Sensitivity analysis and optimization of the whole process of continuous catalytic reforming for Persian gulf star oil company using an optimized data-driven model with tuned parameters

Abstract This paper applies an existing advanced model to improve key outputs in the continuous catalytic reforming (CCR) process for Persian Gulf Star Oil Company. Using tools like Aspen Custom Modeler and Aspen Plus, we focus on optimizing two main results: Research Octane Number (RON) and yield. A design of experiments was conducted to examine the effects of key input variables, including reactor temperatures and the hydrogen-to-hydrocarbon (H₂/HC) ratio, through 256 simulations. Various data fitting methods, including Response Surface Methodology (RSM), Radial Basis Function Network (RNLOOCV), and Artificial Neural Networks (ANN), were applied to describe process behavior. The Akaike Information Criterion (AIC)-optimized ANN model demonstrated the best performance, offering a balanced approach between accuracy and complexity. A sensitivity analysis revealed that increasing reactor temperatures improves RON but reduces yield due to enhanced cracking reactions. The H₂/HC ratio had a minimal impact on RON and yield, primarily serving to limit catalyst coke formation. Optimization using a genetic algorithm confirmed that optimal RON and yield can be achieved within specific temperature ranges. The results provide insights for enhancing CCR efficiency and refinery profitability.