Optimization of activated sludge process in wastewater treatment system using explainable neural network

In this paper, we proposed DO(Dissolved Oxygen) neural network model and DO Control of activated sludge process in wastewater treatment system. Explainable neural network was utilized to decide water qualities which have a much influences in DO biological operation. These water qualities was to be inputs of DO neural network. Also, in regulations, effluent COD, T-N, T-P, pH, SS are hourly to transmitted in Korea Environment Corporation. If these data are exceed the standard, the penalty is given. So, these data are very exact and is controlled by operators critically. So, these data is to be inputs of DO neural network model. DO neural network model is to be utilized for optimal DO set-point which is controlled by blower. As one blower is connected to several aeration tanks, it is difficult to control DO in each aeration tank. Each aeration tank has 2-4 DO sensors which have different values. These are problems in automatic DO control. We also propose practical control solution by valve control logic and DO sensors calibration. The validity of the method is proved by applying to the DO neural network model of activated sludge process which was developed in previous research. The result show that the performance of the proposed model was improved in comparison of previous fuzzy model and conventional neural network models. Also, applicability is proved by field test of DO control in real activated sludge process in wastewater treatment system. In the future, it will be more effective in saving of blower power if this methodology is connected to control of blower valves. © The Korean Institute of Electrical Engineers