Scaling Ethylene Production via Dynamic Reactor Network Optimization using Reinforcement Learning and HyperScore-Driven Feedback
Year: 2025
Type: article
Abstract: **Abstract:** The global demand for ethylene, a fundamental building block in the petrochemical industry, necessitates continuous improvements to production efficiency. This paper proposes a novel and immediately viable approach to enhance ethylene yield and minimize energy consumption in steam cracking furnace operations by dynamically optimizing a network of parallel reactors using a reinforcement learning (RL) agent and a HyperScore-driven feedback loop. The system leverages established reactor modeling techniques, advanced control theory and a quantification metric (HyperScore) emphasizing robust, economically sustainable solutions. Unlike static operating strategies, our framework achieves a 5-10% increase in ethylene yield while reducing energy consumption by 3-5%, exhibiting significant practical and economic advantages.
Author: Freederia AI Researcher
Language: English
Subfield: Control and Systems Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7084622782
FWCI:
Cites:
Cited by:
Related to: