Contents lists available at ScienceDirect Journal of Water Process Engineering journal homepage: www.elsevier.com/locate/jwpe Model identification and control of chlorine residual for disinfection of wastewater Rehab I. Khawaga a , Nabil Abdel Jabbar a, ⁎ , Sameer Al-Asheh a , Mohamed Abouleish b,c a Department of Chemical Engineering, American University of Sharjah, Sharjah, United Arab Emirates b Department of Biology, Chemistry and Environmental Sciences, American University of Sharjah, Sharjah, United Arab Emirates c Gulf Environments Research Institutes (GERI), American University of Sharjah, PO Box 26666, Sharjah, United Arab Emirates ARTICLE INFO Keywords: Chlorination Breakpoint Optimization ANN Fuzzy ABSTRACT In this study, Artificial Neural Network modelling (ANN) was applied to forecast the chlorination behaviour in the secondary wastewater effluent containing ammonia and nitrite. A novel control scheme of chlorination disinfection was then developed through an integration of ANN model with fuzzy logic control (FLC) to optimize the chlorination process by minimizing its cost and maximizing its efficiency while operating within the plant’s budget. The developed FLC platform was applied to a local wastewater treatment plant and was able to improve disinfection quality and reduce chlorine gas consumption by 18%. 1. Introduction Physical and chemical wastewater treatment processes are complex by nature and are often not thoroughly understood [1]. Modelling and controlling such processes turn accordingly as a challenge. Advanced control techniques are rarely found in wastewater treatment facilities, while process control is mainly achieved via either general heuristics or the operators’ knowledge and experience. One of the very last steps and yet the most sensitive amid the wastewater treatment processes, due to its potential acute toxicity effects if not properly applied, is the che- mical disinfection which is typically accomplished through chlorina- tion. Wastewater disinfection aims at providing protection for humans from exposure to pathogenic water born microorganisms [2]. In the absence of ammonia or nitrogenous compounds, water chlorination would have been a lot easier. The reactions involving chlorine, am- monia, and other nitrogen-based compounds produce unique class of compounds known as chloramines. Using role of thumbs to control the appropriate amount of chlorine dosage tolerates an inefficient disin- fection that fails in many cases to conform to the recent regulations of reducing the formation of disinfection by products [1]. Artificial neural networks, (ANN), known as neural networks, are a kind of artificial intelligence, based- to an extent- on the general structure of the brain. The extremely branched system of networks copies the manner in which the brain keeps information by altering the weights of the synapses that link node layers. Neural networks have the ability to learn the pattern from the data, which facilitate mapping complicated input-output connections [3]. ANNs have the capacity to operate a wide range of non-linear interactions and data trends [4–6]. These features have made the implementation of ANNs a favourable option to solve several engineering problems; including forecasting, prediction, as well as process control [7]. Recently, the potential of the ANN methodologies have become increasingly observed in the water treatment processes; this has led to the development of a number of model applications in quality of water and demand forecasting, treatment, and distribution. Various models have been designed in the context of water quality control such as forecasting of chlorine residues in water distribution system [8], sali- nity of source water [9], color of the raw water [7] and water demand [10]. ANN was also successfully employed in modelling physical pro- cesses like alum and polymer dosing for coagulation purposes [1], lime dose softening [11], as well as turbidity and color removal [7]. Modelling chlorine concentrations in the distribution system has been proved as a successful application of ANNs. Rodriguez et al. [3], made use of an ANN model to predict chlorine residuals in a Severn Trent Water Ltd (U.K) circulation system, and noted the great potential of ANNs to replicate the dynamics of chlorine decay within a circulation system. Gibbs et al. [12] made use of three different data-driven techniques, of which two were based on ANN modelling, in order to forecast re- siduals at two main locations in the Hope Valley Distribution System in Southern Australia (2006). Their results showed the capacity of ANNs to foretell chlorine concentrations in a complicated distribution system, https://doi.org/10.1016/j.jwpe.2019.100936 Received 4 April 2019; Received in revised form 29 August 2019; Accepted 30 August 2019 ⁎ Corresponding author. E-mail address: nabdeljabbar@aus.edu (N. Abdel Jabbar). Journal of Water Process Engineering 32 (2019) 100936 2214-7144/ © 2019 Elsevier Ltd. All rights reserved. T