Journal of Geophysics and Engineering

Journal of Geophysics and Engineering (2022) 19, 833–845 https://doi.org/10.1093/jge/gxac052

TOC estimation of shale oil reservoir by combining

nuclear magnetic resonance logging and nuclear
physics logging

Downloaded from https://academic.oup.com/jge/article/19/4/833/6672604 by University of Bahrain user on 01 November 2023

Ziyan Feng1,2 , Cheng Feng 2,
*, Yuntao Zhong1,2 , Zhijun Qin3,4 , Rui Mao4 , Lei Zhao1,2
and Xianghua Zong2
1
School of Geophysics, China University of Petroleum-Beijing, Beijing, 102249, China
2
School of Petroleum, China University of Petroleum-Beijing at Karamay, Karamay, 834000, China
3
School of Geoscience and Technology, China University of Petroleum, Qingdao, Shandong, 266580, China
4
Research Institute of Petroleum Exploration and Development, PetroChina Xinjiang Oilfield Company,
Karamay, 834000, China

*Corresponding author: Cheng Feng. E-mail: [email protected]

Received 22 March 2022, revised 9 May 2022

Accepted for publication 15 June 2022

Abstract
The evaluation of source rock properties has become a vital step in logging interpretation. Total
organic carbon (TOC) content is the key to estimating the quality and hydrocarbon generation
potential of source rocks. In the shale oilfield of the Junggar Basin, the conventional method of
calculating the TOC of hydrocarbon source rocks cannot satisfy logging evaluation
requirements. This paper predominantly deals with a method for the quantitative estimation of
TOC in source rocks via nuclear physics and nuclear magnetic resonance (NMR) logs.
According to this method, the total hydrogen index of the source rock is the sum of the response
of kerogen, clay minerals and fluid, expressed by corrected neutron porosity. The hydrogen index
of fluid and clay minerals is indicated by the effective porosity of NMR and the estimated clay
content, respectively. To eliminate the hydrogen index of fluid, the effective NMR porosity is
subtracted from the corrected neutron porosity. On this basis, a new and overlapping method
suitable for clay-rich rocks and oil reservoirs is proposed. This method was developed by
overlaying the scaled clay content curve on the hydrogen index curve. In non-source rocks, the
two curves regularly overlap. However, in organic-rich rocks the two curves will separate. The
separation distance between the two curves was used to estimate TOC continuously. Possessing
sound application and benefiting from the measured results of sweet spots, this method provides
new insights for TOC quantitative prediction in shale oil reservoirs.
Keywords: shale oil reservoirs, TOC estimation, hydrogen index, NMR log

1. Introduction as source-reservoir integration characteristics (Zou et al.

Shale oil reservoirs, a significant field of unconventional oil 2013). Previous studies show that brittleness, in situ stress
and gas exploration, are an important replacement resource anisotropy, quality of source rock combined with the ‘four
of oil and gas for the world of tomorrow (Stevens et al. properties’ of conventional reservoirs (comprising lithology,
2013; Madi et al. 2015). However, shale oil reservoirs pos- physical property, electrical property and oiliness) form the
sess strong heterogeneous rock physical properties as well ‘seven properties’ for shale oil reservoirs (Kuang et al. 2021).

© The Author(s) 2022. Published by Oxford University Press on behalf of the Sinopec Geophysical Research Institute. This is an Open Access article distributed under the terms of 833
the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium,
provided the original work is properly cited.
Journal of Geophysics and Engineering (2022) 19, 833–845 Feng et al.

Evaluation of source rock properties includes the abundance, 2016), Gaussian sequence simulation (Nezhad et al. 2018)
type and maturity of organic matter. These jointly control and the comprehensive application of a variety of intelligent
and influence the hydrocarbon generation potential of source algorithms (Zhu et al. 2020) have been developed to evalu-
rocks. Total organic carbon (TOC) plays an important role in ate TOC. While highly precise, these intelligent algorithms
estimating the abundance of organic matter. Organic matter maintain high requirements on data sets but lack the theoret-
can be obtained through organic geochemical analysis of core ical guidance of physical models.
samples. The traditional experimental method, however, is Based on the physical volume model of source rocks and
expensive and time-consuming, and produces results that do the typical logging response characteristics of kerogen, this
not explain TOC’s vertical and horizontal variation trend in paper proposes a TOC estimation method combining nu-
the reservoir (Xu & Zhu 2010). clear physics logging (Density logging (DEN), compensated

Downloaded from https://academic.oup.com/jge/article/19/4/833/6672604 by University of Bahrain user on 01 November 2023

