Multi-level sublithology intelligent interpretation model construction method and application thereof

Abstract

The invention relates to the technical field of oilfield development, and discloses a multi-level sublithology intelligent interpretation model construction method and application thereof, and the method comprises the steps: carrying out the preprocessing of an original logging curve; performing multi-level sub-wavelet decomposition according to the preprocessed original logging curve to obtain a plurality of wavelet components with different frequencies; according to the multiple wavelet components with different frequencies, logging curve reconstruction is carried out, and a first reconstructed logging curve and a second reconstructed logging curve are obtained; according to lithology data of a target rock area, an original logging curve, a first reconstruction logging curve and a second reconstruction logging curve, through an integration algorithm and a training model, learning mapping relations between the original logging curve and a composite sand body, between the first reconstruction logging curve and a single sand body and an argillaceous interlayer, and between the second reconstruction logging curve and a thin reservoir and a physical interlayer; and a multi-level lithology intelligent explanation model is obtained. According to the method, well logging fine interpretation of lithology of different thicknesses can be realized, particularly accurate identification of lithology of centimeter-level thickness can be realized, and the well logging interpretation precision is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of oilfield development, and in particular to a method for constructing a multi-level lithology intelligent interpretation model and its application. Background Art

[0002] Identifying the composition and structural characteristics of underground rocks is an important part of geological and geophysical research, and plays a vital role in the exploration and development of oil and natural gas. The data used for lithology interpretation mainly include two categories: core data and well logging data. Correspondingly, the multi-level lithology intelligent interpretation methods are divided into core analysis method and well logging interpretation method. Core analysis method refers to observing, analyzing, and testing core samples to clarify the type and characteristics of rocks. This method is only applicable to coring wells. The cost of coring and analysis in the core analysis method is extremely high, and there are very few coring wells in oil fields, so this method cannot be widely promoted in oil field development. The well logging interpretation method is based on multiple well logging curves, and uses linear or nonlinear methods to construct a mapping relationship between well logging curves and lithology categories, and then identifies lithology. At present, the main well logging interpretation method is to identify lithology by constructing cross plots (such as gamma-density cross plots). However, since the logging curve is a comprehensive response of the underground reservoir and is affected by many factors, this type of method has low recognition accuracy and large errors, especially it is unable to identify siltstone and thin mudstone in thick sandstone, that is, it is unable to identify thin interlayers inside the sand body.

[0003] Therefore, a new multi-level lithology intelligent interpretation method is urgently needed to achieve precise identification of lithology. Summary of the invention

[0004] The present invention provides a method for constructing a multi-level lithology intelligent interpretation model and its application, so as to solve the defect that the prior art cannot realize the fine identification of lithology.

[0005] The present invention provides a method for constructing a multi-level lithology intelligent interpretation model, comprising: Obtain lithology data and original logging curves of the target rock area, where the target rock area includes composite sand bodies, single sand bodies and mud interlayers, thin reservoirs and physical interlayers; Preprocess the original logging curve; According to the pre-processed original logging curve, multi-level sub-wavelet decomposition is performed to obtain multiple sub-wavelet components with different frequencies; Reconstructing the well logging curve according to the wavelet components of multiple different frequencies to obtain a first reconstructed well logging curve and a second reconstructed well logging curve; According to the lithological data of the target rock area as well as the original logging curve, the first reconstructed logging curve and the second reconstructed logging curve, the multi-level lithological intelligent interpretation model is trained through an integrated algorithm to learn the mapping relationship between the original logging curve and the composite sand body, the first reconstructed logging curve and the single sand body and mud interlayer, and the second reconstructed logging curve and the thin reservoir and physical interlayer, so as to obtain the final multi-level lithological intelligent interpretation model.

[0006] In one embodiment, preprocessing includes any one of the following or any combination thereof: rock electrical relocation, outlier processing, and standardization.

[0007] In one embodiment, the multi-level sub-wavelet decomposition is performed based on the pre-processed original logging curve to obtain multiple wavelet components with different frequencies, including: The order of multi-level wavelet decomposition is set according to the thickness of the interlayer in the target rock area; According to the set level of multi-level wavelet decomposition, multi-level wavelet decomposition is performed on the pre-processed original logging curve.

[0008] In one embodiment, the step of setting the levels of the multi-level wavelet decomposition according to the thickness of the interlayer in the target rock area comprises: When the thickness of the interlayer in the target rock area is less than the preset minimum threshold, the level of the multi-level wavelet decomposition is set to the second-level decomposition; When the thickness of the interlayer in the target rock area is between a preset minimum threshold and a preset maximum threshold, the level of the multi-level wavelet decomposition is set to three-level decomposition; When the thickness of the interlayer in the target rock area is greater than a preset maximum threshold, the level of the multi-level wavelet decomposition is set to no decomposition required.

[0009] In one embodiment, performing multi-level wavelet decomposition on the pre-processed original logging curve according to the set level of multi-level wavelet decomposition includes: When the level of the multi-level wavelet decomposition is set to three-level decomposition, the preprocessed original logging curve is subjected to first-level wavelet decomposition to obtain a first wavelet component and a second wavelet component (usually dominated by high-frequency noise), wherein the first wavelet component is a wavelet component within a first preset low-frequency range, and the second wavelet component is a wavelet component within a first preset high-frequency range; Performing secondary wavelet decomposition on the first wavelet component to obtain a third wavelet component and a fourth wavelet component, wherein the second wavelet component is a wavelet component within a second preset low frequency range, and the fourth wavelet component is a wavelet component within a second preset high frequency range, wherein the frequency of the third wavelet component is lower than that of the first wavelet component; The third wavelet component is subjected to three-level wavelet decomposition to obtain a fifth wavelet component and a sixth wavelet component, wherein the fifth wavelet component is a wavelet component within a third preset low frequency range, and the sixth wavelet component is a wavelet component within a third preset high frequency range, wherein the frequency of the fifth wavelet component is less than that of the third wavelet component.

