The effectiveness of secondary and tertiary recovery projects depends heavily on the operator's understanding of the fluid flow characteristics within the reservoir. 3D geo-cellular models and finite element/difference-based simulators may be used to investigate reservoir dynamics, but the approach generally entails a computationally expensive and time-consuming workflow. This paper presents a workflow that integrates rapid analytical method and data-analytics technique to quickly analyze fluid flow and reservoir characteristics for producing near "real-time" results. This fast-track workflow guides reservoir operations including injection fluid allocation, well performance monitoring, surveillance, and optimization, and delivers solutions to the operator using a website application on a cloud-based environment. This web-based system employs a continuity governing equation (Capacitance Resistance Modelling, CRM) to analyze inter-well communication using only injection and production data. The analytic initially matches production history to determine a potential time response between injectors and producers, and simultaneously calculates the connectivity between each pair of wells. Based on the inter-well relationships described by the connectivity network, the workflow facilitates what-if scenarios. This workflow is suitable to study the impact of different injection plans, constraints, and events on production estimation, performance monitoring, anomaly alerts, flood breakthrough, injection fluid supply, and equipment constraints. The system also allows automatic injection re-design based on different number of injection wells to guide injection allocation and drainage volume management for flood optimization solutions. A field located in the Midland basin was analyzed to optimize flood recovery efficiency and apply surveillance assistance. The unit consists of 11 injectors and 22 producers. After optimization, a solution delivering a 30% incremental oil production over an 18-month period was derived. The analysis also predicted several instances of early water breakthrough and high water cut, and subsequent mitigation options. This system couples established waterflood analytics, CRM and modern data-analytics, with a web-based deliverable to provide operators with near "real-time" surveillance and operational optimizations.
Khalaf G. Salem, Khalaf Gad Salem (South Valley Egyptian Petroleum Holding Company) | Abdulaziz M. Abdulaziz, Abdel Aziz Mohamed (Faculty of Engineering, Cairo University) | Abdel Sattar A. Dahab, Abdel Sattar Dahab (Faculty of Engineering, Cairo University)
An accurate estimation of porosity and permeability are extremely essential for designing an ideal and efficient program of an oil and gas field development. Numerous methods have been developed to determine the porosity and permeability including laboratory measurements and log derived models. Artificial neural network (ANN) provides an efficient technique that successfully addressed several engineering and geological challenges. In the present study ANN is applied to help in predicting porosity and permeability in carbonate reservoirs using back propagation neural network (BPNN) with high accuracy on well log data from numerous fields worldwide.
ANN has the ability to understand a highly non – linear relationship and to perform simulation studies in a rapid manner. The BPNN model of porosity and permeability is developed using a set of well logging data as input layers and core porosity and core permeability as output layers. Two scenarios are considered to develop by ANN.
The first scenario considers using all available logging data directly as input for ANN. In the second scenario an additional input, diagenesis parameter, is added as input to ANN which is essentially calculated from logging data. In each scenario two models are developed; the first for porosity prediction and the second predicts permeability in carbonate reservoirs. The optimal learning rate and momentum constant used in the BPNN model are achieved after serial combinative trials. The available data was assigned 80% for training and 20% for verification.
The results of the developed porosity and permeability models are well compared to core data in verification. In the first scenario, cross-plot of the actual porosity versus ANN predicted Porosity exhibited a good match with a correlation coefficient equal to 0.97. Cross-plot of the actual permeability versus ANN predicted permeability exhibited a good match with a correlation coefficient equal to 0.80. In the second scenario, cross-plot of the actual porosity versus ANN predicted porosity exhibited a good match with a correlation coefficient equal to 0.97. Cross-plot of the actual permeability versus ANN predicted permeability exhibited a good match with a correlation coefficient equal to 0.98. Such data indicate that the developed models are successful in predicting the porosity and permeability for carbonate reservoirs.
A typical oil contains various amounts of tens of thousands of different compounds, and the relative abundances of those compounds form a “fingerprint” of that oil. This natural fingerprint can be used to answer some of the most important field-development and production-optimization questions that arise during development of unconventional reservoirs. Applications of this natural fingerprint include:
• Assessment as to whether or not induced fractures have propagated out of the formation containing a lateral and into either an overlying or underlying zone, causing the commingling of production from multiple intervals.
• Quantitative allocation of the contribution of 2-6 individual pay zones to commingled oil or gas production.
Because there are so many different compounds in an oil, even if two oils which occur in adjacent formations are 99% similar in composition, those two oils would still have more than 50 discrete geochemical differences. Any of those geochemical differences between oils from different formations could be used as natural tracers to distinguish the contribution of each reservoir to a commingled production stream.