To overcome the problem of limited core samples and neutron logging (CNL) and gamma logging) and nuclear
to further obtain continuous TOC evaluation results, schol- magnetic resonance (NMR) logging. The two sweet spots in-
ars have carried out numerous studies on TOC estimation terval core data of the Permian Lucaogou Formation shale oil
based on logging data. These mainly involved the single curve reservoir in the Jimsar Sag are applied, showing a comparison
method, natural gamma-ray logging and density logging between estimated results and experimental data. The com-
curves, as well as the multiple regression, carbon/oxygen parison demonstrates that the method is reliable in the TOC
logging, curve overlap, ΔlogR and matrix density overlap quantitative prediction of a shale oil reservoir.
methods. The advantages of the single curve method are
its basic principles and convenient operation. However, it is
2. Geological background
compromised by environmental factors such as well diame-
ter, fluid and lithology (Sehmoker 1981; Fertl & Chilingar As shown in figure 1a and b, the Jimsar Sag, located in
1988). Usually, TOC estimation results calculated in this the southwest of the eastern uplift belt of Junggar Basin, is
way are inaccurate. Compared with the single curve method, bounded by thrust faults in the south, west and north of the
the multiple regression method considers more conventional sag ( Jiang et al. 2015; Cao et al. 2016). The Jimsar Sag has
logging information and is more accurate. However, more experienced Hercynian, Indosinian, Yanshan, Himalayan and
regression parameters mean more complex weight alloca- other multi-stage orogenies, and closed in the late Middle
tion (Tan et al. 2021). Carbon/oxygen logging can exclude Permian, forming a relatively independent lacustrine basin
the influence of formation water salinity (Zhao et al. 1994) environment (Guo et al. 2019). As the main production layer,
but is not suitable for complex mineral types of reservoirs. the Lucaogou Formation experienced multiple periods of la-
The ΔlogR method establishes the relationship among log- custrine regression and transgression during the sedimentary
ging curve, organic matter abundance and kerogen maturity period (Qiu et al. 2016a). As shown in figure 1c, the lithology
(Passey et al. 1990). When the resistivity curve changes fre- of the Lucaogou Formation is diverse, characterized by the
quently, however, it is difficult to determine the baseline po- interbedded distribution of thin bedded clastic and chemical
sition and maturity parameters. To enhance the applicability rocks, including arenaceous dolomite, feldspathic litharenite,
of the ΔlogR method and improve prediction accuracy, some and mudstone, as well as a small amount of limestone (Kuang
scholars have refined the model form and parameters (Wang et al. 2015; Qiu et al. 2016b; Su et al. 2019).
et al. 2020a; Zhu et al. 2019). The matrix density overlap According to lithologic characteristics and logging curves,
method proposed by Jacobi et al. (2008) can effectively elim- Lucaogou Formations can be divided into the upper and
inate the influence of inorganic carbon. However, it needs lower Lucaogou Formation (P2 l2 , P2 l1 ). There are two
elemental capture spectroscopy logging data. Ramirez et al. ‘sweet spots’ (P2 l2 2 , P2 l1 2 ) composed of the interbedding
(2011) proposed a method to estimate TOC by subtract- of fine-silty siltstone, carbonate rock and organic-rich mud-
ing the density porosity from nuclear magnetic total porosity stone. Source rocks of Lucaogou Formation are high quality,
based on the difference between the measurement principles with mature source rocks widely distributed in the Jimsar
of nuclear magnetic and density logging in shale reservoirs. Sag possessing strong hydrocarbon generation potential
However, complex mineral types can lead to instability of (Kuang et al. 2014; Qiu et al. 2016a). Organic geochemical
skeleton density parameters. Zhao et al. (2016) proposed a data show that the kerogen type of shale reservoir in the
new method based on the overlap of natural gamma ray and Lucaogou Formation is mainly type II; siliceous mudstone
clay indicator parameters, making up for the deficiency of the and dolomitic mudstone are predominantly type I and type
ΔlogR method in overmature source rocks. It is not suitable, II, and silky mudstone is type II and type III (Cao et al.
however, for continental shale reservoirs with high potas- 2016; Qiu et al. 2016b). The source rocks have high organic
sium content. In addition to the previous mentioned evalua- matter abundance: the floor of TOC is >1% with the average
tion methods, artificial intelligent algorithms such as artificial TOC reaching >4%. The organic matter abundance of mud-
neural networks (Mahmoud et al. 2017), the support vector stone and dolomite is higher than that of silty mudstone
machine (De et al. 2019), extreme learning machine (Shi et al. (Xiang et al. 2013). At present, the source rocks of
834
Journal of Geophysics and Engineering (2022) 19, 833–845 Feng et al.

Downloaded from https://academic.oup.com/jge/article/19/4/833/6672604 by University of Bahrain user on 01 November 2023

Figure 1. Location of the study area (Cao et al. 2016).

Lucaogou Formation are in the mature stage of the ‘oil

generation window’—vitrinite maturity is between 0.7 and
1%—showing high production potential ( Jiang et al. 2015;
Su et al. 2019).

3. Rock physical volume model

For shale reservoirs, scholars maintain that organic matter is
enriched only in source rock such as mud shale or calcare-
ous mudstone, while only a small or negligible amount ex-
ists in other lithologic reservoirs (Wang & Guo 2000; Liu Figure 2. Physical volume models of non-source rocks (left) and source
rocks (right).
2009). Figure 2 shows the differences between non-source
and source rocks in rock physical volume models (Passy et al.
1990; Lei et al. 2010). Non-source rocks mainly consist of sity and acoustic wave propagation speed is lower than com-
non-clay minerals, clay minerals, pore water and hydrocar- mon minerals (Wang & Guo 2000; Lewis et al. 2004; Ward
bon. In addition to these four elements, the source rocks also 2010). According to the physical volume models of source
contain some kerogen. The evaluation of TOC by logging rocks, the resistivity curve is affected by many factors such as
curve is based on the typical logging response characteristics the mineral type of rock matrix, pore fluid type, pore struc-
of kerogen (figure 3), summarized as ‘four high and one low’. ture, wettability and oil-bearing properties. Thus, the log-
‘Four high’ refers to high natural gamma ray, neutron poros- ging response mechanism is complicated. The three porosity
ity, interval transit time and resistivity, while ‘one low’ refers curves (density, neutron porosity and acoustic time) are the
to low density. comprehensive response of rock matrix, kerogen and pore
Previous studies on the physical and chemical properties fluid. The natural gamma-ray logging signal mainly comes
of kerogen have concluded that kerogen is rich in hydro- from the radioactivity of the source rock matrix. The or-
gen and has high radioactivity and low conductivity. Its den- ganic matter of source rock can absorb uranium ions in the

835
Journal of Geophysics and Engineering (2022) 19, 833–845 Feng et al.

Downloaded from https://academic.oup.com/jge/article/19/4/833/6672604 by University of Bahrain user on 01 November 2023

Figure 3. Typical logging response characteristics of source rock.

