The early 21^st century has witnessed significant discoveries of unconventional oil and gas around the world. In 2018, the global unconventional oil yields accounted for 14% of total petroleum production (44.5×10⁸ t)^[1,2]. Tight oil has become a "hot spot" in global unconventional oil development^[3]. The global tight oil resources have pervasive extension and large potential volume^[4]. The technological recoverable tight oil resources around the world are estimated to be 639.3×10⁸ t^[5].

Tight oil is a type of oil found in tight sandstone and tight carbonate rocks with in situ matrix permeability smaller than or equal to 0.1×10^-3 μm² (or air permeability smaller than 1×10^-3 μm²)^[6]. Jia et al. estimated total geologic resources of tight oil in China at (106.7-111.5)×10⁸ t using the abundance analogy method in 2012^[7]. As of the beginning of 2020, the proved tight oil and shale oil reserves in China were 7.37×10⁸ t, and the probable and possible reserves were 18.3×10⁸ t; the deliverability was established to be over 400×10⁴ t/a^[1]. Continental tight oil resources have been found in the Permian Lucaogou and Fengcheng formations in the Junggar Basin^[8], the 7^th member of the Triassic Yanchang Formation (Chang 7 Member for short) in the Ordos Basin^[9], the Cretaceous Quantou Formation in the Songliao Basin^[10], the Paleogene Kongdian Formation-Shahejie Formation in the Bohai Bay Basin^[11,12], and the Jurassic in the Sichuan Basin^[13]. The Paleogene in the Jiyang depression is rich in tight oil resources. By the end of 2019, the proved reserves, probable reserves, and possible reserves of tight oil were 1.94×10⁸ t, 0.87×10⁸ t, and 1.1×10⁸ t, respectively. The prospective resources were estimated to exceed 5×10⁸ t.

Pores and throats in tight reservoirs are usually of nano-scale (with pore radius smaller than 1 μm)^[14]. Tight oil accumulation and flow are all restricted by such heterogeneous small pores and throats with complicated structures^[15]. In accordance with preceding studies, tight Paleogene reservoirs in the Jiyang depression exhibit complicated porosity-permeability relations. The permeability of reservoirs with the same porosity differs by tens to hundreds of times. The pore throat structure determines the permeability of the reservoir^[16]. Thus, it is necessary to investigate pore throat structures in tight Paleogene reservoirs in the Jiyang depression. At present, researchers have made a lot of progress in the research methods of pore throat structure of tight reservoirs^{[17,18,19,20]}. Qualitative and quantitative characterization of pore throat size and geometry could be accomplished through data analysis and image analysis. The former includes mercury injection, N₂ cryosorption, nuclear magnetic resonance, etc., which are advantageous for accurate quantitative characterization of microscopic tight reservoir pore-throat structures using lab data, but disadvantageous for porosity limitation. The latter includes cast thin section, scanning electron microscopy, focused ion beam microscopy, micron-scale and nano-scale CT scanning, etc. and digital core originated from these techniques^[21,22]. In view of the range of testing and observation, each technique has its advantages and disadvantages. Therefore, some methods may be combined to characterize microscopic pore throat structures in tight reservoirs. Yang et al. adopted physical simulation methods including high-pressure mercury injection, N₂ cryosorption, nuclear magnetic resonance, and centrifuging for all-scale pore throat measurements for tight-oil cores^[23]. Liu et al. used FESEM, constant-rate mercury injection, high-pressure mercury injection, and N₂ cryosorption to qualitatively and quantitatively investigate pore space types and microscopic pore throat diameter distribution in tight sandstone reservoirs in the second and third members of the Upper Cretaceous Qingshankou Formation in the area around Well Long26, the Longhupao oilfield, the Songliao Basin^[24]. According to the geologic setting and data conditions of tight reservoirs in the Jiyang depression, we select routine mercury injection, constant-rate mercury injection, and micron-scale and nano-scale CT scanning to delineate pore throat structure. There are 1100 groups of routine mercury injection data available now related to tight reservoirs in the Jiyang depression. Using these data, we can make a thorough investigation of pore throat size and distribution. As for respective sizes and distributions of pores and throats, it is necessary to perform constant-rate mercury injection using more samples. In view of complicated pore throat structures in tight Paleogene reservoirs in the Jiyang depression and the impact of pore throat connectivity on petrophysical properties, we utilize micron-scale and nano-scale CT scanning to characterize pore throat connectivity.

To investigate pore throat structures in tight Paleogene reservoirs in the Jiyang depression, we use laser grain- size analysis and cast thin sections to examine petrologic composition and rock texture and use routine mercury injection, constant-rate mercury injection, and micron- scale and nano-scale CT scanning to examine pore throat structures in tight reservoirs. Reservoir classification is statistically performed based on pore-throat structure parameters, followed by the assessment of each tight reservoir type.

