Modeling and optimization of direct urea-hydrogen peroxide fuel cell using the integration of artificial neural network and bio-inspired algorithms

In this study, the performance of direct urea-hydrogen peroxide fuel cells (DUHPFCs) was modeled and optimized for the first time by artificial intelligence techniques. Accordingly, an artificial neural network (ANN) model was developed to describe the DUHPFC's voltage relying on basic designing and operating parameters (i.e., anode catalyst properties, urea concentration, KOH concentration, temperature, and feeding flow rate). A two-hidden layer-ANN with 7-10-6-1 topology using the Levenberg-Marquardt algorithm, logistic sigmoid function, and a training data proportion of 80 % was the best suitable model. A mean squared error (MSE) and R-value were estimated to be 0.51 × 10−4 and 0.9993, respectively, indicating a good prediction capability. Subsequently, the bio-inspired algorithms (BIAs), including particle swarm optimization (PSO) and genetic algorithm (GA), were employed to identify the optimal process parameters. Similar optimum results were obtained by both algorithms, although ANN-PSO performed faster than ANN-GA. When Ni0.2Co0.8/Ni-foam was used as the anode catalyst, the calculated maximum power density was 45.6 mW/cm2 under urea concentration of 1.4 M, KOH concentration of 6.2 M, temperature of 70 °C, and flow rate of 5.9 mL/min. However, the electrical energy recovery was only 2.6 % under such optimal conditions, suggesting that other factors, especially novel urea-electrooxidation catalysts, should be investigated further.