poses an estimation method of TOC in source rocks based

on nuclear physics logging (density, neutron and gamma log-
ging) combined with NMR logging.
The compensated neutron logging principle dictates that
the formation deceleration for neutrons is largely dependent
on the hydrogen content (Huang et al. 2008). Compensated
neutron logging converts the thermal neutron counting rate
into hydrogen index (HI) or neutron porosity by calibrating
in water-bearing limestone calibration test pits. In the shale
oil reservoir of the Lucaogou Formation, the main fluid types
include free oil, absorbed oil, free water and bound water.
The hydrogen index of liquid hydrocarbons, however, is not
at all dissimilar to that of fresh water (HI = 1), resulting in
a negligible influence of fluid property change on compen-
Figure 4. The crossplot of TOC and gamma log (R2 = 0.0099). sated neutron logging in shale oil reservoirs. According to the
XRD analysis data in Table 1, the common minerals of the
formation and increase the radioactivity, the result of which Lucaogou Formation are primacy dolomite, calcite, quartz
is the natural gamma-ray curve reflecting the influence of and plagioclase, followed by potash feldspar and clay min-
organic matter. However, previous studies suggest that vol- erals (mainly montmorillonite and illite). The compensated
canic hydrothermal activity is an important geological origin neutron logging responses of these common minerals are
of Permian Lucaogou Formation source rocks (Zhang 2019; shown in Table 2 (Schlumberger 1998; Yong & Zhang 2002).
Ding et al. 2019; Wang et al. 2020b). Therefore, there are Except for clay minerals, the compensated neutron logging
many tuffaceous components composed of alkaline feldspar response values of other minerals (dolomite, calcite, quartz,
and quartz derived from alkaline magma in the Lucaogou plagioclase and potash feldspar) are very small. In summary,
Formation ( Jiang et al. 2015; Zhang 2019), resulting in poor the compensated neutron logging response of source rocks
response sensitivity of natural gamma-ray logging to organic can be simplified into a volume model composed of clay min-
matter (figure 4). erals, kerogen and pore fluid, as
ΦN = Vk Φk + Vcl Φcl + 𝜙Φf , (1)
4. Theory and methodology
where ΦN is the neutron log response (total hydrogen in-
According to logging data collected so far, there is a lack of dex) of shale reservoir, percentages of Φk , Φcl and Φf are
data on natural gamma spectrometry. However, high-quality the hydrogen indices of kerogen, clay mineral and pore fluid,
NMR logging data is sufficient. Consequently, this study pro- respectively; Vk , Vcl and Φ are the volume specific gravity of

836
Journal of Geophysics and Engineering (2022) 19, 833–845 Feng et al.
Table 1. The XRD analysis data of some shale samples in the Jimsar Sag (3100–3450 m, Well JI X)

Sample Dolomite Calcite Quartz Plagioclase K-Feldspar Pyrite Clay

No. (wt.%) (wt.%) (wt.%) (wt.%) (wt.%) (wt.%) (wt.%)

S1 10.6 6.3 30.4 27.6 5.7 0 19.4

S2 14 25.8 20.7 21 3.6 0 14.9
S3 15.2 14 17.7 30.1 3 0 20
S4 40.6 0.5 17.7 23 4.7 0.4 13.1
S5 43.1 8.8 16.4 17.1 2.3 0 12.3
S6 56.2 1.6 13.4 13.9 0.8 0 14.1
S7 18.5 19.5 15.9 23.5 5.3 2.8 14.5
S8 21.8 2.2 20.9 33.6 7.3 0.3 13.9

Downloaded from https://academic.oup.com/jge/article/19/4/833/6672604 by University of Bahrain user on 01 November 2023

S9 29.5 9.7 19.3 23.8 2.3 0 15.4
S10 30.04 3.34 13.35 38.93 11.12 0 3.22
S11 28 4.9 19.5 27.2 5 2.8 12.6
S12 17 14.8 17.1 29.1 5.2 1.4 15.4
S13 30.1 2.5 20 28.3 5.2 0.6 13.3
S14 5 18.3 18.5 33 5.8 3.3 16.1
S15 6.6 13.1 24.6 30.7 6.9 1 17.1
S16 33.6 21.8 16.4 11.4 1.5 1.2 14.1
S17 13.4 16.6 26.6 17.4 5 0.5 20.5
S18 43.3 7.9 18.7 13.3 3.1 0 13.7
S19 1.5 24 9 48.9 4.9 0 11.7
S20 2.4 11.3 13.2 50.2 6 1.5 15.4
S21 0.8 34.1 6.5 47.4 4.5 0 6.7
S22 0.7 30.8 5.5 54.3 3.4 0 5.3
S23 3.6 19.3 6.3 58.7 4.8 0 7.3
S24 1.1 12.1 7.2 67.4 4.6 0 7.6
S25 1.3 10.3 8.2 64.9 6.7 0.5 8.1
S26 1.1 13.9 8.4 62.2 5.1 0 9.3
S27 1 33 6.9 49.5 4 0 5.6
S28 1.2 33.5 8 47.8 3.8 0.4 5.3

Table 2. Compensated neutron logging response empirical values of ging to the kerogen (Vk Φk ) in equation (1), and to reduce
main minerals in source rocks of the Lucaogou Formation the influence of other minerals in the rock matrix on the mea-
Compensated neutron logging surement results, the same calibration standard of CNL and
Minerals response values (porosity unit) DEN logging is used to correct the neutron log response of
illite about 30.0
the shale reservoir.
montmorillonite about 45.0 It is well established that CNL and DEN logs are calibrated
calcite 0.0 in water-saturated limestone standard wells. When the same
dolomite 0.5 calibration is used for both, the CNL and DEN log curves
quartz −2.0 of limestone overlap, as would be the case if neutron poros-
plagioclase about −1.4
ity ranged from 45 to −15% and density ranged from 2.95
potassium feldspar about −1.0
to 1.95 g/cm3 (Yong & Zhang 2002). Dolomite has strong
neutron deceleration ability, leading to high neutron poros-
kerogen, clay minerals and porosity, respectively, as a percent- ity. However, its matrix density is higher than that of lime-
age. stone, resulting in low density porosity calculated via the vol-
ume model. Similarly, sandstone (quartz and feldspar min-
erals) has a weak neutron deceleration ability, leading to low
4.1. The total hydrogen index of shale
neutron porosity. Its matrix density is lower than that of lime-
In Table 2, the neutron deceleration ability of quartz and stone, resulting in higher density porosity calculated via the
feldspar minerals is weaker than that of calcite, resulting in an volume model (Huang et al. 2008).
equivalent hydrogen index of <0. On the contrary, the neu- Based on this, the average value of neutron porosity and
tron deceleration ability of dolomite is stronger than that of apparent density porosity is taken to reduce the influence of
calcite, resulting in the equivalent hydrogen index >0. To fur- minerals on the total hydrogen index of shale. In addition, be-
ther improve the sensitivity of the compensated neutron log- cause density logging instruments consider kerogen response

