Complex lithology identification method, identification system, electronic equipment and storage medium

Abstract

Disclosed are a complex lithology identification method, identification system, electronic device and storage medium. The complex lithology identification method comprises: constructing a two-dimensional convolutional neural network; using the characteristic matrix of each depth point of the logging curve as input, taking the lithology coding result of each depth point as output, training network parameters, and obtaining a lithology automatic identification network model; using the lithology automatic identification network model, performing lithology prediction on the logging data, and obtaining complex lithology identification results of multiple wells. The present invention extracts multiple information of the logging curve, considers the vertical resolution difference of the logging data, establishes the logging information matrix within the depth window as the characteristic matrix, and uses a two-dimensional convolutional neural network to establish a lithology identification model. This method can more accurately identify the lithology of complex lithology reservoirs, and has higher accuracy than other methods such as intersection diagrams, multi-resolution graph clustering, support vector machines, neural networks, etc.

Description

Technical Field

[0001] The present invention relates to the field of oil and gas geological research, and belongs to the field of well logging evaluation technology in oil and gas exploration. More specifically, it relates to a complex lithology identification method based on well logging curves, an identification system, an electronic device and a storage medium. Background Art

[0002] At present, the methods for lithological interpretation using logging data are mainly divided into: spider web diagrams, intersection diagrams, multi-resolution graph clustering, and machine learning methods such as decision trees and random forests.

[0003] The spider diagram is a qualitative method that mainly divides lithology qualitatively based on the differences in logging response characteristics of different lithologies (Zhang Tao et al., 2016). The intersection diagram method can only use part of the logging information to divide a certain lithology, and the accuracy is not high enough for complex reservoirs (Kuang Lichun et al., 2013). The multi-resolution graph clustering method is an unsupervised / semi-supervised method, and the experience of the processing personnel is required when merging lithologies (Mao Yongqiang et al., 2019). The machine learning method has the characteristics of highly automated and intelligent data, and can automatically extract the rock physical characteristics implicit in the logging data, which can be well used for lithology identification.

[0004] However, the commonly used lithology identification methods based on machine learning (such as decision tree (Li Hongqi et al., 2010), support vector machine (Li Hongqi et al., 2009), random forest, neural network, one-dimensional convolutional neural network, etc.) only use the logging curve information of a single depth point, and do not consider the problem of different logging curve resolutions. There are still certain problems in the identification of complex thin interbedded lithologies.

[0005] Therefore, it is necessary to develop a complex lithology identification method, identification system, electronic equipment and storage medium based on well logging curves.

[0006] 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

[0007] The present invention proposes a complex lithology identification method based on well logging curves, an identification system, an electronic device and a storage medium. The present invention extracts multiple information from well logging curves and uses a machine learning method to establish a lithology identification model, and uses the lithology identification model to identify complex lithology.

[0008] According to a first aspect of the present invention, a complex lithology identification method is disclosed, comprising:

[0009] Construct a two-dimensional convolutional neural network;

[0010] Using the characteristic matrix of each depth point of the logging curve as input and the lithology coding result of each depth point as output, the network parameters are trained to obtain the lithology automatic identification network model;

[0011] The lithology automatic identification network model is used to predict the lithology of the well logging data and obtain complex lithology identification results of multiple wells.

[0012] Furthermore, according to the vertical resolution of the logging curve and the sampling interval of the logging instrument, the window length W is selected so that the window length includes the minimum vertical resolution of the selected logging curve, and the characteristic matrix A is established for each depth point:

[0013] A=W*(L+2)

[0014] Among them, W represents the window length and L represents the decomposition scale.

[0015] Furthermore, according to the vertical resolution of different logging curves, a corresponding weight function is established, and the signal of the characteristic matrix is ​​filtered to obtain the characteristic matrix A* of the depth point.

[0016] Furthermore, appropriate wavelet functions and decomposition scale L are selected to perform wavelet packet decomposition on all lithology identification curves, extract wavelet approximation coefficients of different decomposition layers, and reconstruct characteristic curves of different resolutions using wavelet approximation coefficients.

