Deep learning for pH prediction in water desalination using membrane capacitive deionization

Abstract

The pH of a solution has a large influence on the ion removal efficiency of the membrane capacitive deionization (MCDI) process, an electrochemical ion separation process. We developed a convolutional neural network linked with a long short-term memory (CNN-LSTM) model based on an artificial intelligence algorithm to predict the effluent pH of MCDI, as effluent pH is difficult to predict using conventional numerical modeling. The model accurately predicted effluent pH (R2>0.998) based on the analysis of five input variables (current, voltage, influent conductivity and pH, and effluent conductivity) under standard operating conditions of MCDI using either constant-current or constant-voltage conditions. The developed model predicted effluent pH using only limited input variables, current and voltage, with high accuracy (R2>0.997). Thus, the CNN-LSTM model can be used in practical applications as only the current and voltage of MCDI cells are often monitored in field applications.