Robust closed-loop real-time optimization for refinery utility plant with model predictive control for constraint handling

Abstract

This paper presents the results from a very successful Energy Closed-Loop Real Time Optimization (E-CLRTO) application implemented at the Idemitsu Aichi Refinery in Japan. The goal of the application is to optimize the steam, fuel, and electrical needs of a refinery site at minimum cost, taking into consideration process, environmental, and operational constraints as well as daily variations in utilities costs. The paper presents keys to the success of the project, including the handling of the disturbance inputs such as utility load changes, which occur very rapidly, and a design concept that a Model Predictive Control (MPC) would function similarly to a the skilled operator's operation of constraint handling. To accomplish this, we analyzed an operator's optimal actions and then developed the controller model so that the controller outputs would simulate the operator actions. The E-CL-RTO has reliably provided optimal targets to the equipment automatically and continuously, without any need for operator intervention, resulting in increased standardization. Finally, the paper outlines the key benefits derived from the implementation of the E-CL-RTO.