Proxy models for evaluation of permeability, three-phase relative permeability, and capillary pressure curves from rate-transient data

Abstract

This study developed a data-driven forecasting tool that predicts petrophysical properties from rate-transient data. Traditional estimations of petrophysical properties, such as relative permeability (RP) and capillary pressure (CP), strongly rely on coring and laboratory measurements. Coring and laboratory measurements are typically conducted only in a small fraction of wells. To contend with this constraint, in this study, we develop artificial neural network (ANN)-based tools that predict the three-phase RP relationship, CP relationship, and formation permeability in the horizontal and vertical directions using the production rate and pressure data for black-oil reservoirs. Petrophysical properties are related to rate-transient data as they govern the fluid flow in oil/gas reservoirs. An ANN has been proven capable of mimicking any functional relationship with a finite number of discontinuities. To generate an ANN representing the functional relationship between rate-transient data and petrophysical properties, an ANN structure pool is first generated and trained. Cases covering a wide spectrum of properties are then generated and put into training. Training of ANNs in the pool and comparisons among their performance yield the desired ANN structure that performs the most effectively among the ANNs in the pool. The developed tool is validated with blind tests and a synthetic field case. Reasonable predictions for the field cases are obtained. Within a fraction of second, the developed ANNs infer accurate characteristics of RP and CP for three phases as well as residual saturation, critical gas saturation, connate water saturation, and horizontal permeability with a small margin of error. The predicted RP and CP relationship can be generated and applied in history matching and reservoir modeling. Moreover, this tool can spare coring expenses and prolonged experiments in most of the field analysis. The developed ANNs predict the characteristics of three-phase RP and CP data, connate water saturation, residual oil saturation, and critical gas saturation using rate-transient data. For cases fulfilling the requirement of the tool, the proposed technique improves reservoir description while reducing expenses and time associated with coring and laboratory experiments at the same time.

An artificial-intelligence-based petrophysical property predictor for compositional volatile oil reservoir using three-phase production data

“…The neurons are interconnected to enable a variety of all-encompassing tasks. The details of neurons can be found by Zhang and Ertekin (2021).…”

“…The evaluation of reserve and hydraulic fracture qualities has mostly relied on production data (Blasingame et al, 1991; Fetkovich, 1980; Wattenbarger et al, 1998; Zhang and Ayala H., 2016, 2017, 2018). Using proxies, Zhang and Ertekin (2021) evaluated the petrophysical characteristics of black oil reservoirs. To the best of our knowledge, it has not yet been employed to calculate the relative permeability and capillary pressure for compositional volatile oil reserves.…”

A review of tracer testing techniques in porous media specially attributed to the oil and gas industry

“…However, the path of these barriers can be further defined using IWTT. Generally, the coarser clastic reservoirs deposited or matured in a compressive stress regime show the development of deformation bends that can drastically reduce the lateral permeability of the reservoir (Zhang and Ertekin 2019). The tracer technique can help in delineating these zones and associated deformation zones (Agency 2004;Peralta Gomes et al 2018;Pérez-Romero et al 2020;Schümann and Fossen 2018;Wrona et al 2019).…”

Introduction to special issue on advances in multi-phase fluid and multi-scale particulate reactive flow modeling and simulation for petroleum engineering applications

“…Zhang and Ertekin 4 proposed an artificial neural network (ANN) forecasting tool using the production rate and pressure data instead of coring and laboratory measurements to evaluate petrophysical properties, including effective permeability, three-phase relative permeability, and capillary pressure curves. To do this, an ANN structure pool was first generated and trained, covering a wide spectrum of petrophysical properties, and then was validated with blind tests and a synthetic field case.…”