Adaptive neuro-fuzzy approach to sodium chlorate cell modeling to predict cell pH for energy-efficient chlorate production

Abstract

Sodium chlorate production is a highly energy-intensive electro-chemical process. The current efficiency is as low as 66.66% if the pH of the process is not controlled. Hence for energy-efficient sodium chlorate production, it is very important to predict the pH of the cell electrolyte. Though the analytical model of the chlorate cell is available, they are not suitable for pH prediction, as it requires a large number of parameters, which are not easily measurable. This study presents an adaptive neuro-fuzzy inference system model for predicting the pH of sodium chlorate cell from real plant data. This model uses fewer and easily measurable parameters for modeling such as flow rates of brine, HCl, NaOH, cell electrolyte temperature, DC load current, and pH of feed input. An analysis of the accuracy of six models with different combinations of these input parameters is conducted with different model settings for the type of membership functions, epoch number, and number of data sets. The model accuracy is evaluated and compared for six models using the statistical indicators, root mean square error (RMSE), and coefficient of determination (R2). Based on the results obtained, the ANFIS model with HCl flow rate, NaOH flow rate, cell electrolyte temperature, DC load current, and pH of feed as input parameter, is accurate in predicting pH with RMSE 0.1290 and R2 0.8816 at testing stage. The good agreement between the measured pH and estimated pH confirms that the model can be further used to design the controller.