[0010] In one embodiment, the well logging curve is reconstructed according to a plurality of wavelet components with different frequencies to obtain a first reconstructed well logging curve and a second reconstructed well logging curve, including: Reconstructing the well logging curve using a first reconstruction expression according to a plurality of wavelet components with different frequencies to obtain a first reconstructed well logging curve; Reconstructing the well logging curve using a second reconstruction expression according to a plurality of wavelet components with different frequencies to obtain a second reconstructed well logging curve; Among them, the first reconstruction expression is: , The second reconstruction expression is: , In the formula, represents the first reconstructed well log curve, represents the second reconstructed logging curve, , , , Represents the reconstruction coefficients of the first reconstruction expression 1, K1, 1, 0, where the value range of K1 is 1~5, , , Represents the reconstruction coefficients of the second reconstruction expression: 1, K2, 0, where the value range of K2 is 1~5, Respectively represent the third sub-wave component and the fifth sub-wave component, , , They represent the second wavelet component, the fourth wavelet component, and the sixth wavelet component respectively.

[0011] In one embodiment, the integrated algorithm includes at least two of the following: LightGBM (lightweight gradient boosting learning algorithm), MLP (multi-layer perceptron algorithm), Bagging-RF (out-of-stack random forest algorithm), and the multi-level lithology intelligent interpretation model is trained through the integrated algorithm according to the lithology data of the target rock area and the original well logging curve, the first reconstructed well logging curve, and the second reconstructed well logging curve to learn the mapping relationship between the original well logging curve and the composite sand body, the first reconstructed well logging curve and the single sand body and the mud interlayer, and the second reconstructed well logging curve and the thin reservoir and the physical interlayer, so as to obtain the final multi-level lithology intelligent interpretation model, including: A training set is constructed based on the lithology data of the target rock area and the original well logging curve, the first reconstructed well logging curve, and the second reconstructed well logging curve, wherein the composite sand body, single sand body and mud interlayer, thin reservoir and physical interlayer in the lithology data are used as labels of the original well logging curve, the first reconstructed well logging curve, and the second reconstructed well logging curve; Perform multi-label sample balanced sampling on the training set; According to the training set after balanced sampling of multi-label samples, the multi-level lithology intelligent interpretation model is trained to learn the mapping relationship between the original logging curve and the composite sand body, the first reconstructed logging curve and the single sand body and mud interlayer, and the second reconstructed logging curve and the thin reservoir and physical interlayer, so as to obtain the final multi-level lithology intelligent interpretation model.

[0012] A multi-level lithology intelligent interpretation method, comprising: Receiving an original well logging curve of a rock area to be interpreted sent by at least one terminal; According to any of the above-mentioned methods for constructing a multi-level lithology intelligent interpretation model, the original well logging curve of the rock area to be interpreted is preprocessed, multi-level sub-wavelet decomposed, and the well logging curve is reconstructed to obtain a first reconstructed well logging curve and a second reconstructed well logging curve of the rock area to be interpreted; According to the original well logging curve, the first reconstructed well logging curve and the second reconstructed well logging curve of the rock area to be interpreted, the multi-level lithology intelligent interpretation model obtained by the multi-level lithology intelligent interpretation model construction method described in any one of the above items is used to obtain the lithology interpretation result of the rock area to be interpreted; The lithology interpretation result of the rock area to be interpreted is sent back to at least one terminal.

[0013] In one embodiment, the multi-level lithology intelligent interpretation model obtained by any of the above-mentioned multi-level lithology intelligent interpretation model construction methods according to the original well logging curve, the first reconstructed well logging curve and the second reconstructed well logging curve of the rock area to be interpreted, obtains the lithology interpretation result of the rock area to be interpreted, including: According to the original logging curves, the first reconstructed logging curves and the second reconstructed logging curves of the rock area to be interpreted, multiple initial lithological interpretation results are obtained through each item in the integrated algorithm in the multi-level lithological intelligent interpretation model, and combined with the preset algorithm weights, the final lithological interpretation results of the rock area to be interpreted are obtained.

[0014] The present invention also provides a multi-level lithology intelligent interpretation system, comprising: The data receiving module is used to: receive the original logging curve of the rock area to be interpreted sent by at least one terminal; A data processing module is used to: pre-process the original well logging curve of the rock area to be interpreted, perform multi-level sub-wavelet decomposition, and reconstruct the well logging curve according to any of the above-mentioned multi-level sub-lithology intelligent interpretation model construction methods, so as to obtain a first reconstructed well logging curve and a second reconstructed well logging curve of the rock area to be interpreted; A lithology interpretation module, for obtaining a lithology interpretation result of the rock area to be interpreted according to the original well logging curve, the first reconstructed well logging curve and the second reconstructed well logging curve of the rock area to be interpreted and the multi-level lithology intelligent interpretation model obtained by any of the multi-level lithology intelligent interpretation model construction methods described above; The data output module is used to send the lithology interpretation result of the rock area to be interpreted back to at least one terminal.

