Model predictive control of an industrial gasoline hydrodesulphurization unit

This study reviews design and implementation of a model predictive controller algorithm to a gasoline hydrodesulphurization unit in Izmit Refinery, Turkish Petroleum Refineries Corporation which produce gasoline within Euro V specifications. The unit has a splitter column in which high octane low sulphur part of the feed is separated and a reactor part where hydrodesulphurization occurs. The main disadvantage of the process is to lose octane during hydrodesulphurization. A control scheme to optimize operation in both parts is proposed. The main objective is to control product sulphur concentration strictly while minimizing octane loss. In order to obtain an empirical model of the unit, dynamic tests are performed. The model derived is used in controller design. Benefit analysis is performed with six months of data collected before and after MPC implementation. According to the results, standard deviation in product sulphur concentration is decreased and close to high limit operation is achieved for this specification as well as octane loss due to over desulphurization is minimized.