Portal de Periódicos da CAPES

Development of model predictive control system using an artificial neural network: A case study with a distillation column

2020; Elsevier BV; Volume: 277; Linguagem: Inglês

10.1016/j.jclepro.2020.124124

1879-1786

Yeonju Shin, Robin Smith, Sungwon Hwang,

Process Optimization and Integration

Over the past few decades, advanced process control (APC) such as model predictive control (MPC) has been introduced to process industry to enhance its operational efficiency. For this, a linear model has been widely used to reduce the computational burden for iterative simulation and optimization over time, but it caused high inaccuracy of the control system. In this study, an artificial neural network (ANN) model was adopted instead of using the existing linearized model in order to increase the speed of optimization and accuracy of the model. For a case study, a depropanizer was modeled using Aspen HYSYS, and all feasible operation scenarios were considered to generate massive amounts of dynamic simulation data. Then, the accumulated data was implemented to the ANN for training, and it was tested. Once the verification was completed, the model was incorporated with an optimization algorithm in MPC system. For testing its performance, set point change and introduction of disturbances were applied to the model, and efficiency of the MPC was compared with the conventional control such as PID feedback control. The analysis results showed better performance (i.e., shorter settling time and rise time) of the MPC against the PID control. This methodology can be widely used in various types of control systems in the industry.