Volume 8, Issue 1, January – 2023 International Journal of Innovative Science and Research Technology ISSN No:-2456-2165 IJISRT23JAN1267 www.ijisrt.com 1991 Prediction of Performance Efficiency for Wastewater Treatment Plant’s Effluent Biochemical Oxygen Demand Using Artificial Neural Network Samson Balogun Department of Civil Engineering, Faculty of Engineering, University of Abuja P.M.B 117, Abuja Nigeria. Toochukwu Chibueze Ogwueleka Department of Civil Engineering, Faculty of Engineering, University of Abuja P.M.B 117, Abuja Nigeria. Abstract:- This study investigated the application of an artificial neural network (ANN) to predict the performance efficiency of the Abuja-based Wupa WWTP, Nigeria using effluent 5-day biochemical oxygen demand (BOD5) as a performance indicator. Daily data of influent BOD5, pH, total dissolved solids, total suspended solids, chemical oxygen demand, total coliform, Escherichia coliform, and fecal coliform; and effluent BOD5 over a period of five years (2013 to 2017) for the Wupa WWTP was utilized for the plant’s performance efficiency. The four most reliable multilayer perceptron ANN (MLP- ANN) algorithms namely, Levenberg-Marquardt (LM) backpropagation resilient backpropagation, Quasi- Newton backpropagation, and Fletcher-Reeves conjugate gradient backpropagation were adopted; and the most appropriate model was selected following training, validation and testing by altering the number of neurons and activation functions in both the hidden and output layers. The model efficiency was determined using mean square error (MSE) and correlation coefficient (R 2 ). The ML algorithm with Logsig-Tansig activation pairing and architecture [8-1270-1] performed the best in terms of convergence time and prediction error, with MSE and R 2 values of 1.522 and 0.922, respectively. Also, it revealed that the selected ANN model predicted the effluent BOD5 with an overall correlation coefficient of 0.962; thus, demonstrating the efficacy of ANN models for accurate prediction of the Wupa WWTP performance. The novelty of this research is in evaluating the efficiency of the plant over the periods and determining the most precise ANN model for Wupa WWTP, Abuja, Nigerians a study which has never been carried out before now. Keywords:- Artificial Neural Network (ANN); Wastewater Treatment Plant (WWTP); 5-Day Biochemical Oxygen Demand (BOD5); Wupa WWTP; Multilayer Perceptron (MLP) I. INTRODUCTION The treatment and management of wastewater in our environment have increasingly gained attractive attention in the last decade, particularly in the face of the incessantly increasing volume of wastewater owing to population growth; rapid urbanization; increased agricultural; and industrial activities (Abba and Elkiran 2017; Arismendy et al., 2020; Alsulaili and Refaie, 2021). Wastewater treatment plants (WWTPs) are built to clean wastewater and convert it into eco-friendlier water which is released into the environment (Varkeshi et al., 2019). However, due to the wide fluctuation in the quality and quantity of untreated wastewater transported to the treatment plant, the operation of WWTPs can be difficult and challenging (Szeląg et al. 2017). Moreover, many treatment plants are constructed following the conventional activated sludge system which is allegedly riddled with inefficiencies associated with pollutant removal (Ogwueleka and Samson, 2020). In addressing the challenges of the conventional treatment systems, several alternative methods have been proposed, notable amongst which are the advanced oxidation processes (AOPs) (Deng and Zhao, 2015); nanomaterials (Adeleye et al., 2016); microalgae-activated sludge (MAAS) (Ogwueleka and Samson, 2020); microbial electrochemical system (Li et al., 2021). However, many of these emerging methods are still limited to pilot or laboratory scales and are yet to gain widespread practical applications due to a number of reasons such as the initial cost of installation, uncertainties with operations, adaptation and installation of new technologies, etc.; thus, there is still need to seek for means of attaining efficiency, even in the pre-existing installed treatment systems; which can be achieved by attaining and maintaining optimal conditions in WWTPs. Attaining optimal operational conditions in WWTPs even with conventional systems is possible and can be achieved with the use of models to predict the WWTP performance based on previous measurements of major plant parameters (Jami et al., 2012). An important parameter commonly utilized to examine the performance of WWTPs is the 5-day biochemical oxygen demand (BOD5) (Dogan et al. 2008; Araromi et al., 2018). BOD5 is an approximation of the quantity of biochemically degradable organic matter contained in a water sample, defined as the amount of oxygen necessary for the aerobic bacteria present in a sample to oxidize the organic matter to a stable organic form (Dogan et al., 2008). It is, however, difficult to measure, and requires five days for its determination (Dogan et al., 2008; Alsulaili and Refaie, 2021). Therefore, the determination of the output BOD5 of a WWTP as a performance index using predictive tools could achieve