An Integrated Multi-Unit and Multi-Objective Optimization Approach for Enhancing the Efficiency of Ethylene Distillation Process

For complex multi-Unit ethylene distillation process, This work proposes a strategy referred to as knowledge and data-driven hybrid modeling and multi-objective cooperative optimization (KDHM-MOCO). First, the distillation tower group is divided into several units for individual modeling, and a cooperative optimization model is established based on the relationships between tower group variables. In this regard, a three-hidden-layer backpropagation neural network (3h-BPNN) is employed to train proxy models for the individual unit operations. The training datasets are derived from both real-world manufacturing processes and simulated environments. Ultimately, the multi-objective grey wolf optimization (MOGWO) algorithm is employed to achieve the most advantageous solution set, taking into account both product quality and energy cost, as well as the corresponding optimal operating parameters for reflux and reboiler steam rates.