[0017] Furthermore, according to the logging response characteristics and rock physics principles, the sensitive logging curves are combined and the curve gradients are calculated to establish the lithology identification curve.

[0018] Furthermore, the sensitive logging curve is determined by analyzing the logging response characteristics of the logging curve in the study area.

[0019] Furthermore, the method also includes improving the longitudinal resolution of the characteristic curve by reconstructing the signal after enhancing the high-frequency signal.

[0020] According to a second aspect of the present invention, a complex lithology identification system is disclosed, comprising:

[0021] Building blocks for constructing two-dimensional convolutional neural networks;

[0022] The training module uses the characteristic matrix of each depth point of the logging curve as input and the lithology coding result of each depth point as output to train the network parameters and obtain the lithology automatic identification network model;

[0023] The identification module uses the lithology automatic identification network model to predict the lithology of the well logging data and obtain the complex lithology identification results of multiple wells.

[0024] According to a third aspect of the present invention, an electronic device is disclosed, the electronic device comprising:

[0025] A memory storing executable instructions;

[0026] A processor runs the executable instructions in the memory to implement the complex lithology identification method.

[0027] According to a fourth aspect of the present invention, a computer-readable storage medium is disclosed, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the complex lithology identification method is implemented.

[0028] The present invention extracts multiple information of logging curves, takes into account the vertical resolution difference of logging data, establishes the logging information matrix in the depth window as the feature matrix, and uses a two-dimensional convolutional neural network to establish a lithology recognition model. This method can more accurately identify the lithology of complex lithology reservoirs, and is more accurate than other methods such as cross-plots, multi-resolution graph clustering, support vector machines, neural networks, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] 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.

[0030] Figure 1 The figure is a flow chart of the complex lithology identification method of the present invention.

[0031] Figure 2 It is a schematic diagram of a complex lithology identification method based on well logging curves according to an embodiment of the present invention.

[0032] Figure 3 Schematic diagram of wavelet packet decomposition of a logging curve according to an embodiment of the present invention.

[0033] Figure 4 It is a sonic time difference logging curve of a certain reservoir section in an embodiment of the present invention and a curve after wavelet decomposition, reconstruction and standardization.

[0034] Figure 5 Schematic diagram of a weight function according to an embodiment of the present invention.

[0035] Figure 6 It is a normalized acoustic wave time difference and compensated neutron curve within a depth window length and a curve filtered by a weight function in an embodiment of the present invention.

[0036] Figure 7 Schematic diagram of the CNN structure of an embodiment of the present invention.

[0037] Figure 8 This is a lithology identification result diagram of an embodiment of the present invention. DETAILED DESCRIPTION

[0038] 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.

[0039] Lithology identification is an important basis for petroleum geological research and reservoir classification and evaluation. For thin interbedded complex lithology reservoirs, lithology identification is difficult. The present invention provides a complex lithology identification method based on well logging curves, by constructing a 2D convolutional neural network and training network parameters to obtain a lithology automatic identification network model.

[0040] like Figure 1 As shown, the complex lithology identification method of the present invention comprises:

[0041] Construct a two-dimensional convolutional neural network;

[0042] Using the characteristic matrix of each depth point of the logging curve as input and the lithology coding result of each depth point as output, the network parameters are trained to obtain the lithology automatic identification network model;

[0043] The lithology automatic identification network model is used to predict the lithology of the well logging data and obtain complex lithology identification results of multiple wells.

[0044] The complex lithology identification method of the present invention constructs a two-dimensional convolutional neural network, uses the logging information matrix in the depth window as a feature matrix to train network parameters, thereby obtaining a lithology automatic identification network model, and finally uses the lithology automatic identification network model to predict the lithology of the logging data to obtain complex lithology identification results for multiple wells.

