Prediction of hydrocarbon reservoirs within coal-bearing formations

This paper presents a case study on the prediction of hydrocarbon reservoirs within coal-bearing formations of the Upper Palaeozoic. The target reservoirs are low-permeability low-pressure tight-sandstone reservoirs in the Daniudi Gas Field, Ordos Basin, China. The prime difficulty in reservoir prediction is caused by the interbedding coal seams within the formations, which generate low-frequency strong-amplitude reflections in seismic profiles. To tackle this difficulty, first, we undertook a careful analysis regarding the stratigraphy and lithology of these coal-bearing formations within the study area. Then, we conducted a geostatistical inversion using 3D seismic data and obtained reservoir parameters including seismic impedance, gamma ray, porosity and density. Finally, we carried out a reservoir prediction in the coal-bearing formations, based on the reservoir parameters obtained from geostatistical inversion and combined with petrophysical analysis results. The prediction result is accurately matched with the actual gas-test data for the targeted four segments of the coal-bearing formations.


Introduction
This is a case history paper, focusing on the prediction of hydrocarbon reservoirs within coal-bearing formations of the Upper Palaeozoic. The study area is in the Daniudi Gas Field, Ordos Basin, North China block (Şengӧr & Natal'in 1996;Guo et al. 2001;Wang et al. 2005).
The Daniudi Gas Field is tectonically on the nose-like Taba-Temple uplift, in the northeastern part of the Iginholo-Shaanxi slope (figure 1). It is approximately 2000 km 2 , with a square shape spatially (marked in red, figure 1). It is located on the top of the slope of the ancient weathering crust; hence, it has the right development conditions for hydrocarbon reservoirs. However, the gas reservoir in this gas field is a low-permeability low-pressure tight-sandstone reservoir (Gan et al. 2009;Hou & Liu 2012;Yang et al. 2016). The average permeability of reservoirs is about 0.54 × 10 −3 m 2 , and the pressure coefficient of the gas reservoirs is 0.8-1.0. An average single-well production rate is as low as 0.5 × 10 4 m 3 per day. It is indeed a marginal gas field, if measured in terms of economic prospecting.
There were two coal-bearing formations in the Upper Palaeozoic: the Shanxi (S) Formation of the Permian age and the Taiyuan (T) Formation of the Carboniferous age. These two formations may further be divided into four segments, S2, S1, T2 and T1, from the top to the bottom. All these four segments in the Daniudi Gas Field have good reservoir development conditions and all have industrial-scale gas flow potential (Zhang et al. 2011).
Within these coal-bearing formations, coal seams and sand layers lay on top of each other. Such overlaying and interbedding characteristics make both reservoir prediction and fluid identification very difficult. This is especially so when a target reservoir is immediately above or below a coal seam, and the application of seismic information could be very difficult (Wang 2012). This is because of the large differential in seismic impedance between a coal seam and a noncoal layer, which generates a strong reflection on the seismic profile (figure 2). Such a strong reflection of the coal seam will submerge the effective information and thus cause difficulty in the prediction of reservoirs and the fluid contents within the coal-bearing strata.
For instance, the sedimentary facies map (figure 3) indicates that the S1 segment of the Permian is densely covered by the distributary channels and therefore also by the welldeposited sand bodies. On the seismic profile (figure 2), the S1 segment corresponds to the wave peak (indicated by arrows) and the upper edge of the waveform. Influenced by the low-frequency strong-amplitude reflection of coal seams, the seismic resolution is low, and the accuracy of reservoir prediction within the S1 segment is low. As a consequence, the way to conduct hydrocarbon reservoir prediction through coalbearing formations is an urgent problem that needs to be resolved in reservoir geophysics.
In this study, we undertake stratigraphic and lithologic analyses of the Upper Palaeozoic formations in the study area, and a petrophysical analysis of the logging data within Figure 2. Seismic profile of the Daniudi Gas Field, which shows that the S1 segment within the Shanxi Formation corresponds to the wave peak indicated by red arrows and the upper edge. 485 Figure 3. Sedimentary facies of the S1 segment, Daniudi Gas Field. The S1 segment is densely covered by the distributary channels, and therefore was also covered by well-deposited sand bodies. The contours show the thickness of sand deposition. the clastic reservoirs (Liu et al. 2019). Then, we carry out 3D seismic inversion using the geostatistical method to obtain reservoir parameters including seismic impedance, gamma ray, porosity and density. Finally, we conduct reservoir prediction for the four segments sequentially within coal-bearing formations.

