Maximum-Torque-per-Ampere and Magnetization-State Control of a Variable-Flux Permanent Magnet Machine

The variable-flux flux-intensifying interior permanent magnet machine exhibits high efficiency with variable permanent magnet flux properties. Control algorithms for a single magnetization state and during transition between different magnetization states are developed. Maximum-torque-per-ampere control is implemented for loss reduction. However, the proposed prototype machine exhibits sophisticated nonlinear inductance properties, which have a significant impact on precise maximum-torque-per-ampere control. This paper proposes an artificial-neural-network-based maximum-torque-per-ampere control scheme. The well-constructed neural network exhibits an excellent capability in nonlinearity simulations and is suitable for maximum-torque-per-ampere current command generation. In addition, the magnetization state should be manipulated properly according to the machine's working condition. A novel magnetization-state control system that generates magnetization-state switch signals by calculating the dc-link voltage margin ratio is proposed. The combination of the neural-network-based maximum-torque-per-ampere control and the magnetization-state control constitutes the proposed control system. The entire control system is validated experimentally.