[0045] Furthermore, in the method of the present invention, the lithology logging response characteristics can be first analyzed to determine the sensitive logging curve; the sensitive logging curves can be combined and the curve gradients can be calculated to establish a new lithology identification curve. Furthermore, appropriate parameters can be selected for wavelet packet decomposition, and the approximate coefficients and processed detail coefficients can be reconstructed to obtain lithology identification curves with different resolutions. Furthermore, the vertical resolution of the characteristic curve can be improved by reconstructing the signal after enhancing the high-frequency signal.

[0046] Furthermore, in the complex lithology identification method of the present invention, the window length W can be selected according to the vertical resolution of the logging curve and the sampling interval of the logging instrument so that the window length includes the minimum vertical resolution of the selected logging curve, and the characteristic matrix A is established for each depth point:

[0047] A=W*(L+2)

[0048] Among them, W represents the window length and L represents the decomposition scale.

[0049] More preferably, a corresponding weight function can be established according to the vertical resolution of different logging curves, and the signal of the feature matrix can be filtered to obtain the feature matrix A* of the depth point. The feature matrix A* is used to train the network parameters, and the vertical resolution difference of the logging data is further considered, so the recognition accuracy is higher.

[0050] The present invention also provides a complex lithology identification system, comprising:

[0051] Building blocks for constructing two-dimensional convolutional neural networks;

[0052] The training module uses the characteristic matrix of each depth point of the logging curve as input and the lithology coding result of each depth point as output to train the network parameters and obtain the lithology automatic identification network model;

[0053] The identification module uses the lithology automatic identification network model to predict the lithology of the well logging data and obtain the complex lithology identification results of multiple wells.

[0054] The present invention also provides an electronic device, which includes: a memory storing executable instructions; and a processor, which runs the executable instructions in the memory to implement the above-mentioned complex lithology identification method.

[0055] The present invention also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned complex lithology identification method is implemented.

[0056] To facilitate understanding of the solutions and effects of the embodiments of the present invention, four 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.

[0057] Example 1

[0058] like Figure 2 As shown, this embodiment provides a complex lithology identification method based on well logging curves, the method comprising:

[0059] S101: Preprocess the well logging curves in the study area, such as performing environmental correction, analyzing the lithology well logging response characteristics, and determining sensitive well logging curves. Specifically, sensitive well logging curves can be determined by rock physics principles (geological meaning of well logging curves), cross plots, statistical characteristics of well logging responses of different lithologies, etc.

[0060] S102: According to the logging response characteristics and rock physics principles, the determined sensitive logging curves are combined (for example, by mathematical transformations such as difference, quotient, multiplication, and logarithm) and the curve gradient is calculated to establish a new lithology identification curve. In this embodiment, the differential value of each sensitive curve with the depth point is used as the gradient of the curve.

[0061] S103: Select appropriate wavelet function and decomposition scale L, perform wavelet packet decomposition on all new lithology identification curves, extract wavelet approximation coefficients of different decomposition layers, and reconstruct characteristic curves of different resolutions using these approximation coefficients. In this embodiment, each sensitive logging curve can be regarded as a characteristic vector. Figure 3 A schematic diagram of wavelet packet decomposition of logging curves is shown. Figure 4 The acoustic time difference logging curve of a certain reservoir section in this embodiment and the curve after wavelet decomposition, reconstruction and standardization are shown.

[0062] Preferably, in this embodiment, the vertical resolution of the curve is improved by enhancing the high-frequency signal and then reconstructing the signal.

[0063] S104: According to the vertical resolution of the selected logging curve and the sampling interval of the logging instrument, select an appropriate window length W (for example, so that the window length includes the minimum vertical resolution of the selected curve), and establish a feature matrix A=W*(L+2) for each depth point, where W represents the window length and L represents the decomposition scale.

[0064] S105: According to the vertical resolution of different logging curves, a corresponding weight function P is established, and the signal of the characteristic matrix is ​​filtered to obtain the characteristic matrix A* of the depth point. Figure 5 The weight function of this embodiment is given, Figure 6 The standardized acoustic wave time difference and compensated neutron curve within a depth window length in this embodiment and the curve after weight function filtering are shown. Figure 6 The corresponding W=7, the data sampling interval is 0.1 meter, the corresponding minimum vertical resolution is 0.7 meter, and the decomposition scale L=8 used in this embodiment.

