Plant-wide assessment of alternative activated sludge configurations for biological nutrient removal under uncertain influent characteristics.

This study presents an extensive plant-wide model-based assessment of four alternative activated sludge (AS) configurations for biological nitrogen (N) and phosphorus (P) removal under uncertain influent loads and characteristics. Zeekoegat wastewater treatment plant (WWTP) in South Africa was chosen as case study due to its flexible design that enables operation in four different AS configurations: 3-stage Bardenpho (A2O), University of Cape Town (UCT), UCT modified (UCTM), and Johannesburg (JHB). A metamodeling based global sensitivity analysis was performed on a steady-state plant-wide simulation model using Activated Sludge Model No. 2d with the latest extension of physico-chemical processes describing the plant-wide P transformations. The simulation results showed that the predictions of effluent chemical oxygen demand (COD), N and P using the proposed approach fall within the interquartile range of measured data. The study also revealed that process configuration can affect: 1) how influent uncertainty is reflected in model predictions for effluent quality and cost related performances, and 2) the parameter rankings based on variance decomposition, particularly for effluent phosphate, sludge disposal and methane production. The results identified UCT and UCTM as more robust configurations for P removal (less propagated uncertainty and less sensitivity to N load) in the expense of incomplete denitrification. Moreover, based on the results of Monte-Carlo based scenario analysis, the balanced SRT for N and P removal is more sensitive to influent load variation/uncertainty for the A2O and JHB configurations. This gives a more operational flexibility to UCT and UCTM, where a narrow SRT range can ensure both N and P removal.

Integrated Model for Understanding N2O Emissions from Wastewater Treatment Plants: A Deep Learning Approach.

This study aims to demonstrate the application of deep learning to quantitatively describe long-term full-scale data observed from wastewater treatment plants (WWTPs) from the perspectives of process modeling, process analysis, and forecasting modeling. Approximately, 750,000 measurements including the influent flow rate, air flow rate, temperature, ammonium, nitrate, dissolved oxygen, and nitrous oxide (N2O) collected for more than a year from the Avedøre WWTP located in Denmark are utilized to develop a deep neural network (DNN) through supervised learning for process modeling, and the optimal DNN (R2 > 0.90) is selected for further evaluation. For process analysis, global sensitivity analysis based on variance decomposition is considered to identify the key parameters contributing to high N2O emission characteristics. For N2O forecasting, the proposed DNN-based model is compared with long short-term memory (LSTM), showing that the LSTM-based forecasting model performs significantly better than the DNN-based model (R2 > 0.94 and the root-mean-squared error is reduced by 64%). The results account for the feasibility of data-driven methods based on deep learning for quantitatively describing and understanding the rather complex N2O dynamics in WWTPs. Research into hybrid modeling concepts integrating mechanistic models of WWTPs (e.g., ASMs) with deep learning would be suggested as a future direction for monitoring N2O emissions from WWTPs.