Deep learning-based hillside and beach reservoir prediction method and system under well-free condition

Abstract

The invention discloses a deep learning-based hillock and beach reservoir prediction method and system under a well-free condition. The method comprises the following steps: determining seismic data and logging data of an adjacent work area, and matching the seismic data of a research work area and the seismic data of the adjacent work area; extracting seismic attributes of an adjacent work area, and constructing a conversion relation between the seismic attributes and the reservoir porosity of the hillside and beach body through a deep learning algorithm; and seismic attributes of the research work area are extracted, and the mound and beach body reservoir of the research work area is predicted according to the conversion relation. According to the method, the reservoir of the hillock and beach body is predicted according to various attribute characteristics of the seismic data, so that the multiplicity of solutions of a reservoir prediction result is effectively reduced, and the reservoir prediction precision is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of seismic data interpretation, and more specifically, to a method and system for predicting a mound-beach reservoir based on deep learning without a well. Background Art

[0002] The prediction of beach body reservoirs is one of the important goals of deep and ultra-deep marine oil and gas exploration. Due to the relatively low degree of deep and ultra-deep oil and gas exploration, drilling data is often lacking in the target layer. The existing technical solutions generally adopt the phase-controlled reservoir prediction method, which collects the drilling data of the adjacent work area, establishes the conversion relationship between the wave impedance and the porosity of the beach body reservoir, and finds the approximate location of the beach body development in the seismic data according to the geological sedimentary pattern of the beach body. This information is used as a constraint to artificially distinguish the location where the beach body may develop from other areas in the study area, thereby realizing phase-controlled modeling. Finally, the traditional wave impedance inversion technology is used to obtain the wave impedance data body from the seismic data, and the wave impedance data body is converted into porosity data according to the conversion relationship between wave impedance and porosity, thereby realizing the reservoir prediction of the beach body. Due to the strong vertical and horizontal heterogeneity of the beach body reservoir, it is difficult to distinguish between reservoirs and non-reservoirs using only a single seismic wave impedance attribute, resulting in the reservoir prediction results of the existing technical solutions being difficult to accurately describe the spatial distribution characteristics of the beach body reservoir. Reservoir prediction technology based on deep learning can establish the conversion relationship between multiple seismic attributes and the porosity of mound-shoal reservoirs, which is particularly suitable for reservoir prediction under complex geological conditions. However, conventional deep learning technology requires a large amount of known well data for calibration, and it is difficult to implement this technology without wells.

[0003] Therefore, it is necessary to develop a deep learning-based method and system for predicting mound-shoal reservoirs without wells.

[0004] The information disclosed in the background technology section of the present invention is only intended to deepen the understanding of the general background technology of the present invention, and should not be regarded as acknowledging or suggesting in any form that the information constitutes the prior art already known to those skilled in the art. Summary of the invention

[0005] The present invention proposes a method and system for predicting mound-shoal reservoirs based on deep learning under well-free conditions, which can carry out mound-shoal reservoir prediction through multiple attribute characteristics of seismic data, thereby effectively reducing the ambiguity of reservoir prediction results and improving reservoir prediction accuracy.

[0006] In a first aspect, the embodiments of the present disclosure provide a method for predicting a mound-beach reservoir based on deep learning without a well, comprising:

[0007] Determine the seismic data and logging data of the adjacent work area, and match the seismic data of the research work area with the adjacent work area;

[0008] Extracting the seismic attributes of the adjacent work area, and constructing a conversion relationship between the seismic attributes and the porosity of the mound-shoal reservoir through a deep learning algorithm;

[0009] The seismic attributes of the study area are extracted, and the mound-shoal reservoir of the study area is predicted based on the conversion relationship.

[0010] Preferably, matching the seismic data of the research work area with the adjacent work area includes:

[0011] Extracting seismic wavelets within the scope of the target layer to be studied in the adjacent work area;

[0012] Carry out relative wave impedance inversion of the seismic data of the adjacent work area; utilize the extracted seismic wavelet and adopt the conventional wave impedance inversion algorithm to transform the seismic data of the adjacent work area into a relative wave impedance data volume, thereby eliminating the wavelet characteristics in the seismic data.

