Unconventional reservoirs have become the research hotspot (Gao et al., 2018; Cong et al., 2022). As an essential source of unconventional natural gas, the tight sandstone reservoirs have attracted extensive attention (Bai et al., 2013; He et al., 2020). These reservoirs in China have excellent resource potential and are characterized by low porosity, low permeability, and complex pore structure (Li et al., 2017). Stratigraphic lithology classification and identification are essential for evaluating fine modeling reservoir parameters (Li and Li, 2013; Alzubaidi et al., 2021). Accurate evaluation of reservoir properties such as lithology identification and porosity calculation is the basis of exploration and development, as well as the basis for calculation of reservoir parameters (Yuan et al., 2018; Zhou et al., 2018). Their accuracy is very important for predicting, exploring, and developing rich areas of oil and gas resources. However, typical characteristics such as complex lithology, low maturity of sandstone, strong diagenesis, and substantial heterogeneity bring difficulties and challenges to the accurate identification of the lithologic zone and become the main constraints for practical exploration of this type of reservoir (Pang et al., 2019; Xiong et al., 2020; Bai et al., 2021; Wei et al., 2021).

Previous research mainly focused on improving the identification accuracy of lithologic types by using logging curves by various methods (Tan et al., 2017; Liu et al., 2020). Traditional lithologic identification methods include debris logging, core logging, and cross plot (Zhao and Gao, 2003). Gupta et al. (2012) used natural gamma-ray logging, high-resolution resistivity logging, lithologic density logging, and other logging data cross-mapping techniques to evaluate the lithologic and physical property parameters of the reservoir profile from two wells of fractured basalt. Cheng et al. (2016) established the cross-plot method and logging curve calculation method to identify sandstone, siltstone, etc. Zhou et al. (2017) proposed a lithologic classification method based on thin sections, logging curves, and core physical property data. Das and Chatterjee (2018) identified gas sand, saltwater sand, carbonate rock, and shales by analyzing vertical and horizontal impedance diagrams. Khamees et al. (2021) used neutron-density cross plot, acoustic-density cross plot, M−N cross plot, etc. The results show that they are well applied in lithology identification of Jeribe formation in a field in northern Iraq. For the unique strata with strong heterogeneity and complex structure, the traditional lithology identification method is challenging to ensure identification accuracy (Sun et al., 2019).

With the re-emergence of deep learning in recent years, many machine learning methods have been applied to lithology identification using logging data. The combination of machine learning algorithm and lithological classification model mainly includes support vector machine (Sebtosheikh and Salehi, 2015), neural network (Han et al., 2018), and decision tree (Duan et al., 2016). Sebtosheikh et al. (2015) adopted the support vector machine (SVM) classification method to predict the lithology of a heterogeneous carbonate reservoir in Iran based on core analysis data. Zhao et al. (2017) proposed a multivariate membership function discriminant method, which regarded lithology identification as a linear model in the fuzzy domain and obtained target results by establishing a multivariate membership function. Wang et al. (2018) proposed a KNN clustering optimization method based on weighted cosine distance to fit the lithologic model better. Xiang et al. (2020) presented the application of deep learning to establish a deep trust network (DBN) based on logging data in lithology. However, the combination of machine learning methods with geological data is less, and the results may not conform to the general geological laws, resulting in unsatisfactory identification results. Also, most of these methods are limited to lithology identification and rarely combine with reservoir physical property data. Moreover, machine learning methods may be affected by drilling fluid (Chen et al., 2020).

Based on the geological data of the target interval, we classify the lithology of the reservoir by integrating the thin-section information, core analysis data, and well logging data in this paper. The lithology of these reservoirs is mainly coarse-, medium-, and fine-grained quartz sandstone and lithic sandstone. By analyzing the relationship between rock compositions, reservoir parameters, and logging response, we proposed a new method based on the complex reservoir analysis method (CRA) with equivalent components (CRAE). We used the acoustic transit time (AC) log and compensated neutron porosity log (CNL) to estimate formation porosity and set the parameters of the matrix components to make the results agree with the core data. Then, the fractions of the mineral components are calculated. Also, we used the natural gamma-ray log (GR) to calculate the granularity median. By combining the component fractions and granularity median, we can accurately identify the formation lithology of the Xu 2 Member tight gas reservoir in CX depression, West Sichuan Basin. The second part of this article gives the geological background. The third part introduces the data and methods of the study. The fourth and fifth present the application results and comparison and discuss the effect. Finally, conclusions are drawn in the sixth section.

The West Sichuan Depression is located west in the Sichuan Basin, east to the Longmenshan orogenic belt, and northwest to the Yangtze block. It is a part of the western Sichuan foreland basin (Figure 1), and the area is about 3.1 × 10⁴ km². It was formed during the transition from the marine cratonic basin to the continental basin in the Sichuan Basin in the Late Triassic. There is a set of the coal-bearing clastic rock group from bottom to top from marine facies to continental facies and dominated by continental facies. The thickness of this set of formations does not change much, with a slight decrease from west to east. CX depression is a part of the west Sichuan Depression. From bottom to top, it includes Xu 1, Xu 2, Xu 3, Xu 4, and Xu 5. The section of interest of this study is Xu 2.

