Machine Learning-Based Modeling of Capacitive Deionization Efficiency: A Comparative Study of Artificial Neural Network and Decision Tree Approaches

Capacitive deionization (CDI) is a promising technique for desalination, particularly for brackish water. Various operational parameters, including salt concentration, water chemistry, applied potential, and throughput, influence the performance of a CDI system. Optimizing these parameters to achieve a highly efficient CDI system remains a significant challenge. This study aims to develop a machine learning (ML) model capable of providing real-time predictions of desalination efficiency by incorporating diverse experimental parameters. The proposed ML models enable the identification of optimal operational settings, improving system control and overall efficiency. In this work, two different Machine learning (ML) models of artificial neural network and decision tree (DT) have been implemented to predict the salt concentration of CDI with reasonable accuracy based on three input parameters. This work offered the models that estimated the salt removal concentration with good accuracy and precision, with R2 values of 0.99 for both ANN and DT. Both models achieved high prediction accuracy, whereas the DT model displayed better accuracy. The findings of this study pave the way for a more efficient, dependable, and economically viable desalination system in the future.