I did see a similar design once that used four (or two at right angles, i can't recall exactly) piezoelectric sensors that just used an upright rod. From that you can determine both wind strength and direction.
Turbulence around a cylindrical rod may be an issue causing rapid fluctuations in the readings?