Session: 01-05: AI for Energy Sustainability V

Paper Number: 131551

131551 - Machine Learning Based Assessment of the Air Quality Impacts From Natural Gas Production Facilities in Denton County, Texas 

Abstract: 

Advanced machine learning (ML) techniques were employed to interpret the relationship between natural gas production in the Barnett Shale region and changes in ambient air quality levels, particularly VOCs and NOx over a period spanning from 2000 to 2022. The Barnett Shale region, known for its extensive natural gas reserves, has witnessed significant industrial activities, primarily in the form of hydraulic fracturing and natural gas extraction over the past two decades. This study utilized ML algorithms, including time series forecasting and neural network models, to analyze and predict the impact of natural gas production on the observed air quality concentration data at the Denton Airport South site. Data used in this study were publicly obtained from the Texas Commission on Environmental Quality (TCEQ) and from the Texas Railroad Commission (RRC).

The ML models were trained using the comprehensive dataset of monthly averages of VOC and NOx concentration data, along with natural gas production volumes and observed meteorological data from the Denton Airport South site. By incorporating features such as production volumes, meteorological data, and pollution levels, the models provided robust forecasts of air quality impacts under different production scenarios. The neural network approach was effective in capturing the relationships and interactions between multiple variables.

Analyses conducted using Random Forest and XGBoost ML models revealed a good correlation between the gas production activities and observed air quality metrics. The model was fine-tuned through different hyperparameter tuning techniques to optimize within the computational limits. This approach aimed to improve the model's ability to capture the complex relationships between multiple atmospheric variables and natural gas production.

Notably, contrasting air quality levels were observed during the years 2014 and 2018, corresponding closely with fluctuations in natural gas production, as evidenced by peak VOC concentrations in 2014 during the highest production period and lower concentrations in 2018 amidst reduced production. Based on these years, where contrasting air quality levels were observed, the ML models predicted the VOC and NOx concentrations, aligning closely with the actual recorded values. The models were also used to simulate scenarios with varying production levels, offering insights into potential future air quality impacts.

Additionally, the study utilized regression analysis with ML enhancements to interpret temporal patterns and relationships between individual VOC species, NOx levels, and condensate production more accurately. The ML-augmented regression models further examined monthly concentrations of VOC and NOx alongside natural gas production and meteorological data. Strong correlations of VOC with condensate production were noted in 2014 (r=0.80) and 2018 (r=0.55). Additionally, regression analysis was used to interpret temporal patterns in the data, revealing a consistent and statistically significant relationship between the individual VOC species and NOx levels with condensate production (VOC: r=0.51-0.87 in 2014 and r=0.41-0.71 in 2018; NOx: r=0.39-0.59 in 2014 and r=0.30-0.68 in 2018). Analysis based on wind rose plots indicated dominant winds from the south-east and south-west directions influencing the air quality monitoring location. Numerous oil and gas wells are located along this direction from the monitoring site, and it was impacted by pollutant dispersion from proximate oil and gas facilities located in these source-rich regions. The findings from this study suggest that the observed VOC concentration in the area is substantially influenced by local natural gas production activities, thus underlining the need for targeted environmental control at these industrialized facilities to mitigate their impact on urban and regional air quality.

Presenting Author: Kuruvilla John University of North Texas

Presenting Author Biography: Kuruvilla John is a professor in the Department of Mechanical Engineering at the University of North Texas (UNT), Denton, Texas.

Kuruvilla received his B. Tech degree in chemical engineering in 1986 from Anna University in India. He then worked briefly for Asian Paints in Madras before moving to the United States for higher education. He earned his M.S. and Ph.D. degrees in chemical engineering from the University of Iowa, Iowa City, Iowa in 1989 and 1996, respectively.

He has worked as a visiting scientist at IBM’s Bergen Scientific Centre in Norway and as a research associate with the State University of New York at Albany and the New York State Department of Environmental Conservation. In 1995, he moved to Texas to start his academic career with Texas A&M University – Kingsville (TAMUK), where he rose to be a professor of environmental engineering. He also served as the associate dean and interim dean of the Frank H. Dotterweich College of Engineering at TAMUK.

In 2009, he joined the University of North Texas and served as the Associate Dean of Research and Graduate Studies for the College of Engineering until 2016. He helped with the development of new doctoral programs and oversaw a significant growth in the college’s research enterprise. UNT is currently a Tier One Carnegie classified doctoral university with the highest research activity. From 2016 through 2021, Prof. John served as the chair for the second largest department within the college of engineering at UNT.

Prof. John’s research interests are in the area of environmental sustainability with a focus on air quality monitoring, modeling and assessment. He has an active research portfolio and was instrumental in securing 43 research contracts, grants and projects worth over $15 million from various industries and funding agencies including National Science Foundation, Department of Energy, and Texas Commission on Environmental Quality among others. He has served as principal investigator and project director of a National Science Foundation (NSF) funded center for research excellence in science and technology.

As a mentor, Prof. John has supervised 57 M.S. students, 4 Ph.D., and 13 post-doctoral researchers and scientists since 1995. Many of his former students and research scientists are currently pursuing successful careers globally in the environmental arena. With his students, Prof. John has authored over 85 peer-reviewed journal papers, reports, book chapters and conference papers. He has co-edited a book titled “The Changing Climate of South Texas 1900-2100: Problems and Prospects, Impacts and Implications”. He has contributed to the academic and research community globally by serving on multiple advisory and review boards, committees and councils. As an air quality expert, he has spoken to various groups internationally. In 2019, he was selected for the U.S. Speaker Program series sponsored by the Bureau of Education and Cultural Affairs (ECA) of the U.S. Department of State and he visited Kuwait.

Authors:

Jithin Kanayankottupoyil University of North Texas
Kuruvilla John University of North Texas