Partial Least-Squares Regression Based Modeling Research as Applied to Harmonic Source Analysis

摘要

Through the identification and filtering of the useful information in multivariable system, and the distillation of comprehensive variables which can explain the system performance most effectively, we can remove the interference of the overlapping and insignificant information, as well overcome the disadvantage of multiple correlation characteristics in establishing of the system model, which is important for the modeling research in power system with high dimensionality. This paper proposes a new modeling research method, called PLSR(Partial Least-squares Regression), which can solve the above problems effectively. The fundamental principle and thought of the proposed method are introduced in the paper; the comprehensive and filtering functions of the method to multivariable information are analyzed with an example. The proposed method is used to quantitative analysis of the harmonic sources. Utilizing the signals of harmonic voltage and harmonic current measured synchronously, regression coefficients are worked out through PLSR, and then the supply harmonic impedance is got; consequently the harmonic emission level of customer is calculated. The simulation results show that the proposed method is valid and feasible for the evaluation of supply harmonic impedance and harmonic emission level.