[0015] The present invention also provides an electronic device, comprising a processor and a memory storing a computer program, wherein when the processor executes the computer program, it implements any of the above-mentioned multi-level lithology intelligent interpretation model construction methods and / or multi-level lithology intelligent interpretation methods.

[0016] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for constructing a multi-level lithology intelligent interpretation model and / or the multi-level lithology intelligent interpretation method described above is implemented.

[0017] The present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute any of the above-mentioned multi-level lithology intelligent interpretation model construction methods and / or multi-level lithology intelligent interpretation methods.

[0018] The invention provides a method for constructing a multi-level lithology intelligent interpretation model and its application, which can at least bring the following beneficial effects: In view of the problem of low resolution of logging curves and the characteristics of multi-level nested distribution of rock formations, the present invention fully considers the hierarchical nature of lithological interpretation, and performs multi-level wavelet decomposition and logging curve reconstruction in the frequency domain according to the level of multi-level sub-wavelet decomposition determined by the rock formation thickness, thereby obtaining a new logging curve suitable for the interpretation of rock formations of different thicknesses, which can realize nested interpretation of lithologies of different thicknesses and greatly improve the accuracy of logging interpretation.

[0019] The present invention couples multi-level sub-wavelet decomposition and logging curve reconstruction with hierarchical nested lithologic interpretation, and proposes a "reservoir and thin interlayer hierarchical nested logging interpretation method": at the first level, the original logging curve is preferably used to interpret the thick composite sand body; at the second level, the first reconstructed logging curve is preferably used to identify the single sand body and mud interlayer inside the composite sand body; at the third level, the second reconstructed logging curve is preferably used to identify the thin reservoir and physical interlayer inside the single sand body, so as to realize the fine logging interpretation of lithologies of different thicknesses, especially the accurate identification of lithologies with centimeter-level thickness, thereby greatly improving the accuracy of logging interpretation.

[0020] The present invention performs multi-label sample balanced sampling processing on the training data of the multi-level lithology intelligent interpretation model, so that the number of samples of different lithologies reaches the same order of magnitude, which can improve the prediction accuracy of the multi-level lithology intelligent interpretation model and avoid the impact of obvious differences in the number of samples of different lithologies on model training due to the limitation of the number of coring wells.

[0021] The mapping relationship between lithology of different levels (thickness) and logging curves is complex and difficult to identify. A single algorithm model cannot meet the requirements for lithology identification of different levels (especially identification of lithology of centimeter-level thickness). The present invention uses an integrated algorithm to train a multi-level lithology intelligent interpretation model to learn the mapping relationship between the original logging curve and the composite sand body, the first reconstructed logging curve and the single sand body and mud interlayer, and the second reconstructed logging curve and the thin reservoir and physical interlayer, thereby achieving "complementary advantages" between different algorithms, significantly improving the accuracy of lithology identification of different levels, and effectively improving the generalization ability of the multi-level lithology intelligent interpretation model. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0023] Figure 1 A schematic flow chart of a method for constructing a multi-level lithology intelligent interpretation model provided by the present invention.

[0024] Figure 2 A schematic flow chart of a method for constructing a multi-level lithology intelligent interpretation model and its application provided by the present invention.

[0025] Figure 3 Schematic diagram of the process of multi-level sub-wavelet decomposition.

[0026] Figure 4 This is an example diagram of multi-level sub-wavelet decomposition.

[0027] Figure 5An example of well log reconstruction.

[0028] Figure 6 This is one of the flowcharts for training a multi-level sub-lithology intelligent interpretation model.

[0029] Figure 7 This is the second flowchart for training a multi-level sub-lithology intelligent interpretation model.

[0030] Figure 8 This is an example diagram of the lithology interpretation results.

[0031] Fig. 9 A schematic structural diagram of a multi-level lithology intelligent interpretation system provided by the present invention.

[0032] Fig.10 This is a schematic structural diagram of an electronic device provided by the present invention. DETAILED DESCRIPTION

[0033] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings in the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments, and they should not be understood as limitations on the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. In the description of the present invention, it should be understood that the terms used are only for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0034] The existing multi-level lithology intelligent interpretation method is mainly aimed at thicker sand bodies (>1m), without finely identifying the mudstone interlayers and physical interlayers inside the sand body, and cannot take into account the identification of sand bodies of different thicknesses and their internal thin interlayers. In particular, the order of the sand body is not considered, and the coupling relationship between the different frequency information of the logging curve and the different order (thickness) of the rock layer is not considered, resulting in low accuracy of sand body interpretation. In particular, it is impossible to effectively identify the thin interlayers (<1m) inside the sand body, especially the centimeter-level mud interlayers (10cm ~ 30cm) and physical interlayers (10cm ~ 50cm). Therefore, it is urgent to establish a new multi-level lithology intelligent interpretation method.

[0035] Combine the following Figure 1-Figure 10 The multi-level lithology intelligent interpretation model construction method and its application provided by the present invention are described. It should be noted that the execution subject of the multi-level lithology intelligent interpretation model construction method and its application provided by the present invention can be any network side device / terminal side device that meets the technical requirements, such as a lithology interpretation device.

[0036] Figure 1-Figure 2The figure is a flow chart of the method for constructing a multi-level lithology intelligent interpretation model and its application provided by the present invention. Figure 1 The present invention provides a method for constructing a multi-level lithology intelligent interpretation model, which may include: Step S110, obtaining lithology data and original logging curves of the target rock area, wherein the lithology data of the target rock area is data of lithology types contained in the target rock area, and the target rock area includes composite sand bodies, single sand bodies and mud interlayers, thin reservoirs and physical interlayers.

