Machine Learning-Assisted Optimization of Magnetic Field Effects on Hydrogen Production in Water Electrolysis​​​​​​

Hydrogen production through water electrolysis has been a research topic of concern in recent years. However, hydrogen bubbles generated during the electrolysis process tend to adhere to the electrode plates for prolonged periods, resulting in ineffective electrolysis space and reduced efficiency. This study uses magnetohydrodynamics theory(MHD) and places magnets parallel to the external sides of the electrode plates to generate Lorentz forces on the electrode plates, thereby inducing flow in the electrolyzer to accelerate bubble detachment from the electrode surface, enhancing electrolysis efficiency. It has been proven that adding a magnetic field during water electrolysis can create vortices that effectively increase hydrogen production and improve electrolysis efficiency. Nevertheless, the optimal size ratio between the added magnetic field and the electrode plates for maximum electrolysis efficiency remains to be explored.

This study uses machine learning and Latin hypercube sampling methods to sample circular electrode plates and magnets within a fixed range of 30mm electrode spacing. Then a kriging simulator analysis was performed and the commercial software COMSOL was used for simulation. The simulation results show that when the edge effect of a single magnet generates a Lorentz force at a fixed power, an electrode plate with a diameter of 60mm and a magnet with a diameter of 50mm will produce a higher average Lorentz force density, which helps the bubble to move from the electrode plates are detached, improving overall electrolysis efficiency. However, experimental current density measurements show that when the diameters of both the magnet and the electrode plate are 50mm, the highest Lorentz force density is produced, resulting in the maximum current density between the electrode plates, approximately 6% higher than the 60mm-50mm electrode-magnet configuration. This shows that the edge effect of a single magnet acts on a smaller electrode plate area, forming a larger Lorentz force gradient and driving the peripheral bubble motion more effectively. In contrast, as the electrode plate area increases, the effect of the single edge effect is limited, and a higher average Lorentz force density is required to drive more bubbles from the electrode surface.

The optimal configuration of the electrode plate-to-magnet ratio that produces the highest average Lorentz force density was determined through COMSOL software simulations. The results show that a magnet slightly smaller than the electrode plate produces the maximum average Lorentz force density. The relationship between the optimal average Lorentz force density, electrode plate area, and current density will be further verified through subsequent experiments.