Modelling, simulation and multi-objective optimization of industrial hydrocrackers

Abstract

Hydrocracking is a catalytic cracking process for conversion of complex feedstock like heavy vacuum gas oils to valuable middle distillates in hydrogen-rich atmosphere at elevated temperature and pressure. Developments in catalyst and reactor design influence the design and operation of the hydrocracker. This requires choosing design/operating conditions to maximize valuable products and achieve the best resource utilization. Multiple objectives are relevant in such cases. The availability of powerful computational resources and robust evolutionary techniques like Genetic Algorithm (GA), have led to a revolution in the field of multi-objective optimization (MOO). This work describes modeling of industrial hydrocrackers through first principles, data based and hybrid modeling techniques followed by its optimization by GA for single and multiple objectives. First principles model (FPM) adopts discrete lumped model approach to characterize feed and reaction mixture to pseudocomponents. Data based models (DBMs) are developed using industrial data and artificial neural networks. Hybrid models are developed by combining FPMs with DBMs to overcome their limitations. Real-coded non-dominated sorting genetic algorithm (NSGA-II) has been successfully employed for both fine tuning the FPM of hydrocracker and performing MOO of the hydrocracker. FPM, DBM and various hybrid models are compared for their prediction performances before choosing and using one of them for optimization. A generic multi-platform, multi-language environment is also developed for integration of HYSYS with NSGA-II. This platform is useful for simulating and optimizing industrial processes for multiple objectives.