To construct the oil fingerprint, dead oil samples are analyzed by a specialized type of high-resolution gas chromatography. In the average project, 175-250 different natural tracers are quantified in each sample, and the contribution of individual oils to a commingled sample is calculated by a linear-algebra solution of simultaneous equations, where the number of equations is equal to the number of natural tracers.
This paper illustrates these concepts using oils from 6 wells in Mitchell County, Texas in the Eastern Self of the Permian Basin.
More than 50 years ago, Jones and Smith (1965) reported compositional groupings within a set of more than 310 Permian Basin oils collected from reservoirs ranging in age from Cambrian to Cretaceous. Several years later, Kvenvolden and Squires (1967), Chuber and Rodgers (1968), Frenzel (1968), and Holmquest et al. (1968) demonstrated compositional differences among oils from various reservoirs of the Permian Basin; the differences that they noted were differences in whole oil stable carbon isotope values and molecular distributions. The technological constraints of the day prevented those authors from constructing detailed molecular fingerprints of the oils they studied. However, the whole oil isotopic data reported by Kvenvolden and Squires (1967), Chuber and Rodgers (1968), Frenzel (1968) and Holmquest et al. (1968) are still relevant to field development applications of geochemistry in the Permian Basin – even 50 years after the data were originally published.
Acid etching, as the consequence of heterogeneous distribution of petrophysical and compositional properties, results in the conductivity of acid fractures in carbonate reservoirs. Reliable characterization of small-scale formation spatial heterogeneity by use of geostatistical analysis (i.e., variogram analysis) can improve prediction of acid-fracture conductivity significantly. Previous publications suggest that permeability correlation length can be used to assimilate spatial heterogeneity in prediction of acid-fracture conductivity. Well logs are good candidates to provide information about petrophysical and compositional properties of the formation with the required resolution for prediction of acid-fracture conductivity. However, the assessment of permeability and mineralogy from conventional well logs is challenging because of high spatial heterogeneity and complex pore structure. Rock typing has been suggested in the literature to improve permeability assessment in carbonates. Most of the previously introduced rock-typing methods are dependent on core measurements. However, core data are generally sparse and not available with the sampling rate required for prediction of acid-fracture conductivity.
The main objective of this paper is to quantify formation spatial heterogeneity with variogram analysis of well logs and well-log-based estimates of petrophysical and compositional properties in carbonate reservoirs. We introduce an iterative permeability-assessment technique that is based on well logs, which takes into account characteristics of different rock classes in the reservoir. Furthermore, we propose three rock-classification techniques that are based on conventional well logs and that take into account static and dynamic petrophysical properties of the formation as well as mineral composition.
We applied the proposed techniques successfully in two carbonate formations--Happy Spraberry oil field and Hugoton gas field. The petrophysical rock classification is in good agreement with identified core-derived rock classes. The results show approximately 54% improvement in permeability assessment compared with conventional permeability-assessment techniques, which can improve prediction of acid-stimulation jobs significantly. Finally, we investigated the direct application of well logs and well-log-based estimates of petrophysical and compositional properties for variogram analysis required to characterize formation spatial heterogeneity. We conducted variogram analysis in both field examples. The results show that the direct application of well logs and well-log-based estimates of petrophysical/compositional properties is reliable to characterize formation spatial heterogeneity. We also showed that application of well logs can enhance assessment of spatial heterogeneity compared with core measurements.
This study compares reservoir characteristics, completion methods and production for 431 wells in 6 counties producing from the Wichita-Albany reservoir to assess major factors in production optimization and derive ultimate recovery estimates. The purpose of the study is to analyze completion design patterns across the study area by combining public and proprietary data for mining. Integrating several analyses of different nature and their respective methods like statistics, geology and engineering create a modern approach as well as a more holistic point of view when certain measurements are missing from the data set. Furthermore, multivariate statistical analysis allows modeling the impact of particular completion and stimulation parameters on the production outcome by averaging out the impact of all other variables in the system. In addition to completion type, more than 18 predictor variables were examined, including treatment parameters such as fracture fluid volume, year of completion, cumulative perforated length, proppant type, proppant amount, and county location, among others. In this sense, this contribution seems unique in unifying statistical, engineering, and geological perspectives into a singular point of view. This work also provides complementary views for well production consideration.
