In order to improve the accuracy of classification problem in intrusion detection for state grid, a novel intrusion detection model, which is based on the hybrid classifier, is proposed in this paper. The hybrid classifier is composed by kernel principal component analysis (KPCA), back propagation neural network (BPNN) and quantum genetic algorithm (QGA). In the hybrid classifier, KPCA is used to reduce dimensions of data. The classification model is trained by BPNN, of which the parameters are optimized by QGA. Based on the classification model, the data samples are classified by accurate intrusion detection. Compared with the traditional methods, the intrusion detection model based on hybrid classifier has better performance in reducing the calculation errors.