The ongoing and unprecedented transformation of power systems leads to a reduction in the number of conventional power plants, which are the classical actuators of the grid. In addition, this development results in a decreasing system inertia, which is expected to yield faster frequency dynamics. Therefore power-electronics-interfaced units have to take over system control tasks and, in particular, frequency control. For this purpose, accurate and fast estimation algorithms for timevarying frequency signals are needed. Motivated by this fact, we propose a time-varying parameter estimator and a tuning criterion, which for sufficiently small initial estimation errors allows to reconstruct the time-varying frequency signal of a symmetric three-phase waveform in finite time. The proposed estimator is derived by using a time-varying version of the super twisting algorithm and its performance is illustrated via numerical examples.