Application of Neural Networks for Assistance Systems in Wastewater Treatment

Optimisation of the Biological Treatment Stage

Abstract

Berliner Wasserbetriebe (BWB) has set itself the goal of becoming climate-neutral by 2030 and supporting the energy transition. There is great potential for optimising wastewater treatment in the biological treatment stage, which has the highest energy consumption at around 60%. In the "ANNA" cooperation project, the aeration control of the biological treatment stage was analysed for the first time using artificial neural networks (ANN) in a selected process line at the Münchehofe wastewater treatment plant. The aim of the project was to test a real-time assistance system available on the market for reducing energy consumption while simultaneously complying with the limit values under operating conditions. ANNs were used to model the process in order to predict the effluent concentration of the test line. The ANNs were trained with physically relevant historical data. The model-based predictions were used to calculate and implement suggestions for target oxygen concentrations or target air volumes. The assistance system was implemented in stages. In the first step, the use of the assistance system was tested manually. The recommendations for action implemented by the operating personnel led to a reduction in air consumption of approx. 10 %. At the same time, an increase in the ammonium concentration (NH4-N) in the effluent of the test line (2.2 mg/l) compared to the reference line (1.6 mg/l) by 0.6 mg/l. During the test phase, the wastewater treatment in the test and reference line was increased by 10 % in some cases. It was observed that the recommendations for action enabled precise load-based aeration control. The assistance system also recognised rain events at an early stage, whereby the aeration was increased and then reduced again in good time, depending on the freight impact. Based on the positive results, the assistance system was put into automatic operation in order to investigate its influence on the optimisation potential and process quality.