Voltage sag is one of the most perplexing power quality problems in power system. Accurate measurement of its characteristic parameters is the primary premise of voltage sag evaluation and management. However, the current voltage sag detection algorithm has some shortcomings, such as low accuracy of parameters detection, poor real-time performance and lack of phase-angle jumps. Therefore, a new method of voltage sag measurement based on RMS sliding window difference operator and sampling sequence reconstruction is proposed in this paper to achieve fast and accurate detection of characteristic parameters. Firstly, the RMS sliding window difference operator is used to obtain the initial and end time of voltage sag. Next, the original signal is partitioned according to the initial and end time of the voltage sag, and the window interpolation FFT algorithm is performed on the signals in different sections. Then, the sampling sequence is reconstructed and the phase is corrected. Finally, characteristic parameters are calculated, such as the duration of voltage sag, depth of voltage sag and voltage phase-angle jumps. The simulation results under the conditions of simultaneous change of value and phase, different harmonic components and different SNR noise show that the proposed method has the advantages of high accuracy, excellent real-time performance and strong robustness. The test results of the actual hardware test platform meet the requirements of GB/T 30137-2013, such as relative error of voltage amplitude and absolute error of phase-angle jumps are less than 0.2% and 1°, respectively, and absolute error of duration of voltage sag is less than one cycle, which further verifies the effectiveness and feasibility of the proposed algorithm.