A method for predicting reservoir gas content based on depth-domain seismic wave dispersion inversion

By conducting seismic wave dispersion analysis and establishment of deep network models in the depth domain, the problem of prediction of gas content in deep reservoirs is solved, and direct prediction of gas content from seismic wave dispersion attributes to reservoirs is achieved, providing accurate reservoir content Gas volume information provides an important reference for oil and gas exploration and development.

CN115047525BActive Publication Date: 2024-12-03CHENGDU UNIVERSITY OF TECHNOLOGY
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
CN202210615118.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-06-01
Publication Date
2024-12-03
Estimated Expiration
2042-06-01

AI Technical Summary

Technical Problem

The existing reservoir gas content prediction method cannot be effectively carried out under deep conditions and needs to be calculated in the spatial domain, but the time domain calculation results in erroneous results.

Method used

Using a method based on seismic wave dispersion inversion in the depth domain, fine calibration is performed through well data and seismic data, time-domain seismic records are converted into depth domain, depth domain dispersion attributes are extracted, and a nonlinear mapping model is established using the deep network to realize Direct prediction of seismic wave dispersion properties to reservoir gas content.

Benefits of technology

The problem of gas content prediction in deep reservoirs is solved, accurate gas content information is provided, and an important reference basis for oil and gas exploration and development is provided, and the accuracy and accuracy of prediction is improved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115047525B_ABST
    Figure CN115047525B_ABST
Patent Text Reader

Abstract

The present invention provides a method for predicting reservoir gas content based on deep-domain seismic wave dispersion inversion, which belongs to the field of reservoir prediction for oil and gas exploration. The method includes: acquiring well data and seismic data; fine well-seismic calibration; converting time-domain seismic records to the depth domain; performing seismic wave dispersion analysis in the depth domain, and extracting depth-domain dispersion attributes that can characterize the gas content of the reservoir; based on this, using the reservoir gas content information of the well point as a label and the obtained seismic wave dispersion attributes as input, a deep network mapping model for predicting the gas content of the reservoir in the depth domain is established; finally, the seismic wave dispersion attributes of the target layer in the study area are input to realize direct inversion prediction of the reservoir gas content from known to unknown. The present invention is based on deep-domain seismic wave dispersion analysis, and uses a deep network to establish a nonlinear mapping model between dispersion attributes and gas content, which solves the problem of predicting the gas content of deep reservoirs and provides an important reference basis for oil and gas exploration and development.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the field of oil and gas exploration, and in particular to a method for predicting reservoir gas content based on depth domain seismic wave dispersion inversion. Background Art

[0002] Reservoir gas content is a key parameter for determining the amount of resources in natural gas exploration and development, and determines whether the exploration area has economic exploitation value. Therefore, accurate gas content prediction is of great significance to the formulation of reservoir exploitation plans and the prediction of future economic output.

[0003] The estimation of reservoir gas content is rooted in gas content detection, but the existing reservoir gas content detection method cannot be developed into an effective reservoir gas content prediction method under deep conditions. At the same time, gas content prediction requires the "volume" characteristic information of the reservoir. Seismic wave dispersion refers to a phenomenon in which the amplitude or energy of the signal attenuates in the frequency domain when the seismic signal propagates in the underground medium. Recent studies have shown that deep attenuation dispersion is observable and is a "volume" response attribute. As long as the nonlinear mapping relationship between the seismic wave dispersion attribute and the reservoir pore gas content can be accurately established, the seismic wave dispersion can be used to estimate the gas content of the reservoir. However, the current gas content detection work is basically carried out in the time domain, but the accurate reservoir gas content must be calculated in the spatial domain. Because in the same time interval, due to different rock velocities, the actual thickness of the rock layer passed by the seismic wave may vary greatly. At this time, if the dispersion value calculated in the time domain is used to estimate the gas content of the reservoir, an incorrect result will be obtained. Summary of the invention

[0004] The purpose of the present invention is to propose a reservoir gas content prediction method based on deep domain seismic wave dispersion inversion. The method first performs seismic wave dispersion analysis in the depth domain to extract depth domain dispersion attributes that can characterize the reservoir gas content. Then, the reservoir gas content information of the well point is used as a label, and the obtained seismic wave dispersion attributes are used as input. A nonlinear mapping model between the depth domain seismic wave dispersion and the reservoir gas content is established using a deep network. The seismic wave dispersion attributes of the target layer in the study area are input to realize direct inversion prediction of the reservoir gas content from known to unknown, which solves the problem of deep reservoir gas content prediction and provides an important reference basis for oil and gas exploration and development.

