AUTHOR=Ahmad Ashfaq , Qadeer Kinza , Naquash Ahmad , Riaz Fahid , Hasan Mudassir , Qyyum Muhammad Abdul , Lee Moonyong TITLE=Particle Swarm-Assisted Artificial Neural Networks for Making Liquefied Natural Gas Processes Feasible Under Varying Feed Conditions JOURNAL=Frontiers in Energy Research VOLUME=Volume 10 - 2022 YEAR=2022 URL=https://www.frontiersin.org/journals/energy-research/articles/10.3389/fenrg.2022.917656 DOI=10.3389/fenrg.2022.917656 ISSN=2296-598X ABSTRACT=Natural gas (NG) has been widely recognized cleanest fossil fuel compared to other fossil fuels. The reserves of NG are typically located in remote areas and its conditions and composition vary from area to area. The NG from these areas is transported in the form of liquefied natural gas (LNG) which is highly complex process. The LNG process is designed for fixed NG conditions. However, for uncertain and varying NG conditions, the designed LNG process may not perform well. Considering this problem, a study is presented that makes the LNG processes feasible under uncertain NG conditions through artificial intelligence approach rather than conventional robust optimization approaches. This is a first study in this research area that is aimed at to train an artificial neural network (ANN) by using particle swarm optimization (PSO) algorithm as a learning method. The developed PSO-ANN model is then used to predict decision variables of single mixed refrigerant (SMR) LNG process for its feasible design under varying NG conditions. The correctness of predicted set of decision variables (NG conditions) is verified by inputting it to the Aspen Hysys. The output of SMR-LNG process is the overall power at constrained MITA value. i.e., 1.0 ≤ MITA ≤ 3.0. The prediction results of PSO-ANN model are compared with the classical ANN back-propagation learning method. The success rate of the proposed PSO-ANN model was found to be 80%.