In the context of safe control of quadrotors, wind velocity estimation and compensation have a key-role. For this reason, assuming the lack of airspeed sensors and considering sensors noise, in this paper three time-varying parameter estimation algorithms are introduced, studied and merged to estimate the varying wind velocity, using only on-board quadrotor sensors and an inertial tracking position system (e.g. Optitrack camera, GPS). To this end, a detailed quadrotor flight dynamics model is presented using identified aerodynamic coefficients and wind velocity components along the three axes. Then, a decomposition of dynamical equations is performed in known and unknown terms to be estimated. Thanks to this decomposition, the estimation algorithms are built and finally tested and validated in numerical experiments, against the introduced sensors' noise.