At present, voltage sag has become one of the most prominent power quality problems.In order to effectively analyze the impact of lightning stroke on power grid sag, the situation of voltage sag caused by lightning stroke is analyzed in detail.Four types of temporary relief, including lightning induced temporary relief, are identified accurately, which provides an important basis for the rational division of temporary relief liability.Firstly, the difference between the effective waveform of temporary sag caused by lightning stroke fault and the common short circuit fault is analyzed.The characteristics of RMS waveforms of four kinds of sag types, short circuit fault, lightning stroke, transformer switching and induction motor starting, are summarized.Five characteristic indices of sag voltage are introduced and the characteristic matrix of sag type identification is established.Then four types of sags are identified by using decision tree support vector machine (PDT-SVM) classifier based on particle swarm optimization.The training and testing data of the classifier come from the measured sag voltage data of the power grid, which is closely in line with the engineering practice.Finally, the validity and accuracy of the algorithm are verified by the analysis results of an example.