Making holes in the centre of potted coils seems to be proposed regularly but I am not aware of any test confirming the benefit or otherwise, I seriously doubt that it will make any difference without clever fans and ducting.
Like by putting holes inside the magnet ring in ONE side of the rotor and leaving unpotted space between the magnets. As the air is pumped outward between the magnets, some of it will be pulled through the holes to the other side.
Another might be to have one rotor larger than the other, and extend the magnets-as-pump-vanes with some non-magnetic material. The side with the larger diameter "pump" will have a lower pressure, sucking air IN the other side, through the stator, and out the wide side.
ULR suggested not potting the ends of the coils at the outside where they are most likely to see some air flow and if you don't need the potting for weather protection this is likely to be the most effective.
Making holes in the magnet rotors in the correct place and not potting the magnets so that they act as fans seems very sensible and you will induce some radial air flow over the stator. If you don't need weather protection you could combine this with leaving the outer end loops unpotted and you could even spread the end turns a bit.
If you DO spread the turns to improve airflow, or even don't bind them together, you should pot or bind a short section in the middle of 'em. They receive a varying "pinch" when the current flows. If they're not somehow supported against it this will make them vibrate, flex, and fatigue at the point where they emerge from the potting.
By messing about with extra cooling you may be able to maintain power without furling to a slightly higher wind speed but as the power in the wind increases as the cube law you will only hold it off for a few mph. It may make the furling less critical with a bit more in hand for the winding temperature.
The AVAILABLE power in the wind goes up with the cube of wind speed. But absent load effects the RPM and EMF (generated voltage before voltage drop from coil resistance) goes up with the wind speed. If you were driving a resistive load the current and load torque would go up with the voltage and both power delivered and the resistive heating go up with roughly the square of the wind (a little faster because the load torque would be too low to keep the TSR down with increased wind). If you had a max power point controller making the current and load go up to track the wind, the wind speed, RPM, and voltage would go up together, the current with the square, and the heating with the FOURTH POWER of the wind speed. (This is why you MUST get your furling right or have a "max current" override in the controller if you do MPPT.)
For the usual rectifier/battery combo the current doesn't start until cutin RPM, then rises rapidly and linearly with RPM and voltage, putting a load on the blades that, if it weren't for wiring resistance, would stop the RPM from rising at all. Since there is resistance, the current rises with RPM-above-cutin and the heating with the square of that, while the increasing current load causes the RPM to rise more slowly than the wind, and the blades to eventually stall out when the discrepancy becomes so great that the airflow detaches from the back side of the blade. Obnoxious to try to compute this.
If you find a way to match the load then this issue mostly disappears and you can get 3 times more power out of the same stator without dropping below 60% efficiency and the heating problem is very much less.
In particular:
- If your MPPT controller has a max-current-limit function, when the wind is high enough it can let the RPM and voltage climb, tracking "max power I can get without melting the coils" rather than "max power the blades can grab from the wind". (This actually ignores a sliver more you could have grabbed, due to extra cooling from faster wind. But let's not work too close to meltdown.)
- If you use delta or IRP/Jerry-rig, you can get almost twice as much power out as Y for a given amount of stator heating.