Your heating element is going to be a variable resistance. Unless you are using that specific for heating air or water, or something else, you might have better luck with a more linear resistance from large load resistors.
PWM is the same animal no matter whether AC, or DC is on the input. When it is used in the way the Ghurd circuit which is similar to most commercial low power voltage controllers; they operate the MOSFeT as the "dump load" that is disipating the load above the set point voltage.
With the basic cautions Flux raises, your plan is very viable. On the scale of a few hundred Watts designing around IGBT's as your switching devices is still simple. If you are heating air, you can take advantage of the characteristics of Bipolar transistors (inherent internal resistance) in parallel on a heat sink which will heat the air quite well. In either situation, double your expected peak current and power loads. And also bear in mind your mill itself is going to be heating up from the load too. Is your 1500 Watt heating element enough of a load? Will the mill be able to protect itself from overspeeding in the event of circuit failure?
That 1500 Watt rating is also measured at the potential of the circuit it was originally designed for. At a lower voltage, it will not be the same. Think of a toaster element. Take an average toaster for 120 volts. On 32 volts, it has a different load on the circuit. You would need to take the heating elements, and divide them into quarters, and then wire those in parallel to achieve the same performance (ie, wattage) of that toaster when it was on 120 volts. The non-linearity of the resistance is in this situation unimportant because you are looking at the total peak load.
It would not hurt to see if your element exhibits a negative resistive response as it heats.
IMHO, if you reworked your physical plan so your heat sink were on the side of the water tank, with air flowing over the semiconductors, and then directed around the water tank or through it, you may have better results.
Alternately, and still an electrical solution:
I don't know if you looked at any of the tech sheets Zetex, National, ON,etc have on the matter of PWM circuits and designs, if you have or have not: look specifically at the Class "D" audio app notes from Zetex and ON and also the automotive switching app notes and you can come up with a fairly simple circuit with the envelope portion of the signal (audio or 60 Hertz)coming from a 555 series IC set up as your 60 hertz source if your miniprocessor is not suitable to generate a signal. You may end up with the best results coming from a bank of switching transistors (MOSFeTs are cheapest by far, and best thermal characteristics)at the potential of the mill output; with the heat sink still attached to the water tank; driving into the primary side of a MOT to get a stepped up potential to supply the heating element.
Inductive spikes on the generator end can be handled with transient supression, and Littlefuse has a number of good tech articles on using MOV's. Normally you can quench the inductive spikes with a diode, but MOV's are better quench devices if RF noise is a concern for you.
HTH