Lithology of coal-bearing formations
In this study, we select the well E8 area as the target area, since it is in the central part of the gas field. Table 1 lists the stratigraphic and lithologic properties of the Upper Palaeozoic that were encountered by well E8. This well was drilled in October 1992, and it is located at the Hanging-well Ridge, Maogettu Brigade of Taigemiao Town, Iginholo, Inner Mongolia. The coordinates are x = 4 321 567.7 m, y = 19 388 132.1 m, and h = 1333.6 m. Table 1 also lists the depths and the thicknesses of four segments S2, S1, T2 and T1.
The focus of the study is on the two coal-bearing formations, the S and T Formations of the Upper Palaeozoic. The lithology of S Formation is composed of mainly grey medium-coarse sandstone, glutenite, conglomerate, interbedded with black, dark-grey mudstone and black carbonaceous mudstone. The lower part of the S formation contains coal seams and coal sheets. The lithology of the T Formation is mainly dark-grey, black mudstone, carbonaceous mudstone, and coal seams that are interbedded with grey-white medium-coarse sandstone and partially intercalated limestone lenses.
In addition, this gas field also includes the overlying Lower Shihezi Formation and the underlying Ordovician carbonate weathering in the Lower Palaeozoic. The lithology of the Lower Shihezi Formation is mainly light-grey, grey-white glutenite, pebbly coarse sandstone and mediumcoarse sandstone, which are interbedding with grey, brown mudstone. The Benxi Formation of the Carboniferous at the 486 bottom of the Upper Palaeozoic is only 11.8 m in thickness, while the T Formation is 82.5 m in thickness, and interbedded with six coal seams and three layers of carbonaceous mudstone.
Colour is one of the important macroscopic features of sedimentary rocks and this is closely related to the composition of rocks and their depositional environment. Therefore, the original colour of the rock is a good reflection of the physical and chemical conditions of the water body when the rock was formed. In general, the colour was light and oxidized when the rock was formed in a shallow water or oxidation environment, and the colour was dark when the rock was formed in a deep water or redox environment.

Petrophysical properties of coal-bearing formations
We carry out the petrophysical analysis based on cross-plots (Chatterjee & Paul 2012) between the gamma-ray value (the vertical axis, figure 4) and other petrophysical parameters, including the P-wave velocity, P-wave impedance, density and compensated neutron logging (CNL) porosity (the horizontal axis). Then, according to the known gas saturation (colour bar in figure 4), we delineate the upper and lower limits of each parameter in the well E8 area. Figure 4 is a clastic petrophysical analysis of the S and T Formations. The cross-plotting results show that the gammaray value of the clastic gas reservoir in the well E8 area is less than 85 gAPI. The density is between 2.1 and 2.7 g ⋅ cm −3 , but the low density (2.1-2.4 g ⋅ cm −3 ) may be a coal seam, and the relatively high density (2.4-2.7 g ⋅ cm −3 ) may be a gas-bearing layer.
The fourth panel in figure 4 is a cross-plot of the gamma ray value and the CNL porosity. The CNL value measures the hydrogen content of the formation and its value can be approximated as an estimate of formation porosity. The limestone has a CNL value of 0 and the water has a CNL value of 100. For the sandstone, when the porosity is high, the CNL value is also high, and when the porosity is low, the CNL value is also low.
The CNL response to an oil layer is not so obvious. However, for a gas layer, the CNL value will be lower than the 487 Figure 4. Petrophysical analysis of coal-bearing formations (S formation and T formation). The colour bar is the gas saturation (S g ). actual porosity. The larger the volume of gas content of the formation, the lower the CNL value. The fourth panel in figure 4 clearly shows that the CNL value of the clastic formation in the well E8 area is less than 22%. The CNL value for commercial gas flows is generally less than 12%, and sometimes it is as low as 3%, or even less.
Based on the clastic petrophysical analysis here, which was compared against the gas saturation (colour sample), a list of the upper and lower limits of the gas content indication for each parameter is presented in Table 2.

