Steady-state optimization and nonlinear model-predictive control of a reactive distillation process using the software platform do-mpc

Abstract

The increase in the computational power and the development of new efficient algorithms for numerical simulation and dynamic optimization have brought the model-based control of complex industrial chemical processes on the basis of nonlinear large-scale mathematical models within reach. One of the most challenging applications is the control of reactive distillation (RD) processes, due to the high complexity and nonlinearity that results from the tight integration of separation and chemical reactions in one apparatus and the presence of multiple steady states. Model-predictive control of such processes was investigated and shown to improve process performance in several theoretical studies, e.g. [1]. However, the reliable solution of the resulting dynamic optimization problems for such large-scale DAE models in real-time remains a challenge. In this paper we present the realization of nonlinear model predictive control (NMPC) for a RD process described by large DAE models of different levels of detail. The implementation is done by means of the software tool do-mpc [2], which is a development platform for the efficient implementation of dynamic optimal control problems. A two layer control approach is tested, where an evolutionary algorithm is used to determine optimal steady-state points based on a detailed model of the process. Tracking these points with the NMPC in a smooth and real-time feasible fashion is achieved by a second layer. The focus of the paper is on the parametrization and the performance of the numerical solutions of the dynamic optimization problem.