The Xu 2 member in the boreholes of the research area has a depth of 4,300–6,000 m and a thickness of about 400–500 m. It is dominated by gas-bearing lithic sandstone and lithic quartz sandstone. There are primarily argillaceous siltstone and black silty shale, interbedded with different thicknesses and with carbonaceous shale and coal lines. Detrital materials, mostly sedimentary and metamorphic debris with a small number of volcanic debris, are abundant. The reservoirs generally contain clay minerals and calcareous minerals such as calcite and dolomite. They are dense. The porosity distribution of the reservoir is shown in Figure 2, with an average porosity of 3.7% and a permeability of less than 0.1 micro-Darcy (mD). The reservoirs are affected by multiple structures, formation temperature, pressure, and cemented calcium and mud. The frontal sedimentary subfacies of the braided river delta are relatively developed. The porosity is low, and the lithology is complicated, making it challenging to identify the lithology and calculate the porosity accurately.

According to the abovementioned analysis, applications were carried out in the study area. The application results in well 1 are shown in Figure 7. In Figure 7, the first three tracks are the original conventional logs, and the fourth track is the depth track. The fifth track is the lithological component profile we calculated by using the CRAE workflow. The sixth track is the lithological profile from mud logging. The right four tracks are the calculated porosity, quartz, rock debris, and so-called shale content. The black horizontal lines are the core analysis results. It can be seen that the porosity calculated by AC-CNL is consistent with the core porosity. Also, in the last three tracks, the component content of quartz, rock debris, and shale content calculated by the response equation is compared with the thin-section analysis results. It reveals that the calculated quartz content is in good agreement with the core thin-section data. The total volume fraction of the calculated shale content and rock debris is consistent with the rock debris from the core analysis result. Also, the computed shale content corresponds to the composition of sedimentary debris from the core thin-section data. This means there are tiny silty debris and clay minerals in the sedimentary debris, and this part contributes most of the GR radiation. The calculation and discrimination results confirm the correctness of the profile calculated.

According to the sandstone classification standard shown in Table 2, the geological survey and lithologic composition of the study area were classified. The final lithological classification result can be obtained after adding the MD calculation result into the lithology component profile. The MD was calculated by using the ΔGR model. A classification result is shown in Figure 8. The final classification results are compared with the Fisher discriminant and the mud logging results. The last four tracks in this figure are the mud logging, the Fisher discriminant and the calculated lithological profile, and the calculated MD. From the figure, we can see the lithological results of this well are consistent with the lithology in the mud logging profile and more refined. The correct rate is 89.96% of the calculated lithological profile by the CRAE method, and it is higher than the Fisher discriminant (it is 78.62%), which shows that this workflow has applicability in the study area.

The CRA method uses the GR or SP log to obtain the shale content and two other porosity logs to determine the mineral composition of the lithological composition. This method is aimed at carbonate rocks, in which the calcite and dolomite develop. As the minerals of calcite and dolomite have stable physical properties, including acoustic wave velocity and neutron porosity, the calculated mineral composition content is always accurate. However, the CRA method is rarely used in sandstone formations. One reason is that normally in the conventional sandstones, the lithology is simple; the other is the tight sandstones with complex mineral components. In this study area, we propose the CRAE method, in which we treat the tight gas reservoir rock as a mixture of quartz, debris, and pores, ignoring the relatively low volume fraction of the unstable mineral feldspar. The rock debris includes sedimentary, metamorphic, and igneous rocks. The sedimentary rocks can be calculated from the GR log for their radioactivity. Also, the metamorphic and igneous rock debris are combined and set a parameter. Meanwhile, the calculated shale contents in different formations have different meanings. The shale content calculated in the tight sandstone formation is a reflection of the sedimentary debris. Also, in shale and other formations, it represents the actual shale content in the formation.

In summary, this method is simple to select parameters. The accuracy meets production needs, and the workflow is easy to popularize. The determination of parameters is the most crucial step. When this workflow is used in a new study area, we should calibrate with core analysis data to select accurate parameters. In other words, when the lithological composition of the study area is different, the components can be replaced. The matrix parameter value of the lithology can be calculated according to the practical value of the study area and calibrated by core porosity.

The complex geological structure and lithological categories of tight sandstone gas reservoirs have posed challenges for logging lithological identification and porosity calculation. This article introduces a CRAE and MD calculation workflow to identify six different lithologies and estimate porosity. The conclusions are as follows:

1) The GR, CNL, and AC logs contain the lithological information and can be used to reveal the components of the formations.