Encoderless Predictive Control of PMSM Drives Combining Sliding-Mode and Luenberger Observers

A finite-state predictive control scheme of an encoderless permanent magnet synchronous motor (PMSM) drive with neutral-point-clamped (NPC) inverter is suggested in this paper. For accurately estimating the motor speed while avoiding the large discretization errors of the angle derivation at low speeds, a system of a sliding-mode and Luenberger observer is developed. However, the additional mathematical operations of the speed observation, along with the increased inverter complexity, further increase the computation burden for the predictive controller thus requiring larger sampling intervals. On the other hand, if increasing the sampling time, the steady-state and transient performances are degraded in terms of current distortion, torque ripple, as well as response time. Therefore, for online optimizing the system, a minimization algorithm of the switching states is also developed. The suggested algorithm explores only a subset of the available combinations of the NPC inverter thus significantly reducing the needed hardware resources. The effectiveness of the main control routine, the position and speed observers, and the minimization methodology are verified in a PMSM drive.