[0013] Converting the relative wave impedance data volume of the adjacent work area into a reflection coefficient data volume;

[0014] Extract seismic wavelets within the target layer range of the study area;

[0015] Conducting relative wave impedance inversion of seismic data of the study area to obtain a relative wave impedance data volume of the study area;

[0016] Converting the relative wave impedance data volume of the research area into a reflection coefficient data volume;

[0017] The reflection coefficient data volumes of the adjacent work area and the research work area are respectively convolved by wavelet, and the reflection coefficient data volumes are converted into seismic amplitude data.

[0018] Preferably, extracting the seismic attributes of the adjacent work area includes:

[0019] Through the seismic data of the adjacent work area after matching processing, the seismic attribute data body of the adjacent work area is extracted, such as seismic trace integral attributes, seismic vector trace attributes, seismic first-order and second-order derivative attributes, seismic three-moment attributes, seismic main frequency attributes and seismic frequency division attributes.

[0020] Preferably, extracting the seismic attributes of the study area includes:

[0021] The seismic attribute data volume of the seismic data of the study area is extracted through the seismic data of the study area after matching processing.

[0022] Preferably, the seismic attribute data body includes seismic trace integral attributes, seismic vector trace attributes, seismic first-order and second-order derivative attributes, seismic three-moment attributes, seismic main frequency attributes and seismic fractional frequency attributes.

[0023] In a second aspect, the embodiment of the present disclosure further provides a system for predicting a mound-beach reservoir based on deep learning without a well, comprising:

[0024] a memory storing computer executable instructions;

[0025] A processor, wherein the processor runs the computer executable instructions in the memory to perform the following steps:

[0026] Determine the seismic data and logging data of the adjacent work area, and match the seismic data of the research work area with the adjacent work area;

[0027] Extracting the seismic attributes of the adjacent work area, and constructing a conversion relationship between the seismic attributes and the porosity of the mound-shoal reservoir through a deep learning algorithm;

[0028] The seismic attributes of the study area are extracted, and the mound-shoal reservoir of the study area is predicted based on the conversion relationship.

[0029] Preferably, matching the seismic data of the research work area with the adjacent work area includes:

[0030] Extracting seismic wavelets within the scope of the target layer to be studied in the adjacent work area;

[0031] Carry out relative wave impedance inversion of the seismic data of the adjacent work area; utilize the extracted seismic wavelet and adopt the conventional wave impedance inversion algorithm to transform the seismic data of the adjacent work area into a relative wave impedance data volume, thereby eliminating the wavelet characteristics in the seismic data.

[0032] Converting the relative wave impedance data volume of the adjacent work area into a reflection coefficient data volume;

[0033] Extract seismic wavelets within the target layer range of the study area;

[0034] Conducting relative wave impedance inversion of seismic data of the study area to obtain a relative wave impedance data volume of the study area;

[0035] Converting the relative wave impedance data volume of the research area into a reflection coefficient data volume;

[0036] The reflection coefficient data volumes of the adjacent work area and the research work area are respectively convolved by wavelet, and the reflection coefficient data volumes are converted into seismic amplitude data.

[0037] Preferably, extracting the seismic attributes of the adjacent work area includes:

[0038] Through the seismic data of the adjacent work area after matching processing, the seismic attribute data body of the adjacent work area is extracted, such as seismic trace integral attributes, seismic vector trace attributes, seismic first-order and second-order derivative attributes, seismic three-moment attributes, seismic main frequency attributes and seismic frequency division attributes.

[0039] Preferably, extracting the seismic attributes of the study area includes:

[0040] The seismic attribute data volume of the seismic data of the study area is extracted through the seismic data of the study area after matching processing.

[0041] Preferably, the seismic attribute data body includes seismic trace integral attributes, seismic vector trace attributes, seismic first-order and second-order derivative attributes, seismic three-moment attributes, seismic main frequency attributes and seismic fractional frequency attributes.

[0042] Its beneficial effects are:

[0043] 1. The present invention utilizes multiple attribute characteristics of seismic data to carry out reservoir prediction of mound-shoal bodies, thereby effectively reducing the multi-solution of reservoir prediction results and improving the accuracy of reservoir prediction;

[0044] 2. The present invention uses deep learning technology to establish a non-formalized conversion relationship between seismic attributes and the porosity of the mound-shoal reservoir, which can better reflect the heterogeneity characteristics of the mound-shoal reservoir;

[0045] 3. The present invention does not require phase-controlled modeling, thus avoiding the intervention of subjective cognition, thereby achieving pure data-driven prediction of mound-shoal reservoirs.

