Oil and Gas Analytics Breakfast Series
Using Data to Define Models
Date: Thursday, May 10
Speaker: Jerry Jensen, Professor at the University of Calgary
Location: Calgary Petroleum Club
Time: 7:30 am - 9:30 am
Cost: $40 SPE Members, $50 Non-members
Series Discount Cost: $128.00 Members, $160.00 Non-Members
SOLD OUT
Abstract:
“Big data” analysis has shown great potential in many areas of modern life. Machine learning methods have been used to identify relationships and unexpected patterns in economic, social, medical, and other activities. For example, the removal of lead additive in gasoline appears to be responsible for a 50% decrease in US crime rates.
Data analysis is also being widely applied in the oil and gas domain. The oilfield, however, often has a more favorable situation for data analysis than typically is the case for social and biological systems. This is because many oil and gas problems have physical models which give us some insight into which variables might be important and how they may relate to the responses. This means that supervised learning is a preferred method for many oil field data analyses.
Supervised learning is where we provide the input variables and the output variable(s). If we have no model, we may use an artificial neural network to identify the relationship. If we have a model, we might use regression to specify model parameters. Whichever approach we take, we need to be careful when using data because not all data are equally useful to specify the input-output relationship.
In this talk, we will describe some statistical tools for data screening and model specification. The examples come primarily from petrophysics and geoscience, but we also discuss production rate analysis. We show that the concepts of percolation and reduced-physics models are particularly useful for model identification and data selection.
Speaker Bio:
Jerry joined the University of Calgary Chemical and Petroleum Engineering Dept. in 2007 after 10 years in industry (Schlumberger and Gearhart Ind.) and 20 years at academic institutions (Texas A&M and Heriot-Watt Universities). He is coauthor of more than 100 publications, including the books Statistics for Petroleum Engineers and Geoscientists (Elsevier, 2000) and Practical Reservoir Engineering and Characterization (Elsevier, 2015). Jerry was also a 2011-2012 SPE Distinguished Lecturer.
The SPE would like to thank our Platinum sponsor.
Thank you to our participating sponsors.
