Sensorless Speed Control of PMSMs based on an Improved Particle Filter

Abstract

Position sensors in permanent magnet synchronous motor (PMSM) are restrained in some applications because of the space restriction and the reliability of the system. A marginalized particle filter (MPF) can used to staminate the speed of the motor by combining a Kalman filter (KF) with a particle filter (PF). In the MPF algorithm, the rotor position of the permanent magnet synchronous motor (PMSM) is represented by a set of particles, and the rotor speed associated with each particle is estimated by using the KF. The PF here is used to handle the non-Gaussianity and nonlinearity of the system. In this paper, the uniform distribution of particles is proposed to replace the traditional Gaussian distribution of particles in the PF to get better performance at the low-speed range. The motor drive system prototype is built using a TMS320F28335 digital signal processor as a controller core. The proposed improved PF is used in the sensorless speed vector control PMSM system. The experimental results show that the proposed PF with a uniform distribution of initial particles enables more accurate speed estimation in the low-speed range compared to the conventional Gaussian distribution, while increasing number of particles also helps to improve the accuracy.