Regularization-Theory-Based Fast Torque Tracking Method for Interior Permanent Magnet Synchronous Machines

With the rapid growth of interior permanent magnet synchronous machines in electric vehicle applications, there is a need to generate torque tracking look-up tables that can both track the torque command and implement maximum torque per ampere (MTPA)/maximum torque per volt (MTPV). So far, most torq...

Full description

Saved in:
Bibliographic Details
Published inIEEE transactions on industrial electronics (1982) Vol. 70; no. 12; pp. 12113 - 12123
Main Authors Qi, Xing, Aarniovuori, Lassi, Cao, Wenping
Format Journal Article
LanguageEnglish
Published New York IEEE 01.12.2023
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Subjects
Electric vehicles
Global optimization
Interior permanent magnet synchronous machines motor (IPMSM)
look-up table
Lookup tables
Machine learning
maximum torque per ampere (MTPA)
maximum torque per volt (MTPV)
Permanent magnet motors
Permanent magnets
Regularization
regularization theory
Stators
Synchronous machines
Synchronous motors
Table lookup
Testing
Testing time
Torque
torque tracking
Tracking
Workload
Online AccessGet full text

Cover

More Information
Summary:With the rapid growth of interior permanent magnet synchronous machines in electric vehicle applications, there is a need to generate torque tracking look-up tables that can both track the torque command and implement maximum torque per ampere (MTPA)/maximum torque per volt (MTPV). So far, most torque tracking methods require a large amount of test points, giving rise to long test time and workloads. This article proposes a fast torque tracking MTPA/MTPV look-up table generating method to improve the efficiency. The proposed method is based on a machine learning regularization theory, using an L1/L2 regularization to establish a data-driven torque tracking model. Then, a Lagrange dual principle is introduced to solve the unknown parameters, so that the look-up tables of optimal dq -axis currents are yielded by a global optimization solver. Experimental results show that the proposed method can generate the look-up tables with the same accuracy as classical methods, but requires less test points and testing time. As a result, the testing work loads are reduced, as the time cost is only 10%-15% of the classical methods.
ISSN:0278-0046
1557-9948
DOI:10.1109/TIE.2023.3237895