Insights on features' contribution to desalination dynamics and capacity of capacitive deionization through machine learning study

Parameter optimization in designing a rational capacitive deionization (CDI) process is usually performed to achieve both high electrosorption capacity and speed. This necessitates a clear understanding of system behavior and discriminating the features' role on desalination capacity from its dynamic. Machine learning (ML) modeling is widely employed for understanding various systems' behavior as an alternative for physics-based extrapolation models. Herein, various ML models are implemented with reasonable accuracies to unveil CDI electrode and operational features' local and global impacts on equilibrium desalination capacity, speed, and duration. Electrode specific surface area and electrolyte ionic concentration are determined to play the most significant roles in CDI by synergistically enhancing desalination capacity and speed. Increasing electrode micropore volume is detected to inhibit desalination and make ion removal sluggish. According to the established models, electrode nitrogen content extends desalination capacity without improving its dynamic. In addition, unlike the complex impacts from electrodes oxygen content on desalination capacity, it is shown that electrode oxygen content clearly elongates desalination time. This study demonstrates the strong abilities of the established ML models in explaining the underlying complex mechanisms in the CDI process.

Automated summary

This study utilized machine learning models to reveal the impacts of electrode and operational features on desalination capacity, speed, and duration in capacitive deionization process. The findings suggest that electrode specific surface area and electrolyte ionic concentration play significant roles, while electrode oxygen content clearly elongates desalination time.