Multiple Discrete Adaptive Filter Based Position Error Reduction for Sensorless IPMSM Drives

The inverter nonlinearity and the flux spatial harmonics will lead to noticeable $(6k \pm 1)\text{th}$ harmonics in the observed back electromotive force (EMF), causing $(6k)\text{th}$ pulsations in the rotor position estimation error and deteriorating the performance of sensorless interior permanent magnet synchronous motor (IPMSM) drives. To improve the sensorless IPMSM control accuracy, a multiple discrete adaptive filter (MDAF) embedded between sliding mode observer and normalized quadrature phase-locked loop is adopted to solve the above problem in this article. The MDAF consists of a fundamental extraction block and a harmonic cancellation block. Harmonic cancellation works similarly to fundamental extraction, simply by removing the dominant fifth and seventh harmonics from the distorted back-EMF in a frequency multiplicative manner. The proposed algorithm is designed directly in the discrete domain, avoiding the complex discretization process. The open-loop amplitude-frequency characteristic curves show that the method is able to achieve infinite gain exactly at the expected frequency, thus allowing closed-loop zero steady-state error tracking of the back-EMF and providing higher precision for harmonic estimation. The excellent performance of the suggested method is validated based on a 1.5 kW sensorless IPMSM control system.