[0065] S106: construct a 2D convolutional neural network, take the feature matrix A* of each depth point as input, take the lithology coding result B of each depth point as output, train the network parameters, and obtain a lithology automatic identification network model. Figure 7 FIG. 1 shows a schematic diagram of the CNN structure of this embodiment, which is a 2D convolutional network constructed in this embodiment, corresponding to a network model of a depth point. Figure 7In , N represents N lithology identification curves, that is, N input features, and C columns are outputs B, which can be understood as the encoding results of C lithology categories from 1 to C in the B matrix, such as B = [1, 0, ..., 0] or B = [0, 1, ..., 0], etc. Conv1 uses P1 R1*R1 convolution kernels to convolve with the initial data volume, and fills the obtained data volume to make the data dimension consistent with the original data, that is, W*(L+2). Further, the Relu function is used as the activation function to obtain a P1*W*(L+2) data volume. Similarly, Conv2 uses P2 R2*R2 convolution kernels to convolve with the P1*W*(L+2) data volume, fills and uses the Relu activation function to obtain a P2*W*(L+2) data volume. Dropout, standardization and flattening are used on the data volume to obtain a one-dimensional matrix of W*(L+2)*P2. A fully connected layer (Sigmoid is used as the activation function after normalization) is used to process the data body to obtain a P3 one-dimensional matrix, and Softmax is further used as the activation function to obtain the final classification result.

[0066] S107: Use the trained model to predict the lithology of the data in the database to be tested, such as Figure 8 As shown: the first track is the original natural gamma curve and natural gamma curves of different resolutions, the second track is the neutron, density and acoustic time difference curves, the third track is the resistivity curve, the fourth track is the overlap result of the neutron and density curves, the fifth track is the overlap result of the acoustic wave and resistivity curves, the sixth track is the overlap result of the neutron and resistivity curves, the seventh track is the lithology category described by the core, the eighth track FLAG (solid line) is the lithology category described by the core, and FLAG_CNN (dashed line) is the lithology category predicted by the method proposed in this patent. It can be seen from the figure that the lithology category described by the core is consistent with the predicted lithology category.

[0067] Example 2

[0068] This embodiment provides a complex lithology identification system, including:

[0069] Building blocks for constructing two-dimensional convolutional neural networks;

[0070] The training module uses the characteristic matrix of each depth point of the logging curve as input and the lithology coding result of each depth point as output to train the network parameters and obtain the lithology automatic identification network model;

[0071] The identification module uses the lithology automatic identification network model to predict the lithology of the well logging data and obtain the complex lithology identification results of multiple wells.

[0072] Example 3

[0073] This embodiment provides an electronic device, which includes: a memory storing executable instructions; and a processor, which runs the executable instructions in the memory to implement the above-mentioned complex lithology identification method.

[0074] An electronic device according to an embodiment of the present disclosure includes a memory and a processor.

[0075] The memory is used to store non-temporary computer-readable instructions. Specifically, the memory may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may, for example, include random access memory (RAM) and / or cache memory (cache), etc. The non-volatile memory may, for example, include read-only memory (ROM), hard disk, flash memory, etc.

[0076] The processor may be a central processing unit (CPU) or other forms of processing units having data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device to perform desired functions. In one embodiment of the present disclosure, the processor is used to run the computer-readable instructions stored in the memory.

[0077] Those skilled in the art should be able to understand that in order to solve the technical problem of how to obtain a good user experience, the present embodiment may also include well-known structures such as a communication bus and an interface, and these well-known structures should also be included in the protection scope of the present disclosure.

[0078] For detailed description of this embodiment, reference may be made to the corresponding descriptions in the aforementioned embodiments, which will not be repeated here.

[0079] Example 4

[0080] This embodiment provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the complex lithology identification method is implemented.

