Online Multi-Parameter Identification of PMSM Drives Using a Fuzzy PI-Tuned MRAS Observer

Permanent magnet synchronous motors (PMSMs) are widely used in AC drive systems, and their control performance depends strongly on accurate motor parameters. Conventional proportional-integral model reference adaptive system (PI-MRAS) observers use fixed adaptation gains, resulting in a trade-off between rapid convergence and low steady-state fluctuation. To address this limitation, this paper proposes a fuzzy proportional integral (Fuzzy-PI)-tuned MRAS observer for the simultaneous online identification of stator resistance (Rs) and stator inductance (Ls). The parameter-error dynamics are formulated from the PMSM model, and the adaptation laws are derived using Popov hyperstability theory. A fuzzy tuner uses the absolute identification error and its rate of change to schedule the proportional and integral gains online, thereby accelerating transient error convergence when the identification error is large and reducing estimation oscillations during steady-state operation. The method is evaluated through simulation and laboratory experiments involving rated operation, speed variation, parameter perturbation, and load disturbance. Under the investigated conditions, the identification errors of Rs and Ls are 3.8% and 0.18%, respectively. Compared with the conventional PI-MRAS, the reported Rs identification error decreases from 8.1% to 3.8% and the Ls identification error decreases from 0.91% to 0.18%. The results demonstrate an improved identification accuracy and disturbance recovery within the tested operating range. The implementation on an Infineon TC233 platform also demonstrates real-time feasibility, while broader validation under temperature variation, magnetic saturation, inverter nonlinearity, and measurement noise remains necessary.