Exploring Anfis Application Based on Actual Data from Wastewater Treatment Plant for Predicting Effluent Removal Quality of Selected Major Pollutants

Abstract

The aim of this study was to investigate the effectiveness of the Adaptive Neuro-Fuzzy Inference System (ANFIS) model in predicting the removal of major pollutants in a wastewater treatment plant. The parameters were screened using Principal Component Analysis (PCA) and Orthogonal Experiments (OE) were carried out to improve the accuracy of ANFIS. The predictions of pollutant removal were obtained through a normalization transformation, ANFIS for effluent value predictions and then indirect predictions were calculated based on ANFIS predicted results. It was found that there was no one-to-one correspondence between predicted and actual values of pollutants and that the predicted values often fell within a certain range, indicating large errors. To improve accuracy, restrictive statements were added but this did not affect the predicted values. This led to the hypothesis that these restrictive statements must penetrate deeply into the hidden layer without altering the input layer or the predictive logic of the hidden layer and the output values of the output layer.

Keywords: ANFIS, Indirect Prediction, Orthogonal Experiment, Wastewater Treatment Plant, ANFIS predictive logic