Automatic voltage control （ AVC） system may cause misoperation due to its incorrect and inefficient identification for the remote measurement data of power plants. In order to solve this problem,this paper adopts the support vector machine （ SVM） ,a data mining method with excellent pattern recognition ability,to establish a bad data identification model. In this method,first,SVM nonlinear regression is used to perform a curve fitting for the remote measurement data. Then,the classification network is trained with SVM. Finally,the real-time data are input in the curve fitting network and the classification network to judge whether the data are correct. Simulated results show that the proposed model is effective and accurate.