Geostatistical inversion
We perform geostatistical inversion of 3D seismic data to obtain four reservoir parameters including P-wave impedance, density, gamma value and porosity (Bosh et al. 2010). In the seismic inversion, we first use a sparse inversion method to obtain a low-wavenumber background model for P-wave impedance, and then use a geostatistical inversion method to invert for the four reservoir parameters simultaneously.
The parameters of the sparse inversion method include: P-wave impedance uncertainty (=0.02), signal-to-noise ratio (=12), the Lp norm of seismic data residual (P = 0.7) and the cut-off frequency (=9 Hz). The geostatistical inversion method we used is the Gaussian collocated cosimulation (Soares 2001;Nunes et al. 2017). Figure 5 shows the coverage of the well-E8 3D seismic data in the Daniudi Gas Field. The 3D seismic data cube covers a 195 km 2 area of the 3D seismic survey, and well E8 (marked in red colour) is slightly northerly in the centre. There is an arbitrarily selected cross-well line passing through 10 wells: >2.12 <2.72 CNL (porosity, %) <22 Figure 5. The coverage of the well-E8 3D seismic data in the Daniudi Gas Field. Well E8 (marked in red) is slightly more to the north in the centre. There is an arbitrarily selected seismic line connecting 10 wells. Two green stars indicate the two wells annotated in the seismic profile shown in figure 2. Figure 6 displays the geostatistical inversion results following this cross-well line. The inversion results reveal that the P-wave impedance and density of the Ordovician carbonate formation are significantly higher than the overlying clastic formation. The gamma-ray value of the Ordovician carbonate formation is significantly lower than that of the overlying clastic formation. In the inversion results, it is abundantly clear to see that the coal seams have the characteristics of low impedance, low density and high gamma-ray value. But the correspondence between the porosity and either the clastic or carbonate rocks is not obvious (Liu et al. 2018;Li et al. 2019).

