Abstract: Inverter nonlinearity is one of the critical factors that affect the low-speed operation control performance for permanent magnet synchronous motor drives. In this paper, a voltage error self-learning method considering the zero-axis voltage was proposed to reduce the negative effects caused by the inverter nonlinearity. The nonlinear characteristics of the inverter were extracted by injecting a ramp signal into the d-axis, and then the voltage error of the inverter was compensated online. The influence of the inverter nonlinearity on the zero-axis voltage during the coordinate transformation process was clarified, and the nonlinear error voltage characteristics of the inverter were extracted by injecting a ramp current signal into the direct axis. Combined with the characteristics of the zero-axis voltage changing with the electrical angle, the influence of the self-learning process was corrected, so as to realize the offline identification of the inverter nonlinear characteristic curve and the compensation of inverter nonlinearity. Finally, experiments were conducted on the presented method of inverter nonlinearity self-learning, and the results show the effectiveness of the method.