This paper describes a new method for identifying and separating Volterra kernels of nonlinear control systems by use of pseudorandom M-sequence and correlation technique, and its application to model predictive control. By use this identification method, we can obtainVolterra kernels of up to 3rd order of a nonlinear system. M-sequence is applied to a nonlinear system and the crosscorrelation function between the input and output is calculated. then the crosscorrelation function includes all the crossections of Volterra kernels of the nonlinear system. The problem is how to separate these crossections from each other. This paper proposes two methods for separating these crosssections: one is the suitable selection of M-sequence and the other is amplitude variation method. This identification method is applied to nonlinear model predictive control.