Deep learning curve environment correction method and system based on multiple regression samples

By dividing the research wells into high-quality wells and using multiple regression models and deep learning technology to process and correct the data of high-quality wells, the curve distortion problem in the well logging data admission environment is solved, and the measurement is achieved Intelligent correction of well curves and fine exploration evaluation of complex lithologic reservoirs.

CN119003967APending Publication Date: 2024-11-22PETROCHINA CO LTD
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202310552135.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2023-05-16
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

Under the complex conditions of the logging data admission environment, the logging data caused by factors such as diameter expansion have serious distortion problems in the long section. Deep learning and other methods cannot be used for curve environmental correction due to the lack of samples, which in turn affects the logging formation evaluation and Research on well earthquake calibration and other research.

Method used

By dividing the research well into high-quality wells and low-quality wells according to the well diameter curve and rock physics trends, the data of high-quality wells are processed, deep learning sample data are constructed, high-quality well curve database is integrated and generated, and high-quality wells are used as high-quality wells The curve library sample trains and verifies the deep learning model, and the logging data of low-quality wells are input into the deep learning model for logging curve prediction and correction. The specific steps include multivariate regression fitting of high-quality layer segment data of high-quality wells, correction and optimization of distorted layer segment data, and multi-well consistency processing through general matching technology. and curve normalization.

Benefits of technology

Intelligent correction of logging curves in complex logging admission environments is achieved, and serious distortion problem of long curves is solved, providing technical support for the exploration and evaluation of complex lithologic reservoirs and the detailed description of oil and gas reservoirs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119003967A_ABST
    Figure CN119003967A_ABST
Patent Text Reader

Abstract

The invention belongs to the technical field of oil-gas exploration, and particularly relates to a deep learning curve environment correction method and system based on multiple regression samples. The method comprises the following steps: dividing a research well into a high-quality well and a low-quality well according to a well diameter curve and a rock physical trend; processing the data of the high-quality well, and constructing deep learning sample data; processing the deep learning sample data of each high-quality well, and integrating to generate a high-quality well curve library; training and verifying the deep learning model by using a high-quality well curve storage location sample; logging data of a low-quality well are input into the deep learning model for logging curve prediction; and correcting the data of the low-quality well according to the prediction result. According to the method, the problems of serious distortion of a curve long well section, lack of training samples of a deep learning model and the like caused by factors such as serious diameter expansion are solved, intelligent correction of the well logging curve in a complex well logging environment is realized, and technical support is provided for exploration evaluation of complex lithologic reservoirs and fine description of oil and gas reservoirs.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of oil and gas exploration, and in particular relates to a deep learning curve environment correction method and system based on multivariate regression samples. Background Art

[0002] Under the complex conditions of logging data acquisition environment, the logging data has serious distortion problems in long well sections due to factors such as diameter expansion. Methods such as deep learning cannot be used in curve environment correction scenarios due to lack of samples, which makes it impossible to directly use the measured data for research work such as logging formation evaluation and well seismic calibration. Summary of the invention

[0003] In view of the above problems, the present invention provides a deep learning curve environment correction method based on multivariate regression samples, the method comprising:

[0004] The study wells were divided into high-quality wells and low-quality wells based on the wellbore curves and rock physical trends;

[0005] Process the data of high-quality wells and construct deep learning sample data;

[0006] Process the deep learning sample data of each high-quality well and integrate them to generate a high-quality well curve library;

[0007] The deep learning model is trained and validated with high-quality well curve location samples;

[0008] The logging data of low-quality wells are input into the deep learning model for logging curve prediction;

[0009] The data of low-quality wells are corrected based on the prediction results.

[0010] Preferably, the data of the high-quality well includes high-quality layer interval data and distorted layer interval data.

[0011] Preferably, the processing of the data of high-quality wells includes:

[0012] Perform multiple regression fitting on high-quality interval data of high-quality wells;

[0013] Correct and optimize the distorted layer data of high-quality wells.

[0014] Preferably, the distorted layer data of the high-quality wells are corrected and optimized, including: sonic time difference data correction, neutron porosity data correction and density data correction.

