Artificial intelligence technique based performance estimation of solid oxide fuel cells

Abstract

In General, a larger number of experimental data would be required for optimizing the performance efficiency of the electrodes in solid oxide fuel cells (SOFC), as it was likely to involve many complex chemical and physical reactions. Therefore, in this study, we have attempted the artificial intelligence (AI) technique for analyzing the performance of the anode and cathode electrodes of SOFC under various key parameters that normally influences the efficiency of the electrodes. Since the cost of SOFC elements is expensive, computational trials with the aid of AI would not only be curtailing the preparation cost but also ensures less time consumption. Among the variety of AI methods available in the literature, the Support Vector Machine (SVM) appears to be one of the excellent and effective machine learning techniques. Hence, we have developed a SVM model and predicted the maximum current density and power density of Nickel oxide-samarium-doped ceria (NiO-SDC) composite anode and La0.6Sr0.4Co0.2Fe0.8O3-δ (LSCF) cathode electrodes of SOFC. The input parameters such as stack temperature, supply voltage were supplied to the SVM model and obtained maximum current density and power density as output and it decides the working performance of the electrodes. We have also predicted that the maximum current density and power density of 1160 mA cm2 and 225 mW cm2 respectively at 800 °C. To validate the results as predicted from SVM, we have experimented using Yttrium stabilized zirconia electrolyte-supported single cell with our prepared NiO-SDC anode and LSCF cathode which exhibited a maximum current and power density of 1170 mA cm2 and 227 mW cm2 at 800 °C using H2 as fuel. It has been found that that the theoretical predicted data current and power densities of the electrodes from the SVM approach reasonably agrees well with the experimental results.