Application of Artificial Neural Network to Predict TDS Concentrations of the River Thamirabarani, India

Abstract

River water quality modeling is of prime importance in predicting the health of the rivers and in turn warns the human society about the future possibility of water problem in that area. Total dissolved solids is a prominent parameter used to access the quality of the river water. In our current study, artificial neural networking models have been developed to predict the concentrations of total dissolved solids of the river Thamirabarani in India. Neural Network toolbox of the MATLAB 2017 application was used to create and train the models. Monthly data from year 2016 to 2019 at four different sites near Thamirabarani river were procured from Tamilnadu pollution control board. Many artificial neural network architectures were built and the best performing architecture was selected for this study. With several parameters such as pH, chloride, turbidity, hardness, dissolved oxygen as input and the total dissolved solids as output parameter, the model was trained for many iterations and a final architecture was arrived which predicts the futuristic TDS concentrations of Thamirabarani in a more accurate manner. The predicted and the expected values were very close to each other. The root mean square error (RMSE) values for the selected stations such as Papanasam, Cheranmahadevi, Tirunelveli and Punnaikayal were 0.565, 0.591, 0.648 and 0.67 respectively.