Neural network ammonia-based aeration control for activated sludge process wastewater treatment plant

Abstract

The paper proposes an improved effluent control for the operation of a biological wastewater treatment plant using a neural network ammonia-based aeration control. The main advantage of this control method is the simplicity and nonlinear approximation ability that beat the performances of the static-gain Proportional Integral (PI) controller. The trained neural network controller used the measured value of dissolved oxygen and ammonium in compartment 5 of the Benchmark Simulation Model No. 1 (BSM1) to regulate the oxygen transfer coefficient in compartment 5. The effectiveness of the proposed neural network controller is verified by comparing the performance of the activated sludge process to the benchmark PI under dry weather file. Simulation results indicate that Ntot,e, and SNH,e violations are reduced by 22% reduction for Ntot,e, and 4% for SNH,e. The significant improvement in effluent violation, and effluent quality index of the BSM1 confirms the advantage of the proposed method over the Benchmark PI. For future research, the method can also be applied in controlling the nitrate in activated sludge wastewater treatment plant.