837
Journal of Geophysics and Engineering (2022) 19, 833–845 Feng et al.

Downloaded from https://academic.oup.com/jge/article/19/4/833/6672604 by University of Bahrain user on 01 November 2023

Figure 5. Conceptual model showing the correlation between NMR T2 spectrum and various pore fluids.

as part of the pore response, density porosity contains both (interlayer water) and OH− on the surface of clay particles
kerogen and pore fluid effects (Gao et al. 2014). Therefore, (adsorbed water). As the T2 time of the first two forms is
the aforementioned hydrogen index correction method does <0.1 ms, NMR logging failed to obtain relevant information
not greatly influence the hydrogen index of kerogen and pore (Fleury & Romero 2016). The third form can be calculated
fluid. Equations (2) and (3) are used to correct the total hy- by ensuring the cutoff value of the T2 distribution. Hence,
drogen index of shale (the ΦN of equation (1)): the T2 spectrum contains a portion of hydrogen nuclei from
the contribution of clay minerals. To whittle down the ef-
forts of clay minerals on the hydrogen index of pore fluid, the
ΦD = (𝜌ma − 𝜌b )∕(𝜌ma − 𝜌f ) × 100%, (2)
clay-bound water portion needs to be separated from the T2
ΦN + ΦD spectrum. This means that the calculation of NMR effective
Φ= , (3) porosity (ɸe in figure 5) is required.
2
There are two methods to obtain the value of ɸe . The first
where Φ is the neutron log response (total hydrogen index) is experimentally measuring the effective porosity of the core
of shale after correction (the correction of ΦN ), as a per- directly and the corresponding T2 spectrum. The second is
centage; ΦD is apparent density porosity (%); ΦN is neutron to compare the porosity calculated by NMR logging at differ-
porosity (%); 𝜌b is the density logging value (g/cm3 ); 𝜌ma ent T2 times with the effective porosity analysis data of the
is the matrix density of limestone (2.71 g/cm3 ) and 𝜌f is the core. Wang et al. (2019) achieved an accurate calculation of
density of the pore fluid (1.0 g/cm3 ). the effective porosity of shale oil reservoirs in the Lucaogou
Formation of the Jimsar Sag by using an iterative method and
determined that the start time to obtain the effective porosity
4.2. The hydrogen index of pore fluid was 1.7 ms. Effective porosity can be obtained by summing
NMR logging can also measure the hydrogen index but the up the porosity components of the T2 spectrum after 1.7 ms,
measurement objects are different from compensated neu- as
tron logging. As shown in Table 1, paramagnetic minerals ∑
3000ms
𝜙e = 𝜙i , (4)
such as pyrite are rarely developed in the source rocks. We
1.7ms
can regard the NMR response as unaffected by rock skele- where ɸe is the the NMR effective porosity (%) and ɸi is
ton (Deng & Xie 2010; Feng et al. 2017, 2021). On the the porosity component corresponding to different T2 times
other hand, as hydrogen nuclei of organic matter are often at- (%). Due to the hydrogen index of liquid hydrocarbons be-
tached to pore surfaces in solid form, their transverse relax- ing very close to that of fresh water (Φf = 1), the mean-
ation time is very quick in detecting their signals via NMR ing of ɸe can be regarded as the hydrogen index of pore
logging (Ramirez et al. 2011). Thus, the response of NMR fluid (ɸΦf ).
logging is almost unaffected by the solid skeleton and organic
matter of source rocks.
Previous studies have shown that the NMR response char-
4.3. The hydrogen index of clay mineral
acteristics of clay minerals depend on the form and type
of clay (Prammer et al. 1996; Jia 2017). Hydrogen in clay Combining the forms equations (3) and (4) and integrating
minerals mainly takes three forms: OH− of the clay struc- with equation (1), we subtracted ɸe from the total hydrogen
ture (structural water), OH− at the edges of clay platelets index of the source rock. Following subtraction of ɸe, , we can

838
Journal of Geophysics and Engineering (2022) 19, 833–845 Feng et al.

Downloaded from https://academic.oup.com/jge/article/19/4/833/6672604 by University of Bahrain user on 01 November 2023

Figure 6. Cross plots of measured clay content with ΦND (a), ɸe (b), GR (c), AC (d), CNL (e) and DEN (f).

see the sum of hydrogen index contributed to by clay minerals gen was taken as 1.175 g/cm3 (Yao 2019), with the neutron
and kerogen, as porosity of type I and type II kerogen usually at 86 poros-
Φ − 𝜙e = Vk Φk + Vcl Φcl . (5) ity unit (Zhao et al. 2015). The estimated ΦND of kerogen is
∼3.54%, differing considerably from that of clay minerals in
To further discern the contribution of kerogen to the hy- Table 2. In conclusion, ΦND is an important sensitive param-
drogen index (Vk Φk in equation (5)), it is necessary to use eter to characterize clay minerals content:
other log curves and parameters to imply the clay minerals
ΦND = (ΦN − ΦD ), (6)
content and separate it from the residual hydrogen index.
Previous studies have summarized the distribution range of where ΦND is the subtracted result between neutron poros-
neutron porosity and density of four common clay mineral ity and density porosity (%) that will be positively correlated
types (Schlumberger 1998; Mao 2001; Yong & Zhang 2002). with clay content.
These studies showed that the subtracted result between neu- In addition, clay-bound water is related to clay content
tron porosity and density porosity (ΦND ) of clay minerals and type. Thus, ɸe is negatively correlated with clay con-
generally shows a significant positive difference and is much tent (Deng & Xie 2010). Due to the lack of a natural
larger than other mineral types. The average density of kero- gamma-ray spectral log, a multivariate statistical regression