[0005] In order to achieve the above-mentioned purpose and the above-mentioned technical effect, the present invention adopts the following technical scheme:

[0006] A method for predicting reservoir gas content based on depth domain seismic wave dispersion inversion, the method comprising the following steps:

[0007] Step 1: Obtain well data and seismic data in the study area;

[0008] Step 2: Based on the well data and seismic data, combined with the actual geological conditions of the study area, complete the fine calibration of well and seismic data based on quality control, and build a bridge between seismic and well logging;

[0009] Step 3: based on the well-seismic fine calibration results, convert the time domain seismic records to the depth domain;

[0010] Step 4: Based on the depth-domain seismic record, perform seismic wave dispersion analysis in the depth domain to extract depth-domain dispersion attributes that can characterize the gas content of the reservoir;

[0011] Step 5: Based on the depth-domain seismic wave dispersion attribute and the reservoir gas content information of the well point, taking the reservoir gas content information as a label and the extracted depth-domain seismic wave dispersion attribute as input, a deep network nonlinear mapping model between the depth-domain seismic wave dispersion attribute and the reservoir gas content is established;

[0012] Step 6: Based on the established deep network mapping model for predicting reservoir gas content in the depth domain, the seismic wave dispersion attributes of the target layer in the study area are input to realize direct inversion prediction of reservoir gas content from known to unknown.

[0013] Furthermore, the well logging data and seismic data are common knowledge in the art.

[0014] Furthermore, the essence of the well-seismic calibration is to adjust the time-depth relationship, accurately and seamlessly match the seismic information in the time domain with the logging information in the depth domain. First, the seismic of the target layer is aligned with the large set of layers in the synthetic record using the overall time shift, and then the small layer is fine-tuned using local stretching or compression. During the fine-tuning, it is necessary to ensure that the change in the time-depth relationship is reasonable.

[0015] Furthermore, the conversion of the time domain seismic record to the depth domain includes: using logging data to divide the layers and establish a depth-time relationship, and deriving a formula for the wave equation of the reflection coefficient of the multi-layer medium to obtain the reflection coefficient in the frequency domain; using the homomorphic deconvolution method to extract the seismic wavelet from the well bypass record in the time domain to obtain the wavelet spectrum in the frequency domain; using the frequency domain convolution model to calculate the synthetic record, performing an inverse Fourier transform to obtain the time domain synthetic seismic record, and then calculating the depth point corresponding to each time domain sampling point according to the time-depth relationship, and resampling according to the required depth sampling interval to obtain the depth domain synthetic seismic record.

[0016] Furthermore, the seismic wave dispersion attribute in the depth domain is similar to the dispersion attribute in the time domain, and both are special measurements of geometry, kinematics, dynamics and statistical characteristics derived from seismic data. Dynamic attributes reflect the amplitude change of seismic data and the variance, gradient and energy curvature extended from the amplitude, and their meaning in the depth domain is consistent with that in the time domain. Since the imaging points in the depth domain are more accurately located, the spatial position of the calculated attributes is more precise, which can better serve the prediction of reservoir gas content.

[0017] Furthermore, the depth domain dispersion attribute is obtained by quantitatively measuring the dispersion of seismic waves in the depth domain, and is an observable "body" response attribute of the amplitude or energy attenuation change of the signal in the frequency domain when the seismic signal propagates in the underground medium;

[0018] Furthermore, the data for constructing the deep network model includes actual data, synthetic data, or a mixture of actual data and synthetic data, which should be reasonably selected and used depending on the specific conditions of the study area.

[0019] Furthermore, the deep network is a fusion network that integrates the advantages of convolutional neural networks and recurrent neural networks. It uses convolutional networks to establish a nonlinear mapping relationship between the dispersion properties of deep-domain seismic waves and reservoir gas content, and uses meridian networks to incorporate vertical correlation information of dispersion properties.

[0020] Furthermore, the prediction of reservoir gas content is carried out in the spatial domain. Because the thickness of the rock layer actually passed by the seismic waves may vary greatly due to different rock velocities within the same time interval, if the dispersion value measured in the time domain is used to estimate the gas content of the reservoir, an erroneous result will be obtained.

