Editing Estimation of greenhouse gas and odour emissions from cold region municipal biological nutrient removal wastewater treatment processes

Record Editor

Template

Publication
Version 1.0

Advanced Options

Target Template

Overwrite record

Record Title

Publication Additional Information Download

Section 1: Publication

Publication Type

Authorship

Title

Year

Publication Outlet

DOI

ISBN

ISSN

Citation

Abstract

Rising human populations and ever-increasing demand for potable water result in increased municipal wastewater production. The collection, treatment, and management of municipal wastewaters include energy-intensive processes leading to the generation and emission of greenhouse, potentially toxic, and odorous gases. The main goal of this thesis was to advance knowledge of greenhouse gas (including carbon dioxide, CO2; methane, CH4; and nitrous oxide, N2O) and smelly compound (including ammonia, NH3; and hydrogen sulphide, H2S) emissions from typical municipal wastewater treatment plants (MWTPs) to accurately describe their emission rate estimates (EREs) using operating parameters. This research included laboratory and field assessments of greenhouse gas (GHG) and odour emissions in conjunction with monitored operating parameters. Laboratory-scale reactors simulating open-to-air treatment processes including primary and secondary clarifiers and anaerobic, anoxic, and aerobic reactors, were used to monitor gas EREs using wastewater samples taken from the analogous MWTP processes in winter and summer seasons. The Saskatoon Wastewater Treatment plan (SWTP) is a state-of-the-art biological nutrient removal (BNR) type MWTP and a Class IV treatment facility in Canada which was selected as a case study given its highly variable seasonal temperatures from −40 °C to 30 °C and its geographic location near the University of Saskatchewan. The experimental results were then used to develop a variety of novel machine learning models describing gas EREs with further optimization of operating parameters using genetic algorithm (GA). Studied machine learning models were artificial data generation algorithms (including generative adversarial network, GAN) and data-driven models (including artificial neural network, ANN; adaptive network-based fuzzy inference systems, ANFIS; and linear/non-linear regression models). To my knowledge, this is the first application of GAN used for MWTP modelling purposes. Results indicated that anaerobic digestion EREs averagely reached 4,443 kg CH4/d, 9,145 kg CO2/d, and 59.7 kg H2S/d. In contrast, GHG and odour ERE variabilities given ambient temperature changes were more noticeable for open-to-air treatment processes such that the winter EREs were 45,129 kg CO2/d, 21.9 kg CH4/d, 3.20 kg N2O/d, and insignificant for H2S and NH3. The higher temperature for the summer samples resulted in increased EREs for CH4, N2O, and H2S EREs of 33.0 kg CH4/d, 3.87 kg N2O/d, and 2.29 kg H2S/d, respectively, and still insignificant NH3 emissions. However, the CO2 EREs were reduced to 37,794 kg CO2/d, and interestingly, NH3 emissions were still negligible. Overall, the aerobic reactor was the dominant source of GHG emissions for both seasons, and changes in the aerobic reactor aeration rates (in reactor) and BNR treatment configurations (from site) further impacted the EREs. The integration of  field monitoring data with data-driven models showed that the ANN, ANFIS, and regression models provided reasonable EREs using: (1) volatile fatty acids, total/fixed/volatile solids, pH, and inflow rate for anaerobic digestion biogas generations; and (2) hydraulic retention time, temperature, total organic carbon, dissolved oxygen, phosphate, and nitrogen concentrations for aerobic GHG emissions. However, when both model accuracy and uncertainty were considered there appears to be a compromise between these parameters with no model having simultaneously both high accuracy and low uncertainty. Additionally, and interestingly, virtual data augmentation using GAN was found to be a valuable resource in supplementation of limited data for improved modelling outcomes. GA was also coupled with the data-driven models to determine optimal operating parameters resulting in either GHG emission maximization given biogas could be beneficial for energy generation or GHG emission minimization given the aerobic reactor is an open-to-air process that can impact nearby residential neighbourhood air quality. The current study provides a hybrid methodology of mathematical modelling and experiments that can be used to accurately estimate and optimize the GHG and odour EREs from other MWTPs in Canada and worldwide.

Plain Language Summary