[0037] In one embodiment, the target rock area may be a rock area with known core data and logging curve data, and the lithology data of the target rock area may be obtained based on the core data analysis. The target rock area in this embodiment includes five categories: mudstone, medium-fine sandstone, siltstone, conglomerate (containing mud gravel), and calcareous sandstone, among which only medium-fine sandstone is a good reservoir, and siltstone, conglomerate (containing mud gravel) and calcareous sandstone are all physical interlayers.

[0038] In this embodiment, the reservoir refers specifically to the oil and gas reservoir, which refers to the geological unit in the underground rock with the ability to store and flow oil and gas; in clastic rocks, the reservoir usually refers to sandstone with the ability to store and flow oil and gas. Interlayers are impermeable layers or weakly permeable rock layers inside the reservoir, which are relatively thin and can be divided into mud interlayers and physical interlayers. Lithology refers to the material composition and structural characteristics of rocks, reflecting the changes in different geological histories and conditions; in essence, reservoirs and interlayers are specific manifestations of different lithologies. This embodiment mainly distinguishes between mudstone, sandstone (permeable sandstone, i.e., reservoir) and interlayers (mud interlayers and physical interlayers). The nested nature of lithology is also called multi-level. In clastic rock formations, the rock layers have a certain thickness and are nested, that is, the thick composite sand body includes multiple single sand bodies and mudstone interlayers, and the single sand body includes multiple thin reservoirs and physical interlayers.

[0039] Step S120, preprocessing the original logging curve, wherein the preprocessing includes any one of the following items or any combination thereof: rock electrical relocation, outlier processing, and standardization. There is usually a certain systematic error between the core depth and the logging curve depth, and the core depth needs to be matched with the logging depth, i.e., rock electrical relocation. In addition, due to the model differences of the logging instrument, the formation and the measurement environment, etc., the domain value of the logging curve may have certain outliers. Preprocessing the logging curve can improve the accuracy of the original logging curve and help to achieve a detailed interpretation of the lithology later.

[0040] Step S130: Perform multi-level sub-wavelet decomposition according to the pre-processed original logging curve to obtain multiple wavelet components with different frequencies.

[0041] In one embodiment, step S130 can set the level of multi-level wavelet decomposition according to the thickness of the interlayer in the target rock area (the interlayer refers to a low-permeability or non-permeable rock layer existing inside the reservoir), and then perform multi-level wavelet decomposition on the preprocessed original logging curve according to the set multi-level wavelet decomposition level using Daubechies Wavelet basis function or Harr wavelet basis function. Specifically, when the thickness of the interlayer in the target rock area is less than a preset minimum threshold, the level of multi-level wavelet decomposition is set to secondary decomposition; when the thickness of the interlayer in the target rock area is between the preset minimum threshold and the preset maximum threshold, the level of multi-level wavelet decomposition is set to tertiary decomposition; when the thickness of the interlayer in the target rock area is greater than the preset maximum threshold, the level of multi-level wavelet decomposition is set to no decomposition.

[0042] Further, when the level of the multi-level wavelet decomposition is set to three-level decomposition, the preprocessed original logging curve is subjected to first-level wavelet decomposition to obtain a first wavelet component (low frequency) and a second wavelet component (high frequency), wherein the first wavelet component is a wavelet component within a first preset low frequency range, and the second wavelet component is a wavelet component within a first preset high frequency range, which is usually dominated by noise; the first wavelet low frequency component is subjected to second-level wavelet decomposition to obtain a third wavelet component and a fourth wavelet component, wherein the third wavelet low frequency component is a wavelet component within a second preset low frequency range, and the fourth wavelet component is a wavelet component within a second preset high frequency range, wherein the frequency of the third wavelet component is less than that of the first wavelet component; the third wavelet component is subjected to third-level wavelet decomposition to obtain a fifth wavelet component and a sixth wavelet component, wherein the fifth wavelet component is a wavelet component within a third preset low frequency range, and the sixth wavelet component is a wavelet component within a third preset high frequency range, wherein the frequency of the fifth wavelet component is less than that of the third wavelet component.

[0043] This embodiment uses db4 (Dobeshi wavelet function) to perform 2-3 level decomposition on the pre-processed original logging curve ( Figure 3 ): The first level decomposes the original logging curve to obtain cA1 (low frequency) and cD1 (high frequency); the second level decomposes cA1 obtained by the first level decomposition to obtain cA2 (low frequency) and cD2 (high frequency); the third level decomposes cA2 (low frequency) obtained by the second level decomposition to obtain cA3 (low frequency) and cD3 (high frequency). It should be noted that because cD1 is usually high-frequency noise, it should be deleted when the curve is reconstructed; compared with the two high-frequency information cD2 and cD3, cD2 has a higher frequency and is more suitable for predicting thin-layer lithology. Taking the GR logging curve of a well as an example, the results of the three-level decomposition are as follows: Figure 4 It should be noted that the level setting of multi-level wavelet decomposition (i.e., selecting level 2 or level 3 decomposition) needs to take the thickness of the thin layer into consideration. Level 2 decomposition is selected for particularly thin interlayers, and level 3 decomposition is selected for interlayers of moderate thickness. Figure 4Taking the GR curve shown as an example, for interlayers less than 50cm (preset minimum threshold), select 2-level decomposition; for interlayers between 50 and 150cm, select 3-level decomposition; when the thickness is greater than 1.5 meters (preset maximum threshold), generally no decomposition is required. In addition, the level of multi-level sub-wavelet decomposition can be set according to actual conditions, and it is possible to achieve 5-level decomposition.

