(137c) Shifting Real Time Optimization Tasks to Plant-Wide APC Layer By Introducing Unconstrained and Constrained Optimum Concepts

Shifting Real Time Optimization tasks to Plant-wide APC Layer by Introducing Unconstrained and Constrained Optimum Concepts

James Wu

LyondellBasell Industries

Abstract to be submitted to the process control and optimization session in 2019 Spring AIChE meeting

Plant-wide optimization plays a critical role in improving chemical plant and refinery economic performance. Traditionally, steady state rigorous modeling optimization approaches have been used. There are several key challenges, such as using steady state models to deal with dynamic operating conditions and constraints, coordination with APC layer, etc. See a comprehensive RTO review by M.L. Darby et al. [1]. In the recent decades, “grey-box” nonlinear dynamic optimizers have been implemented to replace steady state optimizers. In addition, attempts have been made to move optimization functions to APC layers. One popular approach widely used in olefin plants is to use composite linear programming to coordinate plant-wide APC applications and to push feed against capacity constraints. But, it cannot be used for non-linear yield-related optimization trade-offs[1].

In this paper, a new plant-wide optimization approach is developed for those plants where individual APC applications can be integrated under a composite linear programing application or all key optimization trade-offs are included in one large APC application. It shifts daily tasks related to dynamic optimization trade-offs under constrained conditions from the RTO layer to the plant-wide APC layer. This is done by introducing the concepts of unconstrained and constrained optima. It shows that for many practical applications unconstrained and constrained optima defined here depend mainly on pricing and feed composition changes, but not on real time operating conditions. As a result, the plant-wide APC layer can be used for daily real time optimization based on unconstrained and constrained optima. These optima can be updated using either existing steady state optimizers (with modification and much reduced frequency) or rigorous offline process models. This new approach can significantly improve optimization performance under dynamic capacity constraints, and can also reduce overall optimization implementation and support efforts.

[1] Mark L. Darby, Michael Nikolaou, James Jones, Doug Nicholson, RTO: An overview and assessment of current practice, J. of Process Control 21 (2011) 874-884.