In terms of tectonic setting and sedimentary origin, tight Paleogene reservoirs in the Jiyang depression are classified as tight sandy conglomerate and tight sandstone. The latter mainly refers to tight sandstone occurring in the sub-sags and ramp zones in a half-graben faulted lake basin, which is further classified as tight sandstone of turbidite facies, tight sandstone of deltaic front subfacies, and tight sandstone of shore-shallow lacustrine beach-bar microfacies of inshore shallow lake subfacies. Most of the sandstone is medium- and fine-grained sandstone and silty fine sandstone^[16].

Tight sandy conglomerate, of inshore subaqueous fan facies, fan-delta facies, and sublacustrine fan facies, mainly occurs in the region with abrupt slope in a half-graben faulted lake basin. It lithologically consists of conglomerate, sandy conglomerate, pebbly sandstone, and pebbled sandstone. Gravels are common in tight sandy conglomerate, and sandy particles are mainly of medium and coarse grains. Despite their different sedimentary origins, these rocks are proximal gravity-flow deposits originating in rapid accumulation and thus exhibit similar constituents, texture, pore throat structures, and diagenetic features. All such sedimentary rocks as pebbled sandstone and conglomerate deposited in the abrupt slope zone are entitled tight sandy conglomerate.

Tight reservoirs in the Jiyang depression mainly reside in the third and fourth members of the Paleogene Shahejie Formation (Sha 3 and Sha 4 Members for short). As per petrophysical data of 4897 groups of cores, reservoir porosity is statistically measured to be smaller than 12% with the average of 7% and air permeability is measured to range (0.001-3.000)×10^-3 μm² with the average of 0.64×10^-3 μm². The samples with air permeability below 1×10^-3 μm² account for 77% of the total, and the samples with air permeability of (1-3)×10^-3 μm² account for 23% of the total. The ratio of the number of oil wells with tight reservoirs to the number of all oil wells is calculated to exceed 74% for all evaluation units. According to the national standard "Geological evaluating methods for tight oil"^[6], these reservoirs are assessed to be tight reservoirs.

Owing to the discrepancies in sedimentary environment and diagenetic process, Paleogene tight sandstone and sandy conglomerate in the Jiyang depression exhibit different permeabilities and oil contents at an equal porosity. This means that routine porosity-based reservoir evaluation methods are not feasible for tight reservoir assessment. As per mass data of mercury injection, there is a good correlation between reservoir permeability and pore throat structure. Thus, it is necessary to investigate pore throat structures in the tight reservoirs in the Jiyang depression and accomplish reservoir classification based on pore throat structure.

As per the data of 2013 groups of tight sandstone and sandy conglomerate cast thin sections acquired from the Paleogene in the Jiyang depression, sandstone samples have low quartz content, high feldspar and debris contents, and low component maturity (Fig. 1a). Tight sandstone particles of turbidite facies, deltaic front subfacies, and shore-shallow lacustrine beach-bar microfacies of inshore shallow lake subfacies show highly consistent constituents and have similiar close quartz, feldspar, and debris contents. Pebbly sandstone and pebbled sandstone in tight sandy conglomerate have low quartz content of 30.7% on average and high feldspar and debris contents of 38.5% and 30.8%, respectively (Table 1).

The ratio of quartz content to feldspar-debris content indicates the component maturity of sandstone. For pebbly sandstone and pebbled sandstone in sandy conglomerate, this ratio only ranges 0.40-0.80 with the average of 0.46. But for sandstone samples of deltaic front subfacies, shore-shallow lacustrine beach-bar microfacies inshore shallow lake subfacies, and turbidite facies, the ratio centers on 0.60-1.00 with the average of 0.85, 0.79, and 0.80, respectively for three facies (Table 1).

Paleogene tight sandstone and pebbly sandstone and gravel sandstone of sandy conglomerate in the Jiyang depression show low contents of sedimentary rock debris and high contents of metamorphic and magmatic rock debris (Fig. 1b). Metamorphic rock debris is originated from Archean gneiss and crystalline rocks. Magmatic rock debris is originated from pegmatite and granite. Sedimentary rock debris is originated from Lower Paleozoic carbonate rocks and Mesozoic tuffite and sandstone. Debris constituents of tight sandstone and sandy conglomerate. Pebbly sandstone and pebbled sandstone in tight sandy conglomerate show high metamorphic rock debris content of 77.5% on average, much higher than that of 58.1%-63.3% for tight sandstone samples of deltaic front subfacies, shore-shallow lacustrine beach-bar microfacies, and turbidite facies. Magmatic rock debris in tight sandy conglomerate show a lower content, 11.5% on average, than that in sandstone samples of three facies (Table 1).