[0046] The methods and systems of the present invention have other features and advantages that will be apparent from, or will be described in detail in, the accompanying drawings and subsequent detailed descriptions incorporated herein, which together serve to explain the specific principles of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The above and other objects, features and advantages of the present invention will become more apparent through a more detailed description of exemplary embodiments of the present invention in conjunction with the accompanying drawings, wherein like reference numerals generally represent like components throughout the exemplary embodiments of the present invention.

[0048] Figure 1 A flowchart showing the steps of a method for predicting mound-shoal reservoirs based on deep learning under well-free conditions according to the present invention is shown.

[0049] Figure 2 A flowchart showing the steps of a method for predicting mound-beach reservoirs based on deep learning under well-free conditions according to an embodiment of the present invention is shown.

[0050] Figure 3 A schematic diagram showing seismic data of an adjacent work area and a study work area before matching processing according to an embodiment of the present invention is shown.

[0051] Figure 4 A schematic diagram showing seismic wavelet spectra of an adjacent work area and a study work area before matching processing according to an embodiment of the present invention is shown.

[0052] Figure 5 A schematic diagram showing seismic data of an adjacent work area and a study work area after matching processing according to an embodiment of the present invention is shown.

[0053] Figure 6 A schematic diagram showing seismic wavelet spectra of an adjacent work area and a research work area after matching processing according to an embodiment of the present invention is shown.

[0054] Figure 7 A schematic diagram showing a comparison between the porosity calculated by seismic attributes using a deep learning algorithm and the well logging porosity according to an embodiment of the present invention is shown.

[0055] Figure 8 A schematic diagram showing a result of predicting a mound-shoal reservoir according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0056] The preferred embodiments of the present invention will be described in more detail below. Although the preferred embodiments of the present invention are described below, it should be understood that the present invention can be implemented in various forms and should not be limited to the embodiments set forth herein.

[0057] Figure 1 A flowchart showing the steps of a method for predicting mound-shoal reservoirs based on deep learning under well-free conditions according to the present invention is shown.

[0058] like Figure 1 As shown, the deep learning-based mound-beach reservoir prediction method under the condition of no well includes:

[0059] Step 101, determining the seismic data and logging data of the adjacent work area, and matching the seismic data of the research work area with the adjacent work area;

[0060] Step 102, extracting seismic attributes of the adjacent work area, and constructing a conversion relationship between the seismic attributes and the porosity of the mound-shoal reservoir through a deep learning algorithm;

[0061] Step 103, extracting the seismic attributes of the study area, and predicting the mound-shoal reservoir in the study area based on the conversion relationship.

[0062] In one example, matching seismic data of the study area with those of adjacent areas includes:

[0063] Extract seismic wavelets within the target layer of the adjacent work area;

[0064] Carry out relative wave impedance inversion of seismic data in adjacent work areas; use the extracted seismic wavelets and adopt conventional wave impedance inversion algorithms to convert seismic data in adjacent work areas into relative wave impedance data volumes, thereby eliminating the wavelet characteristics in the seismic data.

[0065] Convert the relative wave impedance data volume of the adjacent work area into a reflection coefficient data volume;

[0066] Extract seismic wavelets within the target layer range of the study area;

[0067] Carry out relative wave impedance inversion of seismic data in the study area to obtain the relative wave impedance data volume of the study area;

[0068] The relative wave impedance data volume of the research area is converted into a reflection coefficient data volume;

[0069] The reflection coefficient data of the adjacent work area and the research work area are convolved by wavelet, and the reflection coefficient data are converted into seismic amplitude data.

[0070] In one example, extracting seismic attributes of a nearby work area includes:

[0071] Through the seismic data of the adjacent work area after matching processing, the seismic attribute data body of the adjacent work area is extracted, such as seismic trace integral attributes, seismic vector trace attributes, seismic first-order and second-order derivative attributes, seismic three-moment attributes, seismic main frequency attributes and seismic frequency fraction attributes.

[0072] In one example, extracting seismic attributes for a study area involves:

[0073] The seismic attribute data volume of the seismic data in the study area is extracted through the seismic data in the study area after matching processing.

[0074] In one example, the seismic attribute data volume includes seismic trace integral attributes, seismic vector trace attributes, seismic first-order and second-order derivative attributes, seismic three-moment attributes, seismic main frequency attributes, and seismic fractional frequency attributes.

