The invention discloses an optimization method of a power enterprise information system anomaly detection scheme based on an improved SVM in the technical field of power information systems, and the method comprises the following specific steps: firstly, understanding the principle of an SVM algorithm, then constructing a network performance index model, then carrying out kernel function selectionand parameter optimization, and finally, determining a confidence interval. By improving the defects of a passive detection method, it is proposed that an SVM mechanism is adopted to participate in the performance evaluation process, real-time dynamic adjustment of baseline parameters is achieved, meanwhile, a training residual error method is used for assisting in determining the confidence interval of network performance indexes, and therefore the reliability and accuracy of threshold value calculation are improved.