Interstitial materials in Paleogene tight sandstone and sandy conglomerate in the Jiyang depression are composed of carbonate cements and argillaceous matrices, buried at 3200-4500 m and in phase B of middle diagenesis. Owing to similar diagenetic fluid environments and diagenetic evolutions, there are mainly carbonate cements, e.g. calcite, dolomite, ferrocalcite, and ankerite cements, and some siderite and zeolite cements in sandstone and sandy conglomerate. Secondary quartz enlargement could be observed occasionally. Matrices mainly consist of argillaceous and argillaceous-dolomitic matrices, the content of which is related to sedimentary facies and sorting.

Sandstone samples of turbidite facies, deltaic front subfacies, and shore-shallow lacustrine beach-bar microfacies of inshore shallow lake subfacies mainly contain fine sandstone, silt-bearing fine sandstone, and silty fine sandstone. Fine sandstone accounts for 70%, and siltstone accounts for 5%-20%. For the three facies, siltstone accounts for 9%, 5%, and 20%, respectively; heterogranular sandstone accounts for 22%, 20%, and 5%, respectively; median grain diameter is 0.20, 0.14, and 0.11 mm, respectively, on average. Tight sandy conglomerate features coarse grains and poor sorting. In pebbly sandstone and pebbled sandstone, heterogranular sandstone and coarse sandstone account for 44.5% and 48.1%, respectively; medium sandstone and fine sandstone only account for 6.9%, and siltstone accounts for 0.5%; median grain diameter is 0.61 mm on average. This differs greatly from the scenarios for tight sandstone samples of three facies.

As per microscopic thin section observation, 90% of tight sandstone and sandy conglomerate in the Paleogene, the Jiyang depression, are grain supported with subangular and angular particles; some samples with high matrix content are matrix-grain supported (Fig. 2a). Porous cementation is the major type of cementation (Fig. 2b, c), and interlocking and poikilitic cementation can be observed occasionally (Fig. 2d). Particles are in point- linear and linear contact with close-by degree of compaction; concavo-convex contact and pressure solution have seldom been observed. Pore space in tight sandstone and sandy conglomerate is composed of residual primary pores (Fig. 2e), secondary dissolved pores (Fig. 2f), grain cracks and grain-boundary fractures (Fig. 2g, h). Secondary dissolved pores mainly occur at the fringe of primary pores or originated from dissolved carbonate cements which choked primary pores. Secondary dissolved pores may improve pore-throat structure to some extent.

In summary, tight sandstone samples of deltaic front subfacies, shore-shallow lacustrine beach-bar microfacies, and turbidite facies in the Paleogene, the Jiyang depression, show similar petrofabric, which differs greatly from that of tight sandy conglomerate.

Routine mercury injection, with the largest injection pressure of 29.7 MPa, is capable of characterizing pores and throats with the radius above 25 nm.

There are routine mercury injection data available for 647 groups of Paleogene tight sandstone and sandy conglomerate samples from the Jiyang depression (Table 2). With respect to pore throat size, the R₀ value of tight sandy conglomerate ranges 0.050-4.669 μm with the average of 1.635 μm. R₀ values are 1.461, 1.298, and 1.459 μm, respectively on average for tight sandstone samples of deltaic front subfacies, shore-shallow lacustrine beach-bar microfacies, and turbidite facies. The average R₀ value of tight sandy conglomerate is 170-340 nm larger than those of tight sandstone samples of three facies. Tight reservoirs of 4 types show close R₅₀ and R_a values. R₅₀ value ranges 0.025-0.661 μm, and the difference is only 20-40 nm for tight reservoirs of 4 types. R_a value ranges 0.025-0.966 μm. The average R_a value is 0.348 μm for tight sandy conglomerate. For sandstone samples of deltaic front subfacies, shore-shallow lacustrine beach-bar microfacies of inshore shallow lake subfacies, and turbidite facies, the R_a values are 0.363 μm, 0.290 μm, and 0.335 μm, respectively on average. With respect to construction coefficient, the value ranges 0.06-9.70 with the average of 2.59 for tight sandy conglomerate and ranges 0.27-8.96 for tight sandstone. The coefficients are 2.92, 2.76, and 2.95, respectively on average for tight sandstone samples of deltaic front subfacies, shore-shallow lacustrine beach-bar microfacies of inshore shallow lake subfacies, and turbidite facies. Compared with tight sandstone, tight sandy conglomerate shows larger variation of construction coefficient and smaller average value; this indicates small tortuosity of pores and throats in tight sandy conglomerate.

The average maximum mercury saturation is 57.1% for tight sandy conglomerate, which is smaller than 67.2%, 60.8%, and 63.8% for tight sandstone samples of deltaic front subfacies, shore-shallow lacustrine beach-bar microfacies, and turbidite facies, respectively.