[0021] The beneficial effects of the present invention are as follows: the present invention first performs seismic wave dispersion analysis in the depth domain to extract the depth domain dispersion attributes that can characterize the gas content of the reservoir, and then based on the depth domain seismic wave dispersion analysis, utilizes a deep network to establish a nonlinear mapping model between the depth domain seismic wave dispersion attributes and the reservoir gas content information, thereby solving the problem of predicting the gas content of deeply buried reservoirs and providing an important reference basis for oil and gas exploration and development. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A flowchart showing the steps of a method for predicting reservoir gas content based on depth domain seismic wave dispersion inversion according to an embodiment of the present invention is shown;

[0023] Figure 2 shows a deep network structure diagram according to an embodiment of the present invention;

[0024] Figure 3 A diagram showing a seismic wave dispersion analysis result of a two-dimensional profile according to an embodiment of the present invention is shown;

[0025] Figure 4 A diagram showing a prediction result of gas content in a certain well according to an embodiment of the present invention is shown;

[0026] Figure 5 A diagram showing the prediction result of gas content in a two-dimensional section reservoir according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0027] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of the specification, not all of the embodiments.

[0028] like Figure 1 As shown, Figure 1 A flowchart of the steps of a reservoir gas content prediction method based on depth domain seismic wave dispersion inversion provided by the present invention, the method comprising:

[0029] Step 101: Obtain well data and seismic data in the study area.

[0030] Among them, well logging data and seismic data are common knowledge in the field and will not be elaborated on.

[0031] Step 102: Based on the well data and seismic data and in combination with the actual geological conditions of the study area, complete the quality control-based well-seismic fine calibration to build a bridge between seismic and well logging.

[0032] The essence of well-seismic calibration is to adjust the time-depth relationship and accurately and seamlessly match the seismic information in the time domain with the logging information in the depth domain. Specifically, firstly, the seismic of the target layer is aligned with the large set of layers of the synthetic record using the overall time shift, and then the small layer is fine-tuned using local stretching or compression. During the fine-tuning, it is necessary to ensure that the change in the time-depth relationship is reasonable.

[0033] Step 103: based on the well-seismic fine calibration result, convert the time domain seismic records into the depth domain.

[0034] Among them, converting the time domain seismic record to the depth domain includes: using logging data to divide the layers and establish a depth-time relationship, deriving a formula based on the wave equation of the reflection coefficient of the multi-layer medium, and obtaining the reflection coefficient in the frequency domain; extracting the seismic wavelet from the well bypass record in the time domain using the homomorphic deconvolution method to obtain the wavelet spectrum in the frequency domain; calculating the synthetic record using the frequency domain convolution model, performing an inverse Fourier transform to obtain the time domain synthetic seismic record, and then calculating the depth point corresponding to each time domain sampling point based on the time-depth relationship, resampling according to the required depth sampling interval, and obtaining the depth domain synthetic seismic record.

[0035] Step 104: Based on the depth-domain seismic records, perform seismic wave dispersion analysis in the depth domain to extract depth-domain dispersion attributes that can characterize the gas content of the reservoir.

[0036] Among them, the seismic wave dispersion attribute in the depth domain is similar to the dispersion attribute in the time domain, and both are special measurements of geometry, kinematics, dynamics and statistical characteristics derived from seismic data. Dynamic attributes reflect the amplitude change of seismic data and the variance, gradient and energy curvature extended from the amplitude. The meaning in the depth domain is consistent with that in the time domain. Since the imaging points in the depth domain are more accurately located, the spatial position of the calculated attributes is more precise, which can better serve the prediction of reservoir gas content.

[0037] Among them, the depth domain dispersion attribute is obtained by quantitatively measuring the seismic wave dispersion in the depth domain. It is an observable "body" response attribute of the amplitude or energy attenuation change of the signal in the frequency domain when the seismic signal propagates in the underground medium.

[0038] Step 105: Based on the depth-domain seismic wave dispersion attributes and the reservoir gas content information of the well point, taking the reservoir gas content information of the well location as a label and the extracted depth-domain seismic wave dispersion attributes as input, a deep network nonlinear mapping model between the depth-domain seismic wave dispersion attributes and the reservoir gas content is established.

