Using data-driven modeling to understand multi-fractured, horizontal Marcellus completions
World Oil, Shaletech Report, May 2015
A data-driven, neural network model was developed to quickly and economically evaluate completion effectiveness for Marcellus shale wells. This model was used to identify significant opportunity to improve production for new wells by modifying completion and frac design. According to the model, geology and reservoir quality dominate Marcellus production. However, controllable contact and conductivity-related parameters are also significant. The number of frac treatments and the amount of proppant used in the completion rank first and second in significance. This is followed by perforation design, fluid volume and treatment rate.
A project was undertaken to evaluate well potential and completion effectiveness for hydraulically fractured, horizontal Marcellus completions in Susquehanna County, Pa. This article summarizes a study of the Marcellus shale's response to hydraulic fractures and identifies performance drivers. How effective are these completions? How would these wells produce, if they were completed and fraced differently? What are the primary controllable production drivers? How significant are geology and reservoir characteristics on well production? This article attempts to answer such questions.
Identification of major performance drivers becomes important in the design and optimization of new completions. They are not just important in enhancing production response and ultimate recoverable reserves, but also prove to be important economic factors in new completion design. This study employs neural network (ANN) modeling techniques to develop a predictive model, to identify performance drivers and evaluate completion effectiveness. Sensitivities performed on the predictive ANN model, developed for this project, indicate that well-to-well variation in reservoir quality and geology has a dominant effect on Marcellus production. Issues, such as fracture spacing, frac volume, perforation distribution, proppant amount and fluid volume, also affect well production.
Author(s): Robert Shelley, Amir Nejad, Nijat Guliyev, Michael Raleigh, David Matz