development and testing of an artificial expert system to design perforation parameters

Open Access
Cengiz, Ugur
Graduate Program:
Energy and Mineral Engineering
Master of Science
Document Type:
Master Thesis
Date of Defense:
July 19, 2012
Committee Members:
  • Turgay Ertekin, Thesis Advisor
  • Ugur
  • Cengiz
  • perforation parameters
  • artificial expert system
Establishing a communication between a producing well and the formation is a crucially important factor that affects the productivity of a well in a most pronounced manner. This communication between the well and the formation is typically established by the implementation of a perforation completion strategy. Use of artificial neural network (ANN) technology in petroleum engineering applications has gained increasingly more attention during the recent years. The principal purpose of these applications is to come up with forecasts for various scenarios under which field is operating. In this study, an ANN based tool is designed and developed to predict the production profiles resulting from vertical wells with various perforation designs. The proposed ANN tool is capable of forecasting the production profiles from single-phase gas reservoir for periods of up to four years. In this forward-looking ANN, the input parameters are reservoir characteristics and perforation design specifics with an output file pertaining to production rates as a function of time. The inverse of the aforementioned tool is the second ANN developed in this study. There are two inverse ANN models that are presented in this study. The first inverse tool receives reservoir characteristics and desired production profiles as input and suggest the specific design parameters for perforations so that the expected production profile can be realized. The second inverse tool receives reservoir characteristics (except permeability, porosity and thickness), perforation design parameters and the observed production profiles as input to forecast the permeability, porosity and thickness values for that reservoir. In vertical wells with perforations, besides reservoir properties, productivity of the well is controlled by the perforation design characteristics including penetration length, shot density, perforation diameter and phase-angle. A numerical model, which is capable of accommodating these characteristics, has also been developed as a part of this study. This numerical model is verified against a commercial model and used to generate the data that is used in training the forward and inverse ANN tools. The ANN tools developed in this study are brought together on a graphical- user- interface (GUI) to ease the practical implementation of the software package developed in this study.