It is important to realize dynamic decoupling control for the torque and the stator flux-linkage of induction motor (IM) drives, which is multi-input and multi-output, nonlinear and strong coupling system. In this paper, an innovative dynamic decoupling control for IM drives is proposed, which is based on inverse system theory. In order to make the induction machine drive operate steadily and attain good dynamic and static performance, it is necessary to compensate the inverse decoupling controller due to parameter variances, disturbance and other uncertainties. A method based on PCA-BP neural network is used to study the dynamic decoupling compensator. The feasibility of inverse decoupling compensation based on PCA-BP is discussed in detail. The simulation results suggested the validity of the proposed scheme.