Joint majorization of waterworks and secondary chlorination points considering the chloric odor and economic investment in the DWDS using machine learning and optimization algorithms

Abstract

The traditional methods of increasing the chlorine disinfectant dosage in the drinking water distribution system (DWDS) to control microorganisms and improve the safety of drinking water quality are subjected to several challenges. One noticeable problem is the unpleasant odor generated by chlorine and chloramines. However, the generally proposed chlorine dosage optimization model ignores the chloric odor distribution in the DWDS. This study proposes a comprehensive multi-parameter water quality model and aims to balance the trade-offs between: (i) minimize the flavor profile analysis (FPA) degree of the chloric odor produced by chlorine and chloramines in the DWDS, and (ii) minimize the economic investment (chlorine dosage and operation cost). EPANET and back propagation (BP) network integrated with the Borg algorithm were employed as innovative approaches to simulate the chlorine, chloramines, and chloric odor intensity in the DWDS. Moreover, the application of the multi-parameter model was demonstrated in a real-world DWDS case study. 0.5 mg-Cl₂/L (mg/L) chlorine at 8 secondary chlorination points was added to the DWDS as an optimized chlorine dosing scheme considering the olfactory and financial objective functions simultaneously. When switching to a superior water source, the FPA of the chloric odor in DWDS increased by a maximum of 1.4 at most if the initial chlorine dosage remained as before. To avoid the occurrence of chloric odor and also control the residual free chlorine (residual chlorine) at a suitable value, the initial and secondary chlorine dosages were optimized to 0.4 mg/L and 0.3 mg/L, respectively. Under this condition, the initial chlorine dosage was reduced by 50% compared to the original operation scheme in City J, China, the qualification rate of the residual chlorine reached 97.2%, basically consistent with that before water source switching, and the chloric odor intensity of the DWDS was controlled below FPA 3.4.

Keywords: BP neural network; Borg algorithm; Chloric odor intensity; Drinking water distribution system (DWDS); Water source switching.