The location of voltage sag source plays an important role in solving the disputes and identifying responsibility between two power supply companies. In view of the situation that using a single location feature does not have high location accuracy, and the voltage sag data with location information is very limited and difficult to obtain, this paper proposes a new method based on semi-supervised Support Vector Machine (SVM) to locate the voltage sag source based on the thought of intelligent classification. Firstly, the location mechanism and criteria of existing source location methods are analyzed and compared. Then this paper combines multiple location features and uses the idea of two classifications to locate upstream and downstream in the high-dimensional location feature spaces. Finally, in order to make full use of a large number of untagged sag monitoring data, the semi-supervised SVM is used for optimizing the upstream and downstream optimal classification planes, and thus can achieve improved sources location performance under a small amount of tag data. The simulation results show that the method has high location accuracy and can locate kinds of sag sources reliably under a small amount of tag data. This work is supported by National Natural Science Foundation of China (No. 51577086).