Evaluating gas production performances in marcellus using data mining technologies

Shale gas development - enabled by the advent of advanced horizontal drilling and hydraulic fracturing technology - has become very important energy resource in the past decades. The estimated ultimate recovery of Marcellus Shale in West Virginia is about 98 to 150 trillion cubic feet (Tcf) (US DOE, 2010). In 2008, 25 billion cubic feet (Bcf) of natural gas was produced from 41 horizontal wells in West Virginia. By 2012, that gas production reached 301.7 Bcf from 631 horizontal wells. However, as mechanism of natural gas production from shale by hydraulic fracture stimulation of horizontal wells with multiple fracture stages is complex, significant uncertainty about production performance in these unconventional reservoirs represents significant risk for whether resource development will lead to favorable return on investment. The objective of this paper is to use post-hoc analysis to identify correlations between gas production performance of a well and attributes of its completion and geological setting, and to identify those factors most important to predicting gas recovery performance. To accomplish this we first characterized the geological attributes of Marcellus Shale in West Virginia through literature review. Then, we down selected from the set of 631 wells to a representative subset of 187 wells for which complete data are available, including well location, completion data, hydraulic fracture treatment data and production data. The wells were classified into five groups based on geological setting. For each geological group, engineering and statistical analysis were applied to study the correlation between well performance and well completion attributes through traditional regression methods. Important factors considered to affect gas production include number of hydraulic fracture stages, lateral length, vertical depth, proppant volume, and fracture fluid volume and treatment rate. The numbers of hydraulic fracture stages and lateral length have relative large influence on well performance. With this analysis results, it was possible to estimate wellscale ultimate natural gas production performance based on known geological conditions and completion parameters. The results lead to a better understanding of the trends in Marcellus Formation well performance. These approaches could, in the future, help to optimize stimulation treatments and well completions and improve resource recovery in the Marcellus, and other unconventional hydrocarbon formations.