Prediction of hydrogen generation from perhydro-dibenzyltoluene empowered with machine learning

Abstract

The perhydro-dibenzyltoluene (H18-DBT) exhibits promising potential as a viable option for hydrogen production purposes. There are several important features for hydrogen generation predictions including dosage of H18-DBT, temperature, concentration of catalyst, and stirring speed. This study presents the Hydrogen Production Prediction System Empowered with Machine Learning (HPPSML), which employs the Scaled Conjugate Approach (SCG) to predict the quantity of hydrogen generated from H18-DBT. The dataset is classified into three categories based on the percentage of produced hydrogen: low class, medium class, and high class. The results elucidate that the accuracy of the proposed HPPSML is higher (96.65 %) for the high class whereas it is 93.20 % and 89.80 % for the low and medium class respectively. The overall performance of the proposed HPPSML using the SCG approach was found to have an accuracy of 89.80 % and a misclassification rate of 10.2 %.