A Dynamic Real-Time Optimization Strategy for the Tennessee Eastman Benchmark

Abstract

Abstract Conventional real-time optimization (RTO) algorithms provide the steady-state set points at which the process would operate economically. However, the process may suffer from severe transient losses going from the nominal steady state to the optimal one. This may also lead to off-specification production during set point transitions. In this paper, a dynamic RTO strategy accounting for the transient production cost is developed for the Tennessee Eastman process. The economic objective function is defined as the integral of sum of two components over a specified prediction horizon: (i) the transient operating cost and (ii) the cost accrued due to unsalability of the off-specification product during transitions. Furthermore, a predictive model is employed to estimate future process outputs and product quality required to evaluate the objective function and constraints. The results demonstrate significant economic savings (up to 28%) of the proposed strategy over a conventional RTO approach that considers steady-state economics only, especially when the process is subjected to sustained disturbances.