A PLS-SVM-based method of general single-phase harmonic load identification

Aiming at the poor precision and universal of current load modeling method problems, combining PLS and SVM, this paper proposes a method based on the voltage and current data for harmonic condition to build the universal single-phase load model. The interaction between input variables is reduced by space mapping, and the orthogonal feature projection is used to solve the co linear effect. In the feature vector selection, emphasizing the input to output explanation prediction removes the useless noise back and reduces the number of model variables. The method improves the accuracy of load model parameters identification and provides a theoretical method for accurate harmonic analysis of power system. Finally, the simulation examples demonstrate the effectiveness of the method.