Full Length ArticleHydrocarbon saturation in shale oil reservoirs by inversion of dielectric dispersion logs
Introduction
Shale oil has gained a significant interest in exploration and development of unconventional resources in the past ten years [1], [2], [3], [4]. In addition to North America, a large number of shale reservoirs are found in the basins in China, including Ordos, Bohai Bay, Junggar, Songliao Basin, etc. that are a huge source of oil production [5], [6]. Although unconventional shale oil plays have been studied for many years, some of the characterization methods especially those focused on fluid properties still need further investigation. Hydrocarbon saturation is a vital petrophysical parameter in play evaluation and reserve estimation. However, it is difficult to calculate hydrocarbon saturation in unconventional shale plays due to variabilities in lithology, high clay content, presence of organic matter and pyrite, low porosity and complex wettability [7], [8]. Dielectric logging is a method that distinguishes reservoir fluid properties through the difference between relative dielectric permittivity of hydrocarbon/ rock minerals and water [9], [10]. The relative dielectric permittivity of hydrocarbon and most rock-forming minerals ranges from 2 to 10, while this value for water varies from 50 to 80 [7], [8]. Therefore, the dielectric permittivity of any reservoir depends largely on the water content of the rock. The Array Dielectric Tool (ADT) developed by Schlumberger provides permittivity and conductivity at four different megahertz range frequencies [11]. Considering the application of ADT logging in shale oil plays, since these reservoirs are extremely tight and mud intrusion in open wells is almost negligible, the saturation in the detection range of dielectric logging would be close to the saturation of an uninvaded formation. This makes dielectric logging suitable for reservoir evaluation and fluid saturation estimation in unconventional reservoirs, in particular.
In recent years, dielectric dispersion logs have been widely used in the evaluation of unconventional reservoir, such as shale oil. Seleznev et al. [12] measured the dielectric permittivity of shale rocks of Green River Formation, further inverted the relative dielectric permittivity of different minerals by (Complex Refractive Index) CRIM method, and applied them to process real logging data. Combining elemental capture spectroscopy (ECS), nuclear magnetic resonance (NMR) with dielectric log data, Musharfi et al. [13] comprehensively evaluated the petrophysical properties of shale gas reservoirs by CRIM methods. Almarzooq et al. [14] measured the dielectric permittivity of shale samples from a certain area located in Saudi Arabia and found that pyrite has a great impact on the dielectric responses of shale. Wang and Popitt [15] believed that at higher frequencies, there is still an evident interfacial polarization in shale reservoirs with high clay content, thus CRIM model may not be applied in such cases. Chen and Heidari [16] studied the pore-scale simulation of dielectric constant for estimating hydrocarbon. In their study, the effects of both organic matter and pyrite content on the relative dielectric constant of the formation as a whole were studied. They found that total organic carbon (TOC) content would affect the dielectric constant, while for pyrite both distribution and content would show a significant impact on the dielectric constant. As a result, they proposed an improved CRIM model by incorporating water distribution in tortuous pathways in the formation. Based on the similarity in dielectric properties between organic matter and hydrocarbon, Zhao et al. [17] improved the pseudo-Archie model and applied it to the saturation calculation of Lucaogou shale oil reservoirs in Jimusar Sag, Junggar Basin.
Pratiksha et al. [18], [19] utilized dielectric dispersion logs measured at four different megahertz range frequencies to evaluate the Wolfcamp and Bakken shale reservoirs, respectively. Based on the Waxman-Smith (WS) equation, CRIM and Stroud-Milton-De (SMD) model [20], Pratiksha [18], [19] obtained water saturation, formation water conductivity, Archie parameters m and n by joint inversion of resistivity and dielectric dispersion logs. They pointed out that the relative error is high in the intervals with higher pyrite content. However, SMD model is more suitable for clean sandstone reservoirs, and the dielectric dispersion is enhanced in pyrite and clay-rich shale formations. In addition, the Levenberg-Marquardt algorithm they used was sensitive for initial values for the unknown parameters. Han et al. [7] calculated the hydrocarbon saturation in a Lower-Paleozoic organic-rich shale gas formation based on the Markov-chain Monte Carlo (MCMC) stochastic inversion of broadband electromagnetic logs with the interfacial polarization IP model [21], [22]. It is noteworthy that defining equations of the IP model would vary based on different grain shapes, however, the real clay or pyrite shapes cannot be characterized by simple shapes. Besides, the MCMC inversion requires prior information about the reservoir properties.
