Using artificial neural network to optimize hydrogen solubility and evaluation of environmental condition effects

Abstract

Hydrogen is a clean energy and has many applications in petroleum refining, glass purification, pharmaceuticals, semiconductors, aerospace applications and cooling generators. Therefore, it is very important to store it in various ways. One of the new and cheap methods to store hydrogen is storing in the brine groundwater. In this method, the hydrogen gas is injected into the brine, in which storing capacity has a direct relationship with the pressure, temperature and salt concentration of the saltwater. In the present study, an artificial neural network (ANN) was used to estimate and optimize the hydrogen solubility (HS) in the saltwater with conventional best algorithms such as the feedback propagation, genetic algorithm (GA) and radial basis function. The optimization is implemented based on available experimental data bank based on the variation of the pressure, working temperature and salt concentration. The results and assessments of different optimization ANN algorithm show that the GA has the most usable and accurate estimation and prediction for HS in the saltwater. Also, the amounts of the relevancy coefficient (Rc) that correspond to the sensitivity of HS on the input parameters demonstrate that the salt concentration and pressure have the minimum and maximum Rc, respectively. That is, the least and most effect on the output values.