[0039] Among them, the deep network model is a fusion network that integrates the advantages of convolutional networks and recurrent networks. The convolutional neural network is used to establish a nonlinear mapping relationship between the dispersion properties of deep-domain seismic waves and reservoir gas content, and the meridian network is used to incorporate the vertical correlation information of the dispersion properties to achieve data-driven direct inversion prediction of reservoir gas content.

[0040] Step 106: Based on the established deep network mapping model for predicting reservoir gas content in the depth domain, the depth domain seismic wave dispersion attribute data of the target layer in the study area is used as input to achieve direct inversion prediction of reservoir gas content from known to unknown.

[0041] Among them, the prediction of reservoir gas content is carried out in the spatial domain. Because in the same time interval, due to different rock velocities, the actual thickness of the rock layer passed by the seismic wave may vary greatly. At this time, if the dispersion value measured in the time domain is used to estimate the gas content of the reservoir, an erroneous result will be obtained.

[0042] Figure 2This is a deep fusion network structure diagram used in an embodiment of the present invention. The relationship between the deep domain seismic wave dispersion attribute and the gas content of the reservoir is complex and nonlinear. In order to obtain accurate prediction results, it is necessary to fully extract the intrinsic characteristic information in the dispersion attribute and the labeled gas content. The three-dimensional seismic dispersion data distributes the dispersion results of each channel horizontally, and there is an intrinsic connection between them. The use of a convolutional network model can better extract this spatial distribution feature. The vertical distribution of the dispersion data is a sequence distribution, and the extraction of sequence data features is the strength of the recurrent network model. For the gas content prediction task based on deep domain seismic wave dispersion, combined with the actual situation of this embodiment, as well as the characteristics and applicability of neural networks, a fusion network model is selected here to construct a complex nonlinear relationship between seismic wave dispersion and gas content.

[0043] Figure 2 In the proposed method, the input is the extracted seismic wave dispersion attribute in the depth domain, and the output is the gas content at the well point. First, the convolutional network is used to establish the nonlinear mapping relationship between the seismic wave dispersion attribute in the depth domain and the gas content in the reservoir. Then, the meridian network is used to integrate the vertical correlation information of the dispersion attribute to accurately establish the complex nonlinear relationship between the seismic wave dispersion attribute in the depth domain and the gas content.

[0044] The actual data of a certain exploration area was used for testing. A 3D data volume of 5×5 channels and 300 meters deep was extracted for each training, and four sections were cut at a 45-degree direction. The well label gas content data and dispersion profile were then input into the model in a "point-window" manner, and iterative learning with an appropriate number of iterations was performed to train a complex nonlinear mapping model between the dispersion attributes of deep-domain seismic waves and the gas content of the reservoir.

[0045] The deep network structure constructed in this embodiment is as follows: 3 convolutional layers, with the number of convolution kernels being 40, 60 and 80 respectively; two fully connected layers, with the number of neurons being 1024 and 512 respectively; two recurrent network layers, with the number of neurons being 256 and 128 respectively; the batch size being 64; and the maximum number of iterations being 500.

[0046] Figure 3 This is a diagram of the seismic wave dispersion analysis results for a two-dimensional profile. By comparing the dispersion analysis results with the gas content in the well, it is observed that there is a strong correlation between the two, indicating that the extracted dispersion attributes can better characterize the gas content of the reservoir.

[0047] Figure 4 This is the gas content prediction result diagram of a well. Figure 4 From the actual gas content and the predicted gas content, it can be seen that the actual gas content is highly correlated with the gas content predicted by using the seismic wave dispersion attributes, and the trends are basically consistent.

[0048] Figure 5This is a two-dimensional section of the reservoir gas content prediction result map, which reflects the gas content distribution of the reservoir well. The results show that the deep network mapping model for predicting gas content in the deep domain reservoir established by using the dispersion attributes that can characterize the gas content of the reservoir extracted by deep domain seismic wave dispersion analysis as input can realize direct inversion prediction of gas content from known to unknown reservoirs, solve the problem of gas content prediction in deep buried reservoirs, and obtain the gas content distribution of the target layer in the study area.

[0049] It should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art should understand that the present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention, and these changes and improvements fall within the scope of the present invention claimed for protection.

