Offline Parameter Self-Learning Method for Low-Impedance Dual Three-Phase PMSMs: Addressing AC Losses and Inverter Nonlinearity

Existing parameter identification methods often neglect the AC losses effect on impedance parameters and exhibit incomplete compensation for low-current-region voltage source inverter (VSI) nonlinearity. As a result, identification failures or inaccuracies occur in low-impedance motors. To address these challenges, this paper presents an impedance-based parameter self-learning method for dual three-phase permanent magnet synchronous motors (DTP-PMSMs). The proposed method utilizes a hybrid current injection sequence that combines xy plane bias current with test-axis sinusoidal current to account for the AC losses effect. The xy bias current saturates the terminal voltage errors, which is then eliminated by differencing the proposed injection states, thereby eliminating the VSI nonlinearity. Additionally, a time-division multiplexed parameter estimation method based on signal processing and recursive least squares (RLS) is introduced to enhance signal-to-noise ratio and identification accuracy. Finally, experiments are carried out on a high-speed, low-impedance DTP-PMSM to verify the proposed method.