The uniformity coefficient ranges 0.25-0.27 for 4 types of tight sandstone. This means that the radius of major pores and throats for fluid flow is 1/4 of the maximum pore throat radius. For tight sandy conglomerate, the value is large; thus, the radius of major pores and throats for fluid flow is also large.

The characteristic structure parameters of tight sandy conglomerate and tight sandstone of turbidite facies are 0.50 and 0.49, respectively, smaller than those of 0.55 and 0.56 for tight sandstone samples of deltaic front subfacies and shore-shallow lacustrine beach-bar microfacies of inshore shallow lake subfacies (Table 2). This indicates poor sorting of pores and throats in tight sandy conglomerate and tight turbidite sandstone of gravity-flow origin.

In summary, tight sandstone samples of deltaic front subfacies, turbidite facies, and shore-shallow lacustrine beach-bar microfacies are close in pore throat size, mean values of pores and throats, and tortuosity. Compared with above rocks of three facies, tight sandy conglomerate features large value of the maximum pore throat radius, small pore throat tortuosity, poor pore throat sorting, and small value of the maximum mercury saturation.

Constant-rate mercury injection is feasible for the measurements of pore size, throat size, pore to throat ratio, and their distributions, especially for ultra-low- permeability and tight sandstone reservoirs with remarkably different pore throat features. It is noted that this technique applies to pores and throats with the radius above 120 nm because the final mercury injection pressure is 6.2 MPa^[25].

As per constant-rate mercury injection data of 11 tight sandstone and sandy conglomerate samples acquired from the Paleogene in the Jiyang depression, the mean pore radius ranges 128.702-150.917 μm with the average of 141.894 μm. The mean throat radius ranges 0.342-1.834 μm with the average of 0.735 μm. The main throat radius ranges 0.055-2.117 μm with the average of 0.411 μm. The pore throat distribution features large pores of micron scale and tiny throats of sub-micron scale to nano scale. The average value of pore-to-throat radius ratio changes greatly from 99.382 to 441.159 (Table 3).

As shown in the frequency distribution diagrams of pore radius and throat radius, the pores in tight sandstone and sandy conglomerate are mainly of 100-150 μm (Fig. 3a) and throat radius exhibits large variation (Fig. 3b), which is closely related to permeability. For the sample B667-1, the pore throat radius mainly ranges 1-2 μm and the permeability is 2.455×10^-3 μm². For the samples Y172-1, S126-2, Y22-22-2, and S548-1, the throat radius ranges 0.4-0.9 μm and the permeabilities are 0.126×10^-3 μm², 0.237×10^-3 μm², 0.112×10^-3 μm², and 0.186×10^-3 μm², respectively. For the samples S126-1 and Y172-2, the throat radius ranges 0.3-0.4 μm and the permeabilities are 0.046×10^-3 μm² and 0.052×10^-3 μm², respectively.

As per the relations between the parameters of constant-rate mercury injection and the permeability, mean throat radius, main throat radius, and the maximum open throat radius correlate positively with permeability; this means that the permeability of tight sandstone and sandy conglomerate is dominated by throat size. The average value of pore-to-throat radius ratio correlates negatively with permeability, indicating that the permeability increases with decreased pore-throat radius difference, and reservoir heterogeneity increases with pore-to-throat radius ratio, when the Jamin effect tends to occur to hinder fluid flow. Relative sorting coefficient correlates positively with permeability, suggesting that permeability increases with pore throat sorting (Fig. 4).

S_kp describes the symmetry of pore throat size distribution. S_kp=0 indicates symmetrical pore throat distribution; S_kp>0 indicates coarse skewness; S_kp<0 indicates fine skewness. K_p represents the steepness of pore-throat frequency distribution curve. K_p=1 indicates normal distribution; K_p>1 indicates a curve with a sharp peak; K_p<1 indicates a curve with a gentle peak or double peaks. For tight sandstone and sandy conglomerate in the area of interest, pore throat distribution features coarse skewness and a sharp peak. In addition, kurtosis is in direct proportion to permeability, and skewness is in inverse proportion to permeability.

X-ray computerized tomography (CT) is a technique of fast full-scale nondestructive scanning and imaging of rock samples using X-ray and numerical reconstruction of 3D pore throat structure using scanning images^[26,27,28]. We used nanoVoxel2000 X-CT scanner made by Sanying Precision Instruments Ltd. and Avizofire, a software system for digital rock analysis, to perform image analysis and parameter calculation. To address the issue of multi-scale characterization of pores and throats in tight sandstone, we used X-CT scanning to observe microscopic pores and throats in cylindrical core samples of 2.54 cm (with image resolution of 10-25 μm as per scanning parameter setting). Mineral images were calibrated using probe analyzer. Two to three cylindrical subsamples of 1.0-5.0 mm in diameter were taken out from the core sample of 2.54 cm in diameter for high-precision X-CT scanning, the theoretical resolution of which may reach 0.5 μm at most. In accordance with the principle of consistent pore segmentation in the images with the equal field of view and different degrees of precision, the low-resolution image, with ambiguous pore-particle boundaries or boundary gray scale higher than pore gray scale, was segmented using a given threshold. The threshold was adjusted until the segmented pore area agreed with the area in the same field of view in the high-resolution image. The final threshold was then used for low-resolution image segmentation to obtain 3D pore distribution in the cylindrical sample of 2.54 cm in diameter. This is the process of multi-scale pore characterization.

