Neural network-inspired performance enhancement of synthetic natural gas liquefaction plant with different minimum approach temperatures

Abstract In this study, a state-of-the-art neural network algorithm (NNA) was explored to improve the overall competitiveness of the single mixed refrigerant (SMR) process for synthetic natural gas (SNG) liquefaction. The NNA approach is inspired by the functions of biological and artificial neural networks. This is the first study to implement the NNA approach, especially to find the energy and cost-saving opportunities in the SMR SNG liquefaction process. Optimized SNG liquefaction processes were analyzed and compared to a recently published SNG liquefaction process optimized by a single-solution-based vortex-search approach. The robustness of the NNA was evaluated against different values of the minimum internal temperature approach (MITA). It is observed that the SMR process corresponding to MITA values of 1.0 °C and 3.0 °C consumes approximately 16% and 2.4% less energy, respectively, compared with the base case. The exergy efficiencies of the optimized process with MITA values of 1.0, 1.5, 2.0, 2.5, and 3.0 °C are 18.52, 13.45, 11.98, 9.60, and 2.24 % higher than the base case, respectively. An economic analysis in terms of total capital investment (TCI) and TAC was also performed. The analysis showed high TCI savings of 3.3% for an MITA value of 3.0 °C compared to the base case, whereas savings in TAC were 6.6%, 7.2%, 8.1%, 7.1%, and 2.7% respectively for MITA values of 1.0, 1.5, 2.0, 2.5, and 3.0 °C. This study will help practitioners design cost-effective liquefaction technologies that would provide clean and affordable energy.