A Predictive Exergy-Adaptive Genetic Algorithm for the Optimization of the DMR Liquefaction Process

Against the backdrop of increasing global energy demands and the need for enhanced efficiency in industrial processes, the dual mixed-refrigerant (DMR) natural-gas liquefaction process involves strong coupling among numerous operating parameters, and conventional optimization methods often fail to reach the global optimum. To address this challenge, we propose an intelligent optimization strategythe Predictive Exergy-Adaptive Genetic Algorithm (PEA-GA)which integrates exergy analysis with a machine-learning surrogate model. A detailed Aspen HYSYS simulation of the DMR process is developed to examine how feed-gas pressure and temperature, as well as precooling and deep-cooling parameters affect specific energy consumption (SEC) and exergy efficiency. By embedding an exergy-loss evaluation mechanism and employing a radial basis function (RBF) neural network to accelerate the search, key operating variables are optimized. Relative to the base case, the optimized scheme lowers SEC from 0.281 kWh kg^-1 LNG to 0.252 kWh kg^-1 and raises exergy efficiency from 31.62% to 35.45%. Minimum temperature differences in both heat exchangers shrink markedly, yielding better composite-curve matching. The results provide theoretical guidance and an engineering reference for energy-saving retrofits and intelligent operation of LNG facilities.