Algorithm for initial position detection of PMSM sensorless control based on Newton's method in conjunction with linear search of Wolfe's criterion

This paper proposes a novel optimization-based algorithm for PMSM initial position detection at full angle range. A cost function is employed for initial position estimation. The minimization problem of the cost function is solved using Newton's method combined with a linear search of Wolfe's criterion to obtain the pole position of the motor. A two-stage current offsets with high frequency voltage injection is applied to the stator winding to identify the pole polarity. Compared with existing initial rotor position detection method, the proposed method avoids the dispersion in non-convex areas of the cost function. Meanwhile, the proposed polarity discrimination avoids the effect of reverse saturation of the stator yoke. The proposed initial position discrimination algorithm is validated by simulation and experiments.