Dynamic Optimization Formulations for Plant Operation under Partial Shutdown Conditions

Abstract

<p>Systematic strategies for optimal plant operation under partial shutdowns were developed. Partial shutdowns are circumscribed process unit shutdowns that permit the rest of the plant to continue operating to some degree. These strategies manipulate the degrees-of-freedom in a plant---during and after a shutdown---to restore plant production in a cost-optimal fashion, while meeting safety and operational constraints. This is accomplished through the adjustments of production rates, recycles and buffer levels.</p> <p>Our multi-tiered dynamic optimization approach allows for the prioritization of multiple objectives and the specification of trade-offs between these objectives. The solution of the optimization problem informs the formulation of inventory management policies. A Model-Predictive-Control (MPC) based partial shutdown algorithm implements these policies under feedback.</p> <p>Parsimonious discrete modeling formulations were presented for handling model discontinuities such as shutdown thresholds, induced shutdowns and minimum shutdown durations. The problem of minimizing restoration time was considered.</p> <p>We investigated the use of state/parameter estimation algorithms to moderate the effects of plant-model mismatch. The algorithms are based on novel configurations of the constrained Unscented Kalman Filter (UKF). Constraints on the estimates are enforced through a simple projection method. A dynamic feasibility tier ensures that terminal constraints and parameters are feasible for the prediction horizon in the control optimization problem.</p> <p>A modeling system (MLDO) was created for prototyping dynamic optimization models. It transforms a mathematical description of a model into code in various computer languages for the purposes of optimization, simulation, visualization and analysis of dynamic optimization problems. Facilities for problem reformulation and transformations are included.</p>