Using Multiple Linear Regression to Model EURs of Horizontal Marcellus Shale Wells

Abstract

Abstract EQT Production has been actively drilling the Marcellus Shale in Pennsylvania and West Virginia since 2008. With over 170 producing horizontal wells in the Marcellus, EQT has a significant data set for evaluation using Multiple Linear Regression. A variety of well design parameters can be compared to production in hopes of finding a correlation and identifying what determines a well’s Estimated Ultimate Recovery (EUR). Some of the most logical parameters to compare include lateral length, lateral azimuth, formation target, formation mineralology and stimulation design. Multiple Linear Regression is a powerful tool for bringing many of these design aspects together for comparison against productivity and can reveal patterns and relationships that may be indiscernible using conventional methods. A regression software package was used for this study following the Least Squares method. Well EUR was the primary dependent variable throughout the analysis. Each individual dataset consists of wells in close proximity to one another (map quad, county, etc.). Dozens of drilling and completion parameters were compared against the EURs, as independent variables. Multiple iterations of removing and adding specific independent variables led to an optimal EUR model for a particular area. Four main areas were the focus for the study, and the results have been conclusive. In all areas, treated lateral length exhibited a near-linear relationship with EUR. More specifically, in Area 1, the model exhibited an R-Squared value of 0.97. This was achieved using only three independent variables: treated lateral length, perforation cluster spacing, and lateral spacing. Realistic models were built, with conclusive results, in Areas 2, 3, and 4 using different criteria. Multiple Linear Regression Models can provide extremely important information for field development. Using MLR models, coefficients are obtained for each of the independent variables (i.e. the quantitative effect each parameter has on a well’s EUR). These values can be extended to new areas, and a more accurate EUR can be predicted, often before the well is drilled. In core areas, when an operator is forced to make decisions based on land and drilling constraints, this method can be used to make the best economic decision for drilling and completion design.