D16.6 - Sensorless Control of SPMSM for Low-Speed Region Based on Physics-Informed Extreme Learning Machine

This digest proposes a sensorless control method for SPMSMs in the low-speed region using a physics-informed extreme learning machine (PI-ELM) trained with motor constraints. Physical relationships are embedded to enable accurate estimation with a reduced input set, resulting in lower computational load, minimized data requirements, and enabling real-time execution on low-cost DSPs. Rotor position and speed are obtained through a lightweight PLL, where a drift at very low speeds is suppressed by periodic addition of sin(θ ̂_r) and cos(θ ̂_r). A stable operation has been demonstrated through experimental results, confirming the feasibility of the method for sensorless operation at as low speed as 0.01 p.u.