839
Journal of Geophysics and Engineering (2022) 19, 833–845 Feng et al.

Downloaded from https://academic.oup.com/jge/article/19/4/833/6672604 by University of Bahrain user on 01 November 2023

Figure 7. TOC was estimated by overlapping Φk+cl curve and Vcl curve.

model for quantitative prediction of clay minerals content At this point, the quantitative relationship can be estab-
was established by using natural gamma ray, neutron poros- lished between TOC and the separation distance of Φk+cl
ity, density, acoustic time difference, ɸe and ΦND (He et al. curve and Vcl curve in source rock, as shown in equations (8)
2020). As shown in figure 6, the single factor correlation anal- and (9) as follows:
ysis above these parameters and the clay minerals content ( )
obtained by the XRD experiment was carried out. From fig- Φk+cl − (Φk+cl )min Vcl − (Vcl )min
Δs = −
ure 6, the correlations between the clay content and ΦND , ɸe (Φk+cl )max − (Φk+cl )min (Vcl )max − (Vcl )min
and GR are relatively high, while the correlations between
×100%, (8)
the clay content and AC, CNL and DEN are relatively low.
Ultimately, we established a multivariate statistical regres-
TOC = k × Δs + b, (9)
sion model for clay content using ΦND , ɸe and GR, as shown
in equation (7), with correlation coefficient is 0.8388:
where Δs is the distance between the Φk+cl curve and the Vcl
Vcl = 0.7152 × ΦND + 0.0589 × GR curve, indicated by the yellow area in figure 7 (%); (Φk+cl ) min
and (Φk+cl ) max are the minimum scale and maximum scale
−0.5277 × 𝜙e + 6.1214, (7)
of the Φk+cl curve, respectively; (Vcl ) min and (Vcl ) max are the
where Vcl is the estimated clay content (%). minimum and maximum scales of Vcl curve, respectively, and
k and b are the slope and intercept of TOC quantitative evalu-
ation model, respectively. The details of the calibration of the
4.4. The new overlap method values of k and b in equation (9) are as follows: first, as shown
For the sake of description, Φ-ɸe in equation (5) is defined in equation (8), (Φk+cl )min , (Φk+cl )max , (Vcl )min and (Vcl )max
as parameter Φk+cl . This includes the contribution of kero- were used to normalize the Φk+cl curve and Vcl curve respec-
gen and clay minerals to the hydrogen index. Vcl is positively tively; next, based on the normalized Φk+cl curve and normal-
correlated with the contribution of clay minerals to the hy- ized Vcl curve, the distance Δs was obtained by deducing the
drogen index. In the non-source rock segment both Φk+cl and difference; finally, the values of k and b in equation (9) are
Vcl mainly reflect the contribution of clay minerals, render- calibrated using linear regression between Δs and the actual
ing their curves very similar. However, in the source rock seg- measured TOC.
ment Φk+cl not only reflects the contribution of clay miner-
5. Application and verification
als but the additional contribution of kerogen is also seen.
The two curves will present different variations. Therefore, The aforementioned methods are applied to the actual well
the curve scale range is adjusted appropriately until the Φk+cl data, with application results shown in figure 8, representing
curve and Vcl curve overlap in the non-source rock segment, the upper sweet spot (figure 8a) and the lower sweet spot
bringing about the separation of the two in the source rock of Well JI A (figure 8b), respectively. In figure 8, the first
segment (figure 7). track is depth, with the second track including lithologic logs
840
Journal of Geophysics and Engineering (2022) 19, 833–845 Feng et al.

Downloaded from https://academic.oup.com/jge/article/19/4/833/6672604 by University of Bahrain user on 01 November 2023

Figure 8. Conventional and NMR logs and evaluation results from the upper sweet spot (a) and the lower sweet spot (b) in Well JI A.

(Gamma ray logging (GR), spontaneous potential logging transitional zone resistivity, and flushed zone resistivity), the
and caliper logging). The third track includes the porosity fifth is the NMR T2 spectrum and the sixth is ΔlogR as well
logs (DEN, CNL and acoustic logging (AC)), the fourth in- as the baselines of AC and RT are 223 𝜇s/m and 10 Ω⋅m
cludes the resistivity logs (True formation resistivity (RT), respectively. The seventh track includes the Φk+cl curve and
841
Journal of Geophysics and Engineering (2022) 19, 833–845 Feng et al.

Downloaded from https://academic.oup.com/jge/article/19/4/833/6672604 by University of Bahrain user on 01 November 2023

Figure 9. Cross plots of actual measured TOC and Δs of the upper sweet spot (a) and the lower sweet spot (b) in Well JI A.

