Composition‐Sensitive Optimization of Mixed Refrigerant Blends for Hydrogen Liquefaction and Storage Using Prediction Based Clustering Method

The growing hydrogen demands have instigated massive developments in the hydrogen liquefaction technology. Hydrogen fuel is potentially seen as a critical aspect toward developing economies and requires substantial improvement in liquefaction mechanisms. The current study explores the hydrogen liquefaction‐storage through mixed refrigerants with the goal of increasing overall efficiency of the process. A hybrid method is applied which makes use of Pearson correlation, prediction modeling, and k‐means clustering to optimize the process operational parameters. The optimization process showed that an optimal mixed refrigerant composition, MR‐6 (Combo 6), which provides an energy consumption of 7.6 kWh/kg, hydrogen storage of 93.2%, and a COP of 3.9. The given composition had strong positive correlations to C3 ( R = 0.884) and n‐C5 ( R = 0.866) with COP. However, a high negative correlation is established with the COP with N 2 ( R = −0.888) and Ne ( R = −0.874). The cluster 2 gave the best performance results, especially for Combination 6. Further, the ANFIS model has shown that the Gaussian function membership function yielded the best predictions (RMSE 0.15, MAE 0.05), whereas the Trapezoidal function has produced the greatest errors (RMSE 0.25, MAE 0.09) for the liquefaction‐based outcomes. The optimum composition is found to be MR‐6 with the composition C1(10%), C2(0%), C2*(10%), C3(5%), n‐C4(5%), i‐C5(5%), n‐C5(5%), R14(10%), He (10%), Ne (10%), N 2 (5%), R23(15%).