Artificial Expert Systems For Rate Transient Analysis Of Fishbone Wells Completed In Shale Gas Reservoirs

Open Access
Bukhamseen, Ibrahim
Graduate Program:
Energy and Mineral Engineering
Master of Science
Document Type:
Master Thesis
Date of Defense:
July 21, 2014
Committee Members:
  • Turgay Ertekin, Thesis Advisor
  • Rate Transient
  • Artificial Neural Networks
  • Fishbone
  • Horizontal
  • Shale Gas Reservoir
  • Multilateral
Reservoir simulation is the most important tool petroleum engineers use to forecast production performance and set production strategies for oil and gas fields. Nonetheless, thorough reservoir simulation is a time consuming process that requires powerful computational capabilities and highly skilled individuals in reservoir engineering, mathematics, and programing. The accuracy of the simulation process is affected by the availability and quality of reservoir characteristics. These characteristics are obtained using different data acquisition methods such as open-hole logging, well testing, and core analysis which is costly and time consuming. Another factor that affects production performance is horizontal and multilateral well configurations. These well completions are used to increases production rate and enhance sweep efficiency. However, the design of horizontal and multilateral wells is a complex task that involves engineering and economic constraints. This thesis demonstrates the development of artificial expert systems that are not only capable of forecasting, but also robust in multilateral well design and reservoir evaluation. The study covers a wide range of well completions and shale gas reservoirs properties. Using artificial neural networks (ANN), the developed approach delivers forward-looking and inverse-looking solutions to relate between reservoir characteristics, well configurations, and production performance. Production Performance Expert System forecasts the gas rate and cumulative gas production of a given well completed in a shale gas reservoir, and thus it is categorized as a forward-looking solution. The average error of the results generated by the first system is 6.5%. The second expert systems provides an inverse-looking solution to propose multilateral well design given a target production of a specific reservoir. Well design parameters are produced within an average error of 5%. The third and final system is capable of predicting shale reservoir properties by analyzing the production performance of a horizontal or multilateral fishbone well completion. The developed system is capable of predicting reservoir characteristics with an average error of 4.5% from the target data. The developed set of expert systems is expected to reduce the long hours required for excessive computations, which are typical in reservoir simulators. In addition, it can reduce costs of reservoir evaluation by working as an auxiliary tool to expensive well tests, logging operations, and core analysis. However, it is worth mentioning that these expert systems do not replace conventional reservoir simulators, rather they help in enhancing the decision making process by making educated estimations of reservoir properties and basic designs of multilateral wells to be used in history matching and field development or infill drilling strategies. A graphical user interface (GUI) has been developed to offer user accessibility to the aforementioned artificial expert systems. It provides easy execution of codes and neural networks without requiring knowledge of the background mechanism. This interface allows the user to provide their own input data to simulate any desired scenario within the dataset limits. The results of each case are shown graphically or numerically.