[0044] Step S140: reconstruct the well logging curve according to the wavelet components of multiple different frequencies to obtain a first reconstructed well logging curve and a second reconstructed well logging curve.

[0045] In one embodiment, step S140 may reconstruct the well logging curve using the first reconstruction expression and the second reconstruction expression according to the plurality of wavelet components with different frequencies, to obtain the first reconstructed well logging curve and the second reconstructed well logging curve; wherein the first reconstruction expression is: , The second reconstruction expression is: , In the formula, represents the first reconstructed well log curve, represents the second reconstructed logging curve, , , , Represents the reconstruction coefficients of the first reconstruction expression 1, K1, 1, 0, where the value range of K1 is 1~5, , , Represents the reconstruction coefficients 1, K2, 0 of the second reconstruction expression, where the value range of K2 is 1~5, Respectively represent the third sub-wave component and the fifth sub-wave component, , , Represent the second sub-wave component, the fourth sub-wave component, and the sixth sub-wave component respectively. It is worth noting that the larger the K1 and K2 values ​​are, the higher the proportion of high-frequency information is, and the more sensitive the response to thin interlayers is, but it is not conducive to the identification of thick sand bodies. It should be optimized according to actual conditions. This example takes the GR logging curve of a certain area as an example. The reconstruction results of coefficient K selected from 1 to 5 are as follows Figure 5 As shown in the figure, when K1=3, the response to the thin interlayer inside the sand body is most obvious.

[0046] Step S150: Based on the lithological data of the target rock area and the original logging curve, the first reconstructed logging curve, and the second reconstructed logging curve, a multi-level lithological intelligent interpretation model is trained through an integrated algorithm to learn the mapping relationship between the original logging curve and the composite sand body, the first reconstructed logging curve and the single sand body and mud interlayer, and the second reconstructed logging curve and the thin reservoir and physical interlayer, so as to obtain the final multi-level lithological intelligent interpretation model.

[0047] In one embodiment, the integrated algorithm includes at least two of the following: LightGBM (lightweight gradient boosting learning algorithm), MLP (multi-layer perceptron algorithm), and Bagging-RF (out-of-stack random forest algorithm). Figure 6 and 7 , step S150 may include: A training set is constructed based on the lithology data of the target rock area and the original well logging curve, the first reconstructed well logging curve, and the second reconstructed well logging curve, wherein the composite sand body, single sand body and mud interlayer, thin reservoir and physical interlayer in the lithology data are used as labels of the original well logging curve, the first reconstructed well logging curve, and the second reconstructed well logging curve; Perform multi-label sample balanced sampling on the training set; According to the training set after balanced sampling of multi-label samples, the multi-level lithology intelligent interpretation model is trained to learn the mapping relationship between the original logging curve and the composite sand body, the first reconstructed logging curve and the single sand body and mud interlayer, and the second reconstructed logging curve and the thin reservoir and physical interlayer, so as to obtain the final multi-level lithology intelligent interpretation model.

[0048] Based on the different lithology data of the target rock area, the distribution of samples of different lithologies is analyzed. The large difference in the number of samples of different lithologies will lead to poor final training effect of the model, especially the low recognition accuracy of samples of relatively few lithology categories. Therefore, this embodiment performs balanced sampling processing on the number of samples of different lithologies (multi-label), the main purpose of which is to supplement the fewer lithology samples to the same level as the most lithology samples.

[0049] Specifically, this embodiment uses multi-algorithm integration to complete the training and optimization of the multi-level lithology intelligent interpretation model between lithology data and well logging curves (including original well logging curves, first reconstructed well logging curves, and second reconstructed well logging curves). This embodiment uses three types of algorithms: LightGBM (lightweight gradient boosting learner model), MLP (multi-layer perceptron model), and Bagging-RF (out-of-stack random forest model). These three single-class models can be defined as weak learners, which can be combined into strong learners through weighting, and each weak learner is based on the "gradient boosting tree" idea to achieve the construction and iteration of multi-level lithology intelligent interpretation models. In addition, this embodiment evaluates the credibility and reliability of the integrated model through cross-validation to ensure the performance of the model. The Bayesian algorithm is further used to optimize the model parameters. If the credibility is greater than 90%, the model is qualified. This embodiment also uses blind wells to test qualified training models to test the prediction accuracy and generalization ability of the model. If the accuracy of the blind well reaches 85%, the model accuracy is qualified, otherwise the model optimization continues. After the final multi-level lithology intelligent interpretation model is obtained, it can be applied to the identification of different levels and lithologies.

