Achieving High Power and Energy Efficiency for Microfluidic Fuel Cells with Flow-through Porous Electrodes at the Desired Cell Voltage and Fuel Flow Rate with Deep Learning-based Multi-target and Criteria Optimization

Achieving both high power output and energy efficiency poses a significant challenge in optimizing membraneless microfluidic fuel cells (MMFCs) due to the inevitable trade-off between fuel utilization and power density. This study aimed to introduce an effective method to enhance power density (P), fuel utilization (Fu), and exergy efficiency (Ee) in an MMFC with flow-through porous electrodes through deep learning-based multi-target and criteria surrogate optimization (MTCO). We developed an experimentally validated 3D numerical model to train a deep-learning artificial neuron network (DNN), achieving high prediction accuracy (R2 = 0.999). An MTCO model was constructed with the latest non-dominated sorting genetic algorithm (NSGA-III) and a comprehensive multiple-criteria decision-making model (PROMETHEE-II). This model identified an optimal combination of input parameters that could achieve high P (121.266 mW/cm2), Fu (98.445%), and Ee (52.483%). Notably, compared with the base model, this model significantly improved the performance of the MMFC, with increases of up to approximately 48% in output parameters at a high cell voltage and various fuel flow rates. The MTCO approach could allow flexible optimization strategies, allowing users to prioritize a desired optimization target with specific criteria according to their personal goals.