Parameters optimization in Chlor-Alkali ion-exchange membrane electrolysis process with artificial neural network

Abstract

Since ion-exchange membrane electrolysis cell has developed for producing Chlor-Alkali products. New higher efficiency and lower consumption technology are released from licensors yearly, which made the process correlation deviate from the original design. The machine learning is used with “Neural Network Fitting Tool (nftool)” in MATLAB. To find a correlation between 5 inputs consisting of current density (CD, KA/m2), operation day (DOL, day), feed brine flow rate (QFB, m3/h), feed caustic flow rate (QHD, m3/h), cell temperature (T, degC) and one output which is cell voltage (CV, V). Datasets were collected from the plant information management system “exaquantum” historian database. The result is shown only on CD as the predictor gives RMSE at 0.0167 V. In 2 predictors, DOL as the second gave RMSE at 0.0065 V, which can conclude that DOL (or clogging factor) has the most impact on CV increasing. In 3 predictors, T as the third gave RMSE at 0.0043 V, from controlled temperature set point change. Developed ANN optimization model can be used to optimize controlled parameters to predict suitable CV after a long run (high DOL) or to compare electrolysis effectiveness by regressing CV for comparing at the same condition.