Search

Abstract Park models, also known as dq-models, are a simple and very efficient way to obtain a faithful image of the behaviour of a machine in transient and steady-state operating conditions to deal with control issues. Their precision is based on the use of saturated direct and quadrature reactances that truly represent the dynamics and the magnitude of the electrical signals at nominal operating point. However, these constant parameters are not sufficient to accurately take into account the saliency of the machine and, therefore, to precisely simulate the harmonic contents of these signals: the latter point can be very problematic for diagnostic purposes. This paper uses simulation results and experimental measurements on a Leroy-Somer 27kVA alternator to clearly show the lack of coherence between the simulated spectra from a dq-model and the experimental data retrieved from a real power alternator. For that reason, a new methodology of machine modelling consisting of a co-simulation process between Flux2D and Matlab software is presented. The use of finite elements enables a fine identification of the inductances of the system. These take into account the geometrical specificities of both the exciter and the main generator and are then used in an analytical model developed in the Matlab environment. Spectral comparisons with experimental data demonstrate the improved accuracy of this new modelling method compared to dq-models in terms of harmonic contents. [ABSTRACT FROM AUTHOR]