Machine Learning Surrogate-Assisted Genetic Algorithm for Energy Optimization in LDPE Tubular Reactors

Abstract

This work introduces a machine learning-driven optimization framework that integrates Genetic Algorithm (GA) with Gaussian Process Regression (GPR) to enhance energy efficiency in low-density polyethylene (LDPE) production. The optimization targets minimizing energy consumption, maximizing conversion (), and improving profitability by tuning key process variables, including FM, , , and . Results indicate that FM and exert the most significant influence on energy consumption, whereas and jacket temperature contribute minimally. The optimization achieved a minimum energy consumption of 603.357 kWh. While adjustments in and led to higher profitability, the overall profit remained negative, underscoring the need for further refinement. The proposed framework offers valuable insights into LDPE process optimization and demonstrates the potential of advanced machine learning-based methods for improving both energy efficiency and economic performance. Future research will focus on refining the optimization strategy and exploring cost-reduction pathways.