[0075] Specifically, collect seismic data and well logging data from adjacent work areas. Since there is no available well logging data in the study area, it is necessary to collect seismic data and well logging data from adjacent work areas with similar sedimentary environment characteristics to provide a data basis for subsequent steps.

[0076] Since there are generally large differences in the acquisition and processing of seismic data in adjacent work areas and in the study area, the energy and spectrum characteristics of seismic data in different work areas are quite different, so seismic data matching processing is required. In order to apply the seismic attributes and the porosity conversion relationship of the mound-shoal reservoir constructed in the adjacent work area to the study area, it is necessary to carry out matching processing between the seismic data in the adjacent work area and the study area. The matching processing steps include:

[0077] 1) Extract the seismic wavelet of the target layer in the adjacent work area. If multiple sets of seismic data of the adjacent work area are collected, extract the seismic wavelet for each set of seismic data respectively;

[0078] 2) Carry out relative wave impedance inversion of seismic data in the adjacent work area, use the extracted seismic wavelet and adopt the conventional wave impedance inversion algorithm to convert the seismic data in the adjacent work area into a relative wave impedance data volume, thereby eliminating the wavelet characteristics in the seismic data;

[0079] 3) Using the mathematical formula between wave impedance and reflection coefficient, the relative wave impedance data volume of the adjacent work area is converted into the reflection coefficient data volume;

[0080] 4) Extract seismic wavelets within the target layer range of the research area;

[0081] 5) Carry out relative wave impedance inversion of seismic data in the study area to obtain the relative wave impedance data volume of the study area, thereby eliminating the wavelet characteristics of the seismic data in the study area;

[0082] 6) Convert the relative wave impedance data volume of the research area into a reflection coefficient data volume;

[0083] 7) Use a unified sub-wave to perform convolution operations on the reflection coefficient data volumes of the adjacent work area and the study work area, respectively, and convert the reflection coefficient data volumes of the adjacent work area and the study work area into seismic amplitude data.

[0084] After the above processing flow, the energy and spectrum characteristics of the seismic data in the adjacent work area and the research work area can be unified.

[0085] Extract seismic attributes of seismic data in adjacent work areas. Using the seismic data in adjacent work areas after matching processing, extract various seismic attribute data bodies of adjacent work areas, such as seismic trace integral attributes, seismic vector trace attributes, seismic first-order and second-order derivative attributes, seismic three-moment attributes, seismic main frequency attributes and seismic frequency division attributes.

[0086] Construct the conversion relationship between seismic attributes and mound-shoal reservoir porosity. Using deep learning algorithms, construct the conversion relationship between seismic attributes of adjacent work areas and logging mound-shoal reservoir porosity data.

[0087] Using the seismic data of the study area after matching processing, various seismic attribute data volumes of the seismic data of the study area are extracted.

[0088] By utilizing the conversion relationship between the seismic attributes of the adjacent work area and the porosity data of the logging mound-shoal reservoir and the seismic attributes of the seismic data in the research work area, the seismic attributes of the research functional area are converted into porosity data, thereby realizing the prediction of the mound-shoal reservoir.

[0089] The present invention also provides a deep learning-based mound-shoal reservoir prediction system under well-free conditions, including:

[0090] a memory storing computer executable instructions;

[0091] A processor executes the computer executable instructions in the memory to perform the following steps:

[0092] Determine the seismic data and logging data of the adjacent work area, and match the seismic data of the research work area with the adjacent work area;

[0093] Extract seismic attributes of adjacent work areas and construct the conversion relationship between seismic attributes and reservoir porosity of mound-shoal bodies through deep learning algorithms;

[0094] The seismic attributes of the study area are extracted, and the hilly-shoal reservoirs in the study area are predicted based on the conversion relationship.

[0095] In one example, matching seismic data of the study area with those of adjacent areas includes:

[0096] Extract seismic wavelets within the target layer of the adjacent work area;

[0097] Carry out relative wave impedance inversion of seismic data in adjacent work areas; use the extracted seismic wavelets and adopt conventional wave impedance inversion algorithms to convert seismic data in adjacent work areas into relative wave impedance data volumes, thereby eliminating the wavelet characteristics in the seismic data.