Vcl curve, while the eighth track includes the TOC predicted Table 3. The summary forecasting statistics contrasting the estimated
by the ΔlogR method (the dotted green line) and the actual results with the core data
measured TOC (red dot). The final track includes the TOC Criteria Upper sweet spot Lower sweet spot
estimated by the proposed method (the solid black line) and
Number 62 40
the actual measured TOC.
Coefficient of determination 0.797 0.846
The experimental measurement process of TOC in cores RMSE 1.410% 0.732%
conforms to the corresponding Chinese national mandatory Average error 8.227% 6.561%
standard (GB/T 19145–2003). The general experimental
process is as follows: first, the sample was crushed into par-
ticles <80 mesh before diluted hydrochloric acid was added is close to the overall trend of measured TOC in cores. How-
to remove the inorganic carbon of carbonate minerals; next, ever, the new method is much more accurate.
the particles were combusted in a high-temperature oxygen Figure 9 parts a and b are the cross plots of actual mea-
stream after drying, releasing CO2 ; finally, the LECO carbon sured TOC and the distance between the Φk+cl curve and Vcl
analyzer was used to measure the TOC. curve (Δs) in the two sweet spots of Well JI A, with correla-
As shown in figure 8b, the section below ‘XX60m’ is the tion coefficients of 0.797 and 0.846, indicating that TOC and
non-source rock zone. This can be distinguished from the Δs maintain sound linear correlation. As shown in figures 8a
source rock zone in two respects. First, the TOC measured in and 9a, the equation to estimate the TOC of the upper sweet
the experiment is low (<1%), indicating that the abundance spot of Well JI A can be expressed as
of organic matter in this section is very low, with the thick- TOC(%) = 0.311 × Δs + 0.0312, R2 = 0.797. (10)
ness of this section stable. Second, the log curves of source In the same way, as shown in figures 8b and 9b, the equation
rocks above ‘XX60m’ are dentiform, with logs characterized to evaluate the TOC of the lower sweet spot of Well JI A can
by higher natural gamma, resistivity, neutron porosity and be expressed as
lower density than those of non-source rocks.
According to the non-source rock section of the lower TOC(%) = 0.2073 × Δs − 0.7109, R2 = 0.846. (11)
member of Well JI A, the scale range of Φk+cl curve and Vcl Table 3 sums up the forecasting statistics from the two sweet
curve were adjusted to ensure they basically overlap. The spot formations for contrasting the evaluated outcome with
confirmed values of (Φk+cl ) min and (Φk+cl ) max are 0 and the experimental data. The average error and root mean
50%, with (Vcl ) min and (Vcl ) max at 0 and 100%. As seen from squared error (RMSE) of the estimated TOC and that of
the final track in figure 8, the estimated results are in line with sealed coring measurement are ∼10 and 3%, respectively,
the actual measured TOC. In section ‘XX42∼XX52m’ in showing sound application effect. Using figure 9 and Table 3,
figure 8a, the measured TOC of the cores is very low. How- the feasibility and accuracy of this novel way of shale oil reser-
ever, the TOC predicted by the ΔlogR method is high. In voirs in the Lucaogou Formation were confirmed.
contrast, section ‘XX20∼XX28m’ in figure 8b shows, mea- As shown in figure 10, to further verify the extensi-
sured TOC of the cores is very high with low TOC predicted bility of equations (10) and (11), the measured TOC of
by the ΔlogR method. TOC predicted by the ΔlogR method another well (Well JI B) was selected to compare with the

842
Journal of Geophysics and Engineering (2022) 19, 833–845 Feng et al.

Downloaded from https://academic.oup.com/jge/article/19/4/833/6672604 by University of Bahrain user on 01 November 2023

Figure 10. Cross plots of measured TOC in cores and TOC estimated by models of the upper sweet spot (a) and the lower sweet spot (b) in Well JI B.

model estimation results. The result shows that the slopes of actual measured data were 0.797 and 0.846. The average er-
the upper and lower sweet spots are 0.894 and 0.904 (close rors and RMSE between the two sweet spots estimated TOC
to a diagonal distribution), with the correlation coefficient at and the experimental data were <10 and 3%, respectively. In
0.780 and 0.862, respectively, proving that the model estima- addition, the introduction of another well and coring data
tion results correspond with the experimental results. validated the extensibility of the new method. This method
corrected the influence of some matrix minerals on TOC
6. Conclusion evaluation results, overcoming the common failure to esti-
mate TOC with abnormal RT and GR. This method is of
Based on accumulated knowledge on the rock physical vol- great significance for the continuous evaluation of the hydro-
ume model of hydrocarbon source rocks and neutron re- carbon generation potential of shale oil reservoirs.
sponse characteristics of common minerals and fluids in
sedimentary rocks, this paper proposes a new overlapping
Acknowledgements
method combining NMR and nuclear physics logs to esti-
mate the TOC. This research was supported by the National Natural Science
First, considering the diversity of shale minerals, neutron Foundation of China (grant no. 42004089), the Natural Science
porosity is corrected by density porosity to obtain the final Foundation of Xinjiang Uygur Autonomous Region (grant no.
total hydrogen content index (Φ); second, the T2 spectrum 2021D01E22) and the Natural Science Foundation of Xinjiang
Uygur Autonomous Region (grant no. 2020D01A141).
after the optimal T2 cutoff value accumulated to obtain the
effective porosity of NMR (ɸe ), characterizing the contribu-
tion of pore fluid to the hydrogen index. The contribution Conflict of interest statement. None declared.
of pore fluid to the total hydrogen index of shale is elimi-
nated by the difference between Φ and ɸe , with the remaining References
hydrogen index (Φk+cl ) forming the sum contribution of clay
Cao, Z. et al. 2016. Lacustrine tight oil accumulation characteristics: Per-
minerals and kerogen. This study developed a suitable esti- mian Lucaogou Formation in Jimusaer Sag, Junggar Basin, International
mated model of clay minerals content (Vcl ) via the multivari- Journal of Coal Geology, 153, 37–51.
ate statistical regression method, the correlation coefficient De, S., Kumar, V. & Sengupta, D., 2019. Application of support vector re-
of which is 0.8388. Except in the case of clay minerals, in gression analysis to estimate total organic carbon content of Cambay
source rock layers, the difference between the Φk+cl curve and shale in Cambay basin, India–a case study, Petroleum Science and Tech-
nology, 37, 1155–1164.
the Vcl curve will expand as Φk+cl reflects the contribution of Deng, K.J. & Xie, R.H., 2010. Nuclear Magnetic Resonance Logging Theory
kerogen. Following calibration with core data, the separation and Application, pp. 1–27, China University of Petroleum Press.
distance between the two curves was used to evaluate TOC Ding, X. J., Qu, J. X., Imin, A., Zha, M., Su, Y., Jiang, Z. F. & Jiang, H., 2019.
sequentially. Organic matter origin and accumulation in tuffaceous shale of the lower
This new method was successfully used in the sweet spots Permian Lucaogou Formation, Jimsar Sag, Journal of Petroleum Science
and Engineering, 179, 696–706.
of a single well. The comparison of the new method with
Feng, C., Feng, J.G., Feng, Z.Y., Zhong, Y.T., Mao, Z.Q. & Ling, K.G., 2021.
the TOC measured by the actual core shows that the new Determination of reservoir wettability based on resistivity index predic-
method achieves sound results. The coefficients of correla- tion from core and log data, Journal of Petroleum Science and Engineering,
tion between the two sweet spots estimated TOC and the 205, 1–10.

