Method for quantitatively recovering paleo-water depth based on preproduct reflection

Abstract

A method for quantitatively recovering paleo-water depth based on preproduct reflection belongs to the field of oil and gas field geological exploration, and comprises the following steps: collecting earthquake, drilling and logging data of a research area; establishing a high-precision isochronous stratigraphic framework of a target interval, and identifying a plurality of sets of front product reflection structures on a plurality of consequent object source profiles; determining the slope crest and the slope toe of the front product reflection structure, and calculating the apparent thickness Hs of each front product body by using a segmented accumulation method; a real thickness Hz is obtained by using a decompaction correction technology; longitudinally connecting the obtained paleo-water depth values to obtain a paleo-water depth change broken line graph; and constructing a deep interpolation neural network model, and smoothing the curve to obtain a lake plane change curve. The paleo-water depth of the lake basin is recovered in a high-precision and quantitative manner by adopting a segmented accumulation calculation method; a deep neural network model is constructed to carry out interpolation calculation and smoothing processing on a broken line graph formed by connecting only a small number of paleo-water depth value points, and a lake plane change curve is rapidly obtained.

Description

Technical Field

[0001] The invention relates to the field of oil and gas field geological exploration, and in particular to a method for quantitatively restoring ancient water depth based on foreset reflection. Background Art

[0002] With the deepening of oil and gas exploration, the industry has gradually formed the understanding that semi-deep water areas and deep water areas control the development of high-quality source rocks. Therefore, the restoration of ancient water depth is an important task in the field of oil and gas geological exploration. It has important guiding significance for studying the sedimentary evolution process of basins, determining the type of sedimentary system, and evaluating and predicting the conditions of source rocks, reservoirs, caprocks, etc. At present, the commonly used ancient water depth restoration methods can be divided into two categories: qualitative and semi-quantitative-quantitative. Qualitative restoration mainly includes sediment color method, bedding type method, geochemical element method and stratigraphic thickness method. For example, patent CN 118296094A discloses a method, equipment and storage medium for quantitative characterization of paleo-geomorphology. Patent CN111580183 B discloses a method for quantitative restoration of ancient lake water depth. Patent CN117851748 B discloses a method and system for calculating ancient water depth in low-exploration areas. In 2009, Volume 45, Issue 6, Journal of Lanzhou University (Natural Science Edition), Pang Jungang and Li Wenhou determined the scope of the deep lake area of ​​the lake basin by combining the sedimentary phases of the Chang 7 of the Yanchang Formation and the core bedding type analysis. In 2017, Volume 31, Issue 6, Modern Geology, Wang Chenjie et al. used the lithology and sedimentary phase characteristics of the lower section of the second section of the Dongying Formation to restore and correct the ancient water depth of the study area. In 2020, Volume 42, Issue 2, Petroleum Experimental Geology, Xiao Kunze et al. used the correspondence between water depth and formation thickness to restore the average depth of ancient seawater in the Miocene of the Yinggehai Basin. These methods can qualitatively restore the bottom structure of the basin and clarify the relative heights of different parts, but cannot obtain quantitative ancient water depth data at each point. In addition, these methods are easily restricted by natural conditions and require a large amount of intensive sampling, which is costly. In recent years, geological prospectors have gradually realized the importance of quantitatively restoring ancient water depths, and new methods have been continuously proposed, including the shoreline trajectory method, the Milankovitch cycle method combined with Co element analysis, the paleontological method, the TOC determination method, and the cobblestone abundance method. For example, patent CN113674806 B discloses a method for restoring the ancient water depth of a sedimentary lake basin based on trace elements and paleontological differentiation. Patent CN113687440 A discloses a method and storage medium for quantitative restoration of ancient water depth based on Milankovitch cycles. Patent CN114624775B discloses a comprehensive quantitative restoration method for ancient water depths of sedimentary lake basins. Patent CN117687113A discloses a quantitative restoration method for ancient water depths. Patent CN118133519A discloses a quantitative restoration method for ancient landforms of faulted lake basins based on quantitative calculation of ancient water depths. Patent CN104932031 B discloses a quantitative calculation method for ancient water depths of lake phase deposits. In the Journal of Paleogeography, Vol. 7, No. 3, 2005, Li Shoujun et al. studied the dominant heterogeneity of ostracods. (s)The relationship between the water depth D of the ancient lake was fitted, and the ancient water depth of the third member of the Paleogene Shahejie Formation in the Dongying Sag was quantitatively restored. In 2016, Volume 18, Issue 2, Journal of Paleogeography, Du Qingxiang et al. restored the relative ancient water depth of the top of the middle sub-member of the first member of the Shahejie Formation in the study area during the deposition period based on the changes in Mn / Fe values, kerogen types, algae fossils, and the discovery of tubercle-like snails. In 2024, Volume 42, Issue 1, Journal of Sedimentary Sciences, Wang Changyong et al. quantitatively restored the paleowater depth of the Early Jurassic using the La-Co method, TOC method, and Th / U ratio method. These methods can semi-quantitatively and quantitatively restore the paleowater depth, which has played a certain role in promoting the restoration of the paleowater depth of lacustrine and marine sediments, but these methods also have some problems in practical application. For example, the shoreline trajectory method is mainly used to restore the paleowater depth of large delta sedimentary areas. The calculation process must consider multiple factors such as decompaction, source supply rate, and differential structural settlement. It has strong uncertainty, and the calculation process is complex and the operability is poor. The paleontological method has high sample requirements and is limited by the diversity of ostracods and foraminifera, so it is only suitable for analyzing the sediment depth of shallow lakes. Trace element methods such as TOC and La-Co require a large amount of trace element test data in the study area, and the experimental cost is relatively high. Due to the inherent living habits of coccoliths, the coccolith abundance method is only suitable for paleo-water depth restoration in areas near the coastline or the coast.