[0098] Convert the relative wave impedance data volume of the adjacent work area into a reflection coefficient data volume;

[0099] Extract seismic wavelets within the target layer range of the study area;

[0100] Carry out relative wave impedance inversion of seismic data in the study area to obtain the relative wave impedance data volume of the study area;

[0101] The relative wave impedance data volume of the research area is converted into a reflection coefficient data volume;

[0102] The reflection coefficient data of the adjacent work area and the research work area are convolved by wavelet, and the reflection coefficient data are converted into seismic amplitude data.

[0103] In one example, extracting seismic attributes of a nearby work area includes:

[0104] Through the seismic data of the adjacent work area after matching processing, the seismic attribute data body of the adjacent work area is extracted, such as seismic trace integral attributes, seismic vector trace attributes, seismic first-order and second-order derivative attributes, seismic three-moment attributes, seismic main frequency attributes and seismic frequency fraction attributes.

[0105] In one example, extracting seismic attributes for a study area involves:

[0106] The seismic attribute data volume of the seismic data in the study area is extracted through the seismic data in the study area after matching processing.

[0107] In one example, the seismic attribute data volume includes seismic trace integral attributes, seismic vector trace attributes, seismic first-order and second-order derivative attributes, seismic three-moment attributes, seismic main frequency attributes, and seismic fractional frequency attributes.

[0108] Specifically, collect seismic data and well logging data from adjacent work areas. Since there is no available well logging data in the study area, it is necessary to collect seismic data and well logging data from adjacent work areas with similar sedimentary environment characteristics to provide a data basis for subsequent steps.

[0109] Since there are generally large differences in the acquisition and processing of seismic data in adjacent work areas and in the study area, the energy and spectrum characteristics of seismic data in different work areas are quite different, so seismic data matching processing is required. In order to apply the seismic attributes and the porosity conversion relationship of the mound-shoal reservoir constructed in the adjacent work area to the study area, it is necessary to carry out matching processing between the seismic data in the adjacent work area and the study area. The matching processing steps include:

[0110] 1) Extract the seismic wavelet of the target layer in the adjacent work area. If multiple sets of seismic data of the adjacent work area are collected, extract the seismic wavelet for each set of seismic data respectively;

[0111] 2) Carry out relative wave impedance inversion of seismic data in the adjacent work area, use the extracted seismic wavelet and adopt the conventional wave impedance inversion algorithm to convert the seismic data in the adjacent work area into a relative wave impedance data volume, thereby eliminating the wavelet characteristics in the seismic data;

[0112] 3) Using the mathematical formula between wave impedance and reflection coefficient, the relative wave impedance data volume of the adjacent work area is converted into the reflection coefficient data volume;

[0113] 4) Extract seismic wavelets within the target layer range of the research area;

[0114] 5) Carry out relative wave impedance inversion of seismic data in the study area to obtain the relative wave impedance data volume of the study area, thereby eliminating the wavelet characteristics of the seismic data in the study area;

[0115] 6) Convert the relative wave impedance data volume of the research area into a reflection coefficient data volume;

[0116] 7) Use a unified sub-wave to perform convolution operations on the reflection coefficient data volumes of the adjacent work area and the study work area, respectively, and convert the reflection coefficient data volumes of the adjacent work area and the study work area into seismic amplitude data.

[0117] After the above processing flow, the energy and spectrum characteristics of the seismic data in the adjacent work area and the research work area can be unified.

[0118] Extract seismic attributes of seismic data in adjacent work areas. Using the seismic data in adjacent work areas after matching processing, extract various seismic attribute data bodies of adjacent work areas, such as seismic trace integral attributes, seismic vector trace attributes, seismic first-order and second-order derivative attributes, seismic three-moment attributes, seismic main frequency attributes and seismic frequency division attributes.

[0119] Construct the conversion relationship between seismic attributes and mound-shoal reservoir porosity. Using deep learning algorithms, construct the conversion relationship between seismic attributes of adjacent work areas and logging mound-shoal reservoir porosity data.

[0120] Using the seismic data of the study area after matching processing, various seismic attribute data volumes of the seismic data of the study area are extracted.

[0121] By utilizing the conversion relationship between the seismic attributes of the adjacent work area and the porosity data of the logging mound-shoal reservoir and the seismic attributes of the seismic data in the research work area, the seismic attributes of the research functional area are converted into porosity data, thereby realizing the prediction of the mound-shoal reservoir.

