Prediction of Air Blower Flow Setpoint in Wastewater Treatment Plants Based on the LSTM Model

Effective wastewater treatment is key to protecting the environment, wastewater treatment plants are responsible for treating domestic sewage in built-up areas. The process of wastewater treatment is a long-term, complex, non-linear reaction, consuming chemicals and electrical energy to satisfy effluent standards. In the Anaerobic-Anoxic-Oxic process, Dissolved Oxygen (DO) needs to be accurately controlled to reduce treatment costs. Air blowers provide oxygen via the aeration system, and DO adjustment by changing the air blower flow setpoint. Most previous studies have considered single variables and linear control, and can not take comprehensive consideration of all effecting parameters. In this study, the Long Short-Term Memory model is used to predict the air blower flow setpoint, taking into account the effects of multiple variables. In addition, the Long Short-Term Memory (LSTM) model has lower Mean Absolute Error (MAE = 21.00), Root Mean Square Error (RMSE = 28.83) and Mean Absolute Percentage Error (MAPE = 0.47%) than the Support Vector Regression (SVR) and Feedforward Neural Networks (FNN).