Optimizing configurable parameters of model structure using genetic algorithms

Fractionation product properties of crude distillation unit (CDU) need to be monitored and controlled through feedback mechanism. Due to inability of on-line measurement, soft sensors for product quality estimation are developed. Soft sensors for kerosene distillation end point are developed using linear and nonlinear identification methods. Experimental data are acquired from the refinery distributed control system (DCS) and include on-line available continuously measured variables and laboratory assays. In present work development of AutoRegressive Moving Average with eXogenous inputs (ARMAX) and Nonlinear AutoRegressive model with eXogenous inputs (NARX) are presented. To overcome the problem of selecting the best model set parameters by trial and error procedure, genetic algorithms were used for optimizing the best set of model parameters. Genetic algorithms were approved to be suitable method for optimizing ARMAX and NARX model structure in a way to find the best fits for given parameters range. Based on developed soft sensors it is possible to estimate fuel properties continuously by embedding model in DCS on site as well as applying the methods of inferential control.