Adaptive Nonlinear Model Based Predictive Control: Application to an Esterification Semi Batch Reactor

This paper deals with the identification and the control of nonlinear processes described by input -output models, such as parametric Volterra models. In particular, we extend an adaptive predictive algorithm without taking into account constraints. The calculation of the control law can be posed as a thirdorder nonlinear program. The building algorithm is based on a new approach using a convolution product. The self-tuning, on-line of the predictive regulator parameters is assured via the weighted recursive least square algorithm WRLS. The nonlinear optimization problem will be reformulated in case of presence of constraints and resolved by a nonlinear programming method as the Lagrange multipliers. The main developed results are applied to an esterification reactor. Key-Words: Predictive ControlEsterification Reactor Parametric Volterra Model Constrained Optimization.