Using micron-scale CT scanning and image processing software, we achieved 3D quantitative characterization of pore throat structure (Fig. 5a). Based on the skeleton model of pores and throats, we obtained pore radius distribution, throat radius distribution, and some structure parameters, e.g. throat length and coordination number (Fig. 5b).

According to micron-scale CT scanning and quantitative analysis of pore throat structure parameters for 7 samples (B667-2, Y22-22-1, Y22-22-2, S548-1, S548-2, S126-1, and Y172-2), Paleogene tight sandstone and sandy conglomerate in the Jiyang depression feature pores of micron scale and throats of sub-micron scale to nano scale. The throat radius ranges 1.20-4.35 μm on average. The throat length ranges 23.9-67.7 μm at most and 5.60-14.11 μm on average. The coordination number ranges 1-5 on average. Throats are of small length and poor connectivity in general; but local connectivity could be good and the largest coordination number of the samples may reach 9-34 (Table 4). The theoretical spatial resolution of the instrument we used in the study is 500 nm, and in fact pores and throats with their radius above 1-2 μm could be characterized. This means that the average throat radius obtained from CT scanning is larger than that from mercury injection.

We compared 3D CT-reconstructed pore throat structures of two samples at the same section with the same sedimentation type and different permeabilities. For the sample Y22-22-1 with the porosity of 3.6% and the permeability of 0.099×10^-3 μm² (Fig. 5c), pores and throats which can be identified using micron-scale CT scanning occur locally in isolation with poor pore-throat interconnectivity. For the sample Y22-22-2 with the porosity of 5.9% and the permeability of 0.112×10^-3 μm² (Fig. 5d), pores and throats which can be identified using micron-scale CT scanning occur homogeneously in the sample with good pore-throat interconnectivity. Pore throat structure dominates oil-bearing capacity of two samples. The sample Y22-22-2 with good pore-throat interconnectivity and homogeneous pore throat distribution shows the oil-bearing grade of oil cut, while Y22-22-1 shows the oil-bearing grade of no oil.

Different techniques and methods give rise to different results of pore throat distribution in tight reservoirs. This is attributed to the discrepancies in measuring range and accuracy. Routine mercury injection can be used to characterize pores and throats with the radius above 25 nm, and constant-rate mercury injection is feasible for the study of pores and throats with the radius exceeding 120 nm. Owing to resolution restriction, micron-scale CT scanning is applicable to more limited pores and throats. Some parameters, e.g. average pore throat radius, obtained using the last two techniques are obviously higher than those derived from routine mercury injection. But to get such structure parameters as coordination number, pore-to-throat volume ratio, pore-to-throat radius ratio, and spatial distribution of pores and throats, we still need constant-rate mercury injection and CT scanning data.

As per the stipulations of tight clastic reservoir classification in "Geological evaluating methods for tight oil", grade-I reservoirs are defined with the porosity above 12% and the in situ permeability above 0.1×10^-3 μm²; grade-II reservoirs are defined with the porosity of 8-12% and the in situ permeability above (0.01-0.10)×10^-3 μm²; grade-III reservoirs are defined with the porosity of 5%-8% and the in situ permeability above (0.001-0.010)× 10^-3 μm². These stipulations are the basis of petrophysical property evaluation for tight oil reservoirs^[6]. But tight oil yield is dominated by many factors, e.g. formation fluids and pore throat structure of the reservoir; thus, feasible classification criteria should be established in accordance with reservoir properties and production in the area of interest. Many efforts focused on reservoir classification and evaluation based on pore throat structure^{[18, 22, 29-33]}. Gao et al. used a large amount of mercury injection data for cluster analysis and accomplished the classification of low-permeability and tight sandstone reservoirs with the air permeability smaller than 15×10^-3 μm^2[16]. We also used a large amount of mercury injection data and hierarchical cluster analysis, together with oil testing data, to perform the classification of ultra-low-permeability and tight sandstone and sandy conglomerate with the air permeability smaller than 3×10^-3 μm².

