Neural network approach on loss minimization control of a PMSM with core resistance estimation

Abstract

Abstract:Permanent magnet synchronous motors (PMSMs) are often used in industry for high-performance applications.Their key features are high power density, linear torque control capability, high efficiency, and fast dynamic response.Today, PMSMs are prevalent especially for their use in hybrid electric vehicles. Since operating the motor at high efficiency values is critically important for electric vehicles, as for all other applications, minimum loss control appears to be an inevitable requirement in PMSMs. In this study, a neural network-based intelligent minimum loss control technique is applied to a PMSM. It is shown by means of the results obtained that the total machine losses can be controlled in a way that keeps them at a minimum level. It is worth noting here that this improvement is achieved compared to the case with I d set to zero, where no minimum loss control technique is used. Within this context, hysteresis and eddy current losses are primarily obtained under certain conditions by means of a PMSM finite element model, initially developed by CEDRAT as an educational demo. A comprehensive loss model with a dynamic core resistor estimator is developed using this information. A neural network controller is then applied to this model and comparisons are made with analytical methods such as field weakening and maximum torque per ampere control techniques. Finally, the obtained results are discussed.

A robust high-speed sliding mode control of permanent magnet synchronous motor based on simplified hysteresis current comparison

“…Other nonlinear methods that have been used to control the PMSM includes a neural network loss minimization control [18], generalized predictive control (GPC), sliding mode controller (SMC) [19]- [20], fuzzy logic controller (FLC) [21]. Generally, when compared to the linear control using PID controllers, nonlinear control methods, since the PMSMs are complex nonlinear systems, are more suitable in achieving better systems dynamics and steady state performance.…”

The decoupled vector-control of PMSM based on nonlinear multi-input multi-output decoupling ADRC

“…In 2013, Sun et al 7 used a neuralnetwork inverse speed observer to achieve a sensorless vector control of an induction motor. In 2017, Erdogˇan and Ö zdemir 8 applied a neural-network based minimum-loss control to the PMSM. A comprehensive loss model with a dynamic core resistor estimator was developed with the neural-network.…”