[0122] To facilitate understanding of the solutions and effects of the embodiments of the present invention, two specific application examples are given below. Those skilled in the art should understand that the examples are only for facilitating understanding of the present invention, and any specific details thereof are not intended to limit the present invention in any way.

[0123] Example 1

[0124] Figure 2 A flowchart showing the steps of a method for predicting mound-beach reservoirs based on deep learning under well-free conditions according to an embodiment of the present invention is shown.

[0125] Figure 3 A schematic diagram showing seismic data of an adjacent work area and a study work area before matching processing according to an embodiment of the present invention is shown.

[0126] like Figure 2As shown in the figure, collect seismic data and well logging data from adjacent work areas. Since there is no well logging data available in the study area, it is necessary to collect seismic data and well logging data from adjacent work areas with similar sedimentary environment characteristics to provide a data basis for subsequent steps. Figure 3 shown.

[0127] Figure 4 A schematic diagram showing seismic wavelet spectra of an adjacent work area and a study work area before matching processing according to an embodiment of the present invention is shown.

[0128] Since there are generally large differences in the acquisition and processing of seismic data in the adjacent work area and the study area, the energy and spectrum characteristics of seismic data in different work areas are quite different, so seismic data matching processing is required. In order to apply the seismic attributes and the porosity conversion relationship of the mound-shoal reservoir constructed in the adjacent work area to the study area, it is necessary to carry out matching processing between the seismic data of the adjacent work area and the study area. The seismic wavelet spectra of the adjacent work area and the study area before matching processing are as follows Figure 4 As shown, the matching process steps include:

[0129] 1) Extract the seismic wavelet of the target layer in the adjacent work area. If multiple sets of seismic data of the adjacent work area are collected, extract the seismic wavelet for each set of seismic data respectively;

[0130] 2) Carry out relative wave impedance inversion of seismic data in the adjacent work area, use the extracted seismic wavelet and adopt the conventional wave impedance inversion algorithm to convert the seismic data in the adjacent work area into a relative wave impedance data volume, thereby eliminating the wavelet characteristics in the seismic data;

[0131] 3) Using the mathematical formula between wave impedance and reflection coefficient, the relative wave impedance data volume of the adjacent work area is converted into the reflection coefficient data volume;

[0132] 4) Extract seismic wavelets within the target layer range of the research area;

[0133] 5) Carry out relative wave impedance inversion of seismic data in the study area to obtain the relative wave impedance data volume of the study area, thereby eliminating the wavelet characteristics of the seismic data in the study area;

[0134] 6) Convert the relative wave impedance data volume of the research area into a reflection coefficient data volume;

[0135] 7) Use a unified sub-wave to perform convolution operations on the reflection coefficient data volumes of the adjacent work area and the study work area, respectively, and convert the reflection coefficient data volumes of the adjacent work area and the study work area into seismic amplitude data.

[0136] Figure 5A schematic diagram showing seismic data of an adjacent work area and a study work area after matching processing according to an embodiment of the present invention is shown.

[0137] Figure 6 A schematic diagram showing seismic wavelet spectra of an adjacent work area and a research work area after matching processing according to an embodiment of the present invention is shown.

[0138] After matching, the seismic data of the adjacent work area and the study area and the seismic wavelet spectrum are as follows: Figure 5-Figure 6 As shown in the figure, after the above processing flow, the energy and spectrum characteristics of the seismic data in the adjacent work area and the research work area can be unified.

[0139] By using the seismic data of the adjacent work area after matching processing, various seismic attribute data bodies of the adjacent work area are extracted, such as seismic trace integral attributes, seismic vector trace attributes, seismic first-order and second-order derivative attributes, seismic three-moment attributes, seismic main frequency attributes and seismic frequency fraction attributes.

[0140] Using deep learning algorithms, a conversion relationship between the seismic attributes of adjacent work areas and the porosity data of well logging mound and shoal reservoirs is constructed.

[0141] Figure 7 A schematic diagram comparing the porosity calculated by seismic attributes using a deep learning algorithm and the well logging porosity according to an embodiment of the present invention is shown, wherein the dark line represents the well logging porosity curve and the light line represents the porosity curve calculated by seismic attributes using a deep learning algorithm.

