In order to solve the problems of frequent voltage sag faults in modern power grid construction as well as insufficient accuracy and speed of traditional voltage sag detection algorithms, a new voltage sag detection method is proposed. First, the single-phase voltage is filtered by an adaptive composite morphological filter at optimal scale. Then, the single-phase voltage is delayed by a small angle to construct mutually perpendicular and orthogonal αβ two-phase voltages. After αβ-dq transformation, characteristic quantities such as voltage transient amplitude, phase jump and duration are obtained. The new voltage sag detection algorithm simulation model is built in Matlab/Simulink. It is compared with traditional filtering algorithms and voltage sag detection algorithms. Based on a TMS320F28069DSP processor, a high-speed data acquisition and processing circuit is designed, and a voltage sag detection platform is built, The proposed method has passed engineering verification. The simulation results show that the proposed method has high accuracy and sensitivity under harmonic noise interference, voltage sag, phase jump and so on, and the overall response time of the measured system is less than 5 ms. The proposed method is feasible and effective. This work is supported by Industrial Technology Research and Development Project of Jilin Province Development Reform Commission (No. 2019C057-3).