[0015] Preferably, the acoustic wave time difference data correction comprises the following steps:

[0016] Using high-quality layer data, a multivariate regression model for the acoustic time difference curve is established;

[0017] The acoustic time difference curve is synthesized by using the multivariate regression model of the acoustic time difference curve to correct the distorted layer interval acoustic time difference data.

[0018] Preferably, the neutron porosity data correction comprises the following steps:

[0019] Using high-quality interval data, a multivariate regression model for neutron porosity curves is established;

[0020] The neutron porosity curve is synthesized by using the multivariate regression model of the neutron porosity curve to correct the distorted neutron porosity data.

[0021] Preferably, the density data correction comprises the following steps:

[0022] Using high-quality layer data, a multivariate regression model for density curves is established;

[0023] The multi-regression model of density curve is used to synthesize density curve and correct the distorted layer density data.

[0024] Preferably, the processing of the deep learning sample data of each high-quality well includes:

[0025] The non-geological data differences in deep learning sample data are removed through multi-well consistency processing using pattern matching technology;

[0026] The logging curves of each high-quality well are normalized.

[0027] Preferably, the deep learning model is trained and verified with high-quality well curve location samples, including: constructing a deep learning model structure and determining machine learning parameters.

[0028] Preferably, the data of the low-quality wells are corrected according to the prediction results, specifically by replacing the distorted curve of the low-quality wells with the predicted curve.

[0029] The present invention also proposes a deep learning curve environment correction system based on multivariate regression samples, the system comprising a classification module, a sample generation module, a data processing module, a model verification module, a prediction module and a correction module;

[0030] Wherein, the classification module is used to classify the studied wells into high-quality wells and low-quality wells according to the wellbore curve and the rock physical trend;

[0031] The sample generation module is used to process the data of high-quality wells and construct deep learning sample data;

[0032] The data processing module is used to process the deep learning sample data of each high-quality well and integrate and generate a high-quality well curve library;

[0033] The model verification module is used to train and verify the deep learning model with high-quality well curve storage location samples;

[0034] The prediction module is used to input the logging data of low-quality wells into the deep learning model to predict the logging curve;

[0035] The correction module is used to correct the data of low-quality wells according to the prediction results.

[0036] Preferably, the data of the high-quality well includes high-quality layer interval data and distorted layer interval data.

[0037] Preferably, the sample generation module is used to process the data of high-quality wells, including:

[0038] Perform multiple regression fitting on high-quality interval data of high-quality wells;

[0039] Correct and optimize the distorted layer data of high-quality wells.

[0040] Preferably, the distorted layer data of the high-quality wells are corrected and optimized, including: sonic time difference data correction, neutron porosity data correction and density data correction.

[0041] Preferably, the data processing module is used to process the deep learning sample data of each high-quality well, including:

[0042] The non-geological data differences in deep learning sample data are removed through multi-well consistency processing using pattern matching technology;

[0043] The logging curves of each high-quality well are normalized.

[0044] Preferably, the model verification module is used to train and verify the deep learning model with high-quality well curve location samples, including: constructing a deep learning model structure and determining machine learning parameters.

[0045] Preferably, the correction module is used to correct the data of the low-quality well according to the prediction result, specifically: replacing the distorted curve of the low-quality well with the prediction curve.

[0046] The present invention has the following beneficial effects:

[0047] The present invention divides the research wells into high-quality wells and low-quality wells according to the well diameter curve and rock physical trend, processes the data of the high-quality wells, constructs deep learning sample data, processes the deep learning sample data of each high-quality well, integrates and generates a high-quality well curve library, trains and verifies the deep learning model with the high-quality well curve library samples, inputs the logging data of the low-quality wells into the deep learning model for logging curve prediction, and corrects the data of the low-quality wells according to the prediction results, thereby solving the problems of serious distortion of long well sections of the curve caused by factors such as serious diameter expansion and lack of training samples for the deep learning model, and realizes intelligent correction of logging curves in a complex logging acquisition environment, providing technical support for the exploration and evaluation of complex lithology reservoirs and the detailed description of oil and gas reservoirs.