[0142] Using the seismic data of the study area after matching processing, various seismic attribute data volumes of the seismic data of the study area are extracted. The porosity and logging porosity calculated by the deep learning algorithm through seismic attributes are as follows: Figure 7 As shown, the comparison results show that the dark line and the light line have a high consistency, which reflects the accuracy of the conversion relationship between the seismic attributes of the adjacent work area and the reservoir porosity of the logging mound beach constructed by the deep learning algorithm.

[0143] Figure 8 A schematic diagram showing a result of predicting a mound-shoal reservoir according to an embodiment of the present invention is shown.

[0144] By using the conversion relationship between the seismic attributes of the adjacent work area and the porosity data of the logging mound-shoal reservoir and the seismic attributes of the seismic data of the research work area, the seismic attributes of the research functional area are converted into porosity data, thereby realizing the prediction of the mound-shoal reservoir, such as Figure 8 shown.

[0145] Example 2

[0146] The deep learning-based mound-shoal reservoir prediction system under well-free conditions includes:

[0147] a memory storing computer executable instructions;

[0148] A processor executes the computer executable instructions in the memory to perform the following steps:

[0149] Determine the seismic data and logging data of the adjacent work area, and match the seismic data of the research work area with the adjacent work area;

[0150] Extract seismic attributes of adjacent work areas and construct the conversion relationship between seismic attributes and reservoir porosity of mound-shoal bodies through deep learning algorithms;

[0151] The seismic attributes of the study area are extracted, and the hilly-shoal reservoirs in the study area are predicted based on the conversion relationship.

[0152] In one example, matching seismic data of the study area with those of adjacent areas includes:

[0153] Extract seismic wavelets within the target layer of the adjacent work area;

[0154] Carry out relative wave impedance inversion of seismic data in adjacent work areas; use the extracted seismic wavelets and adopt conventional wave impedance inversion algorithms to convert seismic data in adjacent work areas into relative wave impedance data volumes, thereby eliminating the wavelet characteristics in the seismic data.

[0155] Convert the relative wave impedance data volume of the adjacent work area into a reflection coefficient data volume;

[0156] Extract seismic wavelets within the target layer range of the study area;

[0157] Carry out relative wave impedance inversion of seismic data in the study area to obtain the relative wave impedance data volume of the study area;

[0158] The relative wave impedance data volume of the research area is converted into a reflection coefficient data volume;

[0159] The reflection coefficient data of the adjacent work area and the research work area are convolved by wavelet, and the reflection coefficient data are converted into seismic amplitude data.

[0160] In one example, extracting seismic attributes of a nearby work area includes:

[0161] Through the seismic data of the adjacent work area after matching processing, the seismic attribute data body of the adjacent work area is extracted, such as seismic trace integral attributes, seismic vector trace attributes, seismic first-order and second-order derivative attributes, seismic three-moment attributes, seismic main frequency attributes and seismic frequency fraction attributes.

[0162] In one example, extracting seismic attributes for a study area involves:

[0163] The seismic attribute data volume of the seismic data in the study area is extracted through the seismic data in the study area after matching processing.

[0164] In one example, the seismic attribute data volume includes seismic trace integral attributes, seismic vector trace attributes, seismic first-order and second-order derivative attributes, seismic three-moment attributes, seismic main frequency attributes, and seismic fractional frequency attributes.

[0165] Those skilled in the art should understand that the purpose of the above description of the embodiments of the present invention is only to exemplarily illustrate the beneficial effects of the embodiments of the present invention, and is not intended to limit the embodiments of the present invention to any given examples.

[0166] The embodiments of the present invention have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments.

Claims

1. A method for predicting mound-beach reservoirs based on deep learning without wells, characterized in that: include: Determine the seismic data and logging data of the adjacent work area, and match the seismic data of the research work area with the adjacent work area; Extracting the seismic attributes of the adjacent work area, and constructing a conversion relationship between the seismic attributes and the porosity of the mound-shoal reservoir through a deep learning algorithm; The seismic attributes of the study area are extracted, and the mound-shoal reservoir of the study area is predicted based on the conversion relationship.

