After production ceases, the company exploiting an oil or gas well has legal obligation to clean up the site, a process called land reclamation. Should such cleanup not occur, e.g. if the company is bankrupt, the province is responsible for the cost. Hundreds of wells are classified as orphan in various regions, including British Columbia, Alberta, and numerous states in the US. To decrease the financial risk associated with land reclamation, an NSERC Alliance grant was awarded to fund a postdoctoral fellow position.
The project would explore the variables that may increase the probability of a well ceasing the production of oil and gas. Using decades of historical data available on well activities, a predictive model will be built. The project’s challenge is to balance model accuracy with explainability, interpretability, and fairness in a sensitive environment that includes oil and gas companies, environmental activities, and Indigenous lands. Addressing such challenges requires new predictive models running on an efficient computer architecture. Thus, the project seeks to:
Models under considerations include traditional learning models (e.g. explainable random forests, decision trees), and deep neural networks.
Oil and gas operators must apply to the provincial regulator authorities (the BC Oil and Gas commission) for approvals before beginning work; examples of considerations for approval include the impact on the environment, Indigenous Nations, forestry, archaeology, etc. Efficacy of the Commission is critical for the economic dynamism of the sector. In 2016, the Commission implemented a massive change to their application processes that allowed oil and gas operators to bundle multiple activities (e.g. wells, facilities, pipelines, roads, etc.) into a single application and obtain a decision for the entire application instead of multiple independent decisions for each planned activity. While the new application process brings efficiencies for the applicant, no two applications are the same (as the number and type of activities can vary), rendering data analysis more complex. To improve services for BCOGC stakeholders, an NSERC Alliance grant was awarded to fund a postdoctoral fellow position and benchmark both application processing timelines and resource forecasting.
The project would explore the variables that may increase the complexity of an application and its timeline, as well as forecasting application volume in prevision of resources management. Using several years of data available on application activities, predictive models will be built. The project’s challenge is to balance models accuracy with explainability, interpretability, and fairness in a sensitive environment that includes oil and gas companies, environmental activities, and Indigenous lands. Addressing such challenges requires new predictive models running on an efficient computer architecture. Thus, the project seeks to:
Models under considerations include traditional learning models (e.g. explainable random forests, decision trees), and deep neural networks.
The successful candidate will have completed, or be nearing completion, of a Ph.D. in Computer Science, Engineering, Mathematics, Statistics, or a related discipline. They must have previous experience in developing machine learning models and implementing them successfully (for the second project, knowledge of business process mining is an asset). In addition, as this is a time sensitive client project, the following soft skills are desirable experience that should be highlighted with the application letter: communication skills with industrial collaborators, project management skill, supervision and mentoring of undergraduate and master’s students.
To apply, submit a Cover Letter and CV to yves.lucet@ubc.ca. The deadline for application is Feb 28, 2023 although applications will keep getting reviewed until the positions are filled. The preferred starting date is June 1, 2023, or as soon as possible thereafter. Funding is already secured so an earlier start date is possible.
The cover letter should clearly indicate, and justify, which areas of experience are applicable. In addition, arrange for 3 reference letters to be sent directly to the above.
The position will be based at UBC Okanagan (Kelowna, BC) and hosted in collaboration with several research laboratories including the Center for Optimization, Convex Analysis, and Nonsmooth Analysis (COCANA). Alternatively, the position could be hosted at the University of Victoria (Victoria, BC).
Each position is a 1-year position with renewal for a 2nd year if warranted. Only successful candidates will be contacted.