Predicting hydrogen production from formic acid dehydrogenation using smart connectionist models

Hydrogen is a promising clean energy source that can be a promising alternative to fossil fuels without toxic emissions. It can be generated from formic acid (FA) through an FA dehydrogenation reaction using an active catalyst. Activated carbon-supported palladium (Pd/C) catalyst has superior activity properties for FA dehydrogenation and can be reused after deactivation. This study focuses on predicting the FA conversion to H2 (%) in the presence of Pd/C using machine learning techniques and experimental data (1544 data points). Six different machine learning algorithms are employed, including random forest (RF), extremely randomized trees (ET), decision tree (DT), K nearest neighbors (KNN), support vector machine (SVM), and linear regression (LR). Temperature, time, FA concentration, catalyst size, catalyst weight, sodium formate (SF) concentration, and solution volume are considered as the input data, while the FA conversion to H2 (%) is the target value. Based on the train and test outcomes, the ET is the most accurate model for the prediction of FA conversion to H2 (%), and its accuracy is assessed by root mean squared error (RMSE), R-squared (R2), and mean absolute error (MAE), which are 3.16, 0.97, and 0.75, respectively. In addition, the results reveal that solution volume is the most significant feature in the model development process that affects the amount of FA conversion to H2 (%). These techniques can be used to assess the efficiency of other catalysts in terms of type, size, weight percentage, and their effects on the amount of FA conversion to H2 (%). Moreover, the results of this study can be used to optimize the energy, cost, and environmental aspects of the FA dehydrogenation process.