant.

For the true pwm method you need to measure power, make a change, see whether it increases or decreases and make a new change and so on. If you can do this faster than the wind changes you might get it to work.

For a given machine on a given load you can approximate the curve and once you have it right you can again reference the pwm angle to prop speed by a look up table or far simpler ( at least to me) by analogue means.

What I do on the boost converters will probably work near enough. Set your pwm to match at cut in, use a current signal to phase the pwm forwards with increasing load current and just set the current amplifier gain to phase fully forward at full load. You will almost certainly be close enough without logarithmic amplifiers.

I think my approach to a heating controller would be to use a 3 phase mixed ( half ) bridge with one heater on the dc side and control the rectifier firing angle. Standard fixed frequency firing circuits won't work but I think that problem can be got round.

I don't think you need anything complicated but it will be a big mistake to attempt to start on load, that will throw you way off compared with switching in at about 7mph.