[0081] According to the computer-readable storage medium of the embodiment of the present disclosure, non-transitory computer-readable instructions are stored thereon. When the non-transitory computer-readable instructions are executed by a processor, all or part of the steps of the method of each embodiment of the present disclosure are executed.

[0082] The above-mentioned computer-readable storage media include, but are not limited to: optical storage media (e.g., CD-ROM and DVD), magneto-optical storage media (e.g., MO), magnetic storage media (e.g., magnetic tape or mobile hard disk), media with built-in rewritable non-volatile memory (e.g., memory card) and media with built-in ROM (e.g., ROM box).

[0083] 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.

[0084] 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 complex lithology identification method, characterized in that: include: According to the logging response characteristics and rock physics principles, the sensitive logging curves are combined and the curve gradients are calculated to establish the lithology identification curve; the differential value of each sensitive curve with the depth point is used as the gradient of the curve; Select appropriate wavelet function and decomposition scale L, perform wavelet packet decomposition on all lithology identification curves, extract wavelet approximation coefficients of different decomposition layers, and use these wavelet approximation coefficients to reconstruct characteristic curves of different resolutions; each sensitive logging curve is regarded as a characteristic vector; According to the vertical resolution of the logging curve and the sampling interval of the logging instrument, the window length W is selected so that the window length contains the minimum vertical resolution of the selected logging curve, and the characteristic matrix A is established for each depth point: A=W*(L+2) Among them, W represents the window length, and L represents the decomposition scale; According to the vertical resolution of different logging curves, the corresponding weight function is established, and the signal of the characteristic matrix is ​​filtered to obtain the characteristic matrix A* of the depth point; Construct a two-dimensional convolutional neural network; Using the characteristic matrix of each depth point of the logging curve as input and the lithology coding result of each depth point as output, the network parameters are trained to obtain the lithology automatic identification network model; The lithology automatic identification network model is used to predict the lithology of the well logging data and obtain complex lithology identification results of multiple wells.

2. The complex lithology identification method according to claim 1, characterized in that: The sensitive logging curve is determined by analyzing the logging response characteristics of the logging curve in the study area.

3. The complex lithology identification method according to claim 1, characterized in that: The method further includes improving the longitudinal resolution of the characteristic curve by reconstructing the signal after enhancing the high-frequency signal.

4. A complex lithology identification system, characterized in that: include: Feature matrix calculation module, used for: According to the logging response characteristics and rock physics principles, the sensitive logging curves are combined and the curve gradients are calculated to establish the lithology identification curve; the differential value of each sensitive curve with the depth point is used as the gradient of the curve; Select appropriate wavelet function and decomposition scale L, perform wavelet packet decomposition on all lithology identification curves, extract wavelet approximation coefficients of different decomposition layers, and use these wavelet approximation coefficients to reconstruct characteristic curves of different resolutions; each sensitive logging curve is regarded as a characteristic vector; According to the vertical resolution of the logging curve and the sampling interval of the logging instrument, the window length W is selected so that the window length contains the minimum vertical resolution of the selected logging curve, and the characteristic matrix A is established for each depth point: A=W*(L+2) Among them, W represents the window length, and L represents the decomposition scale; According to the vertical resolution of different logging curves, the corresponding weight function is established, and the signal of the characteristic matrix is ​​filtered to obtain the characteristic matrix A* of the depth point; Building blocks for constructing two-dimensional convolutional neural networks; The training module uses the characteristic matrix of each depth point of the logging curve as input and the lithology coding result of each depth point as output to train the network parameters and obtain the lithology automatic identification network model; The identification module uses the lithology automatic identification network model to predict the lithology of the well logging data and obtain the complex lithology identification results of multiple wells.

5. An electronic device, characterized in that: The electronic device comprises: A memory storing executable instructions; A processor, wherein the processor runs the executable instructions in the memory to implement the complex lithology identification method according to any one of claims 1 to 3.

6. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the complex lithology identification method according to any one of claims 1 to 3 is implemented.

Images