Neural Network Modeling and Development of an Application for Optimizing Flare Gas Recovery Module in Skikda Refinery RA1K
Loading...
Date
2025-07-06
Journal Title
Journal ISSN
Volume Title
Publisher
20th August 1955 University – SKIKDA Faculty of Technology Department of Petrochemistry
Abstract
This memory presents the modeling and optimization of the Flare Gas Recovery (FGR)
module, focusing on the GTK unit at Skikda Refinery RA1K, using artificial neural networks
(ANNs) and process simulation. The goal is to develop a predictive model and a MATLAB
application to make the unit smarter, helping engineers reduce gas flaring, improve energy
efficiency, and make better operational decisions. The operational goal is to maximize LPG
recovery in unstabilized naphtha, while minimizing energy losses and emissions related to
flaring. The GTK module was simulated with ASPEN HYSYS to understand its behavior and
generate reference data. Real operational data from the DCS were used to train and test the
ANN model. The developed application integrates this model with a user-friendly interface for
prediction and scenario analysis. Results confirm the accuracy of the approach (R² > 0.99) and
its potential to minimize flaring losses. This work shows how combining simulation and AI
tools can support more sustainable refinery operations