Articles & Issues

Conflict of Interest: In relation to this article, we declare that there is no conflict of interest.

This is an Open-Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/bync/3.0) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

Articles in press

강화학습을 이용한 천연가스 액화 공정 최적화에 관한 연구

A Study on Optimization of Natural Gas Liquefaction Process Using Reinforcement Learning

Jieun Lee¹ Kyungtae Park^1†

¹Sookmyung Women’s University Department of Chemical & Biological Engineering, ¹숙명여자대학교 화공생명공학부

In Press, Journal Pre-proof, Available online 1 February 2025

Abstract

본 연구에서는 강화학습 방법론 중Deep Q-Network(DQN)와 Advantage Actor-Critic(A2C)알고리즘을 이용하여 천연가스 액화공정 중 단일혼합냉매 공정을 최적화하고 각 알고리즘에 따른 에너지 소모량 결과를 유전 알고리즘(Genetic algorithm, GA)을 통한 최적화 결과와 비교 분석하였다. 그 결과 DQN 최적화 결과가 A2C보다 낮은 에너지 소모량을 보였으며 학습 시간은 A2C 알고리즘이 짧은 것을 확인하였다. 그러나 GA와 비교분석 결과 GA 최적화 결과가 가장 좋았으며, 강화학습을 통한 공정의 최적화를 위해 연속적인 변수를 다루는 행동 지정에 대한 연구가 필요함을 제시하였다.

In this study, Deep Q-Network and Advantage Actor-Critic algorithms among reinforcement learning methodologies were used to optimize the single-mixed refrigerant process for a natural gas liquefaction. And optimization results using these algorithms were compared with the results of genetic algorithm(GA). The results showed that the optimization results using the DQN algorithm had lower energy consumption than A2C, and the learning time was shorter for the A2C algorithm. However, the comparison analysis with the genetic algorithm(GA) showed that the GA had the best performance, suggesting that research on specifying actions that deal with continuous variables is necessary for optimizing the process through reinforcement learning.

Keywords

Liquefied natural gas Single mixed refrigerant process Reinforcement Learning Optimization Deep Q-Network Advantage actor-critic algorithm