Predictive control using artificial neural network

The first aim of this study is to introduce a technique to predict the product qualityof fractionated gases using artificial neural networks to help the operator to get the rightdecision in gas fractionation process, which is one of a very complicated process due to thevariety of the parameters, that affects this process and the interdependence between theseparameters. This in order to be used to control the process in-spite of fast variations in thisprocess to get better quality.The predictive control model consists of multi layer neural network with 16 inputsand one output. The inputs are the feeding gas mixture parameters (flow rates, pressures,and temperatures), hot oil parameters, the operating conditions of the fractionation towerand the product gases analysis. The output of the model is the product quality.The model has then used in a predictive control algorithm. The objective of thecontrol is to avoid the escaping of the lighter components of hydrocarbons in LPG product,thus to maintain constant product quality and also, to optimize the productivity of theprocess. Traditionally this is done by taking samples and analyzing them in laboratory,which is a time consuming, manual process, which does not fully control the process.Traditional control will fail because of the time gap between products entering and leavingthe column. This study presents a predictive control system that addresses those problemsin order to achieve better gas quality.