2. The method for predicting mound-beach reservoirs based on deep learning without wells according to claim 1, wherein: The seismic data matching between the study area and the adjacent area includes: Extracting seismic wavelets within the scope of the target layer to be studied in the adjacent work area; Carry out relative wave impedance inversion of the seismic data of the adjacent work area; utilize the extracted seismic wavelet and adopt the conventional wave impedance inversion algorithm to transform the seismic data of the adjacent work area into a relative wave impedance data volume, thereby eliminating the wavelet characteristics in the seismic data. Converting the relative wave impedance data volume of the adjacent work area into a reflection coefficient data volume; Extract seismic wavelets within the target layer range of the study area; Conducting relative wave impedance inversion of seismic data of the study area to obtain a relative wave impedance data volume of the study area; Converting the relative wave impedance data volume of the research area into a reflection coefficient data volume; The reflection coefficient data volumes of the adjacent work area and the research work area are respectively convolved by wavelet, and the reflection coefficient data volumes are converted into seismic amplitude data.

3. The method for predicting mound-shoal reservoir based on deep learning without wells according to claim 1, wherein: Extracting the seismic attributes of the adjacent work area includes: Through the seismic data of the adjacent work area after matching processing, the seismic attribute data body of the adjacent work area is extracted, such as seismic trace integral attributes, seismic vector trace attributes, seismic first-order and second-order derivative attributes, seismic three-moment attributes, seismic main frequency attributes and seismic frequency division attributes.

4. The method for predicting mound-shoal reservoir based on deep learning without wells according to claim 1, wherein: Extracting seismic attributes of the study area includes: The seismic attribute data volume of the seismic data of the study area is extracted through the seismic data of the study area after matching processing.

5. The method for predicting mound-beach reservoir based on deep learning without wells according to claim 3 or 4, wherein: The seismic attribute data body includes seismic trace integral attributes, seismic vector trace attributes, seismic first-order and second-order derivative attributes, seismic three-moment attributes, seismic main frequency attributes and seismic frequency fraction attributes.

6. A deep learning-based mound-beach reservoir prediction system without wells, characterized in that: The system includes: a memory storing computer executable instructions; A processor, wherein the processor runs the computer executable instructions in the memory to perform the following steps: Determine the seismic data and logging data of the adjacent work area, and match the seismic data of the research work area with the adjacent work area; Extracting the seismic attributes of the adjacent work area, and constructing a conversion relationship between the seismic attributes and the porosity of the mound-shoal reservoir through a deep learning algorithm; The seismic attributes of the study area are extracted, and the mound-shoal reservoir of the study area is predicted based on the conversion relationship.

7. The deep learning-based mound-shoal reservoir prediction system without wells according to claim 6, wherein: The seismic data matching between the study area and the adjacent area includes: Extracting seismic wavelets within the scope of the target layer to be studied in the adjacent work area; Carry out relative wave impedance inversion of the seismic data of the adjacent work area; utilize the extracted seismic wavelet and adopt the conventional wave impedance inversion algorithm to transform the seismic data of the adjacent work area into a relative wave impedance data volume, thereby eliminating the wavelet characteristics in the seismic data. Converting the relative wave impedance data volume of the adjacent work area into a reflection coefficient data volume; Extract seismic wavelets within the target layer range of the study area; Conducting relative wave impedance inversion of seismic data of the study area to obtain a relative wave impedance data volume of the study area; Converting the relative wave impedance data volume of the research area into a reflection coefficient data volume; The reflection coefficient data volumes of the adjacent work area and the research work area are respectively convolved by wavelet, and the reflection coefficient data volumes are converted into seismic amplitude data.

8. The deep learning-based mound-shoal reservoir prediction system without wells according to claim 6, wherein: Extracting the seismic attributes of the adjacent work area includes: Through the seismic data of the adjacent work area after matching processing, the seismic attribute data body of the adjacent work area is extracted, such as seismic trace integral attributes, seismic vector trace attributes, seismic first-order and second-order derivative attributes, seismic three-moment attributes, seismic main frequency attributes and seismic frequency division attributes.

9. The deep learning-based mound-shoal reservoir prediction system without wells according to claim 6, wherein: Extracting seismic attributes of the study area includes: The seismic attribute data volume of the seismic data of the study area is extracted through the seismic data of the study area after matching processing.

10. The deep learning-based mound-shoal reservoir prediction system without wells according to claim 8 or 9, wherein: The seismic attribute data body includes seismic trace integral attributes, seismic vector trace attributes, seismic first-order and second-order derivative attributes, seismic three-moment attributes, seismic main frequency attributes and seismic frequency fraction attributes.

Images