Predicting Petrophysical Properties From Rate-Transient Data: An Artificial Intelligence Application

Open Access
Zhang, Zhenzihao
Graduate Program:
Petroleum and Mineral Engineering
Doctor of Philosophy
Document Type:
Date of Defense:
June 07, 2017
Committee Members:
  • Turgay Ertekin, Dissertation Advisor
  • Turgay Ertekin, Committee Chair
  • John Yilin Wang, Committee Member
  • Hamid Emami-Meybodi, Committee Member
  • Ming Xiao, Outside Member
  • Relative Permeability
  • Capillary Pressure
  • Permeability
  • Artificial Neural Network
  • Rate-Transient Data
  • Petrophysical Properties
Traditional methods to obtaining relative permeability curves and capillary pressure curves were expensive and time consuming. The methods rely on cores which are typically retrieved only for a small fraction of wells. This study proposes a new methodology for predicting petrophysical properties from rate-transient data. In this study, we develop artificial-intelligence-based tools that predict characteristics of three-phase relative permeability, characteristics of capillary pressure for water-oil system and liquid-gas system and formation permeability using rate-transient data. Petrophysical properties are related to rate-transient data since these properties govern the fluid flow dynamics in reservoirs. An artificial neural network (ANN) can mimic any functional relationship with a finite amount of discontinuities. This study develops ANNs for predicting petrophysical properties for four different kinds of reservoirs-a circular black-oil reservoir, a rectangular black-oil flow unit, a circular volatile-oil reservoir, and a circular naturally fractured black-oil reservoir. The well setting for the black-oil rectangular flow unit is horizontal well while those for the other three types of reservoir are vertical wells. In a very rapid manner, the developed ANNs infer characteristics of relative permeability for three phases, characteristics of capillary pressure curves for water-oil system and liquid-gas system, horizontal permeability and vertical permeability. With the predicted characteristics, relative permeability and capillary pressure data can be generated. The generated data as well as the predicted permeabilities can be applied in history matching and reservoir modeling. This ANN tool can be used along with production practices to improve our ability in reservoir description. This tool can fill in the spot for predicting petrophysical properties for uncored wells and provide reference for modeling drainage area including the wells.