Support vector machine based dynamic load model using synchrophasor data

Abstract

Load modeling remains a challenging task in planning, operation and control of power grids. In this paper, a support vector machine (SVM) based machine learning method is proposed for dynamic load modeling of large scale power systems using synchrophasor data recorded by Phasor Measurement Units (PMUs). The difference equation based dynamic load model structure is recommended, however, if a traditional transfer function based model format is preferred, it can be directly obtained from difference equation based model. Case studies are conducted using PMU data recorded in a large power grid in North America. The accuracy of the developed load models is verified by comparing the simulated load model dynamic response with real PMU data. The proposed method not only provides an accurate dynamic load model, parameters of the load model can also be easily updated using new synchrophasor data for either on-line or off-line applications.