Data-Driven-Based Vector Space Decomposition Modeling of Multiphase Induction Machines

doi:10.1109/TEC.2023.3255792

2023-09-01

Mohamed A. Abu-Seif

Mohamed M. Ahmed *

Mohamed Y. Metwly

Ayman S. Abdel‐Khalik

Mostafa S. Hamad

Shehab Ahmed

Noha A. Elmalhy

DOI：10.1109/TEC.2023.3255792

原文链接

原文求助

摘要

En 中文

For contemporary variable-speed electric drives, the accuracy of the machine's mathematical model is critical for optimal control performance. Basically, phase variables of multiphase machines are preferably decomposed into multiple orthogonal subspaces based on vector space decomposition (VSD). In the available literature, identifying the correlation between states governed by the dynamic equations and the parameter estimate of different subspaces of multiphase IM remains scarce, especially under unbalanced conditions, where the effect of secondary subspaces sounds influential. Most available literature has relied on simple RL circuit representation to model these secondary subspaces. To this end, this paper presents an effective data-driven-based space harmonic model for n-phase IMs using sparsity-promoting techniques and machine learning with nonlinear dynamical systems to discover the IM governing equations. Moreover, the proposed approach is computationally efficient, and it precisely identifies both the electrical and mechanical dynamics of all subspaces of an IM using a single transient startup run. Additionally, the derived model can be reformulated into the standard canonical form of the induction machine model to easily extract the parameters of all subspaces based on online measurements. Eventually, the proposed modeling approach is experimentally validated using a 1.5 Hp asymmetrical six-phase induction machine.

Keyword:

Multiphase machine

induction motor

machine learning

parameter estimation

six-phase machines