Cluster analysis originates in variance analysis. The sum of squares of deviations increases when uniting two classes, and thus the principle of clustering is to unite two classes with the minimum increment. We adopted Ward hierarchical clustering, and sample-to-sample distance was measured using squared Euclidean distance. Cluster analysis dealt with 10 parameters derived from mercury injection, i.e. mean pore throat radius (R_a), maximum pore throat radius (R₀), median pore throat radius (R₅₀), uniformity coefficient, coefficient of variation, construction coefficient, maximum mercury saturation, efficiency of mercury withdrawal, lithology factor, and characteristic structure coefficient.

Through cluster analysis, we classified the samples into several types. We investigated the permeability and intrusive mercury curve for each type of tight reservoirs and then united similar types^[24]. As a result, we classified Paleogene tight sandstone and sandy conglomerate reservoirs of different types in the Jiyang depression as grade-I, grade-II, grade-III, and grade-IV reservoirs.

As per the statistical results, grade-I to grade-IV reservoirs exhibit different R_a and R₅₀. For grade-I reservoirs, average R_a is 0.6 μm and average R₅₀ is 0.25 μm. For grade-II reservoirs, average R_a is 0.40 μm and average R₅₀ is 0.14 μm. For grade-III reservoirs, average R_a is 0.30 μm and average R₅₀ is 0.10 μm. For grade-IV reservoirs, average R_a is only 0.15 μm and average R₅₀ is only 0.06 μm. Thus, we could use R₅₀ and R_a as the indicators to classify pore throat structure types.

As per pore throat structure parameters of tight sandstone and sandy conglomerate reservoirs of different types, grade-I to grade-IV tight sandstone reservoirs of deltaic front subfacies, shore-shallow lacustrine beach- bar microfacies of inshore shallow lake subfacies, and turbidite facies are close in pore throat radius parameters. The differences in R₅₀ and R_a are less than 0.1 μm for the same grade of sandstone reservoirs with different sedimentation types. Compared with the other three types, grade-I, grade-II, and grade-III tight sandy conglomerate reservoirs show higher R₀ and R_a and close-by R₅₀. This means that with respect to a specific permeability of tight sandy conglomerate reservoirs, the maximum pore throat radius and mean pore throat radius are high, and median pore throat radius is close. This indicates strong heterogeneity of pore throat distribution (Fig. 6).

For example, for tight sandstone of turbidite facies, grade-IV reservoirs show R₀ below 0.8 μm, R_a below 0.2 μm, and R₅₀ below 0.15 μm; grade-III reservoirs exhibit bimodal R₀ which mainly distributes within 0.2-0.8 μm and 1.4-1.6 μm, R_a of 0.2-0.4 μm, and R₅₀ of 0.05-0.15 μm; grade-II reservoirs also exhibit bimodal R₀ distribution with two major intervals of 1.4-1.6 μm and 2.0-2.6 μm, R_a of 0.4-0.7 μm, and R₅₀ of 0.05-0.20 μm; grade-I reservoirs show R₀ within 1.4-1.6 μm and 2.0-2.6 μm, R_a of 0.6-1.0 μm, and R₅₀ of 0.15-0.55 μm (Fig. 7). The distributions of pore throat structure parameters of tight sandstone reservoirs of shore-shallow lacustrine beach-bar microfacies and deltaic front subfacies are similar to those of tight sandstone reservoir of turbidite facies; thus, it is unnecessary to discuss the details.

For tight sandy conglomerate, grade-IV reservoirs similarly show R₀ below 0.8 μm, R_a below 0.2 μm, and R₅₀ below 0.1 μm; grade-III reservoirs exhibit bimodal R₀ distribution with its dominant peak at 1.2-1.6 μm and secondary peak at 2.2-2.6 μm, R_a of 0.2-0.4 μm, and R₅₀ below 0.15 μm; grade-II reservoirs show bimodal R₀ distribution at 1.2-1.6 μm and 2.2-3.2 μm, R_a of 0.4-0.7 μm, and R₅₀ of 0.05-0.25 μm; grade-I reservoirs show R₀ of 2.2-3.2 μm, R_a of 0.5-1.0 μm, and bimodal R₅₀ distribution mainly at 0.05-0.15 μm and 0.30-0.55 μm (Fig. 8).

According to the results of hierarchical cluster analysis, we establish a tight reservoir classification scheme based on pore throat structure parameters (Table 5). R_a is an important indicator in this scheme and thus is taken as the dominant parameter of pore throat structure classification; R₅₀ and R₀ are taken as the secondary parameters. Using these parameters, grade-III and grade-IV reservoirs can be clearly categorized; but there is an overlap 200 nm of pore throat structure parameters for grade-I and grade-II reservoirs.