[0050] An application of a multi-level lithological intelligent interpretation model can form a multi-level lithological intelligent interpretation method. Specifically, after the multi-level lithological intelligent interpretation model is trained, the original logging curve of the rock area to be interpreted sent by at least one terminal can be received, and then the original logging curve of the rock area to be interpreted is preprocessed, multi-level sub-wavelet decomposition, and logging curve reconstruction are performed according to the above-mentioned multi-level lithological intelligent interpretation model construction method to obtain the first reconstructed logging curve and the second reconstructed logging curve of the rock area to be interpreted, and then the original logging curve, the first reconstructed logging curve and the second reconstructed logging curve of the rock area to be interpreted are input into the above-mentioned multi-level lithological intelligent interpretation model to obtain the lithological interpretation results of the rock area to be interpreted. Among them, the multi-level lithology intelligent interpretation model will first obtain the initial lithology interpretation results corresponding to each algorithm according to the original logging curve, the first reconstructed logging curve and the second reconstructed logging curve of the rock area to be interpreted, that is, multiple initial lithology interpretation results, and then combine the preset algorithm weights of each algorithm (the preset algorithm weights can be set according to actual conditions) to obtain the final lithology interpretation results of the rock area to be interpreted; finally, the lithology interpretation results of the rock area to be interpreted are sent back to at least one terminal. An example of a hierarchical nested lithology interpretation result output by the multi-level lithology intelligent interpretation model is shown in the figure below: Figure 8 As shown, the logging interpretation results of the composite sand body, the logging interpretation results of the single sand body and mud interlayer inside the composite sand body, and the logging interpretation results of the thin reservoir and physical interlayer inside the single sand body are shown in sequence.

[0051] The invention provides a method for constructing a multi-level lithology intelligent interpretation model and its application, which can at least bring the following beneficial effects: In view of the problem of low resolution of logging curves and the characteristics of multi-level nested distribution of rock formations, the present invention fully considers the hierarchical nature of lithological interpretation, and performs multi-level wavelet decomposition and logging curve reconstruction in the frequency domain according to the level of multi-level sub-wavelet decomposition determined by the rock formation thickness, thereby obtaining a new logging curve suitable for the interpretation of rock formations of different thicknesses, which can realize nested interpretation of lithologies of different thicknesses and greatly improve the accuracy of logging interpretation.

[0052] The present invention couples multi-level sub-wavelet decomposition and logging curve reconstruction with hierarchical nested lithologic interpretation, and proposes a "reservoir and thin interlayer hierarchical nested logging interpretation method": at the first level, the original logging curve is preferably used to interpret the thick composite sand body; at the second level, the first reconstructed logging curve is preferably used to identify the single sand body and mud interlayer inside the composite sand body; at the third level, the second reconstructed logging curve is preferably used to identify the thin reservoir and physical interlayer inside the single sand body, so as to realize the fine logging interpretation of lithologies of different thicknesses, especially the accurate identification of lithologies with centimeter-level thickness, thereby greatly improving the accuracy of logging interpretation.

[0053] The present invention performs multi-label sample balanced sampling processing on the training data of the multi-level lithology intelligent interpretation model, so that the number of samples of different lithologies reaches the same order of magnitude, which can improve the prediction accuracy of the multi-level lithology intelligent interpretation model and avoid the impact of obvious differences in the number of samples of different lithologies on model training due to the limitation of the number of coring wells.

[0054] The mapping relationship between lithology of different levels (thickness) and logging curves is complex and difficult to identify. A single algorithm model cannot meet the requirements for lithology identification of different levels (especially identification of lithology of centimeter-level thickness). The present invention uses an integrated algorithm to train a multi-level lithology intelligent interpretation model to learn the mapping relationship between the original logging curve and the composite sand body, the first reconstructed logging curve and the single sand body and mud interlayer, and the second reconstructed logging curve and the thin reservoir and physical interlayer, thereby achieving "complementary advantages" between different algorithms, significantly improving the accuracy of lithology identification of different levels, and effectively improving the generalization ability of the multi-level lithology intelligent interpretation model.

[0055] The multi-level lithology intelligent interpretation system provided by the present invention is described below. The multi-level lithology intelligent interpretation system described below and the multi-level lithology intelligent interpretation method described above can be referenced to each other.

[0056] Reference Fig. 9 The present invention provides a multi-level lithology intelligent interpretation system, which may include: The data receiving module is used to: receive the original logging curve of the rock area to be interpreted sent by at least one terminal; A data processing module is used to: pre-process the original well logging curve of the rock area to be interpreted, perform multi-level sub-wavelet decomposition, and reconstruct the well logging curve according to any of the above-mentioned multi-level sub-lithology intelligent interpretation model construction methods, so as to obtain a first reconstructed well logging curve and a second reconstructed well logging curve of the rock area to be interpreted; A lithology interpretation module, for obtaining a lithology interpretation result of the rock area to be interpreted according to the original well logging curve, the first reconstructed well logging curve and the second reconstructed well logging curve of the rock area to be interpreted and the multi-level lithology intelligent interpretation model obtained by any of the multi-level lithology intelligent interpretation model construction methods described above; The data output module is used to send the lithology interpretation result of the rock area to be interpreted back to at least one terminal.

[0057] Fig.10 An example of a physical structure diagram of an electronic device is shown in FIG. Fig.10 As shown, the electronic device may include: a processor 810, a communications interface 820, a memory 830 and a communication bus 840, wherein the processor 810, the communications interface 820 and the memory 830 communicate with each other via the communication bus 840. The processor 810 may call the logic instructions in the memory 830 to execute the multi-level lithology intelligent interpretation method, which includes: Receiving an original well logging curve of a rock area to be interpreted sent by at least one terminal; According to any of the above-mentioned methods for constructing a multi-level lithology intelligent interpretation model, the original well logging curve of the rock area to be interpreted is preprocessed, multi-level sub-wavelet decomposed, and the well logging curve is reconstructed to obtain a first reconstructed well logging curve and a second reconstructed well logging curve of the rock area to be interpreted; According to the original well logging curve, the first reconstructed well logging curve and the second reconstructed well logging curve of the rock area to be interpreted, the multi-level lithology intelligent interpretation model obtained by the multi-level lithology intelligent interpretation model construction method described in any one of the above items is used to obtain the lithology interpretation result of the rock area to be interpreted; The lithology interpretation result of the rock area to be interpreted is sent back to at least one terminal.

