Modeling and optimizing of an actual municipal sewage plant: A comparison of diverse multi-objective optimization methods

Abstract

The activated sludge process of an actual municipal sewage treatment plant was systematically modeled, calibrated, and verified in this study. Identified multi-objective optimization (MOO) methods were employed to optimize the process parameters of the validated model, and the optimal MOO algorithm was obtained by comparing Pareto solution sets. The optimization model consisted of three key evaluation indicators (objective functions), which are the average effluent quality (AEQ), overall cost index (OCI), and total volume (TV) of the biochemical tank, along with 12 more process parameters (decision variables). Three optimization algorithms, i.e., adaptive non-dominated sorting genetic algorithm III (ANSGA-III), non-dominated sorting genetic algorithm II (NSGA-II), and particle swarm algorithm (PSO), were adopted using MATLAB. The comparison of these algorithms demonstrated that the ANSGA-III algorithm had better Pareto solution sets under the triple objective optimization, and the effluent quality of COD, TN, NH4+-N, and TP after optimization decreased by 2.22, 0.47, 0.13, and 0.02 mg/L, respectively. Additionally, the simulated AEQ was reduced by 13% compared to the original effluent, and the OCI and TV decreased from 21,023 kWh d−1 and 17,065 m3 to 20,226 kWh d−1 and 16,530 m3, respectively. The reported ANSGA-III algorithm and the proposed multi-objective method have a promising ability for energy conservation, emission reduction, and upgrading of municipal sewage treatment plants.