AI-Accelerated Catalyst Selection and Operating Conditions Optimization for Hydrocracking

Hydrocracking is a critical refining technology for upgrading heavy oils, where catalyst selection and operating condition adjustment are crucial for enhancing catalytic performance and product quality. Currently, this matching process relies heavily on the experimental method, which is time-consuming and resource-intensive. Data-driven methods provide a solution for this problem. However, the application of data-driven methods demands specialized data science expertise. This work utilized GPT-4 as an AI assistant to facilitate the development and interpretation of data-driven models for hydrocracking catalysis, establishing the relationship between catalyst properties, feedstock characteristics, operating conditions, and hydrocracking tail oil properties. Gradient-weighted class activation mapping was employed to identify key factors influencing the properties of tail oil. Based on the model’s prediction, the impacts of replacing catalysts and adjusting operating conditions on tail oil properties were explored. The framework in this study is expected to reduce experimental iterations by 60%, highlighting the potential of AI in optimizing hydrocracking processes and offering valuable insights for industrial applications.