WATER QUALITY MODELLING OF RIVER PERIYAR USING ARTIFICIAL NEURAL NETWORK AND ADAPTIVE NEURO-FUZZY INFERENCE SYSTEM

Abstract

Water is requisite on earth for life to survive. The study area of the present work is the river Periyar, the longest river in Kerala, India. The overall objective of the present work is to compare the accuracy and performance of the artificial neural network (ANN) and adaptive neuro-fuzzy inference system (ANFIS) in the training and prediction of dissolved oxygen (DO) concentration and biochemical oxygen demand (BOD) in the river. The models were used to examine secondary data from the past for four water quality metrics generated at five monitoring stations located along the river Periyar to predict DO concentrations and BOD. The performance of the models was examined using root mean square error (RMSE) and coefficient of correlation (R) values. This study revealed that ANN is superior to ANFIS in most scenarios. In most situations, ANFIS appears to get the superiority in training, however, ANN seemed to have the dominant position in testing and validation. The inputs having the highest relevance in the case of DO prediction were identified to be nitrate and phosphate concentrations, whereas total solids and COD were seen to be the most impactful in the case of BOD forecasting employing sensitivity analysis.