Claims

1. A method for predicting reservoir gas content based on depth domain seismic wave dispersion inversion, characterized in that: The gas content prediction method comprises: Acquisition of well and seismic data; Carry out fine well-seismic calibration on well data and seismic data; Based on the well-seismic calibration, the time domain seismic records are converted to the depth domain; Conduct seismic wave dispersion analysis in the depth domain to extract the depth domain dispersion attributes that can characterize the gas content of the reservoir; Using the reservoir gas content information of the well point as the label and the extracted seismic wave dispersion attribute as the input, a deep network mapping model for predicting reservoir gas content in the depth domain is established; Input the seismic wave dispersion attribute of the target layer in the study area to achieve direct inversion prediction of gas content from known to unknown reservoirs and obtain the gas content distribution of the target layer in the study area; The essence of the well-seismic calibration is to adjust the time-depth relationship, accurately and seamlessly match the seismic information in the time domain with the logging information in the depth domain. First, the seismic of the target layer is aligned with the large set of layers of the synthetic record by using the overall time shift, and then the small layer is fine-tuned by using local stretching or compression. During the fine-tuning, it is necessary to ensure that the change of the time-depth relationship is reasonable. The method of converting the time domain seismic record into the depth domain seismic record is specifically as follows: using the well logging data to divide the layers and establish a depth-time relationship, and obtaining the reflection coefficient in the frequency domain according to the wave equation derivation formula of the multi-layer medium reflection coefficient; extracting the seismic wavelet from the well bypass record in the time domain by using the homomorphic deconvolution method to obtain the wavelet spectrum in the frequency domain; calculating the synthetic record by using the frequency domain convolution model, performing inverse Fourier transform to obtain the time domain synthetic seismic record, and then obtaining the depth point corresponding to each time domain sampling point according to the time-depth relationship, and resampling according to the required depth sampling interval to obtain the depth domain synthetic seismic record; The depth domain dispersion attribute is obtained by quantitatively measuring the dispersion of seismic waves in the depth domain. It is an observable "body" response attribute of the amplitude or energy attenuation change of the signal in the frequency domain when the seismic signal propagates in the underground medium. The deep network model is a fusion network that integrates the advantages of convolutional neural networks and recurrent neural networks, wherein the convolutional network is used to establish a nonlinear mapping relationship between the seismic wave dispersion attribute in the depth domain and the gas content of the reservoir, and the meridian network is used to integrate the vertical correlation information of the seismic wave dispersion attribute; The data for establishing the deep network model include actual data, synthetic data, or a mixture of actual data and synthetic data, and should be reasonably selected and used depending on the specific conditions of the study area.

2. The method for predicting reservoir gas content based on depth domain seismic wave dispersion inversion according to claim 1, characterized in that: The depth domain imaging point positioning is more accurate, and the measured attribute spatial position is more precise, which can better serve the reservoir gas content prediction work.

3. The method for predicting reservoir gas content based on depth domain seismic wave dispersion inversion according to claim 1, characterized in that: Seismic wave dispersion analysis is performed in the depth domain to extract the depth domain dispersion attributes that can characterize the gas content of the reservoir. On this basis, the depth domain seismic wave dispersion attribute data at the well location is extracted, and the corresponding relationship between the depth domain seismic wave dispersion attributes and the reservoir gas content is established, thereby obtaining a database for constructing a deep network model.

4. The method for predicting reservoir gas content based on depth domain seismic wave dispersion inversion according to claim 1, characterized in that: The reservoir gas content prediction is carried out in the spatial domain. In the same time interval, due to different rock velocities, the actual thickness of the rock layer passed by the seismic wave may vary greatly. At this time, if the dispersion value measured in the time domain is used to estimate the gas content of the reservoir, an erroneous result will be obtained.

5. The method for predicting reservoir gas content based on depth domain seismic wave dispersion inversion according to claim 1, characterized in that: Based on the deep-domain seismic wave dispersion analysis, a nonlinear mapping model between dispersion attributes and gas content was established using a deep network, which solved the problem of predicting gas content in deep reservoirs and provided an important reference for oil and gas exploration and development.

Citation Information

Patent Citations

  • Depth domain stratal configuration inversion method based on geological information mapping

    CN106094019A

  • Reservoir gas-bearing evaluation method based on depth domain seismic wave dispersion analysis

    CN109283581A

  • Reservoir feature inversion method based on seismic scattered wave field analysis

    CN113589385A

  • Reservoir prediction method based on deep hybrid neural network

    CN113608265A

  • Analytics and machine learning method for estimating petrophysical property values

    US20200379135A1