According to the lithologic, physical and electrical characteristics of the interlayer under Ng III down, the interlayer can be divided into two types, namely argillaceous and physical interlayer. Through the analysis of the logging response characteristics of the interlayer, the logging identification standard of the interlayer is established by using the intersection diagram method (Table 1). The logging response characteristics of the muddy interlayer are: high gamma, high density and low resistivity. The response characteristics of physical interlayer are: low gamma, high density and low resistivity. It is distinguished from argillaceous interlayer by relatively low gamma value and high density.

According to the principle of approximate equal elevation and equal thickness comparison of fluvial facies, according to the distribution characteristics and development needs of interlayer, the top oil-bearing interval of NgIII down formation is subdivided into three single stage braided flow zones, corresponding

to three small layers respectively. Two stages of relatively stable interlayer are developed between the small layers (Figure 2). Among them, the interlayer in the first stage is mainly filled with flood plain mud and abandoned river mud, with relatively stable distribution. The thickness above the well is 1.0 - 4.2 m, with an average of 2.1 m. There are various genetic types of interlayer in phase 2, including flood plain mud, inter dam mud, ditch mud filling and scouring mud gravel at the bottom of the river channel. Due to the frequent change of river channel scouring and hydrodynamic force, the thickness is thin and changes rapidly laterally, with poor contrast. The thickness above the well is 0.5 - 2 m, with an average of 1.2 m.

Based on the classification of rock elastic attribute and logging lithology, the elastic attribute sensitive to lithology is expressed. Its purpose is to study the corresponding relationship between elastic attribute and lithology by using the elastic parameters of logging. Considering that there are certain differences in the petrophysical characteristics of each small layer, the statistical analysis of petrophysical parameters by small layers can effectively eliminate the “illusion” of P-wave impedance superposition caused by differential compaction. Figure 3 shows the statistical histogram of multi well P-wave impedance and lithology divided by sub layers. It can be seen that the P-wave impedance of each sub layer has a good distinction between sand and mudstone, sandstone has low impedance and mudstone has high impedance, which lays a petrophysical foundation for seismic inversion.

The optimization of vertical variation function in traditional geostatistics inversion mainly depends on well data statistics, and the horizontal variation function mainly comes from dense well pattern. In this study, the longitudinal variation range of variation function mainly depends on well data and prediction demand, and the transverse variation function mainly comes from deterministic inversion and quantitative knowledge base of sandy braided river deposition.

Predecessors [6] [8] [10] [15] summarized the geological knowledge base such as the geometric shape, vertical and horizontal distribution range of sandy braided river interlayer by using the methods of dense well pattern anatomy and field outcrop actual measurement. Combined with the actual geological characteristics of the study area, they summarized the geological knowledge base of different genetic reservoirs and interlayer of NgIII down lower oil formation (Table 2) as the constraint conditions of variation function.

According to the above well statistical analysis, the sedimentary microfacies of each small reservoir of NgIII down lower oil formation are mainly bar, and the thickness of a single stage is 6 - 7 m. According to the empirical formula of sedimentary configuration unit scale of sandy braided river summarized by predecessors, it is calculated that the single core beach dam is 200 - 260 m wide, 900 - 1100 m long and the length width ratio is 3.5 - 5.2. Comprehensively considering the distribution characteristics and scale of reservoir and interlayer, it is finally determined that the longitudinal variation range of reservoir of the first layer in NgIII down is 2 ms, the main variation range is 400 m and the secondary variation range is 1200 m. The vertical range of the Phase I interlayer is 1 ms, the main range is 1200 m and the secondary range is 2300 m (Figure 4). Finally, this method is used to determine the variable range parameters in all directions of the reservoir and interlayer of all the layers, providing constraints for statistical inversion.

Based on the optimization of variation function, the sequential Gaussian random simulation method is used to simulate the reservoir parameters, integrate the logging data and deterministic inversion, establish a high-resolution reservoir model, and obtain the lithofacies probability volume through comprehensive operation.

