Modelling COD And DO Concentration By Using Adaptive Neuro-Fuzzy Inference System (ANFIS)

Dissolved oxygen (DO) & COD is a parameter frequently used to evaluate the water quality on different rivers. The aim of the present study is to investigate applicability of artificial intelligence techniques such as ANFIS (Adaptive Neuro-Fuzzy Inference System) in water quality DO & COD prediction for the case study, Mahi river at Khanpur in Thasara Taluka of Kheda District in Gujarat State, India. The proposed technique combines the learning ability of neural network with the transparent linguistic representation of fuzzy system. ANFIS models with various input structures and membership functions are constructed, trained and tested to evaluate efficiency of the models. Statistical indices such as Root Mean Square Error (RMSE), Correlation Coefficient (R), Coefficient of Determination (R2) and Discrepancy Ratio (D) are used to evaluate performance of the ANFIS models in forecasting DO & COD. ANFIS model is used for the estimation of DO & COD concentration. INTRODUCTION Using the surface waters for the purpose of suitable and safely requires the determination of the water quality. To determine water quality is very important issue for drinking and irrigation water, and many other purposes. A mathematical model is a set of mathematical expressions, relationships and logic rules that mimic the behavior of a physical process. Modern artificial intelligence methods such as neuro-fuzzy systems can be used for forecasting. These methods provide fast, reliable and low-cost solutions. Another advantage of these methods is that they can handle dynamic, non-linear and noisy data, especially when the underlying physical relations are very complex and not fully understood. The purposes of this study are to investigate the applicability of ANFIS in predicting Water Quality parameters like COD , DO in the Mahi River. DO & COD are a parameter frequently used to evaluate the water quality on different rivers. To estimate COD & DO accurately, inclusion of all past data is essential in this ANFIS model. In the present study to develop ANFIS model, past discharge, pH, temperature, EC, suspended solids, DO and COD of study area ANFIS models with various input structures and membership functions are constructed, trained and tested to evaluate efficiency of the models. Statistical indices such as Root Mean Square Error (RMSE), Correlation Coefficient (R), Coefficient of Determination (R2) and Discrepancy Ratio (D) are used to evaluate performance of the ANFIS models in estimation of COD & DO concentration. STUDY AREA The study area of Mahi river basin is located near village Khanpur in Thasara taluka of district Kheda in Gujarat State, India. The latitude and longitude of the study area are 22o 53’ N and 73o 13’E, respectively. The type of bed of river is rocky covered with sand. The basin is comprised of two sub-basins: Mahi upper sub basin (65.11% of total basin area) consisting of 41 watersheds and Mahi lower sub basin (34.89% of total basin area) consisting of 22 watersheds. The Mahi river and its tributaries constitute an inter-state river system flowing through the states of Madhya Pradesh, Rajasthan and Gujarat. Mahi river is comprised of several tributaries on both the banks, viz. Som, Anas, Panam and others. For this study, monthly discharge, pH, temperature, EC, suspended solids, DO & COD and data are collected for estimation of DO & COD of Mahi River. METHODOLOGY Adaptive Network-Based-Fuzzy Inferences System (ANFIS) approach was employed in this study. The ANFIS architecture consists of 5 layers such as input layer, fuzzification layer, inferences process, defuzzification layer, and summation as final output layer. Typical architecture of ANFIS is shown by Figure 1. Figure 1: Typical Architecture of ANFIS In above architecture the process flows from layer 1 to layer 5. It is started by giving a number of sets of crisp values as input to be fuzzyfied in layer 1, passing through infer-