AI-based decision support tools for short-term forecasting and aeration control in wastewater treatment plants

Introduction Effective operation of wastewater treatment plants (WWTPs) requires timely and informed decisions to maintain process stability and energy efficiency. This paper presents two artificial intelligence (AI) tools designed to support WWTP operators in short-term decision-making.Materials The tools were developed and evaluated using historical operational data from the Alcantarilla municipal wastewater treatment plant in Murcia, Spain. The dataset included measurements of influent water quality parameters and aeration process-related variables collected over a 14-year period (2012–2026).Methods The first tool utilizes Long-Short Term Memory (LSTM) based recurrent neural networks (RNNs) to generate daily forecasts of influent water quality parameters, for 24 h and 48 h horizons. The second tool replicates the plant’s existing aeration control logic while incorporating additional control functionalities and short-term trend estimation for key aeration-related parameters, such as ammonium and nitrate. This second tool implements the control logic as a state‑machine model with operational states and uses an LSTM-based RNN for trend estimation.Results The infulent forecasting tools achieved a prediction accuracy of R2 ≥ 0.8 for most water quality parameters across both the 24h and 48h horizons. The replica of the aeration control logic showed a 98% exact coincidence with the plant’s actual logic, while the additional controls indicated potential for reducing blower activity. Furthermore, the trend estimation accuracy for most parameters exceeded an R2 of 0.8.Conclusions The integration of these two AI tools provides a robust predictive framework that enhances situational awareness, supports proactive control actions, and contributes to more stable and energy-efficient WWTP performance.