Ziming Xu

6-361 DICE, 9211 116th St, Edmonton, AB, CA • ziming4@ualberta.ca

Ph.D. Candidate in Petroleum Engineering

Hello, my name is Ziming Xu, and I am a fourth-year Ph.D. student at the University of Alberta. My research is dedicated to pioneering advancements in production forecasting and optimization through the application of advanced deep learning and numerical methods. I am deeply passionate about machine learning and its potential to solve critical subsurface problems. My work focuses on pushing the boundaries of knowledge in these areas, aiming to make a significant impact on both the academic and industrial fronts.

English/中文

Education

University of Alberta

Ph.D. in Petroleum Engineering • 2021 — Present

Edmonton, Alberta, CA

University of Alberta

M.Sc. in Petroleum Engineering • 2019 — 2021

Edmonton, Alberta, CA

University of Alberta

B.Sc. in Petroleum Engineering • 2015 — 2019

Edmonton, Alberta, CA

Experience

University of Alberta

Teaching Assistant (PETE375; PETE475) • Sep, 2022 — Present

Assisted in grading assignments and reports.
Facilitated engaging classroom discussions and responded to student inquiries.
Provided remote one-on-one guidance and addressed student concerns.

University of Alberta

Research Assistant • Sep, 2019 — Present

Conducted research and data analysis on unconventional reservoirs.
Collaborated with professors on publications.

Projects

Physics-Informed Deep Learning for Shale Gas Production Forecasting

• April 2023 — Present

Developed neural network solution for 3D Diffusivity equation.
Conducted field-scale experiments with a combined physics and data-driven model.

Graph Convolution for Capturing Well Interference in Production Forecasting

• Dec. 2022 — Present

Developed novel ST-GraphSAGE model for generalizing unseen nodes and computation of evolving graphs.
Incorporated a GRU encoder-decoder model to capture cross-time neighborhood information effectively.

Enhanced RNN-Based Production Forecasting for Shale Gas Reservoirs

• Sep. 2021 — April 2023

Developed novel MRA (Masked Recurrent Alignment) architecture, enabling inputs sequence with varied lengths.
Proposed novel RUF (Recursively Updated Forecasting) architecture, enabling real-time updates with production data.
Generated multi-stage hydraulic fracture synthetic dataset using DFNs and geomechanical simulation.

Impacts of Meshing and Discretization on DPDK Flow Simulation

• Jan. 2019 — Dec. 2021

Proposed quick analytical workflow for generating non-planar twisted fractures, improving fracture representation.
Developed novel workflow to restore connectivity between disconnected fractures after discretization.
Upscaled DFN into DPDK and conducted DPDK Flow simulation on fractured reservoir using a black oil simulator.