A complex Sigma point Kalman filtering（CSPKF） algorithm for the improvement of frequency estimation of distorted power system signals in the presence of random noise and harmonic disturbance is presented. The 3-phase voltage signal of a power system is transformed into a complex voltage signal for the processing of the proposed algorithm. CSPKF outperforms the popular extended complex Kalman filter（ECKF） in terms of accuracy and stability. Moreover， it's not necessary to reset the covariance matrix for CSPKF as it is for all Kalman filter algorithms to track sudden parameter changes after initial convergence， thus significantly improving its tracking speed. Numerical simulations show that the proposed CSPKF algorithm possesses an excellent dynamic frequency tracking characteristic while maintaining a considerably low tracking error.