(409e) Large-Scale Optimization Strategies for the Operation of Low-Density Polyethylene Tubular Reactors

There is substantial economic interest to optimize the operations of low-density polyethylene (LDPE) tubular reactors. Due to the high complexity of these units, systematic optimization techniques need to be used for this. One of the main limitations associated to this is the high dimensionality and complexity of the multi-zone tubular reactor model which is described by a large set of partial, ordinary and algebraic equations (PDAEs). In this work, we demonstrate that a simultaneous full-discretization approach coupled to a full-space nonlinear programming (NLP) solver results in an efficient strategy to cope with these limitations. We exploit these advantages in the analysis of different scenarios arising in the operation of LDPE reactors. In particular, we propose moving horizon estimation and nonlinear model predictive control strategies to track and compensate for persistent disturbances such as fouling in order to keep the reactor temperature profile and final properties of the polymer at targets. Finally, we show that the controller can easily be extended to incorporate economic decisions in the objective and we illustrate the potential benefits and bottlenecks of this approach.