[0003] After analyzing the above existing technologies, the following problems were found: (1) At present, the restoration of paleowater depth is mainly based on qualitative analysis, and the research on quantitative restoration of paleowater depth is still relatively lacking. Previous researchers have focused on semi-quantitative analysis by comprehensive use of paleontological markers, geochemical analysis, trace element testing and other means. However, these methods require a large amount of core sample data and analytical test data when restoring paleowater depth, which is costly and is mostly suitable for areas with rich drilling data. (2) Although the shoreline trajectory method for large delta deposits can quantitatively calculate paleowater depth, due to too many influencing factors in the calculation process, the calculation process is complicated and the labor cost is high, which is not suitable for large-scale paleowater depth restoration work. (3) At present, there is no method that can simply and quickly quantitatively restore paleowater depth based on only a small amount of exploration data, and then obtain the lake level change curve to provide guidance for studying the sedimentary filling evolution process of the basin. Summary of the invention

[0004] The purpose of the present invention is to provide a method for quantitatively restoring paleo-water depth based on foreset reflection, so as to solve the problem that the prior art cannot quickly and accurately restore paleo-water depth and lake level change curve. The method combines seismic data, core data and logging data, and quantitatively calculates paleo-water depth by identifying the foreset reflection structure on the seismic profile. Not only does the de-compaction correction technology solve the influence of formation differential compaction on sediment thickness, but also the deep neural network model is constructed to interpolate and smooth the line graph connected by only a small number of paleo-water depth value points to obtain the lake level change curve. This method requires little data, and the calculation process is simple and fast. The paleo-water depth can be quickly and quantitatively restored only through simple geometric calculations, and the lake level change curve is generated using deep neural network technology. In addition, the trained deep interpolation neural network model can be reused, which reduces the operating cost. This method is suitable for the recovery of paleo-water depth and the generation of lake level change curves over a wide range.

[0005] The technical solution adopted by the present invention is:

[0006] A method for quantitatively restoring paleo-water depth based on foreset reflection comprises the following steps:

[0007] Step 1: Collect seismic data, drilling data and logging data in the study area;

[0008] Step 2: Explain the seismic data in the study area in detail, establish a high-precision isochronous stratigraphic framework for the target layer, and flatten the top surface on multiple seismic sections along the source to identify the delta progradation reflection structure; the high-precision isochronous stratigraphic framework here is a general term, without specific standards, and is a recognized term in the industry. It is generally believed that when a stratigraphic framework is constructed using two-dimensional or three-dimensional seismic data with good data quality, it can be called high-precision, and high precision can ensure that the model can truly reflect the geological history and sedimentary environment.

