School of Earth, Energy and Environmental Sciences
Showing 1-10 of 23 Results
-
Jingfan Wang
Ph.D. Student in Energy Resources Engineering
Student Employee, Vice Provost for Teaching and Learning - Stanford Center for Professional DevelopmentCurrent Research and Scholarly InterestsNatural gas leaks waste money, reduce energy availability, induce sea-level rise, and result in both local air quality and global climate impacts. The climate impacts of leaked gas are particularly important due to the high global warming potential of methane (36 times more potent per kg than CO2 over 100 years). In addition to environmental concerns, the economic impacts of gas leakage are clear: lost natural gas costs nearly $2 billion per year at current prices.
Current EPA estimates suggest that about 1.5% of the natural gas produced in the U.S. is lost in leaks, while recent studies suggest that potential emissions from the gas system may be higher. Currently many natural gas leak detection and repair (LDAR) technologies exist. These methods include manually-operated flame ionization detectors and manually operated infrared (IR) video cameras for real-time optical gas imaging (OGI). EPA has recently released proposed regulations that codify the use of manually operated optical gas imaging as the standard leak detection technique.
Despite the current dominance of OGI, a number of fundamental problems with the technology exist: (1) labor costs for IR surveys are high, (2) continuous monitoring with IR is infeasible and (3) IR surveys cannot provide information about whether there is a leak happening in the real time.
To tackle the problems, we proposed an interdisciplinary project that harnesses the potential of computer science advances to allow for the rapid and automatic detection of methane leaks and estimation of their sizes. There are four main parts of my study: 1. We built a large video dataset of gas leaks for deep learning training purposes - GasVid, which includes a large number of labeled videos of methane leaks with representative leak sizes from different leak locations and imaging distances. 2. We developed a convolutional neural network (CNN) model to identify and detect leaks from imagery and systematically examined the efficiency of automatic system with data from the real world. 3. We are developing models based on both CNN and Recurrent Neural Network (RNN) to classify the leak size. 4. We will then perform economic and policy analysis to elaborate the benefits of automating pollution detection. -
Lijing Wang
Ph.D. Student in Geological Sciences
BioLijing is a first year Ph.D. student in Geological Sciences. Her interests include Bayesian Inference, Geometric and Topological Data Analysis, Geostatistics and Deep Neural Networks. Her research goal is to develop statistical methods to quantify uncertainty and make decisions in energy resources and environments. She is currently working on groundwater management and landslides hazards.
-
Ziyan Wang
Ph.D. Student in Energy Resources Engineering
BioI am a PhD student in Ilenia Battiato’s group at the Department of Energy Resources Engineering in Stanford University. My research focuses on developing hybrid multiscale methods for unconventional hydrocarbon recovery. Currently, I am conducting simulations of reactive transport during hydraulic fracturing. This research helps to reuse hydraulic fracture fluid efficiently, which can reduce the environmental impacts of unconventional hydrocarbon production.
