Under the background of "dual carbon", a large number of distributed generations connected to the power grid cause power quality disturbances, which pose challenges to the stable operation of the power grid. With the integration of nonlinear and high-speed loads, power quality disturbance will exhibit non-stationary characteristics in the novel power system. By analyzing the energy distribution characteristics of different types of power quality disturbance in the fractional domain, an improved de-noising algorithm based on fractional Fourier transform (FRFT) is proposed. In addition, the estimation method of the optimal fractional transform angle is discussed. Based on the fourth-order origin moment of fractional spectrum, the calculation complexity can be effectively reduced by using one-dimensional peak search to determine the best transform angle. The experimental results show that the improved algorithm can effectively realize the de-noising of both stationary and non-stationary transient disturbances and retain the location information of transient disturbance.