Artificial neural network for pressure- concentration- temperature (P-C-T) curves of metal hydrides.
Simelane, Ziphezinhle
Simelane, Ziphezinhle
Citations
Altmetric:
Abstract
Hydrogen shows very interesting features for its use on-board applications as fuel cell vehicles. This paper presents the modelling of a tank with a metal hydride alloy for onboard applications, which provides good performance under ambient conditions. The metal hydride contained in the tank is AB2 + AB5 mixture. A two-dimensional model has been performed for the refuelling process (absorption) and the discharge process (desorption). For that, individual models of hydrogen gas behavior has been modelled. The model has been developed under Matlab / Simulink© environment. Finally, individual models have been integrated into a global model, and simulated under ambient conditions. With the aim to analyse the temperature influence on the state of charge and filling and emptying time, other simulations were performed at different temperatures. The obtained results allow to conclude that this alloy offers a good behaviour with the discharge process under normal ambient conditions. This thesis investigate the use of metal hydrides for applications including hydrogen
storage for fuel cell vehicles and metal hydride hydrogen compressors. There is increasing demand for the development of a low carbon emission economy. Continuing growth in population coupled with ever-increasing fears over the environmental and health effects of climate change has meant that reducing our dependency on fossil fuels is one of the greatest challenges over the coming decades. The use of hydrogen as an energy vector may form part of the solution to this mounting energy challenge (Pickering, 2013). The prediction for initial discharge capacity of AB2 + AB5 based alloy was tested by employing the simulated annealing method in artificial neural network. Metal Hydrides are currently studied for the purpose of being used for hydrogen storage and compression in various stationary and mobile hydrogen and in fuel cell applications. Equilibrium of hydrogen absorption and desorption were studied by PCT diagram. ANN modeling tool was used to obtain relations between pressure, temperature, and concentration of a metal hydride. The aim of this study was to develop faster calculation algorithm for pressure-concentration isotherms of metal hydrides (AB2 + AB5) mixture. The curves were first plotted using Lototskyy model; where the model parameters were obtained by the fitting of experimental PCT data for the AB2 and AB5 components but took significant time. The application of ANN as a faster tool for modelling fuel cells for hydrogen storage was considered. Each dataset (absorption and desorption) containing 407 points, 285 (70%) were used for training, 61 (15%) for validation and the remaining 61 (15%) for testing. The training was performed with Levenberg-Marquardt function using backpropagation algorithm. The default performance criteria used in the assessment of the training and testing efficiency were the Mean Square Error (MSE) and goodness of fit (R2). A satisfactory fit was obtained for configuration ANN [2-14-1-1] (14 hidden neurons) and further improved to
ANN [2-18-1-1] (18 hidden neurons). The model had an overall MSE of 0.4311 and R2 of 0.9939 for absorption and an overall MSE of 0.3311 and R2 of 0.9981 for desorption. It was observed that the ANN model adequately reproduces faster PCT modelling results and was able to calculate one of three parameters (equilibrium pressure, concentration, and temperature) in 0.07s which was 2.5 – 5 times faster than Lototskyy model. The increase in number of hidden neurons allows to improve accuracy of the PCT curve representation. The incorporation of the developed neural network in the heat and mass transfer model allowed to significantly increase speed of simulations.
Description
Submitted in partial fulfilment of the requirements for the degree Master of Engineering: Chemical Engineering (MECE17) in the Department of Chemical and Metallurgical Engineering within the Faculty of Engineering and Built the Environment at the Tshwane University of Technology.
Date
2023-01-01
Journal Title
Journal ISSN
Volume Title
Publisher
Tshwane University of Technology
Research Projects
Organizational Units
Journal Issue
Keywords
Artificial neural networks, Fuel cell, AB2 and AB5, Metal hydrides, Modeling, PCT curves, Absorption, Desorption, Lototskyy model