Reservoir prediction of coal-bearing formations
For coal-bearing formations in the Upper Palaeozoic, we conduct reservoir prediction interactively, based on the crossplot of the gamma ray and density values. It is noteworthy that we use the density rather than the porosity in the prediction of these clastic reservoirs. We perform in this manner to avoid the effect of coal seams and to achieve that by setting the density to be greater than 2.4% as a threshold. Meanwhile, we set the gamma-ray value to be less than 85 gAPI. Figure 7 shows the results of the reservoir prediction for the S2, S1, T2 and T1 segments, respectively. The colour scale indicates the reservoir thickness, which is quantified in terms of seismic time in units of ms.
The S2 segment is the sub-facies of the braided river delta plain. The sedimentary microfacies are the distributary channels and depressions and swamps between channels. In this area, the S2 segment is well-developed with respect to the distributary channels, which exhibit the characteristics of extending from north to south and of multi-level distributaries and convergence. The thickness of sand body is 4.5 m on average. The sand body formed by the sedimentary microfacies of the distributary channels has a large thickness, a coarse grain size and a good sorting property, so the sandstone has a good porosity and permeability and is the most favourable reservoir layer (Chen & Wang 2010). The sand body that formed at the edge of river channels is thin, fine in grain size, poor in physical properties and weak in its ability to store natural gas. The prediction result in figure 7a shows that the sand body in some individual depressions and swamps might reach a thickness of 15 ms (about 30 m).
The S2 segment discussed basically contains no coal seams, while the S1 segment is a coal-bearing stratum. Although the layer thickness of the S1 segment is about 50-70 m on average, and it is relatively large, the sandstone layers are much more dense vertically and the grain size is relatively coarse (Dong et al. 2009) because of the presence of coal seams in the S1 segment. The effective reservoir thickness within the S1 segment is relatively small, as shown in figure 7b, compared to the effective reservoir thickness of the S2 segment.
The sediment of the T Formation is mainly composed of a barrier-type coastal facies. This coastal facies is characterized by a set of interbedded shelf and paralic sediments, including lake-marsh and littoral mudstones, carbonaceous mudstones, limestones, coal seams and coastal-littoral sandstones with varying degrees of development. A rhythmic deposition of marine carbonate-mudstone-thin coal seam is a typical representative of the middle-late sedimentation of the T Formation. Tidal flats, lagoons and barrier islands are sedimentary facies that developed in the T Formation. Sedimentary microfacies mainly include barrier sand dams, sand flats and mixed rafts. The reservoir rock types are mainly composed of pebbly coarse sandstone, coarse sandstone, medium sandstone, fine sandstone and siltstone. The reservoir pore type is mainly secondary, the throat is mainly fine and the pore throat connectivity is medium.
The rock type of the T2 segment is mainly medium-coarse quartz sandstones. The volume fraction of quartz in the 489 sandstone crumb structure reaches 71-98% and the average debris is 13.4%. The sorting of granule rounding is generally fine and good, and the rounding is mainly in a round form. The cementing type is mainly regenerative pores and primary pores. The cements contain authigenic quartz, calcite, dolomite and a small amount of siderite.
The rock type of the T1 segment is mainly coarse-grained quartz sandstone. Quartz volume fraction is 92.3-99.8%, and the debris is 0-6.5% with almost no feldspar. The granule sorting is between medium and good, and the rounding is dominated by a sub-angular shape. The cementing type is mainly regenerative pores and primary pores, and there is a small number of compaction pores.
In the T2 segment, the class II gas reservoirs (mediumyield gas reservoirs, with unobstructed flow greater than 1.0 × 10 4 m 3 per day and less than 5.0 × 10 4 m 3 per day) had the widest distribution and were mainly developed in the middle of the sand dams. Class III gas reservoirs (low-yield gas reservoirs, with unobstructed flow less than 1.0 × 10 4 m 3 per day and greater than 0.5 × 10 4 m 3 per day) were secondarily developed on the edge of sand dams. Class I gas reservoirs (high-yield gas reservoirs with unobstructed flow above 5.0 × 10 4 m 3 per day) were distributed in the right centre of the barrier sand dams.
The class I gas reservoirs in the T2 segment were more widely distributed than the T1 segment. The lithology of these areas in a large portion of the T2 segment is the coarse and coarser sandstone. These areas in the T2 segment generally have a good physical property, a large thickness of sandstone and fewer interbeds. However, the reservoir predictions show that the reservoir thickness of both T2 and T1 segments within the Taiyuan Formation is relatively thin (figures 7c and 7d).

Conclusion
We have conducted the stratigraphic and lithologic analyses of the coal-bearing formations and the petrophysical analysis of the logging data, and delineated the upper and lower limits of each petrophysical parameter according to the known gas saturation. The cross-plotting results show that the clastic gas reservoirs in this area have the gamma-ray value between 25 and 85 gAPI, a density value between 2.1 and 2.7 g ⋅ cm −3 and CNL porosity value below 22%.
Geostatistical inversion of the 3D seismic data has produced elastic and reservoir parameters including P-wave 491 impedance, density, gamma-ray value and porosity. The inversion results show that the P-wave impedance and density of the overlying clastic formations are significantly lower than those of the underlying Ordovician carbonate rock. The gamma ray value of the overlying clastic formations is significantly higher than that of the underlying Ordovician carbonate. The clastic formations have the obvious characteristics of low impedance, low density and high gamma-ray value. The correspondence between porosity and either clastic or carbonate rocks is not obvious.
In this study, we have set a threshold value for the density to effectively exclude the interference of coal seams. We have compared the reservoir prediction result with the geological characteristics such as the sedimentary facies and reservoir types, and found that the prediction result is finally matched with the actual gas-test data for each segment of the coalbearing formations.