Date of Graduation
Statler College of Engineering and Mineral Resources
Lane Department of Computer Science and Electrical Engineering
Unconventional reservoirs are full of uncertainty when dealing with conventional methods of modeling and analysis. The objective of this work is to use a trained artificial intelligent (AI) model to compare actual production data to AI predicted possible well production. Data-driven models based on AI are efficient tools for optimizing the production, stimulation, and completion design of wells and can be very beneficial when determining the success or failure of wells based on production. An AI model for production predictions require both native and design parameters, which include well characteristics, completion design, and stimulation design parameters.;Data from over 100 Marcellus Shale wells are used to train and test an AI model for production predictions. Feature selection algorithms are used to determine the most influential input parameters pre and post modeling for both increased model accuracy and quality assurance. Post modeling, Monte-Carlo simulation and Type Curves are used to assess the performance of each well based on the AI generated well production values. AI model generation is a very useful tool for predicting production performance of existing wells, which can be used to optimize design characteristics and reservoir production. Generating AI predictive models in fields with low amount of cases to train and test the artificial neural network require very delicate and careful considerations in order to maximize the effect and accuracy of the predictive model.;This study will be able to give an underlying method of applying these artificially intelligent solutions to a complex petroleum engineering problem. The ability to correctly apply these techniques will allow for the optimization of completion and stimulation designs within complex, unconventional reservoir.
Yingling, Daniel, "Application of Artificial Intelligent Predictive Modeling for Completion Optimization and Refracture Candidacy" (2018). Graduate Theses, Dissertations, and Problem Reports. 7009.