843
Journal of Geophysics and Engineering (2022) 19, 833–845 Feng et al.

Feng, C., Shi, Y.J., Hao, J.F., Wang, Z.L., Mao, Z.Q., Li, G.R. & Jiang, Z.H., Nezhad, Y., Moradzadeh, A. & Kamali, M.R., 2018. A new approach to eval-
2017. Nuclear magnetic resonance features of low-permeability reser- uate organic geochemistry parameters by geostatistical methods: a case
voirs with complex wettability, Petroleum Exploration and Development, study from western Australia, Journal of Petroleum Science and Engineer-
44, 252–257. ing, 169, 813–824.
Fertl, W.H. & Chilingar, G.V., 1988. Total organic carbon content deter- Passey, Q.R., Creaney, S., Kulla, J.B., Moretti, F.J. & Stroud, J.D., 1990.
mined from well logs, SPE Formation Evaluation, 3, 407–419. A practical model for organic richness from porosity and resistivity
Fleury, M. & Romero, S.M., 2016. Characterization of shales using T1 –T2 logs, American Association of Petroleum Geologists Bulletin, 74, 1777–
NMR maps, Journal of Petroleum Science and Engineering, 137, 55–62. 1794.
Gao, Y.W., Zhao, Y.J., Li, G.L., Hu, T.T. & Fan, X. Q., 2014. Total organic Qiu, Z., Tao, H., Zou, C., Wang, H., Ji, H. & Zhou, S., 2016a. Lithofacies
carbon evaluation of tight oil source rocks in Junggar basin based on nu- and organic geochemistry of the Middle Permian Lucaogou Formation
clear magnetic logging and density logging, Petroleum Reservoir Evalua- in the Jimsar Sag of the Junggar Basin, NW China, Journal of Petroleum
tion and Development, 4, 5–8. Science and Engineering, 140, 97–107.

Downloaded from https://academic.oup.com/jge/article/19/4/833/6672604 by University of Bahrain user on 01 November 2023

Guo, X. G. et al. 2019. Evaluation and application of key technologies of Qiu, Z. et al. 2016b. Geological characteristics of source rock and reservoir
“sweet area” of shale oil in Junggar Basin: case study of Permian Lucao- of tight oil and its accumulation mechanism: a case study of Permian
gou Formation in Jimsar Depression, Natural gas Geoscience, 30, 1168– Lucaogou Formation in Jimsar sag, Junggar Basin, Petroleum Exploration
1179. and Development, 43, 1013–1024.
He, F., Rao, Y., Wang, W. & Wang, Y., 2020. Prediction of hydrocarbon Ramirez, T.R., Klein, J.D., Bonnie, R.J. & Howard, J.J., 2011. Compara-
reservoirs within coal-bearing formations, Journal of Geophysics and En- tive study of formation evaluation methods for unconventional shale-
gineering, 17, 484–492. gas reservoirs: application to the Haynesville shale, in Proceedings the
Huang, L.J., Chu, Z.H., Xiao, L.Z. & Gao, J., 2008. Nuclear Logging, Chapter North American Unconventional Gas Conference and Exhibition, 14-16
3, Geophysical Logging Methods and Principles, pp. 134–146, Petroleum June, Texas Woodlands, USA.
Industry Press. Schlumberger logging company, & Wu, Q., 1998. Handbook of Common
Jacobi, D.J., Gladkikh, M., LeCompte, B., Hursan, G., Mendez, F., Longo, J. Minerals in Log Interpretation, p. 204, Petroleum Industry Press.
& Shoemaker, P., 2008. Integrated petrophysical evaluation of shale gas Sehmoker, J., 1981. Determination of organic-matter content of Appalae-
reservoirs, in Proceedings of the Canadian International Petroleum Confer- hian Devonian shales from gamma- ray log. AAPG Bull, 65, 2165–2174.
ence/SPE Gas Technology Symposium Joint Conference, Calgary, Alberta, Shi, X., Wang, J., Liu, G., Yang, L., Ge, X. & Jiang, S., 2016. Application of
Canada, 16-19 June. DOI: 10.2118/114925-MS. extreme learning machine and neural networks in total organic carbon
Jia, Z.J., 2017. Study of Magnetic Resonance Relaxation Mechanism and Mea- content prediction in organic shale with wire line logs, Journal of Natural
surement Method on Shale, China University of Petroleum, Beijing City. Gas Science and Engineering, 33, 687–702.
Jiang, Y.Q. et al. 2015. Characteristics and origin of tuff-type tight oil in Stevens, S.H., Moodhe, K.D. & Kuuskra, V.A., 2013. China shale gas and
Jimusaer sag, Junggar Basin, NW China, Petroleum Exploration and De- shale oil resource evaluation and technical challenges, in Proceedings of
velopment, 42, 810–818. the SPE Asia Pacific Oil and Gas Conference and Exhibition, 22-24 October,
Kuang, L.C., Gao, G., Xiang, B.L., Wang, X., Wang, C. & Liu, G., 2014. Low- Jakarta, Indonesia.
est limit of organic carbon content in effective source rocks from Lucao- Su, Y., Zha, M., Ding, X., Qu, C., Gao, C., Jin, J. & Iglauer, S., 2019. Pet-
gou Formation in Jimsar Sag, Petroleum Geology & Experiment, 36, 224– rographic, palynologic and geochemical characteristics of source rocks
229. of the Permian Lucaogou formation in Jimsar Sag, Junggar Basin, NW
Kuang, L.C., Hou, L.H., Yang, Z. & Wu, S.T., 2021. Key parameters and China: origin of organic matter input and depositional environments,
methods of lacustrine shale oil reservoir characterization, Acta Petrolei Journal of Petroleum Science and Engineering, 183, 1–16.
Sinica, 42, 1–14. Tan, J.J., Wu, K.J., Li, Y.H., Zhao, X.Y., Yan, C.M., Huang, Z.H., Du, Z.K.
Kuang, L.C., Wang, X.T., Guo, X.G., Chang, Q. S. & Jia, X.Y., 2015. & Li, S., 2021. Application of logging prediction TOC method in shale
Geological characteristics and exploration practice of tight oil of reservoir evaluation, Progress in Geophysics, 36, 258–266.
Lucaogou Formation in Jimsar Sag, Xinjiang Petroleum Geology, 36, Wang, G. & Guo, R., 2000. Logging Study of Source Rock and Cap Rock,
629–634. Chapter 8, Well Logging Geology, p. 225, Petroleum Industry Press.
Lei, T., Zhou, W., Yang, Y., Deng, H.C., Zhang, G.H. & Zhao, A.K., 2010. Wang, W., Zhao, Y., Mao, R., Sun, Z.C. & Mu, L. W., 2019. Determination
Logging evaluation method for marine source rocks in middle Tarim of the starting time for measurement of NMR effective porosity in shale
Basin, Lithologic Reservoirs, 22, 89–94. oil reservoir: A case study of the Permian Lucaogou shale oil reservoir,
Lewis, R., Ingraham, D., Pearcy, M., Williamson, J., Sawyer, W. & Frantz, J., Jimusaer sag, Oil and Gas Geology, 40, 550–557.
2004. New evaluation techniques for gas shale reservoirs, in Proceedings Wang, X., Ma, J., Zhang, X., Wang, Z., Wang, F., Wang, H. & Li, L., 2020a.
of the Schlumberger Reservoir Symposium, 29-30 July, Denver, Colorado. Prediction of total organic content by a generalized ΔlogR method con-
Liu, G., 2009. Oil and Gas Generation and Source Rocks, Chapter 4, Petroleum sidering density factors: illustrated by the example of deep continental
Geology, pp. 157–164, Petroleum Industry Press. source rocks in the southwestern part of the Bozhong sag, Progress in
Madi, J.A. & Belhadj, E.M., 2015. Unconventional shale play in Oman: pre- Geophysics, 35, 1471–1480.
liminary assessment of the shale oil /shale gas potential of the Silurian Wang, J., Zhou, L., Liu, J., Zhang, X.J., Zhang, F. & Zhang, B.Z.,
hot shale of the Southern Rubal-Khali Basin, in Proceedings of the SPE 2020b. Acid-base alternation diagenesis and its influence on shale
Middle East Unconventional Resources Conference and Exhibition, 26-28 reservoirs in the Permian Lucaogou Formation, Jimsar Sag, Jung-
January, Muscat, Oman. gar Basin, NW China, Petroleum Exploration and Development, 47,
Mahmoud, A.A.A., Elkatatny, S., Mahmoud, M., Abouelresh, M., Abdul- 898–912.
raheem, A. & Ali, A., 2017. Determination of the total organic carbon Ward, J.A., 2010. Kerogen density in the Marcellus shale, in Proceedings
(TOC) based on conventional well logs using artificial neural network, of the SPE Unconventional Gas Conference, 23-25 February, Pittsburgh,
International Journal of Coal Geology, 179, 72–80. Pennsylvania. doi: 10.2118/131767-MS.
Mao, Z.Q., 2001. The physical dependence and the correlation character- Xiang, B.L., Liao, J.D., Zhou, N., Yu, Q.L., Ma, W.Y., Gao, X.W. & Wang,
istics of density and neutron logs, Petrophysics, 42(5), 438–443. M., 2013. Organic geochemical features of Permian source rock from

