A multivariable nonlinear decoupling control strategy based on least squares support vector machine (LS-SVM) inverse system is presented for a L-lysine fermentation process. The invertibility of dynamic model for L-lysine fermentation process is verified and is obtained offline using a LS-SVM with Gaussian kernel, which connects in series with the original system and makes the complicated nonlinear multivariable system decoupled into several independent single-input-single-output (SISO) pseudo- linear subsystems. The simulation results demonstrate the applicability and superiority of the presented method.