[0009] Step 3: Determine the top and foot of the delta prosedation reflection structure, and use the segmented cumulative summation method to calculate the distance between the top and foot of the slope to determine the apparent thickness of the prosedation body; the top of the slope refers to the dividing point between the top accumulation layer and the foresed layer in the prosedation reflection structure, and the bottom of the slope refers to the lowest point of the foresed layer on the profile, that is, the dividing point between the foresed layer and the bottom accumulation layer.

[0010] Step 4: Combine drilling and logging data to perform decompaction correction on the apparent thickness obtained in step 3 to obtain the true thickness of the foreset body, i.e., the paleo-water depth;

[0011] Step 5, repeating steps 3 to 4, quantitatively calculating the thickness of the foreset body of different periods on the source seismic section, obtaining the paleo-water depth values ​​of different periods, and connecting them vertically to obtain a line graph of paleo-water depth changes;

[0012] Step 6: construct a deep interpolation neural network model, set model parameters, repeatedly train to obtain the best model, and use the best model to interpolate the paleo-water depth calculated in step 5;

[0013] Step 7: Draw the interpolated curve and use the CubicSpline function to smooth the curve to obtain a smooth paleo-water depth change curve, that is, a lake level change curve.

[0014] Furthermore, in step 2, the process of identifying the foreset reflection structure includes the following steps:

[0015] S201: According to the sequence stratigraphic theory, a set of coal seams with strong peaks and strong troughs in seismic response in the target layer section is identified as the top marker layer, and a set of widely and stably developed black shale is identified as the bottom marker layer;

[0016] S202: Combine well and seismic data to calibrate synthetic seismic records, identify, track and interpret target layer segments, and establish a high-precision isochronal stratigraphic framework;

[0017] S203: On multiple seismic sections along the source, flatten the top surface and finely identify and depict the multi-stage foreset bodies of the target layer.

[0018] Furthermore, in step 3, the front volume thickness (H s ) method comprises the following steps:

[0019] S301: According to the drilling data in the study area, the acoustic time difference logging curve of the well closest to the slope top and which has been calibrated with synthetic seismic records is selected, abnormal data values ​​are eliminated, the logging curve data of the target layer is counted, and the acoustic wave velocity V corresponding to each depth point is calculated according to formula (1) i , where i corresponds to different depth points, i = 1, 2, ... n, n represents the number of depth points;

[0020] S302: According to the difference in logging phases between sandstone and mudstone, the acoustic wave velocity of the sandstone section and mudstone section with a thickness of not less than 5m is counted respectively; the acoustic wave time difference of the mudstone section is higher than that of the sandstone, and the whole section is irregular and sharp;

[0021] S303: Calculate the average acoustic wave velocity values ​​of different sandstone and mudstone intervals according to formula (2) Where j = 1, 2, ... m represents different intervals, and m is the number of intervals;

[0022] S304: Read the time ΔT from the top of the slope to the foot of the slope in segments on the seismic profile j , substitute into formula (3) to calculate the apparent thickness H of the forebody s ;

[0023]

[0024] Where: V i is the sound wave velocity at the i-th depth point, m / us; Δt i is the acoustic time difference value of the i-th depth point, us / m; is the average sound wave velocity of the jth interval, m / us; ΔT j is the time difference of the jth time period read on the seismic profile, ms; n j is the number of depth points in the jth interval.

[0025] Among them, the segmented cumulative summation method uses the acoustic logging velocity instead of directly using the seismic logging velocity. This is because compared with the average formation velocity directly obtained by seismic logging, the average velocity obtained by acoustic logging represents the true velocity of the formation, and because it can be measured continuously, the receiving distance is small, and the layer velocity can be finely divided, which better reflects the lithological characteristics of the formation, and the results are more accurate and more in line with the actual geological significance. In addition, acoustic logging can be carried out at the same time as electrical logging, which is more efficient and convenient.

[0026] In addition, the acoustic time difference logging curve of the well closest to the top of the slope and which has been calibrated with synthetic seismic records is used. Because the distance on the well and the time on the seismic profile can be one-to-one correspondence for the wells with synthetic seismic records and stratum calibration, the formation velocity and the time of reading the seismic profile can be calculated in sections, and the cumulative summation can be used to calculate the pre-deposition volume thickness (H s ).

