Deep learning-based optimization of a microfluidic membraneless fuel cell for maximum power density via data-driven...

Area of Science:

• Energy Conversion and Storage
• Computational Modeling
• Microfluidics

Background:

• Membraneless microfluidic fuel cells (MMFCs) offer potential for portable power but require performance optimization.
• Traditional optimization methods can be computationally intensive and time-consuming.

Purpose of the Study:

• To develop and validate a hybrid artificial neural network (ANN) and genetic algorithm (GA) approach for optimizing MMFC performance.
• To significantly reduce the computational resources and time required for MMFC design optimization.

Main Methods:

• A validated 3D multiphysics model (R ² =0.976) was used to generate training data for the ANN.
• An ANN model (R ² =0.999) was developed for rapid performance prediction, with an execution time of 0.041 s.
• The ANN model was integrated with a GA to efficiently search for optimal MMFC design parameters.

Main Results:

• The ANN-GA method identified optimal parameters (e.g., microchannel dimensions, temperature, cell voltage) for maximum power density.
• The optimized MMFC achieved a maximum power density of 0.263 mWcm^-2 (current density of 0.852 mAcm^-2).
• The ANN-GA predicted maximum power density closely matched numerical calculations, with a difference of only 0.766%.

Conclusions:

• The hybrid ANN-GA approach provides a highly accurate and computationally efficient method for optimizing MMFC performance.
• This method accelerates the design and development cycle for advanced microfluidic fuel cell technologies.
• The findings demonstrate the potential of deep learning for enhancing energy device optimization.