According to mercury injection data for 101 tight reservoir layers with oil testing at 66 wells, there is a positive correlation between mean pore throat radius and testing yield; in other words, oil and gas production increases with R_a. For tight sandstone of turbidite facies, beach-bar microfacies of inshore shallow lake subfacies, and deltaic front subfacies, grade-I reservoirs may yield economic oil flow without stimulations, e.g. fracturing, with average R_a exceeding 0.6 μm (or even exceeding 0.8 for 84% of oil testing layers). Grade-II and grade-III reservoirs were tested to be low-yield oil layers or dry layers, which should be fractured to yield economic oil flow; the average R_a ranges 0.3-0.8 μm, or reaches 0.4-0.6 μm for 64% of oil testing layers. Grade-IV reservoirs cannot yield economic oil flow using current stimulation techniques, e.g. fracturing, with the average R_a below 0.2 μm (Fig. 9a).

For tight sandy conglomerate, grade-I reservoirs may yield economic oil flow without stimulations, e.g. fracturing, with the average pore throat radius exceeding 0.8 μm. Grade-II and grade-III reservoirs were tested to be low-yield oil layers or dry layers, which should be fractured to yield economic oil flow, with the average R_a ranging 0.3-0.8 m. Grade-IV reservoirs cannot yield economic oil flow using current stimulation techniques, e.g. fracturing, with the average R_a below 0.2 μm.

As per the R_a value of sandstone in the oil testing layers and clustering results, grade-I reservoirs are believed to yield economic oil flow without stimulations, e.g. fracturing, with the lower limit of R_a set to be 0.6 μm for grade-I tight sandstone reservoirs of turbidite facies, shore-shallow lacustrine beach-bar microfacies of inshore shallow lake subfacies, and deltaic front subfacies, and 0.8 μm for grade-I tight sandy conglomerate reservoirs. Grade-II and grade-III reservoirs may yield economic oil flow after stimulations, e.g. fracturing, with the lower limit of R_a set to be 0.2 μm. Sandstone reservoirs with R_a below 0.2 μm were tested to be ineffective reservoirs which cannot be stimulated using current fracturing techniques. Fracturing mentioned above was implemented using total fracturing fluids of 103.60-441.46 m³ and sands of 5.3-51.5 m³.

We investigated some tight reservoir classification and evaluation schemes ^{[18, 22, 29-33]}. Compared with the criteria for tight Cretaceous sandstone in the Songliao Basin and tight Chang 7 sandstone in the Ordos Basin, our threshold values of classification are higher. With respect to tight Chang 7 sandstone in Longdong, the Ordos Basin, Xu et al.^[33] presented the criteria of permeability above 0.25× 10^-3 μm² and R_a above 0.3 μm for grade-I reservoirs and permeability of (0.10-0.25)×10^-3 μm² and R_a of 0.15-0.30 μm for grade-II reservoirs, which are equivalent to our standards for grade-III and grade-IV reservoirs, respectively. They also suggested permeability below 0.1×10^-3 μm² and R_a below 0.15 μm for grade-III reservoirs. After investigation, we found that our high thresholds of classification were related to fluid mobility. Reservoir classification has usually been based on petrophysical properties and seldom involved fluids. But for porous media and fluid system of nano-scale, molecular dimension and structure of fluids have considerable impacts on porous media^[34]. With respect to fluid mobility, the Paleogene oil sources in the Jiyang depression are Sha 3 and Sha 4 brackish source rocks, the R_o of which is below 0.9% nowadays. Due to the influence of kerogen, crude oil generated features large density mostly exceeding 0.85 g/cm³, high wax content, and large long-chain alkane molecules. This results in poor oil mobility in nano-scale pores and throats; consequently, the lower limit of pore throat radius should be large enough to yield economic oil flow. This is why that our thresholds of pore throat radius for Paleogene tight sandstone reservoir classification are higher than those for the Chang 7 Member in the Ordos Basin.

The above scheme of reservoir classification was applied to the optimization of tight reservoirs to be fractured and sweet spot pinpointing in the Paleogene, the Jiyang depression.

With respect to the optimization of tight reservoirs to be fractured, we used log data to evaluate pore throat structure parameters and selected well intervals with grade-I and grade-II reservoirs for fracturing. This improved per-well post-frac deliverability. In YX233 Block, for example, 3 exploratory wells were interpreted to have 3 tight sandy conglomerate oil layers, i.e. Well YX232 with the layer at 4470-4480 m, Well YX229 with the layer at 4266.0-4273.2 m, and Well YX233 with the layer at 3670.0-3678.5 m. Log interpretation showed close porosities of 7.1%, 7.7%, and 7.8% for these 3 tight sandy conglomerate layers. But for pore throat structure parameter, R_a was estimated to be 1.01, 0.54, and 0.38 μm, respectively. As per classification criteria, these layers were evaluated to be grade-I, grade-II, and grade-III reservoirs, respectively. Thus, different oil testing methods were adopted. For the grade-I reservoir at 4470-4480 m in Well YX232, the choke of 5 mm in diameter and orifice plate of 22 mm in diameter were used to obtain daily oil output of 24.5 t and daily gas output of 3473 m³. For the grade-II reservoir at 4266.0-4273.2 m in Well YX229, which was a low-yield oil layer with the initial daily oil output of 2.89 t, CO₂-enhanced fracturing was performed and increased daily oil output to 43.5 t (total fracturing fluids of 472 m³, sands volume of 44 m³, and maximum displacement of 5.58 m³/min). For the grade-III reservoir at 3670.0-3678.5 m in Well YX233, which produced no oil and gas before stimulation, large-scale CO₂-enhanced fracturing was performed and resulted in daily oil output of 5.27 t (total fracturing fluids of 988 m³, sands volume of 50 m³, and maximum displacement of 9.5 m³/min). Post-frac oil yields of these 3 tight sandy conglomerate reservoirs agreed with the prediction based on the classification scheme. This scheme was applied to the optimization of tight layers to be fractured in 41 wells, among which 37 wells successfully yielded economic oil and gas flow.

