Synthetic Well Log Generation for Complex Well Architectures Using Artificial Intelligence Based Tools

Open Access
Ozdemir, Ibrahim
Graduate Program:
Energy and Mineral Engineering
Master of Science
Document Type:
Master Thesis
Date of Defense:
July 09, 2015
Committee Members:
  • Turgay Ertekin, Thesis Advisor
  • Oil Field Development
  • Artificial Neural Networks
  • Synthetic Well Log
  • Oil Production Profile
  • Shale Index
The need for energy has been increasing in the world. Among all, oil is the most widely used primary energy source in the world. Oil reservoirs are characterized for field development purposes. Conventional reservoir characterization methods require hard computing techniques, which are time and labor intensive. This thesis aims at developing artificial neural network based systems to characterize oil reservoirs to build powerful field development strategies. First set of tools developed in this study addresses the need for synthetic well logs. This tool set is capable of generating 5 different types of synthetic well logs at any location within the seismic boundaries. In addition, shale content tool is devised by using the gamma ray tool. Furthermore, log surfaces and shale content surfaces are created at desired depths for the entire reservoir. In the development stage of these expert systems; seismic attributes extracted from 3D seismic data, 5 different types of well logs, and well coordinates are utilized. Second set of tools are devised in order to predict the oil flow rates, cumulative oil productions at the end of each year, and perforation intervals at any desired well location in the reservoir. Moreover, production surfaces are created at the end of each year. The data utilized in these tools are seismic attributes, synthetic well logs, perforation intervals, and well coordinates. Finally, tools that are capable of predicting the well logs and shale content along complex well architectures are developed. These tools can be used in any type of well architecture including vertical, slanted and horizontal well structures. The resolutions of these predictions can be set at the desired values. The artificial expert systems devised in this study enable to make a fast evaluation of the oil reservoir. These tools allow detection of sweet spots in the reservoir for infill drilling strategies. They also predict the perforation intervals for desired well locations. The results presented in this study show that neural network predictions and field data are in good agreement.