Reliable permeability and mineralogy estimates in carbonate formations can significantly improve the prediction of acid fracture conductivity in these challenging reservoirs. Well logs are good candidates to provide information about petrophysical properties of the formation with the required resolution for prediction of acid fracture conductivity. The assessment of permeability and mineralogy from well logs, however, has been highly dependent on core measurements in carbonate formations. Conventional well-log-based permeability assessment techniques including porosity-permeability correlations are not reliable in the case of carbonate formations. High spatial heterogeneity, variable lithology, and complex pore structure result in a poor correlation between permeability, porosity, and irreducible water saturation in carbonates. Rock typing has been suggested in the literature to be used to improve permeability assessment in carbonates. Most of the introduced rock typing methods are dependent on core measurements. However, core data are generally sparse and not available with the sampling rate required for prediction of acid fracture conductivity.
In this paper we propose an iterative permeability assessment technique based on well logs, which takes into account rock types. We also introduce three rock classification techniques based on conventional well logs including (a) a log-derived analytical factor, (b) unsupervised artificial neural network, and (c) supervised artificial neural network. The first two techniques are independent of core measurements for rock classification. However, the third technique is highly dependent on core measurements in the field. We successfully applied the proposed techniques in two carbonate formations, Happy Spraberry oil field and Hugoton gas field. The petrophysical rock classification is in a good agreement with identified core-derived rock classes. The results show approximately 54% improvement in permeability assessment compared to conventional permeability assessment technique, which can significantly improve prediction of acid stimulation jobs.
Due to quick development in horizontal drilling and fracturing technologies, shale gas, formerly considered very difficult if not impossible to recover, has become one of the hottest energy topics. As a significant number of studies focus on fracture stimulation, drilling and completion optimization, only a limited number of studies have been carried out aiming at the investigation of shale structure at nanometer scale due to limited access to scanning electron microscopy (SEM), transmission electron microscopy (TEM) and atomic force microscopy (AFM), and also due to less experience using these instruments in studies of formation rocks.
SEM and TEM are all capable of revealing nano-scale structures of rock samples. AFM also offers some possibilities. In order to obtain images of high quality and ultra-high magnification, sample preparation is the initial and probably most important step in the whole imaging process. Images from low and ultra low permeability formations in the Western Canada Sedimentary Basin are compared with images from other low and ultra low permeability formations reported in the literature. As powerful as these nanoscale capable microscopes are, they all have limitations due to the high instrument cost, limited access and time consuming imaging process. As a result it is difficult and impractical to obtain statistical significant data. To make these instruments of practical value to geoscientists and petroleum engineers, methods and models need to be developed to build a bridge from detailed nanoscale structures to typical lab/field measurements and practical geoscience and engineering-scale applications. In this paper, we propose the concept of a multiple porosity model for evaluation of shale formation based on our observations of the nanopore structure of shales. This includes a model for determining the porosity exponent m in shales and water saturation evaluation. Water saturation curves based on this model are plotted on Pickett Plots for both tight gas and shale formations.
Continuous studies in this area are needed to further explore shale structure in fine detail in order to understand the role of nanopores on shale gas production.
This paper presents a model to estimate permeability from well logs in carbonate reservoirs. The model relates permeability to interparticle porosity, makes special accommodation for separatevug porosity, and includes a rock-fabric classification scheme with an important dual petrophysical-geological significance. The dual significance of the rock-fabric classification provides an important link to geological models for use in distributing permeabilities between wells. Porosity and permeability are highly variable and are difficult to predict spatially in most carbonate reservoirs, but rock-fabric changes tend to be organized systematically within a sequence stratigraphic framework.
A fundamental component in the construction of most reservoir performance models is an empirical relationship between permeability as measured in a limited number of cored wells and other petrophysical properties measured in well logs. This paper presents a permeability model specially designed for carbonates. The model relates permeability to interparticle porosity, makes special accommodation for separate-vug porosity, and includes a rock-fabric classification scheme with an important dual petrophysical-geological significance. Methods to estimate the separate-vug porosity from sonic logs and the rock-fabric from initial saturation are presented.
The dual petrophysical-geological significance of the rock-fabric classification is important for providing a link to geological models for use in distributing permeabilities between wells. Porosity and permeability are highly variable and difficult to predict spatially in most carbonate reservoirs, but rock-fabric changes tend to be systematically organized in a predictable manner within a sequence stratigraphic framework.
Reservoir characterization and modeling is primarily a problem of understanding the 3D spatial arrangement of petrophysical properties. Petrophysical measurements must be linked to spatial information when building a reservoir model, and geologic models contain vital spatial information to be applied in interwell areas where direct petrophysical measurements are difficult. The link is best accomplished through the integration of geologic rock-fabric descriptions and petrophysical measurements.
