Study On Lng Process Optimization Using Artificial Neural Network

Growing demand for liquefied natural gas(LNG)has intensified research on process optimization of liquefaction technology so as to cater for the increasing demand at enhanced production efficiency and reduced costs.However,the optimization of an LNG process is often set back by nonlinearity resulting from inter-dependence of the process variables.To this end a two step scheme is proposed for the optimization of Single Mixed Refrigerant(SMR)LNG process targeting energy minimization while upholding high liquefaction ratio.Process data collected from an LNG plant in northern China was used to model the SMR process in Aspen Hysys software.The first optimization step involved the application of quadratic scheffe mixture regression to model the dependence of compressor power consumption and natural gas liquefaction ratio on composition of the mixed refrigerant(MR).The models demonstrated great significance and accuracy with P-values less than 0.0001 and R~2values greater than 0.99 and were simultaneously optimized to obtain 17.7%savings in compressor duty and 2.7%increase in liquefaction ratio relative to the base case.More so,the exergy efficiency of the optimized process improved by 26.9%.Then,in order to ensure robust performance amidst process uncertainties,the MR composition was fixed at the optimum values and artificial neural network was used to model the dependence of compression power consumption and liquefaction ratio on certain SMR process conditions.The network predicted compression power with an accuracy score of 99%and liquefaction ratio at an accuracy score of 97.6%,with normalized RMSE values of 0.0045 and 0.0268 respectively.Sensitivity analysis carried out on the test data revealed that the flow rate of the mixed refrigerant stream is critical to the performance of an SMR process.Hence,it must be carefully chosen as a trade-off between low energy demand and high productivity in order to ensure robust process performance.The findings from this study will be useful for the design and optimization of SMR LNG processes.