Advanced algorithm for rotor position saliency extraction in induction machines using PWM-induced transient excitation

For speed sensorless control at zero electrical stator frequency inherent machine saliencies need to be exploited. One way to obtain saliency information is to use the voltage step excitation of the inverter and evaluate the current reaction by estimating the derivative of the current from the measured current slope. The initial transient oscillation of the current slope prohibits estimation of the derivative from short signal length. To overcome this problem, a robust and fast derivative estimation technique is presented in this paper. The proposed algorithm is capable to estimate the current derivative from short signal lengths, which also contain the initial transient oscillation process. A fast Fourier transformation with subsequent linear transformation is used. The paper compares the proposed algorithm with already practically applied standard evaluation methods like difference calculation and linear regression methods, at different sampling rates and signal lengths. To grade the performance the statistical properties were compared, based on measurement results taken from a 15kW induction machine.