[0027] Furthermore, in step 4, the decompaction correction includes the following steps:

[0028] S401: Statistically analyze the measured porosity and corresponding depth values ​​of different lithologies in the target layers of multiple wells in the study area, and fit the porosity Ф-burial depth Z relationship curve of different lithologies according to the exponential function distribution:

[0029] φ=φ0e -C*Z (4)

[0030] Where: Ф is the porosity at depth Z; Ф0 is the surface porosity; C is the compaction coefficient; Z is the burial depth, m;

[0031] S402: Based on the principle that the volume of the rock skeleton particles remains unchanged before and after compaction, and given the current top boundary depth (Z1) and bottom boundary depth (Z2) of the formation, the rock skeleton thickness integral mathematical model (Formula 5) is used to iteratively calculate the top interface burial depth Z'1 and the bottom interface burial depth Z'2 of the corresponding original formation, and then calculate the original formation thickness; then, the compaction rate t of the target layer section of the single well is calculated based on the ratio of the original formation thickness to the current formation thickness (Formula 6):

[0032]

[0033] Where: Z1, Z2 are the top and bottom depths of the current strata, m; Z'1, Z'2 are the top and bottom depths of the original strata, m; t is the compaction rate;

[0034] S403: Compare the compaction rate of the target layer section of the single well corresponding to the different foreset bodies with the apparent thickness H of the foreset body calculated in step 3 s Multiply them to get the true thickness H after de-compaction correction z , that is, the paleo-water depth:

[0035] H z =t·H s (7)

[0036] Where: H z is the true thickness of the foreset body (paleo-water depth), m; H s is the apparent thickness of the foreset volume, m.

[0037] Furthermore, in step six, the interpolation calculation includes the following steps:

[0038] S601: Extract the depth column and calculate the actual thickness H of the pre-integrated volume at each stage z (i.e. ancient water depth);

[0039] S602: Perform normalization preprocessing on the extracted data: Use MinMaxScaler to scale the data to the interval [0,1] to improve the efficiency and accuracy of neural network training;

[0040] S603: Construct a deep interpolation neural network model: the model includes an input layer, three hidden layers and an output layer, each hidden layer contains 32 neurons, and the activation function is ReLU;

[0041] S604: Conduct model training: Use the mean square error (MSE) loss function to train the model, and use the Adam optimizer for optimization, where the learning rate is set to 0.01 and the training rounds (epochs) are 2000; stop training when the MSE reaches 0.01, and obtain the best deep interpolation model;

[0042] S605: Use the trained best model to perform interpolation calculation and draw the interpolation result.

[0043] Beneficial effects of the present invention:

[0044] (1) Based on the identification of the foreset reflection structure on the seismic section, the present invention calculates the thickness of the foreset body by using the segmented cumulative summation method. At the same time, considering the influence of differential compaction on the stratum thickness, the target layer is decompacted and corrected, and the thickness of the foreset body of each period is converted into the original sedimentary thickness, which quantitatively characterizes the change trend of the lake level in the lake basin, and the recovery result is accurate and reliable.

[0045] (2) Compared with other methods that directly use the average formation velocity on the seismic profile to calculate the thickness of the progradation body, the segmented cumulative summation method fully considers the differences in the formation velocities of different lithologies and is more in line with the actual geological significance. Different lithology intervals are accurately divided according to the logging phase, the average formation velocity of each interval is calculated, and then the apparent thickness of the progradation body is accumulated and calculated, making the calculation results more accurate and the overall operation simpler.

[0046] (3) Compared with previous studies, the calculation of the thickness of the foreset body (paleo-water depth) requires a large amount of analytical test data and exploration logging data, and the calculation process is complicated. The present invention only requires a small amount of seismic data, drilling data and the most basic logging curve data (AC, CNL, DEN, GR, SP), and obtains the thickness of the foreset body (paleo-water depth) in different periods through the segmented cumulative summation method, and fully considers the influence of differential compaction. The overall calculation process is simple and fast, the results are more accurate, and it has good promotion and application value.