Pore throat structure parameters for each individual well could be established using the log evaluation model of pore throat structure parameters derived from pseudo-capillary pressure curve inverted from transverse relaxation time spectrum of nuclear magnetic resonance log or routine log curves calibrated using mercury injection data. Combining per-well pore throat structure parameters with our classification criteria, we may calculate per-well net pay of grade-I and grade-II reservoirs and predict spatial extension of grade-I+II reservoirs within the tight oil evaluation unit. Using this method, we accomplished reservoir classification and evaluation based on pore throat structures for 19 blocks with possible tight oil reserves and 23 blocks with probable tight oil reserves in the Paleogene System, the Jiyang depression. Twelve blocks with grade-I+II reservoirs accounting for more than 70%, e.g. YX233 Block with upper Sha4 tight sandy conglomerate, T764 Block with upper Sha4 tight sandy conglomerate, and XX49 Block with lower Sha3 tight deltaic front sandstone, were selected for evaluation and deployment. From 2015 to 2019, 16 exploratory wells, e.g. YX236, were deployed. New probable reserves were be 3880.5×10⁴ t, and new proved reserves were 408.8×10⁴ t.

Tight reservoirs in the Paleogene System, the Jiyang depression include tight sandstone and tight sandy conglomerate. The former is classified into turbidite facies, shore-shallow lacustrine beach-bar microfacies of inshore shallow lake subfacies, and deltaic front subfacies. Pebbly sandstone and gravel of tight sandstones feature small particle size, low component maturity and textural maturity, and major interstitial materials of matrix and carbonate cements. Tight sandy conglomerate features large particle size and lower component maturity and textural maturity of sandstone.

As per routine mercury injection, constant-rate mercury injection, and micron-scale CT scanning, Paleogene tight reservoirs in the Jiyang depression features large pores and tiny throats with greatly varied radius. Tight sandstone samples of turbidite facies, shore-shallow lacustrine beach-bar microfacies of inshore shallow lake subfacies, and deltaic front subfacies exhibit very close-by pore throat size, homogeneous distribution of pores and throats, and pore throat tortuosity. Tight sandy conglomerate features large value of the maximum pore throat radius, small pore throat tortuosity, poor pore throat sorting, and small value of the largest mercury saturation.

In terms of routine mercury injection results of tight sandstone and sandy conglomerate, the mean pore throat radius is taken as the dominant parameter of pore throat structure classification; the median pore throat radius and maximum pore throat radius are taken as the secondary parameters. In accordance with these parameters and oil testing results, grade-I reservoirs are believed to yield economic oil flow without stimulations, e.g. fracturing. The lower limit of mean pore throat radius is set to be 0.6 μm for grade-I tight sandstone reservoirs of turbidite facies, shore-shallow lacustrine beach-bar microfacies of inshore shallow lake subfacies, and deltaic front subfacies and 0.8 μm for grade-I tight sandy conglomerate reservoirs. Grade-II and grade-III reservoirs may yield economic oil flow after stimulations, e.g. fracturing, with the lower limit set to be 0.2 μm. Grade-IV reservoirs with mean pore throat radius below 0.2 μm are taken as ineffective reservoirs, which cannot be stimulated using current fracturing techniques.

The establishment of the above criteria provides a basis for the classification and evaluation of Paleogene tight reservoirs in the Jiyang depression. This classification scheme has been successfully applied to the optimization of tight reservoirs to be fractured, sweet spot evaluation, and exploration and assessment of tight oil reserves.

K_p—kurtosis of pore throat radius distribution curve derived from mercury injection, dimensionless;

R₀—maximum pore throat radius derived from mercury injection, μm;

R₅₀—median pore throat radius derived from mercury injection, μm;

R_a—mean pore throat radius derived from mercury injection, μm;

S_kp—skewness of pore throat radius distribution curve derived from mercury injection, dimensionless.