Robert Balch, Petroleum Recovery Research Center - New Mexico Tech
The objective of this project is to use data mining to analyze well completion data to determine if trends or interesting patterns exist between well completion and stimulation methods and subsequent production. While a multitude of completion and stimulation techniques exist there are few objective, systematic examinations of their relative utility. A 370 well subset of basin Dakota wells drilled during 1994-2004 was examined to determine the feasibility of this concept with 58 additional wells drilled during 2004-2006 used for testing. Comparison of companies showed statistically significant differences in average production which could not be explained resolved by studies of land position via K-means clustering. Data mining was able to determine variances in first years gas production based on fracture fluid gallons, fracture fluid type, fracture interval thickness and sand lbs. A forward model was generated that could accurately estimate first years gas production based on these input parameters