The wire size and the number of turns per coil are linked together. This means that a certain space is available for a coil and this gives a certain cross sectional copper area of a coil. So the more turns per per coil, the smaller the wire diameter must be. The optimum number of turns per coil depends on the wanted open voltage for a certain rotational speed. The wanted open voltage depends on the matching in between rotor and generator. Matching is explained in chapter 8 of my free public report KD 35. To optimize the matching, you need the optimum cubic line of the rotor and the measured Pmech-n curves of the generator for a range of different voltages for a certain winding. The formula for the optimum cubic line is given as formula 8.1 in KD 35. To use this formula, you need the maximum Cp of the rotor at the design tip speed ratio lambda.
Assume the generator will be used for 24 V battery charging. The average charging voltage for a 24 V battery is about 26 V. Assume that the PM-generator has a 3-phase winding which is rectified in star. The product of U * I gives the supplied electrical power Pel. The product of the torque Q times the angular velocity gives the supplied mechanical power Pmech. So you have to measure the torque Q which requires a rather sophisticated test rig.
For the generator you make a choice for the wire thickness and you make the coils such that you have the maximum possible wire thickness. Assume that you have a test winding with 200 turns per coil with 1 mm wire. Next you measure the generator on a test rig for a range of constant DC voltages of for instance 14, 16, 18, 20, 22, 24, 26, 28, 30, 32, 34, 36 and 38 V. For every voltage you get a certain Pmech-n curve. Only one curve will give the optimal matching with the cubic line if the rotor. Optimal matching means that you have two points of intersection which are not lying far apart. Assume that optimum matching is found for 34 V. This is a factor 34/26 = 1.308 higher than 26 V.
So for the final winding, the number of turns per coil has to be reduced by a factor 1,308 and becomes 200 / 1.308 = 153. The cross sectional area of a wire is proportional with the square of the diameter so it has to be increased by the root of 1.308 and becomes 1.144 * 1 = 1.144 mm. If you make the final winding this way it will have the same Pmech-n curve for 26 V as the test winding for 34 V and so the matching for the final winding will be optimal too. There is only a little chance that you have made that choice for the number of turns per coil of the test winding that the test winding can be used as final winding.
So you see that finding of the optimal winding for a certain PM-generator and a certain windmill rotor is not an easy job an that you need a sophisticated test rig. Measuring of only the electrical power isn't enough as matching is done by comparing the mechanical power of the rotor and the mechanical power of the generator.
If the winding is a given fact, like if you use the housing of an asynchronous motor, you can measure the Pmech-n curve of the generator for this winding for 26 V star and design a rotor such that the optimum cubic line of this rotor matches with the measured generator curve. Measured generator curves for different loads of a radial flux generator made from an asynchronous motor frame size 90 are given in my free public report KD 78.