Load modeling plays a vital role in the operation and control of power system. It is an important and practical significance to find out an effective and accurate strategy for identifying load modeling parameters. In this paper

the support vector machine (SVM) method is used to model the load of power system. The Gaussian radial basis function is used as the kernel function and the basic structure of the support vector machine model is selected. During the model identification period

the actually measured current

voltage and phase angle data of the PMU are used as input variable and the active and reactive power as output variables. The measurement-based load modeling method uses a leave-one-out cross-validation (LOOCV) optimized load model parameter identification strategy. In the model the SVM is used to set up the load model

which avoids the local minimum problem that plagues the neural network. The relatively considerable model fitting degree is obtained by using the load model parameters optimized by the leave-one-out cross validation method. The optimal comprehensive load model parameters are obtained through LOOCV optimization method so that the model can better fit the power curve. It is shown through the analysis of the measured modeling parameter identification of the PMU that the SVM load model based on the leave-one-out cross-validation method (LOOCV_SVM) to optimize the parameters proposed in this paper has better performance than the SVM load model based on the particle swarm optimization algorithm (PSO_SVM) to optimize the parameters and the SVM load model based on the bootstrap method (B_SVM) to optimize the parameters.