Through deterministic inversion and 90 degree phase shift trend comparison, the rationality of geostatistical inversion results is analyzed. Figure 5 shows the original seismic, 90 degree phase shift and geostatistical lithologic probability body profile passing through well C1. It can be seen that the statistical inversion has the same description of the overall reservoir structure, and the vertical resolution is higher than 90 degree phase shift, which has a fine description of the discontinuous thin interlayer.

In order to analyze the quality control of the prediction results, the statistical analysis of the error of blind wells not involved in the inversion operation (Figure 6)

shows that for the interlayer with a thickness of more than 1.5 m, the relative error within 20% accounts for more than 75%, and only a few thin interlayer above the well (thickness less than 1.5 m) have a large error, indicating that the overall prediction result of the interlayer has high reliability.

The NgIII down is the deepest development layer series in the oilfield, and there are few drilled wells. Therefore, rich dynamic production data are used to verify the interlayer inversion results. Taking well D19H as an example, the trajectory of the horizontal section of the well is located in a small layer. The production characteristics are that the water cut rises slowly, and the water cut slowly breaks through to 60% after one and a half years of production. The water cut curve is of “slow climbing” type, which is significantly different from the non shielding type, indicating that there is a large range of interlayer shielding in the lower part, and the water cut breaks through after the bottom water flows around. The dynamic response is consistent with the inversion result (Figure 7 and Figure 8).

There are obvious differences between the two main interlayers at the top of the NgIII down due to hydrodynamic changes, different sedimentary genetic types and different degrees of undercut erosion and erosion of the river channel. Among them, the distribution of interlayer layers in the first stage is relatively stable, which is mainly formed by flood mud. The plane is distributed in blocks in northeast direction, with a wide distribution area and a thickness of 1 - 4 m, forming a wide range of shielding. It is the main interlayer the NgIII down (Figure 8). In the second stage, the interchannel interlayer has small distribution area and thin thickness. It is mainly filled with inter dam mud and abandoned/late stage river mud. Due to the development of “skylight” and the formation of seepage channel, the braided river reservoir presents the structural characteristics of “overall connectivity and internal complexity”.

Based on the prediction research on the sedimentary stages and distribution range of interlayer, the interlayer in phase 1 has a wide distribution range and large thickness, resulting in strong seepage shielding, which can effectively inhibit the coning of bottom water. In order to avoid the gas cap, the horizontal production wells deployed in the high part of the structure and the gas cap area in the early stage of development are relatively deep in the horizontal section, mainly producing small layers 2 and 3, resulting in the formation of residual oil enrichment area in small layer 1 due to large-area interlayer shielding, which is a favorable area for adjusting and tapping the potential. According to this understanding, 12 development wells were deployed in the lower part of NgIII down from 2019 to 2020, most of which belong to inter well densified wells in small layer 1 at the high part (Figure 8). The average daily oil production in the initial stage is 120 m³/day and the water cut is also lower than that of the surrounding old wells. Significant results have been achieved in tapping the potential of bottom water reservoirs with high water cut and high recovery. At the same time, further potential tapping can be further intensified in the future to improve oil recovery.

Using core, logging and seismic data, according to the genetic type and sedimentary period of interlayer sedimentation, the distribution range of thin interlayers between thick braided rivers is quantitatively predicted by the geostatistical inversion method.

1) Based on the core and logging data and combined with the field outcrop of a sandy braided river, the sedimentary genetic types and sedimentary periods of thick sandstone interlayer of NgIII down are studied, and two stages of interlayer are identified and divided. The interlayer in the first stage is mainly formed by flood mud, which is relatively stable in distribution and is the main interlayer of this layer.

2) On the basis of deterministic inversion and qualitative prediction, the geostatistical inversion method guided by logging constraints and geological knowledge base is used to predict the distribution range of thin interlayer. Through the verification of blind good error and production well reservoir performance, it shows that the prediction result of interlayer greater than 1.5 m has high reliability, which also provides a new idea and method for the prediction of thin reservoir or interlayer of the same type of sedimentation.

3) Based on the interval division and prediction results of the interlayer, the oil-bearing interval at the top of NgIII down is divided into three small layers. The practice shows that the remaining oil enrichment area is formed in the high part of the structure due to the shielding of large-scale interlayer in the layer 1.