[0058] In addition, the logic instructions in the above-mentioned memory 830 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0059] On the other hand, the present invention further provides a computer program product, the computer program product comprising a computer program, the computer program can be stored in a non-transitory computer-readable storage medium, when the computer program is executed by a processor, the computer can execute the multi-level sub-lithology intelligent interpretation method provided by the above methods, the method comprising: Receiving an original well logging curve of a rock area to be interpreted sent by at least one terminal; According to any of the above-mentioned methods for constructing a multi-level lithology intelligent interpretation model, the original well logging curve of the rock area to be interpreted is preprocessed, multi-level sub-wavelet decomposed, and the well logging curve is reconstructed to obtain a first reconstructed well logging curve and a second reconstructed well logging curve of the rock area to be interpreted; According to the original well logging curve, the first reconstructed well logging curve and the second reconstructed well logging curve of the rock area to be interpreted, the multi-level lithology intelligent interpretation model obtained by the multi-level lithology intelligent interpretation model construction method described in any one of the above items is used to obtain the lithology interpretation result of the rock area to be interpreted; The lithology interpretation result of the rock area to be interpreted is sent back to at least one terminal.

[0060] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented when the computer program is executed by a processor to execute the multi-level sub-lithology intelligent interpretation method provided by the above methods, the method comprising: Receiving an original well logging curve of a rock area to be interpreted sent by at least one terminal; According to any of the above-mentioned methods for constructing a multi-level lithology intelligent interpretation model, the original well logging curve of the rock area to be interpreted is preprocessed, multi-level sub-wavelet decomposed, and the well logging curve is reconstructed to obtain a first reconstructed well logging curve and a second reconstructed well logging curve of the rock area to be interpreted; According to the original well logging curve, the first reconstructed well logging curve and the second reconstructed well logging curve of the rock area to be interpreted, the multi-level lithology intelligent interpretation model obtained by the multi-level lithology intelligent interpretation model construction method described in any one of the above items is used to obtain the lithology interpretation result of the rock area to be interpreted; The lithology interpretation result of the rock area to be interpreted is sent back to at least one terminal.

[0061] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0062] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0063] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for constructing a multi-level lithology intelligent interpretation model, characterized in that: include: Obtain lithology data and original logging curves of the target rock area, where the target rock area includes composite sand bodies, single sand bodies and mud interlayers, thin reservoirs and physical interlayers; Preprocess the original logging curve; According to the pre-processed original logging curve, multi-level sub-wavelet decomposition is performed to obtain multiple sub-wavelet components with different frequencies; Reconstructing the well logging curve according to the wavelet components of multiple different frequencies to obtain a first reconstructed well logging curve and a second reconstructed well logging curve; According to the lithological data of the target rock area as well as the original logging curve, the first reconstructed logging curve and the second reconstructed logging curve, the multi-level lithological intelligent interpretation model is trained through an integrated algorithm to learn the mapping relationship between the original logging curve and the composite sand body, the first reconstructed logging curve and the single sand body and mud interlayer, and the second reconstructed logging curve and the thin reservoir and physical interlayer, so as to obtain the final multi-level lithological intelligent interpretation model.

2. The method for constructing a multi-level lithology intelligent interpretation model according to claim 1, characterized in that: The method of performing multi-level sub-wavelet decomposition according to the pre-processed original logging curve to obtain multiple sub-wavelet components with different frequencies includes: The order of multi-level wavelet decomposition is set according to the thickness of the interlayer in the target rock area; According to the set multi-level sub-wavelet decomposition level, the pre-processed original logging curve is subjected to multi-level sub-wavelet decomposition; Preferably, the step of setting the levels of the multi-level wavelet decomposition according to the thickness of the interlayer in the target rock area comprises: When the thickness of the interlayer in the target rock area is less than the preset minimum threshold, the level of the multi-level wavelet decomposition is set to the second-level decomposition; When the thickness of the interlayer in the target rock area is between a preset minimum threshold and a preset maximum threshold, the level of the multi-level wavelet decomposition is set to three-level decomposition; When the thickness of the interlayer in the target rock area is greater than a preset maximum threshold, the level of the multi-level wavelet decomposition is set to no decomposition required.

3. The method for constructing a multi-level lithology intelligent interpretation model according to claim 2, characterized in that: The step of performing multi-level wavelet decomposition on the pre-processed original logging curve according to the set level of multi-level wavelet decomposition includes: Performing a first-level wavelet decomposition on the preprocessed original logging curve to obtain a first wavelet component and a second wavelet component, wherein the first wavelet component is a wavelet component within a first preset low-frequency range, and the second wavelet component is a wavelet component within a first preset high-frequency range; Performing secondary wavelet decomposition on the first wavelet component to obtain a third wavelet component and a fourth wavelet component, wherein the second wavelet component is a wavelet component within a second preset low frequency range, and the fourth wavelet component is a wavelet component within a second preset high frequency range, wherein the frequency of the third wavelet component is lower than that of the first wavelet component; The third wavelet component is subjected to three-level wavelet decomposition to obtain a fifth wavelet component and a sixth wavelet component, wherein the fifth wavelet component is a wavelet component within a third preset low frequency range, and the sixth wavelet component is a wavelet component within a third preset high frequency range, wherein the frequency of the fifth wavelet component is less than that of the third wavelet component.