A method for linking basic rock-fabric descriptions and petrophysical properties has been proposed by Lucia.1,2 Carbonate pore space is divided into interparticle, which includes both intergrain and intercrystal, and vuggy pore space (Fig. 1). Vuggy pore space is subdivided into separate and touching vugs on the basis of vug interconnection. Separate vugs are connected through the interparticle pore space (grain molds, for example), and touching vugs form an interconnected pore system independent of the interparticle pore space (caverns and fracture pore space, for example). Interparticle pore space is subdivided into rock-fabric classes on the basis of geologic descriptions of particle size and sorting.
In this paper we present an approach to permeability modeling in carbonates on the basis of this rock-fabric classification. The paper is organized into five main sections: (1) the carbonate rock-fabric classification is summarized and its relationship to porosity and permeability is presented; (2) exponential and power-law porosity-permeability models are compared, and a generalized power-law model relating porosity, permeability, and rock fabric is presented; (3) the generalized permeability model is compared with three others from the literature; (4) rock-fabric based methods for permeability prediction from well logs are summarized; and finally (5) an approach to 3D modeling of carbonate permeability taking advantage of the geological link provided by the rock-fabric method is described.
Carbonate Rock-Fabric Petrophysical Classification
Permeability and capillary properties of interparticle pore space can be related to interparticle porosity and geologic descriptions of particle size and sorting called rock fabrics.1,2 These rock fabrics were initially grouped into three categories called rock-fabric petrophysical classes on the basis of porosity, permeability, and capillary properties1 (Fig. 2):
Class 1 is composed of grainstones, dolograinstones, and large crystalline dolostones.
Class 2 is composed of grain-dominated packstones, fine and medium crystalline, grain-dominated dolopackstones, and medium crystalline, mud-dominated dolostones.
Class 3 includes mud-dominated limestones and fine crystalline, mud-dominated dolostones.
South Wasson Clear Fork field produces from two reservoirs, the middle Clear Fork, with a seal located within the upper Clear Fork Formation, and the lower Clear Fork, with a seal located in the Tubb Formation. Six sequences have been defined on the basis of facies succession, seismic interpretation, and outcrop analog studies in Apache Canyon, Sierra Diablo Mountains, West Texas. Rock-fabric/petrophysical studies have shown that a single porosity-permeability transform and porosity-saturation-capillary- pressure model can be used to calculate permeability and water saturation in uncored wells. Five rock fabrics have been described, and all plot in the petrophysical class 1 field. This surprising result is related to the presence of large volumes of poikilotopic anhydrite, which reduces porosity but has little effect on pore size or perme-ability. Permeability values calculated from porosity logs are distributed in the interwell environment within a high-frequency-cycle (HFC) stratigraphic framework. Analog outcrop studies demonstrate that the Clear Fork Formation can be characterized by upward shallowing (HFC's). Identifying HFC's, however, is made difficult because of the high uranium content of the Clear Fork and the lack of a relationship between water saturation and rock fabric. A statistical relationship between porosity and rock fabric is developed, however, that allows porosity to be used as a surrogate for rock fabric, and vertical increases in porosity are interpreted as mud-dominated to grain-dominated fabric successions and used to define HFC's. The HFC's are divided into two rock-fabric flow layers, an upper, grain-dominated and a lower, mud-dominated layer, in order to preserve high- and low-permeability intervals. Permeability data from the outcrop analog in Apache Canyon are used to demonstrate that the vertical variability seen in the log calculations within the rock-fabric flow units does not represent permeable strata but is statistical variability at a near-random scale.
South Wasson Clear Fork field, located in Yoakum County, West Texas, on a small structural high (fig. 1), produces from two thick reservoirs—the middle and lower Clear Fork. The top of the lower Clear Fork reservoir is the base of the Tubb Formation. This reservoir appears to reach a thickness of 500 ft. The top of the middle Clear Fork reservoir is defined by the change from low water saturation to high water saturation, which corresponds to the top of Leonardian sequence 4, not to the top of the upper Clear Fork Formation (fig. 2). The middle Clear Fork reservoir obtains a thickness of about 700 ft. The interval from the top of the middle Clear Fork to the top of the upper Clear Fork Formation appears to be at residual oil saturation, according to core and production data, and perhaps represents a third reservoir that has remigrated. This study has concentrated on the middle and lower Clear Fork reservoirs in a 1-mi2 area in the middle of the field (fig. 1).
A sequence stratigraphic framework has been constructed for the Clear Fork on the basis of geologic descriptions of nine cores and guided by the sequence stratigraphic framework developed at Apache Canyon in the Sierra Diablo Mountains, West Texas (fig. 2). Unfortunately, the Clear Fork section exposed in Apache Canyon represents only the lower Clear Fork, Tubb, and basal part of the upper Clear Fork in the subsurface. Nevertheless, the outcrop observations provide important guidelines for understanding the sequence and cycle stratigraphy in South Wasson1.