A new algorithm for synchrophasor estimation is proposed to improve estimation accuracy under dynamic conditions. The real component of a constantly rotating vector and a low-frequency band-limited vector, which is linearized by Taylor series, is employed to model supplied signals in power systems. And dynamic characteristics of power systems are described in form of Taylor series' derivatives, which are calculated through consecutive phasors. The accuracy of phasor estimation was dramatically improved by the reassignment of estimation of short-time Fourier transform (STFT) with knowledge of these derivatives. Simulation results allow us to make the conclusion that the proposed algorithm improves the accuracy of phasor estimations under dynamic conditions dramatically, such as frequency ramping, low frequency oscillation, etc., at the cost of the calculation burden is a minimal increase whilst still meeting the requirements of the application for power systems.