Model predictive control and neural network predictive control of TAME reactive distillation column

Model predictive control (MPC) is an advantageous methodology to control the nonlinear processes such as tert-amyl methyl ether (TAME). Multiple reactions of the system make the synthesis of the TAME process more complicated which exhibits highly nonlinear behavior. The need to handle such difficult control problem has led to use neural network in MPC. In the present work, three different control strategies, viz., conventional PID control, model predictive control and neural network predictive control (NNPC) are implemented to a TAME reactive distillation column (RDC). All these controllers are compared and it is found that NNPC and MPC give smoother and better control performance than the PID controller for both set point change and ±10% load change in feed flow rate of methanol.

► PID, MPC and NNPC have been used for controlling a TAME reactive distillation column.
► Dynamic model for the synthesis of TAME in RDC has been formulated.
► A MATLAB® program was written to solve the resulting ODE's.
► All control strategies were compared for set point as well as for load changes.