In windcharging applications with a Y connection in an alternator, the Y point typically isn't brought down. (Why bother with a fourth wire when you don't need it?) A delta connection doesn't even have a neutral. In a three-phase alternator -> bridge rectifier -> battery setup the grounded side of the battery provides the ground reference. If the mill is turning fast enough that some diodes are always conducting the effective neutral point will cycle near half the battery voltage, while if it's slow enough that they're not conducting it can wander around but is still between the phase voltages, which are clamped between the voltages of the two sides of the battery. That's more than adequate for static protection of the wire insulation, by an order of magnitude or more.
I'd be inclined to check for phase balance by using phase-to-phase, rather than phase-to-neutral measurements. While they aren't quite as easy to interpret, they don't depend on finding a good, alternator-winding-referenced, neutral. The voltage of an under-loaded phase, or one that's disconnected downstream of where you measure, will be high, while an over-loaded one or one that is disconnected upstream of your measurement point will be low. All you're interested in here is current balance, so if the currents aren't balanced the three phase-to-phase voltages won't be approximately equal.
While an unbalanced load will produce a lot of vibration, at twice the generated frequency, due to cyclic variations in the load's drag on the shaft, there are a couple other noise sources to consider.
If the frequency of the sound doesn't track the rotation rate of the blades, it's not an imbalance among the three phases. Whistling notes sound more like vortex shedding or other air-blade interactions. (You might be able to do something about that by making adjustments to the shape of the blades, if you want to change them. But that way also leads to potential trouble.)
Now if the fundamental frequency tracks the rotation rate - at six times the alternator's output frequency - but the harmonic content (as speed goes up) is very high at the low speed where it starts, and drops with speed rise, you might be listening to rectification noise. When your voltage gets just up to cut-in, the rectifier only conducts (and thus only loads the shaft) at the voltage peaks, so you get six short, sharp, pulses of drag per cycle. As the mill speeds up the generated voltage rises. So conduction is both progressively greater and occurs for a larger percentage of the cycle. So at first the sound gets louder but the high frequencies fade out, then (as the conduction begins to be more than half-time) the fundamental note at six times the AC also dies down.
(Note that unbalance noise might also include some rectification noise, too.)
And then there's bearing noise and/or some other thing rubbing...
