Multiobjective Nonlinear Model Predictive Control of a Class of Chemical Reactors

Many problems in science and engineering can be posed as multiobjective optimization problems where several objectives must be met simultaneously. Commonly such multiobjective optimization problems are reduced to a single optimization problem by merging all the involved objective functions by using rather subjective weighting functions. This solution procedure can produce suboptimal solutions, and it is not a systematic method since the choice of the weighting functions is up to the designer. Process control is one of the engineering fields where multiobjective optimization control problems frequently emerge because such problems normally involve conflicting objective functions such as economical profit and environmental concerns. Because the optimal value of the conflicting objective functions cannot be simultaneously achieved one has to compute a trade-off solution that best suits, in a given sense, all the objective functions. Moreover, additional requirements, beyond upset rejection and set-point tracking, such as the determination of optimal operating conditions should also be handled by dynamic real time optimal control approaches. In this work we propose a novel multiobjective optimization and control approach able to get target points while simultaneously computing optimal operating conditions. The proposed approach can be applied to nonlinear dynamic systems and does not require the specification of arbitrary weighting functions to handle conflicting multiobjective optimization problems. Several case studies using chemical reactors of varying nonlinear behavior are deployed to illustrate the practical application of the proposed dynamic real time optimization approach.