Optimal Selection of Measurements and Manipulated Variables for Production Control

Abstract

The main objective in a chemical plant is to improve profit while assuring products meet required specifications and satisfy environmental and operational constraints. A sub-objective that directly affects profit (main objective) is to improve the control performance of key economic variables in the plant, such as production rate and quality. An optimal control-based approach is proposed to determine a set of measurements and manipulated variables (dominant variables) and to structure them to improve plant profitability. This approach is model-based, and it uses optimal control theory to find the dominant variables that affect economic variables in the plant. First, the measurements and manipulated variables that affect product flow and quality are identified. Then, a decentralized control structure is designed to pair these measurements with the manipulated variables. Finally, a model predictive control (MPC) is built on top of the resulting control structure. This is done to manipulate the set point of these loops in order to change the production rate and product quality.

Another sub-objective that affects the profit in the plant is to improve the control of inerts. In general, the inventory of the inerts is controlled using a purge. A new methodology to optimally control inerts is presented. This methodology aims to reduce the losses that occur throughout the purge by solving an optimization problem to determine the maximum amount of inert that can be handled in the plant without having shut down of the plant due to inert accumulation. The methodology is successfully applied to the Tennessee Eastman Plant where the operating cost was reduced approximately 4%.

This methodology solves an approximation to an optimal economic problem. First, it improves the control performance of key economic variables in the plant. Therefore, tighter control of these economic variables is achieved and the plant can be operated closer to operational constraints. Second, it minimizes purge which is a variable that generally causes significant costs in the plant. This approach is applied to the Tennessee Eastman and the Vinyl Acetate Processes. Results demonstrating the effectiveness of this method are presented and compared with the results from other authors.