[0048] Other features and advantages of the present invention will be described in the following description, and partly become obvious from the description, or be understood by implementing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures pointed out in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. 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.

[0050] Figure 1 A diagram showing a method for deep learning curve environment correction based on multivariate regression samples in an embodiment of the present invention is shown;

[0051] Figure 2a An example diagram of low-quality wells in the well quality classification of a certain area in the Tarim Basin according to an embodiment of the present invention is shown;

[0052] Figure 2b An example diagram of high-quality wells in the well quality classification of a certain area in the Tarim Basin according to an embodiment of the present invention is shown;

[0053] Figure 3 A flow chart of high-quality well environment correction based on a multivariate regression method in an embodiment of the present invention is shown;

[0054] Figure 4 An example diagram of high-quality well environment correction based on a multiple regression method in an embodiment of the present invention is shown;

[0055] Figure 5 An example diagram of deep learning sample consistency processing in an embodiment of the present invention is shown;

[0056] Figure 6A schematic diagram of the structure of a machine learning model in an embodiment of the present invention is shown;

[0057] Figure 7 A diagram showing the acoustic time difference training of the logging curve deep learning model in an embodiment of the present invention is shown;

[0058] Figure 8 A neutron porosity training diagram of a well logging curve deep learning model in an embodiment of the present invention is shown;

[0059] Fig. 9 A density training diagram of a deep learning model for well logging curves according to an embodiment of the present invention is shown;

[0060] Fig.10 A diagram showing the prediction and correction of acoustic wave time difference of a logging curve deep learning model in an embodiment of the present invention is shown;

[0061] Fig.11 The prediction and correction of neutron porosity in the deep learning model of well logging curves in an embodiment of the present invention are shown;

[0062] Fig.12 The density prediction and correction of the deep learning model of the well logging curve in the embodiment of the present invention are shown;

[0063] Fig.13 A diagram of a deep learning curve environment correction system based on multivariate regression samples in an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0064] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. 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.

[0065] Through the quality analysis of well logging data, the well logging data are divided into low-quality wells and high-quality wells. The multivariate regression model is used to optimize the correction of single-well curves for high-quality wells. The multi-well consistency processing and curve normalization processing techniques are further combined to realize deep learning sample processing. The deep learning model obtained by training high-quality wells as samples is used for curve correction of low-quality wells, realizing the curve environment correction of wells with severe distortion in long well sections, which is of great significance for exploration and development in complex logging data acquisition environments.

[0066] like Figure 1 As shown, the present invention proposes a deep learning curve environment correction method based on multivariate regression samples, the method comprising:

[0067] (1) Based on the wellbore curve and rock physics trend, the research wells are divided into high-quality wells and low-quality wells.

[0068] like Figure 2a and Figure 2b As shown, Figure 2a This is a typical low-quality well in a certain area of ​​the Tarim Basin. Figure 2b It is a typical high-quality well in a certain area of ​​the Tarim Basin. Figure 2a and Figure 2b In the figure, the first column on the right is the borehole diameter curve, which reflects the expansion and contraction of the borehole. Usually, the degree of curve distortion is proportional to the degree of expansion and contraction and the roughness of the borehole diameter curve. The three columns on the left are the lithology curve column, the depth curve column and the electrical curve column. Figure 2a and Figure 2b The fourth to sixth columns on the left side are the acoustic time difference curve, neutron porosity curve and density curve of the three porosity curves. Figure 2a and Figure 2b The larger the diameter expansion, the greater the difference between the porosity curve and the lithology and electrical curves, and the more serious the distortion. Figure 2b For the high-quality wells represented by the curve, the distorted intervals caused by limited expansion are short, and there are a large number of non-distorted intervals. Figure 2a The low-quality wells represented by the data have long well section distortion problems and no long well section non-distorted layer sections. For high-quality wells, the non-expanded layer section data can be used to establish a multivariate regression relationship to achieve the prediction and correction of the expanded layer section data; for low-quality wells, due to the lack of non-expanded layer sections, it is impossible to train data models such as multivariate regression, and correction is extremely difficult.

