Black box nonlinear model predictive control using recurrent neural network
A black box Nonlinear Model Predictive Control (NMPC) based on a Recurrent Neural Network (RNN) is implemented to solve two nonlinear benchmark examples: a Continuous Stirred Tank Reactor (CSTR) and Quadruple Tank Process (QTP). The RNN model is trained by a set of input and output data from the pla...
| Main Authors: | , , |
|---|---|
| Format: | Conference or Workshop Item |
| Language: | English |
| Published: |
2013
|
| Subjects: | |
| Online Access: | http://irep.iium.edu.my/32567/ http://irep.iium.edu.my/32567/ http://irep.iium.edu.my/32567/1/Paper_30166_-_Camera_ready_Moumen_Mohammad.pdf |
| Summary: | A black box Nonlinear Model Predictive Control (NMPC) based on a Recurrent Neural Network (RNN) is implemented to solve two nonlinear benchmark examples: a Continuous Stirred Tank Reactor (CSTR) and Quadruple Tank Process (QTP). The RNN model is trained by a set of input and output data from the plant. A nonlinear observer based on Extended Kalman Filter (EKF) is used for the state estimation process. The implementation of successive linearization technique in NMPC shows an improvement in handling the plant nonlinearity and in the same time preserves all the important features of linear model predictive control (LMPC) with quadratic optimization function. To demonstrate the improvement, the NMPC performance is compared with LMPC based on identified state space model. Both examples show the superiority of the NMPC over LMPC. |
|---|