Linear model predictive control has been widely accepted in industry as an important tool for the operation of difficult interacting processes. Linear identification and control techniques are well developed and well understood. In the industry, it is rare to find a system that is truly linear. While for many systems linear modeling and control can approximate their performance in certain regions, there exist some systems whose nonlinearity is great enough that an approximate linear model and control scheme cannot yield the desired accuracy. In order to control these more complex nonlinear systems, significant research has been dedicated to extending model predictive control to nonlinear systems. The problem of implementing nonlinear model predictive control can be split into two main tasks: making the nonlinear model and calculating control inputs. The significant contributions of this dissertation are in the area of identification of nonlinear Volterra models from input-output data. Historically, the identification of Volterra models has been limited to lower order models because of the large amount of model parameters that need to be identified. By using the Laguerre polynomials, the number of model parameters can be greatly reduced, which limits the required input-output tests. The goal of this dissertation is to move nonlinear multivariable control closer to industrial application by addressing practical model identification questions. Results from three test cases are presented and discussed. The results have shown a decrease in parameters of as much as 99% without a significant loss in model fidelity.