St. James Ballroom
Evaluation of the Effect of Well Parameters on Oil Production (303912)
Alberto Giussani, Texas Tech University
Eshaan Nasir, Baker Hughes a GE Company
Marshall C. Watson, Texas Tech University
Peter H. Westfall, Texas Tech University
*Marshal E. Wigwe, Texas Tech University
Keywords: regression analysis, decision trees, EDA, data analysis, big data, statistical techniques
Operators in the oil and gas industry are faced with different economic decisions relating to an unconventional oil wells. With the popularity of data science and big data analytics tools, a petroleum engineer applies statistical techniques to analyze oil and gas data. We use regression analysis and decision tree in R to evaluate the effect of various well parameters on oil production. Our data-set has 5704 horizontal oil wells located in the six most productive counties in North Dakota. Two formations present are bakken and three forks. Initial EDA shows that, on average, operators are applying the same drilling and completion techniques across both formations as indicated in a comparative boxplot and two-sample t-test. Linear and "loess" bivariate fit indicates that higher completion parameters lead to higher production. Recursive partitioning trees also support this finding. However, we see reduction in oil production with these parameters if we model production per the different variables. The average well costs in the bakken increased from $6 – $6.5 million in 2008 to over $9 - $10 million in 2011. More stages or proppant does not necessarily equate to more oil, but more cost.