The importance of first-principles, model-based steady-state gain calculations in model predictive control-a refinery case study

This paper addresses the development and application of a first-principles, steady-state modeling framework in multivariable control applications. A rigorous approach based on detailed nonlinear models calibrated with reconciled online measurements is presented. Sensitivity analysis of this model is then applied in order to generate steady-state gain (inferential) models used in a DMC-based control application of a refinery unit. The benefits of using open-equation based inferential models to account for online product quality control are demonstrated in the context of a real-time model predictive control system, applied to a refinery. Finally, the direct economic impact of this application is assessed in a detailed quantitative manner and offered along with the relevant business process changes and operational practice recommendations for sustaining the benefits achieved.