The identification of characteristic parameters of voltage sag is a prerequisite for evaluating the severity of voltage sag and analyzing its impact on equipment. In this paper, a complex voltage sags parameter identification method is proposed, which combines phase space reconstruction and atomic decomposition. Firstly, it reconstructs the phase space of the voltage sag disturbance sequence, divides the disturbance waveform into segments according to the phase trajectory characteristics of the sag signal, and determines the start and end time of the disturbance section. Then the method builds a sineweight atom library, makes full use of the waveform segmentation results, realizes the orderly matching of atomic parameters in a specific way, and obtains the characteristic parameters of each segment of the voltage sag waveform. Thereby the number of atomic matching parameters and the search range is reduced as well as the computational load of the atomic decomposition algorithm to ensure the accuracy of the algorithm. The results of the calculation example show that the method can accurately extract the characteristic parameters of the voltage sag signal.