Formulation of model predictive control algorithm for nonlinear processes
Process control is essential in any chemical plant. For the past forty years, the conventional PID controller has governed the process control industry. It is the sole selection although many other sophisticated control algorithms have been developed largely because it is able to deliver satisfact...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Monograph |
| Language: | English |
| Published: |
Universiti Teknologi Malaysia
2004
|
| Subjects: | |
| Online Access: | http://eprints.utm.my/id/eprint/2892/1/71992.pdf http://eprints.utm.my/id/eprint/2892/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Universiti Teknologi Malaysia |
| Language: | English |
| Summary: | Process control is essential in any chemical plant. For the past forty years, the conventional PID controller has governed the process control industry. It is the sole selection although many other sophisticated control algorithms have been developed largely because it is able to deliver satisfactory performance for most control problems when properly tuned and installed. However, with faster computing technology, the industry is now demanding a tighter advanced control strategy. To fulfil all these objectives, Model Predictive Control (MPC), an optimal model based control algorithm is definitely the best choice among all the advanced control algorithms available to date. The most significant feature that distinguishes MPC from other control algorithms is its long range prediction concept. MPC will perform the prediction over the future horizon and this will enable current computations to consider future dynamic events and hence allow it to overcome the limitation from the process dead-time, nonminimum phase and slow process dynamic. This research explores the capability of MPC in controlling a highly nonlinear, iterative process. Two case studies are explored. For the first case study, linear MPC is applied on a continuous solution copolymerization reactor with promising results. For the second case study, linear and Nonlinear MPC is applied on a high purity distillation column. This is to determine if there is superiority of one over the other. An unconstrained MIMO DMC and nonlinear MPC (NNMPC) algorithms were developed using a step response model and two feedforward neural networks respectively. Additionally, the comparison between DMC, NNMPC and PI controller based on IAE tuning rules was conducted. Overall, NNMPC control scheme shows a superior performance over the DMC and PI controllers by presenting a smaller overshoot and shorter settling time. |
|---|