4. The method for constructing a multi-level lithology intelligent interpretation model according to claim 3, characterized in that: The method of reconstructing the well logging curve according to the wavelet components of multiple different frequencies to obtain a first reconstructed well logging curve and a second reconstructed well logging curve includes: Reconstructing the well logging curve using a first reconstruction expression according to a plurality of wavelet components with different frequencies to obtain a first reconstructed well logging curve; Reconstructing the well logging curve using a second reconstruction expression according to a plurality of wavelet components with different frequencies to obtain a second reconstructed well logging curve; Among them, the first reconstruction expression is: , The second reconstruction expression is: , In the formula, represents the first reconstructed well log curve, represents the second reconstructed logging curve, , , , Represents the reconstruction coefficients of the first reconstruction expression 1, K1, 1, 0, where the value range of K1 is 1~5, , , Represents the reconstruction coefficients 1, K2, 0 of the second reconstruction expression, where the value range of K2 is 1~5, Respectively represent the third sub-wave component and the fifth sub-wave component, , , They represent the second wavelet component, the fourth wavelet component, and the sixth wavelet component respectively.

5. The method for constructing a multi-level lithology intelligent interpretation model according to claim 4, characterized in that: The integrated algorithm includes at least two of the following: lightweight gradient boosting learning algorithm, multi-layer perceptron algorithm, and off-stack random forest algorithm. The integrated algorithm is used to train a multi-level lithology intelligent interpretation model based on the lithology data of the target rock area and the original well logging curve, the first reconstructed well logging curve, and the second reconstructed well logging curve to learn the mapping relationship between the original well logging curve and the composite sand body, the first reconstructed well logging curve and the single sand body and the mud interlayer, and the second reconstructed well logging curve and the thin reservoir and the physical interlayer, so as to obtain the final multi-level lithology intelligent interpretation model, including: A training set is constructed based on the lithology data of the target rock area and the original well logging curve, the first reconstructed well logging curve, and the second reconstructed well logging curve, wherein the composite sand body, single sand body and mud interlayer, thin reservoir and physical interlayer in the lithology data are used as labels of the original well logging curve, the first reconstructed well logging curve, and the second reconstructed well logging curve; Perform multi-label sample balanced sampling on the training set; According to the training set after balanced sampling of multi-label samples, the multi-level lithology intelligent interpretation model is trained to learn the mapping relationship between the original logging curve and the composite sand body, the first reconstructed logging curve and the single sand body and mud interlayer, and the second reconstructed logging curve and the thin reservoir and physical interlayer, so as to obtain the final multi-level lithology intelligent interpretation model.

6. A multi-level lithology intelligent interpretation method, characterized in that: include: Receiving an original well logging curve of a rock area to be interpreted sent by at least one terminal; According to the method for constructing a multi-level lithology intelligent interpretation model according to any one of claims 1 to 5, the original well logging curve of the rock area to be interpreted is preprocessed, multi-level sub-wavelet decomposition is performed, and the well logging curve is reconstructed to obtain a first reconstructed well logging curve and a second reconstructed well logging curve of the rock area to be interpreted; According to the original well logging curve, the first reconstructed well logging curve and the second reconstructed well logging curve of the rock area to be interpreted, the multi-level lithology intelligent interpretation model obtained by the multi-level lithology intelligent interpretation model construction method according to any one of claims 1 to 5 is used to obtain the lithology interpretation result of the rock area to be interpreted; The lithology interpretation result of the rock area to be interpreted is sent back to at least one terminal.

7. The multi-level lithology intelligent interpretation method according to claim 6, characterized in that: The multi-level lithology intelligent interpretation model obtained by the multi-level lithology intelligent interpretation model construction method according to any one of claims 1 to 5 based on the original well logging curve, the first reconstructed well logging curve and the second reconstructed well logging curve of the rock area to be interpreted, obtains the lithology interpretation result of the rock area to be interpreted, including: According to the original logging curves, the first reconstructed logging curves and the second reconstructed logging curves of the rock area to be interpreted, multiple initial lithological interpretation results are obtained through each item in the integrated algorithm in the multi-level lithological intelligent interpretation model, and combined with the preset algorithm weights, the final lithological interpretation results of the rock area to be interpreted are obtained.

8. A multi-level lithology intelligent interpretation system, characterized in that: include: The data receiving module is used to: receive the original logging curve of the rock area to be interpreted sent by at least one terminal; A data processing module is used for: preprocessing, multi-level sub-wavelet decomposition, and logging curve reconstruction of the original well logging curve of the rock area to be interpreted according to the multi-level sub-lithology intelligent interpretation model construction method according to any one of claims 1 to 5, so as to obtain a first reconstructed well logging curve and a second reconstructed well logging curve of the rock area to be interpreted; A lithology interpretation module, for obtaining a lithology interpretation result of the rock area to be interpreted based on the original well logging curve, the first reconstructed well logging curve and the second reconstructed well logging curve of the rock area to be interpreted and the multi-level lithology intelligent interpretation model obtained by the multi-level lithology intelligent interpretation model construction method according to any one of claims 1 to 5; The data output module is used to send the lithology interpretation result of the rock area to be interpreted back to at least one terminal.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the multi-level sub-lithology intelligent interpretation model construction method as described in any one of claims 1 to 5 and / or the multi-level sub-lithology intelligent interpretation method as described in claim 6 or 7 is implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for constructing a multi-level lithology intelligent interpretation model as described in any one of claims 1 to 5 and / or the method for constructing a multi-level lithology intelligent interpretation as described in claim 6 or 7 is implemented.

Images