SVM-Based Parameter Identification for Static Load Modeling

Abstract

Load modeling is critical to power system studies. This paper proposes a parameter identification technique for the ZIP load model by leveraging the support vector machine (SVM) approach. The ZIP load model is represented as a linear regression expression. To improve the accuracy of parameter identification, one filter, i.e., Hampel filer, is used to preprocess measurements to reduce noises. The data after noise reduction are used as training data of the regression model, which is handled by the SVM approach. Several case studies show that the SVM with filters can identify the parameters for the static load model with high accuracy.