844
Journal of Geophysics and Engineering (2022) 19, 833–845 Feng et al.

Lucaogou Formation in Ji 174 well of Jimsar Depression, Science Tech- Zhao, P.Q., Mao, Z.Q., Huang, Z.H. & Zhang, C., 2016. A new method for
nology and Engineering, 13, 9630–9640. estimating total organic carbon content from well logs, AAPG Bulletin,
Xu, S. & Zhu, Y., 2010. Well logs response and prediction model of or- 100, 1311–1327.
ganic carbon content in source rocks: a case of WenChang Forma- Zhao, Y.C., Ma, Z. & Yao, G.Q., 1994. C/O log—a potential mea-
tion in the Pearl Mouth Basin, Petroleum Geology and Experiment, 32, surement to evaluate source rock, Well Logging Technology, 18,
290–295. 240–247.
Yao, B., 2019. Log Evaluation Method for TOC in the Lucaogou High-Oil For- Zhu, L.Q., Zhang, C., Zhang, C.M., Zhang, Z.S., Zhou, X.Q., Liu, W.N. &
mation in the Jimsar Sag, China University of Petroleum, QingDao City. Zhu, B.Y., 2020. A new and reliable dual model- and data-driven TOC
Yong, S. & Zhang, C., 2002. Logging Data Processing and Comprehensive In- prediction concept: a TOC logging evaluation method using multiple
terpretation, pp. 122–154, China University of Petroleum Press, DongY- overlapping methods integrated with semi-supervised deep learning,
ing City. Journal of Petroleum Science and Engineering, 188, 1–16, article number:
Zhang, S.M. et al. 2019. Characterization of lacustrine mixed fine-grained 106944.

Downloaded from https://academic.oup.com/jge/article/19/4/833/6672604 by University of Bahrain user on 01 November 2023

sedimentary rocks using coupled chemostratigraphic-petrographic Zhu, L.Q., Zhang, C.M., Zhang, Z.S., Zhou, X.Q. & Liu, W.N., 2019. An
analysis: a case study from a tight oil reservoir in the Jimsar Sag, Junggar improved method for evaluating the TOC content of a shale formation
Basin, Marine and Petroleum Geology, 99, 453–472. using the dual-difference ΔlogR method, Marine and Petroleum Geology,
Zhao, P.Q., Mao, Z.Q., Jin, D., Zhao, P.H., Sun, B.D., Sun, W. & Pang, X., 102, 800–816.
2015. Investigation on log responses of bulk density and thermal neu- Zou, C.N. et al. 2013. Formation mechanism, geological characteristics
tron in coal bed with different ranks, Journal of Geophysics and Engineer- and development strategy of nonmarine shale oil in China, Petroleum
ing, 12, 477–484. Exploration and Development, 40, 14–25.

845