Simulated annealing (SA) algorithm is a general probabilistic algorithm, which is used to find the global optimal solution of a proposition in a large search space [23]. It has been widely used in geosciences and engineering field [24], [25], [26], [27]. Compared with Markov-chain Monte Carlo (MCMC) stochastic inversion, it is not sensitive to the initial guess and does not require any prior information. In addition, inspired by the SMD model, Han et al. [28] proposed a shaly sandstone (SHSD) model that considers the effect of shale on dielectric dispersion. This model provides more accurate hydrocarbon saturation in shaly sandstone reservoirs.
Based on what was above, in this paper, a method that combines SHSD and CRIM models to calculate the hydrocarbon saturation of shale oil reservoirs by the simulated annealing algorithm of dielectric dispersion logs is proposed. The method is applied to the Triassic Yanchang Formation, Ordos Basin, China, which is a prominent shale oil reservoir. The effectiveness and reliability of the proposed model is verified by synthetic log responses and core data from the lab.
Section snippets
Data
For this study, the conventional, NMR and dielectric dispersion logs along with data from core plus retrieved from two wells were collected and analyzed. Conventional and NMR logs are used to calculate accurate porosity and mineral volume contents. The dielectric dispersion logs provide four permittivity and conductivity, which were measured at 22 MHz, 100 MHz, 350 MHz and 960 MHz. The core data includes porosity, and hydrocarbon and water saturations. The core and logging data were taken from
Interpretation models for dielectric dispersion logs
CRIM is a commonly used model for interpreting the complex dielectric constant of dielectric dispersion logs at around 1 GHz. It does not include the variation of texture because the texture affects the interfacial polarization but the interfacial polarization around 1 GHz is negligible. The formulation of the CRIM for a three-phase model is [11]:where is the complex relative permittivity of the rocks; is the total porosity of the rock; Sw is water saturation;
Proposed inversion method to synthetic logs with assumed parameters
The proposed inversion method is first applied to the synthetic logs to test its robustness. We synthesize four permittivity and four conductivity logs at 22 MHz, 100 MHz, 350 MHz, 960 MHz by assuming formation properties close to the Chang 7 member summarized in Table 1. The porosity is set as 10%, wet-clay dispersive phase parameters pd, are set as 0.05, 5, and 650, respectively, while the salinity is assumed 20ppk. These values are utilized to generate the permittivity and conductivity
Conclusions
In this paper, we proposed a method utilizing two interpretation models combined with simulated annealing algorithm to process four dielectric permittivity and four conductivity logs measured at four frequencies in shale oil reservoirs. The output parameters are the hydrocarbon saturation, rock textual parameter, volume fraction of clay-induced dispersive phase, and water salinity. The proposed method is applied to shale oil reservoirs of Triassic Yanchang Formation, Ordos Basin, China. The
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
This paper is supported by the National Science and Technology Major Project of China (2016ZX05050-008) and the National Natural Science Foundation of China (41574121, 51874262).
References (39)
- et al.
Formation mechanism, geological characteristics and development strategy of nonmarine shale oil in China
Petrol Explor Dev
(2013) - et al.
Exploration potential of shale oil in Chang7 member, upper Triassic Yanchang formation, Ordos Basin, NW China
Petrol Explor Dev
(2016) - et al.
Hydrocarbon saturation in a Lower-Paleozoic organic-rich shale gas formation based on Markov-chain Monte Carlo stochastic inversion of broadband electromagnetic dispersion logs
Fuel
(2019) - et al.
Petrophysical parameters prediction and uncertainty analysis in tight sandstone reservoirs using Bayesian inversion method
J Nat Gas Sci Eng
(2018) - et al.
Upper Paleozoic petroleum system, Ordos Basin, China
Mar Petrol Geol
(2005) - et al.
Mesozoic structural evolution of the Hangjinqi area in the northern Ordos Basin, North China
Mar Petrol Geol
(2015) - et al.
The dissolution characteristics of the Chang 8 tight reservoir and its quantitative influence on porosity in the Jiyuan area, Ordos Basin, China
J Nat Gas Geosci
(2018) - et al.
Factors controlling the reservoir accumulation of Triassic Chang 6 Member in Jiyuan-Wuqi area, Ordos Basin, NW China
Petrol Explor Dev
(2019) - et al.
Training ANFIS structure using simulated annealing algorithm for dynamic systems identification
Neurocomputing
(2018) - et al.
Very fast simulated re-annealing (VFSA) approach for land data assimilation
Comput Geosci-UK
(2004)