Hide/Show Apps

Nonlinear model predictive controller using neural network

Karahan, O
Ozgen, C
Halici, U
Leblebicioğlu, Mehmet Kemal
In this paper, a non-linear Model Predictive Control (MPC) algorithm is proposed which extends the capacities of Linear Model Predictive Controllers to control non-linear systems. A Neural Network (NN) is used to model the deviation of the non-linear system from its linear MPC model. Proposed algorithm is tested in control of an industrial multi-component high purity distillation column by simulation. Results of NN-MPC show high improvement in control of system over linear MPC algorithm.