Artificial Neural Network Ensembles in Hybrid Modelling of Activated Sludge Plant

Keskitalo, Jukka; Leiviskä, Kauko

doi:10.1007/978-3-319-11313-5_60

Cited by 3 publications

(2 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The co-application of machine learning and mechanistic models in serial or parallel configurations have already been demonstrated in the wastewater sector. For instance, studies have shown that artificial neural network (ANN) models trained with the residuals of ASM-type models can enhance the prediction and generalization capabilities of the ASM models for highly daily variable influent pollutant concentrations and flow-rates that are usually observed at WWTPs (Fang et al, 2010;Keskitalo and Leiviskä, 2015). Additionally, detailed simulations of variables that are not conventionally monitored in WWTPs (i.e.…”

Section: Limitations and Future Researchmentioning

confidence: 99%

A decade of nitrous oxide (N2O) monitoring in full-scale wastewater treatment processes: A critical review

et al. 2019

View full text Add to dashboard Cite

Direct nitrous oxide (N 2 O) emissions during the biological nitrogen removal (BNR) processes can significantly increase the carbon footprint of wastewater treatment plant (WWTP) operations. Recent onsite measurement of N 2 O emissions at WWTPs have been used as an alternative to the controversial theoretical methods for the N 2 O calculation. The full-scale N 2 O monitoring campaigns help to expand our knowledge on the N 2 O production pathways and the triggering operational conditions of processes. The accurate N 2 O monitoring could help to find better process control solutions to mitigate N 2 O emissions of wastewater treatment systems. However, quantifying the emissions and understanding the long-term behaviour of N 2 O fluxes in WWTPs remains challenging and costly. A review of the recent full-scale N 2 O monitoring campaigns is conducted. The analysis covers the quantification and mitigation of emissions for different process groups, focusing on techniques that have been applied for the identification of dominant N 2 O pathways and triggering operational conditions, techniques using operational data and N 2 O data to identify mitigation measures and mechanistic modelling. The analysis of various studies showed that there are still difficulties in the comparison of N 2 O emissions and the development of emission factor (EF) databases; the N 2 O fluxes reported in literature vary significantly even among groups of similar processes. The results indicated that the duration of the monitoring campaigns can impact the EF range. Most N 2 O monitoring campaigns lasting less than one month, have reported N 2 O EFs less than 0.3% of the N-load, whereas studies lasting over a year have a median EF equal to 1.7% of the N-load. The findings of the current study indicate that complex feature extraction and multivariate data mining methods can efficiently convert wastewater operational and N 2 O data into information, determine complex relationships within the available datasets and boost the long-term understanding of the N 2 O fluxes behaviour. The acquisition of reliable full-scale N 2 O monitoring data is significant for the calibration and validation of the mechanistic models of-describing the N 2 O emission generation in WWTPs. They can be combined with the multivariate tools to further enhance the interpretation of the complicated full-scale N 2 O emission patterns. Finally, a gap between the identification of effective N 2 O mitigation strategies and their actual implementation within the monitoring and control of WWTPs has been identified. This study concludes that there is a further need for i) long-term N 2 O monitoring studies, ii) development of data-driven methodological approaches for the analysis of WWTP operational and N 2 O data, and iii) better understanding of the trade-offs among N 2 O emissions, energy consumption and system performance to support the optimization of the WWTPs operation.

show abstract

Section: Limitations and Future Researchmentioning

confidence: 99%

A decade of nitrous oxide (N2O) monitoring in full-scale wastewater treatment processes: A critical review

et al. 2019

View full text Add to dashboard Cite

show abstract

“…In Alavi et al (2014), a neural network model is incorporated within the mass balance model of an adiabatic fixed-bed reactor to reduce the computation load. Keskitalo & Leiviskä (2015) combined several artificial neural networks into ensembles to account for the difference between the mechanistic model outputs and the real process values. Chaffart & Ricardez-Sandoval (2018) proposed a hybrid model for the simulation of thin film deposition.…”

Section: Introductionmentioning

confidence: 99%

A comprehensive hybrid first principles/machine learning modeling framework for complex industrial processes

Sun

Yang

Wang

et al. 2020

Journal of Process Control

View full text Add to dashboard Cite

Modeling Based on LS-SVM Method and Predicting Dewatering Ratio in Sludge Process

Luo¹,

Qiao²,

Liao³

2018

Proceedings of the 2017 3rd International Forum on Energy, Environment Science and Materials (IFEESM 2017)

View full text Add to dashboard Cite

Abstract. The sludge produced by treating the urban sewage needs thickening and dewatering and it is important for improving environment to reduce the quantity of discharging sludge by increasing dewatering ratio in sludge process. This paper proposes a novel method that can model least squares support vector machines (LS-SVM) by Soft-sensing to meet the characteristics of multivariable, nonlinearity, large time delay, and time-varying in sludge thickening and dewatering process，and predict the dewatering ratio. The results indicate that the proposed method can not only improve prediction accuracy, but also efficiently get high dewatering ratio by controlling thickening chemicals comparing to the cases using different methods.

show abstract

Artificial Neural Network Ensembles in Hybrid Modelling of Activated Sludge Plant

Cited by 3 publications

References 30 publications

A decade of nitrous oxide (N2O) monitoring in full-scale wastewater treatment processes: A critical review

A decade of nitrous oxide (N2O) monitoring in full-scale wastewater treatment processes: A critical review

A comprehensive hybrid first principles/machine learning modeling framework for complex industrial processes

Modeling Based on LS-SVM Method and Predicting Dewatering Ratio in Sludge Process

Contact Info

Product

Resources

About