Formaldehyde production process control improvement using multivariable model predictive control

Formaldehyde is chemical substance that is used in adhesive industry. PT X is formaldehyde producer in East Java which is using proportional-integral control system. This conventional controller has several weaknesses. Multivariable model predictive control (MMPC) is used to increase the performance of control system at PT X. Empirical model is made with process reaction curve followed by first order plus dead time calculation. Four manipulated variables and four controlled variables will construct 16 empirical models. Calculation of MMPC parameter, which include sample time (T), prediction horizon (P), and control horizon (M), is done with Shridhar and Cooper method (1998) and optimized by fine-tuning method. Performance of MMPC is tested by set point (SP) tracking and disturbance rejection. Four controllers tested are evaporator pressure control (PIC-101), liquid percent level control (LIC-101), steam flow control (FIC-102), and air temperature control (TIC-101). The optimized parameter of MMPC which include T, P, and M are 3, 62, and 2 respectively. Multivariable model predictive control can increase control performance in SP tracking with average number of 33.24% for IAE and 42.93% of ISE. Meanwhile, in disturbance rejection, there is an increase in average of 33.485 for IAE and 58.08% for ISE.