Robust Wastewater Characterisation for the Development of a Biokinetic-Artificial Intelligence Hybrid Model to Reduce Nitrous Oxide Emissions

Nitrous oxide (N2O), is considered a potent and very harmful greenhouse gas (GHG); in addition, N2O is also considered to contribute to the depletion of the ozone layer in the stratosphere. With the global warming potential of N2O being as high as 298 times greater than that of CO2 on a 100-year time scale, identifying and mitigating the anthropogenic sources of N2O is crucial in curbing its harmful environmental effects.

In the past decade, wastewater treatment plants (WWTPs) are increasingly considered to be one of the potent sources of N2O, and therefore, advance wastewater treatment technologies and operational strategies to reduce the generation of the harmful N2O gas are being investigated. For that purpose, a clear understanding and representation of the process conditions in the system is necessary in order to investigate mitigation measures. One such approach that can be utilized is in the form of a biokinetic model, using widely known Activated Sludge Models (ASMs) that have been extended to include the N2O production pathways.

In parallel, Artificial Intelligence (AI) models have also been utilised in the prediction of key wastewater parameters including N2O in data-rich systems. However, in the N2O field, hybrid models comprising the biokinetic and AI models have not yet been developed. Such hybrid models could allow for the biokinetic model features to provide process insights while the AI models can enable (near) real-time estimation and control of the treatment processes in order to achieve the reduction of N2O emissions. In our present research, as a prelude to the development of the hybrid model, a biokinetic model is being built and calibrated for the Amsterdam West WWTP using the EnviroSim software, BioWin®.

In order to acquire the data necessary to input into the model, a comprehensive and intensive sampling campaign was conducted. The campaign’s duration was for 8 days. Flow proportional daily composite samples and diurnal sampling were taken for the raw influent and effluent wastewater where the following parameters were monitored – CODtotal, CODfiltered, BODtotal, BODfiltered, TKN, NH4, NO3, Total P, Ortho-PO4, ISS and TSS. The raw influent was also periodically monitored for pH, Alkalinity, Ca, Mg and Total Sulphide. The diurnal sampling was conducted on one weekday and weekend day. Numerous grab samples were also taken at various locations including the anaerobic, anoxic and aerobic zones of the bioreactor, sludge treatment lines and filtrate streams. The grab sampled data provide a sanity check for the outputs provided by the biokinetic model. With the execution of the comprehensive monitoring campaign, the sampling data can now characterise the common quality measurements (COD, TKN, TP, TSS, etc.) into their fractions as required for use in the biokinetic models and can now be mapped into the BioWin ASDM inputs. Additionally, Amsterdam West WWTP is a large plant, with a capacity of 1.1 million population equivalent. Therefore the logistics in conducting a monitoring campaign of such scale was complex, time consuming and labour intensive. Another key insight is that sampling during periods of normal and stable plant operation provides the most reliable results of wastewater characteristics. In addition, increasing the number of samples can help partially overcome the adverse impacts on sampling results from occasional period of unusual plant operation and control.

Source: Seshan, S. 2021. Robust Wastewater Characterisation for the Development of a Biokinetic-Artificial Intelligence Hybrid Model to Reduce Nitrous Oxide Emissions. Treatment and monitoring of Water, Reuse, Recycle & Recover. AIWW 2021

Image credits

Icon image: Wikimedia Commons - Wastewater