A Computationally Efficient Sequential RLS Estimator for PMSM Parameters: Algorithm Design and Experimental Verification

Accurate knowledge of electrical parameters is crucial for high-performance field oriented control (FOC) of permanent magnet synchronous motors (PMSMs). However, parameters vary significantly due to thermal effects and magnetic saturation. This study proposes a computationally efficient sequential recursive least squares (SRLS) algorithm for real-time parameter identification. Unlike conventional methods, proposed approach decouples the d- and q-axis dynamics to reduce the matrix operation complexity. Identification is achieved under persistent excitation provided by the active torque current (i_q) and the standard FOC scheme at a 20 kHz sampling frequency. Experimental validation is conducted on a 500 W motor operating within a speed range of 200–1000 RPM under constant load conditions at room temperature. Results show that the SRLS algorithm achieves a relative estimation error of 2.59 % for resistance and less than 5% for inductance. The proposed algorithm executes in only 26.5µs on an STM32F427 microcontroller, resulting in a 53% CPU load, which represents a 43% reduction in computational burden compared to the standard RLS method.