This paper documents follow-on work to an original data mining study of horizontal wells in the North Texas Barnett Shale play. In this study, the authors have analyzed well and production data beginning with over 15,000 producing Barnett wells. Study wells were grouped for similar map-based reservoir properties, normalized for the effects of well architecture, and normalized for production. The study used statistical and data mining techniques plus Geographical Information System pattern-recognition techniques to aid in interpretation of production result trends. The principal focus of the follow-on study is in the areas of completion and hydraulic fracturing. It is intended to demonstrate lessons-learned from careful analysis of large volumes of data, some of which can only be examined by proxy.

Data sets from public and proprietary data sources were collected and merged into a common database. Data elements were subjected to statistical quality control methods in order to minimize the potential for inaccurate data to be included in the analysis. Short-term production proxies were studied and decided upon in order to adequately compare well productivity values over many years of horizontal well drilling across the Fort Worth Basin.

Previous work showed that the application of practical data mining methods to a large Shale Gas data set resulted in learning key lessons that were not apparent from small data sets. The previous work also showed that certain popular beliefs did not hold up to examination of the available data. Similar to past work, this work is significant in the use of merged reservoir quality proxies, well architecture data, completion data, and stimulation data, against which production results are placed in geographical perspective for improved interpretation. This work is also significant in the application of advanced data mining methods beyond routine statistical work.

You can access this article if you purchase or spend a download.