[0047] (4) The present invention proposes an interpolation calculation method based on a deep neural network model, which interpolates and smoothes a line graph composed of a small number of paleo-water depth value points to obtain a smooth lake level change curve. Compared with the traditional one-time interpolation method, the interpolation technology based on deep learning can achieve better interpolation effects through optimization of model parameters and repeated training. This method is particularly suitable for interpolation calculations of discrete data points such as paleo-water depth, thereby generating a more accurate and smoother interpolation curve. The lake level change curve after smoothing can better reflect the trend of basin sedimentary evolution, and provides new research ideas and technical means for the study of the sedimentary evolution process of ancient lakes and the prediction of various types of lake-phase high-quality reservoirs. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a flowchart of the method of the present invention;

[0049] Figure 2 This is a schematic diagram of the research area and the source seismic survey line of the present invention;

[0050] Figure 3 A schematic diagram of a high-precision isochronous stratigraphic framework established by the present invention taking the A-A' seismic profile as an example;

[0051] Figure 4 This is a schematic diagram of restoring paleo-water depth based on the foreset reflection structure of the present invention;

[0052] Figure 5 This is a schematic diagram of the porosity-depth relationship of sandstone and mudstone fitted in the decompaction correction of the present invention;

[0053] Figure 6 The sand-shale structure model and compaction rate calculation of each single well in the present invention;

[0054] Figure 7 It is a flow chart of the deep interpolation neural network algorithm of the present invention;

[0055] Figure 8 This is a structural diagram of the deep interpolation neural network model of the present invention;

[0056] Fig. 9 The paleo-bathymetric map restored for the present invention;

[0057] Fig.10 This is the lake level change curve restored by the present invention. DETAILED DESCRIPTION

[0058] In this example, through the systematic analysis of the three-dimensional seismic data in the Qingcheng area of ​​the Ordos Basin, a high-precision isochronous stratigraphic framework of the Yanchang Formation was established, the foreset reflection structure was identified, and the top and foot of each foreset body were determined. The apparent thickness of the foreset body was calculated using the segmented cumulative summation method, and the decompaction correction technology was used to restore the original sedimentary thickness, and the apparent thickness was converted into the true thickness (i.e., paleowater depth). Finally, a deep neural network model was constructed to interpolate a small number of paleowater depth values ​​to obtain a smooth lake level change curve. Based on the identification of the delta proset reflection structure, the present invention only requires a small amount of data to quickly and quantitatively calculate the paleowater depth, and takes into account the influence of differential compaction of the strata, so that the result is more accurate. In addition, in view of the characteristics of the small number of paleowater depth numerical points and the unevenness of the resulting curve, an interpolation calculation method based on a deep neural network model is proposed. Specifically, the following steps are included, such as Figure 1 As shown:

[0059] Step 1: Collect 3D seismic data, drilling data and logging data in Qingcheng area of ​​Ordos Basin.

[0060] Step 2: Finely interpret the 3D seismic data, establish a high-precision isochronous stratigraphic framework for the Yan9-Chang7 segment, and flatten the top surface on multiple seismic sections along the source to identify the delta progradational reflection structure, including the following steps:

[0061] 1) According to the sequence stratigraphic theory, multiple sets of coal seams are widely developed in Yan7-Yan9, and the seismic profile shows a set of response characteristics of strong wave peaks and strong wave valleys, so Yan9 is taken as the top surface marker layer. Chang7 shale shows the seismic response characteristics of "one black and two red, one peak and two valleys", so Chang7 is taken as the bottom surface marker layer.

[0062] 2) Combine well and seismic data to calibrate synthetic seismic records and calibrate the main seismic reflection interfaces y9, Tj, C1, C2, C3, and C7 顶 、C7 底 and C8 identification, tracing and interpretation to establish a high-precision isochronous stratigraphic framework.

[0063] 3) Combined with the well location distribution in the 3D work area, five seismic profiles A-A', B-B', C-C', D-D', E-E' were made in the direction of the source (southwest) (see Figure 2 ), flatten the top surface marker layer Yan9, and finely identify and depict the foreset body. Taking the A-A' seismic profile as an example, six foreset bodies F1 to F6 were identified (see Figure 3 ).

