The greenhouse gases emitted from the excessive use of fossil fuels severely impact the global climate, and many effects are already being observed. Therefore, transitioning to a climate-neutral society is crucial, and hydrogen will play an essential role in this transition.

A promising technology for the production of green hydrogen is alkaline water electrolysis. However, for it to be competitive with conventional fossil fuel-based production, this technology needs to be optimised to increase efficiency, reduce costs and facilitate scalability. One approach to achieving this is using a flow through the electrolysis cell. This flow provides better temperature control, ensures a faster removal of gas bubbles and supplies new reactants, increasing efficiency and allowing larger and thinner cells to be built. This can be realised without the need for additional energy by using the buoyancy of the produced gas bubbles to induce the flow. This is called natural recirculation.

This research investigates how the flow rate and velocity of the recirculating electrolyte depend on the geometry and operating conditions of the cell. Several models have previously been made to predict this. However, they have different methods and results, and a general method of modelling natural recirculation has not yet been accepted. An experimental setup consisting of an alkaline electrolysis cell that uses natural recirculation has been designed and built to assess the validity of these models.

First, the recirculation flow rate was measured for multiple electrode-wall distances and current densities. These experiments showed that increased current density leads to higher recirculation flow rates. The relationship can be approximated with a power function with an exponent ranging from 0.3 to 0.6. Larger electrode-wall distances initially also lead to higher recirculation flow rates until a distance of 5.4 mm. Beyond this distance, the effects of backflow cause a slight decrease in the recirculation flow rate.

The pressure drop in the downcomer was measured as a function of flow rate. This function could then be inserted into the models, and the predicted recirculation flow rate was compared to the experimental data. Although the 1D models did not account for backflow and did not accurately estimate the pressure drop in the riser, they showed better results than the 2D models. The 2D models attempted to include backflow but overestimated its effect as it is challenging to predict its behaviour. Currently, the laminar 1D model most accurately predicts the recirculation flow rate.

The pressure drop in the downcomer was controlled using a valve, and the recirculation flow rate was measured to investigate the effects of backflow. Even though further research is required to understand the effects of backflow fully, this experiment did show that backflow reduces the recirculation flow rate.

Finally, the bubble plume and backflow were filmed using a high-speed camera, and PIV was used to analyse the behaviour of the backflow and determine the plume width.

To summarise, this research studied the relationship between the recirculation flow rate and the electrode-wall distance and current density. The experimental results were then used to assess the validity of different models.