Neural network-based ammonium aeration control of wastewater treatment plant

Abstract

Due to the expensive operation and stringent effluent requirements of wastewater treatment plants, the wastewater treatment operator has been forced to find an alternative to improve the current control strategy, particularly for those using conventional activated sludge systems. The goal of this research is to design a controller capable of reducing aeration energy while improving effluent quality. The objectives are met through a technique known as ammonium-based aeration control (ABAC). In this study, neural network (NN) – ABAC was designed and proposed for the Benchmark Simulation Model No. 1. The simulation results were compared to those of the proportional-integral (PI) controller and PI ABAC control configurations. During the NN training, a dropout layer was added to improve NN generalization. The simulation results show that the dropout layer successfully reduced the complexity of the NN while maintaining a good mean squared error and regression value. When compared to PI, the proposed NN – ABAC is more effective in terms of energy efficiency by lowering aeration energy by up to 23.86%, improving effluent quality by up to 1.94%, and lowering the total overall cost index by up to 4.61%. The findings suggest that the NN – ABAC has the potential to improve the performance of the activated sludge system.