[0064] Step 3: Determine the top and foot of each foreset reflection structure, and use the segmented cumulative summation method to calculate the distance between the top and foot of the slope, and determine the apparent thickness of each foreset body. See the model diagram for details. Figure 4 As shown, the specific steps are:

[0065] 1) According to the drilling data in the study area, the acoustic time difference logging curve of the well (X1) closest to the slope top of the foreset body F1 is selected to eliminate abnormal data values. The logging curve data of the target layer is counted, and the acoustic velocity V corresponding to each depth point is calculated according to formula (1). i , where i corresponds to different depth points;

[0066] 2) According to the difference in logging phases between sandstone and mudstone, the acoustic velocity of the sandstone section and mudstone section with a thickness of not less than 5 m was counted respectively; the acoustic time difference of the mudstone section was higher than that of the sandstone, and the whole section was irregular and pointed (see Figure 6 );

[0067] 3) According to formula (2), the average acoustic velocity values ​​of different sandstone and mudstone intervals are calculated: Where j = 1, 2, ... m represents different intervals, and m is the number of intervals (see Table 1)

[0068] 4) Read the time ΔT from the top of the slope to the foot of the slope in sections j , substitute into formula (3) to calculate the apparent thickness Hs of the foreset body (see Table 1);

[0069] Table 1 Apparent thickness of the forebody F1 obtained by the segmented cumulative summation method

[0070]

[0071] Step 4: Combine drilling and logging data to make a decompaction correction for the apparent thickness obtained in step 3. Specific steps:

[0072] 1) The measured porosity and corresponding depth values ​​of sandstone and mudstone in Chang 3 to Chang 7 sections of 6 wells in Qingcheng work area were counted, and the porosity Ф-burial depth Z relationship curves of sandstone and mudstone were fitted according to the exponential function distribution (see Figure 5 ).

[0073] The fitted relationship equation between the porosity Φ and burial depth Z of sandstone and mudstone is as follows:

[0074] Sandstone: Ф=46.5121e -0.00079Z (8)

[0075] Mudstone: Ф=62.9984e -0.00099Z (9)

[0076] 2) Based on the principle that the volume of the rock skeleton particles remains unchanged before and after compaction, the original top depth (Z'1) and bottom depth (Z'2) of the sandstone and mudstone of the Chang 3 to Chang 7 sections of the single well X1 are iteratively calculated using the rock skeleton thickness integral mathematical model (Formula 4) when the current top depth (Z1) and bottom depth (Z2) of the formation are known; then, based on the ratio of the original formation thickness to the current formation thickness, the compaction rate (t) of Chang 3 to Chang 7 of the single well X1 is calculated to be 1.19 (see Figure 6 ).

[0077] 3) Compare the compaction rate 1.19 of the single well X1 corresponding to the foreset F1 with its apparent thickness (H s )186.47m to obtain the true thickness after de-compaction correction (H z ) is 221.89m, which is the ancient water depth.

[0078] H z =t·H s =1.19×186.47=221.89m

[0079] Where: t is the compaction rate; H z is the true thickness of the foreset body (paleo-water depth); H s is the apparent thickness of the foreset volume;

[0080] Step 5: Repeat steps 3 to 4 to quantitatively calculate the thickness of the progradation body at different stages on each seismic section along the provenance direction to obtain the paleo-water depth values ​​(H z )(see Table 2), and the vertical connection gives the line graph of paleo-water depth variation (see Fig. 9 ).

Claims

1. A method for quantitatively restoring paleo-water depth based on foreset reflection, characterized in that: The following steps are involved: Step 1: Collect seismic data, drilling data and logging data in the study area; Step 2: Finely interpret the seismic data in the study area, establish a high-precision isochronous stratigraphic framework for the target layer, and flatten the top surface on multiple seismic sections along the source to identify the delta progradational reflection structure; Step 3: Determine the top and foot of the delta progradation reflection structure, and use the segmented cumulative summation method to calculate the distance between the top and the foot of the slope to determine the apparent thickness of the progradation body; Step 4: Combine drilling and logging data to perform decompaction correction on the apparent thickness obtained in step 3 to obtain the true thickness of the foreset body, i.e., the paleo-water depth; Step 5, repeating steps 3 to 4, quantitatively calculating the thickness of the foreset body of different periods on the source seismic section, obtaining the paleo-water depth values ​​of different periods, and connecting them vertically to obtain a line graph of paleo-water depth changes; Step 6: construct a deep interpolation neural network model, set model parameters, repeatedly train to obtain the best model, and use the best model to interpolate the paleo-water depth calculated in step 5; Step 7: Draw the interpolated curve and use the CubicSpline function to smooth the curve to obtain a smooth paleo-water depth change curve, that is, a lake level change curve.

