Hi! I’m Meng Jia (贾萌 in Chinese), a PhD student at Colorado School of Mines in the Department of Applied Mathematics and Statistics, working with my advisor Dr. Dorit Hammerling. I’m also a graduate student researcher at the Payne Institute for Public Policy and at Energy Emissions Modeling and Data Lab (EEMDL). My research focuses on methane emission detection, localization, and quantification on oil and gas (O&G) facilities using statistical and machine learning methods.
Before my current role, I worked in NASA’s InSight Project, where I contributed to characterizing Mars’ interior structure using the first-ever in-situ seismic data from Mars.
Originally from Beijing, China, I spent three years living in Florida before relocating to Colorado. My leisure pursuits include a passion for travel, reflected in my collection of nearly a hundred magnets from various destinations. Additionally, I enjoy playing basketball and board games as a way to unwind and relax.
Education
Ph.D. Statistics, Colorado School of Mines, in progress
M.S. Data Science, Colorado School of Mines, 2020
M.S. Geophysics, University of Florida, 2018
B.S. & M.S. Geophysics, Peking University, 2015
My research
Gaussian puff model
I’ve collaborated with Ryker Fisher and Will Daniels to develop a computational efficient implement of the Gaussian puff model. This model has demonstrated enhanced accuracy in characterizing atmospheric transport of air pollutants, surpassing the commonly used Gaussian plume model in precision. For more detailed insights into our work, please refer to our preprint.
Sensor placement optimization
Together with Troy Sorensen, I am currently engaged in developing a data-driven algorithm aimed at optimizing sensor placement for continuous monitoring on oil and gas sites. Our approach leverages on-site wind data and practitioner-provided emission characteristics to simulate a range of emission scenarios. Concurrently, we identify potential sensor locations based on the site’s geometry and operational guidelines. To determine the most effective sensor placement within a specified budget, we are employing evolutionary algorithms within a Pareto optimization framework. This strategy is designed to maximize detection efficiency across the varied emission scenarios we’ve modeled. Find this preprint for more details.
Physics-informed neural netowrks
Looking ahead, my future work includes plans to employ Physics-Informed Neural Networks (PINNs) for solving the advection-diffusion partial differential equation (PDE), which is pivotal in describing the atmospheric transport of methane. Furthermore, by conceptualizing the PINN as an inverse problem, we aim to extract critical information about the emission source. This includes determining the start and end times, location, and intensity of the emissions. Find this conference presentation for more details.
Contact
You can reach me at m{my last name}@mines.edu.