Using machine learning algorithms improve the performance of both fuel cells and lithium-ion batteries.

Fuel cells generally produce electricity by converting hydrogen and oxygen into the water for sustainable energy solutions. If the fuel cell is powered by pure hydrogen, it can be up to 80 percent efficient. This means it converts 80 percent of the energy substance of the hydrogen into electrical energy. Researchers in this field, with the development of machine learning, can now explore designs for the microstructure of fuel cells and lithium-ion batteries. The performance of both devices relies majorly on their microstructures – pores or holes inside their electrodes. These microstructures can affect how much power fuel cells generate, and how quickly the lithium-ion batteries charge and discharge.

As the pores are quite small in size, studying them for cell performance has found challenging. In an effort to overcome this challenge, a team of researchers at the Imperial College London have devised a solution to study virtually. The team used a machine learning algorithm to explore possible designs for the microstructure of fuel cells and lithium-ion batteries, before running 3D simulations that assist them to make changes to improve cell performance.

In a statement, lead author Andrea Gayon-Lombardo, of Imperial’s Department of Earth Science and Engineering said, “Our technique helps us zoom right in on batteries and cells to see which properties affect overall performance. Developing image-based machine learning techniques like this could unlock new ways of analysing images at this scale.”

Machine Learning to Gain Fuel Cell Efficiency

To run 3D simulations to foresee cell performance requires a voluminous amount of data for an adequate statistical representation of the cell. However, obtaining large volumes of microstructural image data at the required resolution is difficult. Thus, the researchers trained their code to generate either much larger datasets that have all the same properties or create structures that would deliver an effective outcome in better-performing batteries.

In this way, they employed a new machine learning technique called deep convolutional generative adversarial networks (DC-GANs). They used this technique to generate 3D image data of fuel cell’s microstructure based on training data from nano-scale imaging performed a synchrotron. A synchrotron is a type of circular particle accelerator that works by accelerating charged particles (electrons) through sequences of magnets until they reach almost the speed of light.

Project supervisor at Imperial’s Dyson School of Design Engineering, Dr. Sam Cooper said, “Our team’s findings will help researchers from the energy community to design and manufacture optimized electrodes for improved cell performance. It’s an exciting time for both the energy storage and machine learning communities, so we’re delighted to be exploring the interface of these two disciplines.”

Through the study, researchers expect their models produce results that are currently feasible to manufacture optimized electrodes for next-generation fuel cells, as well as hope to implement their algorithms into manufacturing and designing.