Modeling of hydrogen liquefaction process parameters using advanced artificial intelligence technique

Abstract

The main subject of this work is the application of advanced artificial intelligence (AI) techniques to accurately predict the parameters of the hydrogen liquefaction process. This study employs a comparative analysis of the most reliable AI techniques: Artificial Neural Networks (ANNs), Adaptive Neuro-Fuzzy Inference Systems (ANFIS), support vector machines (SVM), perturbed chain statistical associated fluid theory (PCSAFT) equation of state and Hybrid technique based on the combination of ANN model and perturbed chain statistical associated fluid theory (AI-PCSAFT). The training and validation strategy focuses on using a validation agreement vector, determined through linear regression analysis of the predicted versus reference outputs, as an indication of the predictive ability of the studied models. A dataset collected from scientific papers containing hydrogen liquefaction process data was utilized in the modeling process. The modeling strategy is performed using the temperature (T), pressure (P), and mass flow rate (m) as input parameters and the stream energy (E) as output parameters.

The results show high predictability of the optimized ANFIS model followed by AI-PACSAFT model compared to ANN, SVM models and PCSAFT equation of state with coefficient of correlation (R) and absolute relative deviation (AARD) equal to 0.9988 and 0.98% respectively.