Artificial Intelligence Techniques Applied in the Control of the Composition of Liquefied Petroleum Gas
Fuzzy, Set Points Generation, Natural Gas Processing, LPG, Distillation Column
In the present work, artificial intelligence techniques were applied in a simulated natural gas processing plant, composed of two distillation columns: a deethanizing column and a debuthanizing column. In this process, the background product of the deethanizingngl, known as NGL, flows to the debuthanizing column, where it is fractionated. The lighter components are evaporated giving rise to LPG (Liquefied Petroleum Gas), while the heavier fractions, called C5 +, remain in the liquid state. Ideally, LPG is composed only of butane and propane, but in practice this is not the case, since contaminants such as pentanes and ethanes are always present. In this work it is proposed to regulate simultaneously the amount of pentane and ethane in the LPG composition, through the dynamic generation of set points (SP) of controllers present in the regulatory control of the columns. For this purpose, a multivariable fuzzy system is used that will adjust the values of these SPs, from the comparison of the molar fraction of pentane and ethane present in LPG and its desired quantities. The measurement of pentane and ethane is considered difficult due to the high cost, long measurement intervals and low reliability of the equipment when operating directly on the production line. For this reason, in this work also investigates the possibility of the combinated use of the fuzzy systems proposed and inference systems, based on artificial neural networks.