Surrogate modeling for mixed refrigerant streams in the refrigeration cycle of an LNG plant

Given the importance of liquefaction processes in the LNG value chain, it is necessary to model the complexity of such process. A key stage is the mixed refrigerant (MR) cycle used to liquify the natural gas in the liquefaction plant. The MR refrigeration cycle consists of compressors and heat exchangers in different compression stages that affect the MR properties in terms of temperature and pressure. In this work, the use of surrogate models is addressed for the compressor's power consumption and efficiency formulations along with the heat exchanger's performance in terms of heat duty after each compression stage. A training data set containing 500 points is used for building the surrogates, while a testing data set of 500 points verifies their accuracy. The surrogates built herein are shown to be sufficiently accurate to be further employed in decision-making industrial applications such as simulation, optimization, and control.