2. A method for quantitatively restoring paleo-water depth based on foreset reflection as claimed in claim 1, characterized in that: In step 2, the process of identifying the foreset reflection structure includes the following steps: S201: According to the sequence stratigraphic theory, a set of coal seams with strong peaks and strong troughs in seismic response in the target layer section is identified as the top marker layer, and a set of widely and stably developed black shale is identified as the bottom marker layer; S202: Combine well and seismic data to calibrate synthetic seismic records, identify, track and interpret target layer segments, and establish a high-precision isochronal stratigraphic framework; S203: On multiple seismic sections along the source, flatten the top surface and finely identify and depict the multi-stage foreset bodies of the target layer.

3. The method for quantitatively restoring paleo-water depth based on foreset reflection according to claim 1, characterized in that: In step 3, the method for determining the front volume thickness includes the following steps: S301: According to the drilling data in the study area, the acoustic time difference logging curve of the well closest to the slope top and which has been calibrated with synthetic seismic records is selected, abnormal data values ​​are eliminated, the logging curve data of the target layer is counted, and the acoustic wave velocity V corresponding to each depth point is calculated according to formula (1) i , where i corresponds to different depth points, i = 1, 2, ... n, n represents the number of depth points; S302: According to the difference in logging phases between sandstone and mudstone, the acoustic wave velocity of the sandstone section and mudstone section with a thickness of not less than 5m is counted respectively; the acoustic wave time difference of the mudstone section is higher than that of the sandstone, and the whole section is irregular and sharp; S303: Calculate the average acoustic wave velocity values ​​of different sandstone and mudstone intervals according to formula (2) Where j = 1, 2, ... m represents different intervals, and m is the number of intervals; S304: Read the time ΔT from the top of the slope to the foot of the slope in segments on the seismic profile j , substitute into formula (3) to calculate the apparent thickness H of the forebody s ; Where: V i is the sound wave velocity at the i-th depth point, m / us; Δt i is the acoustic time difference value of the i-th depth point, us / m; is the average sound wave velocity of the jth interval, m / us; ΔT j is the time difference of the jth time period read on the seismic profile, ms; n j is the number of depth points in the jth interval.

4. The method for quantitatively restoring paleo-water depth based on foreset reflection according to claim 1, characterized in that: In step 4, the decompaction correction includes the following steps: S401: Statistically analyze the measured porosity and corresponding depth values ​​of different lithologies in the target layers of multiple wells in the study area, and fit the porosity Ф-burial depth Z relationship curve of different lithologies according to the exponential function distribution: φ=φ0e -C*Z (4) Where: Ф is the porosity at depth Z; Ф0 is the surface porosity; C is the compaction coefficient; Z is the burial depth, m; S402: Based on the principle that the volume of the rock skeleton particles remains unchanged before and after compaction, and given the current top and bottom boundary depths of the formation, the rock skeleton thickness integral mathematical model is used to iteratively calculate the top interface burial depth Z'1 and the bottom interface burial depth Z'2 of the corresponding original formation, and then the original formation thickness is calculated; then the compaction rate t of the target layer section of a single well is calculated based on the ratio of the original formation thickness to the current formation thickness: Where: Z1, Z2 are the top and bottom depths of the current strata, m; Z'1, Z'2 are the top and bottom depths of the original strata, m; t is the compaction rate; S403: Compare the compaction rate of the target layer section of the single well corresponding to the different foreset bodies with the apparent thickness H of the foreset body calculated in step 3 s Multiply them to get the true thickness H after de-compaction correction z , that is, the paleo-water depth: H z =t·H s (7) Where: H z is the true thickness of the foreset body, m; H s is the apparent thickness of the foreset volume, m.

5. The method for quantitatively restoring paleo-water depth based on foreset reflection according to claim 1, characterized in that: In step six, the interpolation calculation includes the following steps: S601: Extract the depth column and calculate the actual thickness H of the pre-integrated volume at each stage z ; S602: Perform normalization preprocessing on the extracted data: use MinMaxScaler to scale the data to the interval [0,1]; S603: Construct a deep interpolation neural network model: the model includes an input layer, three hidden layers and an output layer, each hidden layer contains 32 neurons, and the activation function is ReLU; S604: Conduct model training: Use the mean square error loss function to train the model, and use the Adam optimizer for optimization, where the learning rate is set to 0.01 and the number of training rounds is 2000; stop training when the MSE reaches 0.01, and obtain the best deep interpolation model; S605: Use the trained best model to perform interpolation calculation and draw the interpolation result.

Images