Error-triggered on-line model identification in economic model predictive control

Abstract

Economic model predictive control (EMPC) is a feedback control technique that employs real-time dynamic optimization to find optimal control actions with respect to a cost function representing the plant economics. The model used in EMPC must be able to capture the important dynamics of the plant. In industry, it may be difficult in many applications to obtain a first-principles model of the process, in which case a linear empirical model constructed using process data may be used as the process model within an EMPC. However, linear empirical models may not capture the nonlinear dynamics over a wide region of state-space, restricting an EMPC to operate in a small region within which the potential of EMPC for improving process profit is not realized. For this reason, we present a scheme for expanding the level sets used to design state constraints in Lyapunov-based economic model predictive control (LEMPC) with linear empirical models to improve the process profit, incorporating on-line updates of the linear empirical model triggered by model prediction errors quantified by a moving horizon error detector. A chemical process example illustrates that the proposed LEMPC can maintain closed-loop stability of the process and bring the performance close to that which would be obtained if the nonlinear process model were used in LEMPC.