[0069] (2) Processing high-quality well data to construct deep learning sample data

[0070] Different curves have different measurement depths. The shallower the logging curve, the more easily it is affected by the logging environment. According to the measurement depth and diameter expansion, the quality of the corrected acoustic time difference curve (DT) is usually better than that of the neutron porosity curve (CNL), and the quality of the neutron porosity curve (CNL) is better than that of the density curve (DEN). In order to increase the stability of the multivariate regression model, the following design is used: Figure 3 The iterative correction process shown is that for each high-quality well, the acoustic time difference curve (AC), neutron porosity curve (CNL) and density curve (DEN) are corrected in sequence according to the idea of ​​iterative correction.

[0071] The processing of the data of the high-quality wells includes: performing multivariate regression fitting on the high-quality layer data of the high-quality wells; and correcting and optimizing the distorted layer data of the high-quality wells.

[0072] Correction and optimization of distorted layer data of high-quality wells, including: sonic time difference data correction, neutron porosity data correction and density data correction.

[0073] The acoustic time difference data correction includes the following steps: using high-quality layer data, establishing a multivariate regression model for the acoustic time difference curve:

[0074]

[0075] in, is the regression coefficient calculated by multivariate regression of the acoustic time difference curve AC, GR is the natural gamma curve, RT is the resistivity curve, AC syn is the acoustic time difference curve synthesized by multiple regression, and log is the logarithmic operator with base 10;

[0076] The acoustic time difference curve is synthesized by using the multivariate regression model of the acoustic time difference curve to correct the distorted layer interval acoustic time difference data:

[0077] if CAL>n and AC syn <AC,AC cor =AC syn ; else AC cor =AC;

[0078] Among them, CAL is the wellbore curve, AC is the measured acoustic wave time difference curve, AC syn The time difference curve of the acoustic wave synthesized by multiple regression, AC cor is the corrected acoustic time difference curve, n is the diameter expansion threshold;

[0079] The neutron porosity data correction includes the following steps: using high-quality interval data, establishing a multivariate regression model for the neutron porosity curve:

[0080]

[0081] in, is the regression coefficient calculated by CNL multivariate regression of neutron porosity curve, GR is the natural gamma curve, RT is the resistivity curve, AC cor is the corrected acoustic time difference curve, CNL syn is the neutron porosity curve synthesized by multiple regression, and log is the logarithmic operator with base 10;

[0082] The neutron porosity curve is synthesized by using the multivariate regression model of the neutron porosity curve to correct the distorted neutron porosity data:

[0083] if CAL>n and CNL syn <CNL,CNL cor=CNL syn ; else CNL cor =CNL;

[0084] Among them, CAL is the wellbore curve, CNL is the measured neutron porosity curve, and CNL syn The neutron porosity curve synthesized by multivariate regression, CNL cor is the corrected neutron porosity curve, n is the expansion threshold;

[0085] The density data correction includes the following steps: using high-quality layer data, establishing a multivariate regression model for the density curve:

[0086]

[0087] in, is the regression coefficient calculated by multivariate regression of density curve DEN, GR is the natural gamma curve, RT is the resistivity curve, AC cor is the corrected acoustic time difference curve, CNL cor is the corrected neutron porosity curve, DEN syn is the density curve synthesized by multiple regression, and log is the logarithmic operator with base 10;

[0088] The multi-distance regression model of the density curve is used to synthesize the density curve and correct the distorted layer density data:

[0089] if CAL>n and DEN syn <DEN,DEN cor =DEN syn ; else DEN cor =DEN;

[0090] Among them, CAL is the well diameter curve, DEN is the measured density curve, and DEN syn is the density curve synthesized by multiple regression, DEN cor is the corrected density curve, and n is the diameter expansion threshold.

[0091] Figure 4This is an example of the environmental correction effect of a high-quality well based on the multivariate regression method. In the figure, the first to third columns on the left are the lithology curve column, depth curve column and electrical curve column, and the first column on the right is the wellbore curve. The fourth to ninth columns on the left are the forward acoustic time difference curve (solid line) of the multivariate regression model, the corrected acoustic time difference curve (solid line), the forward neutron porosity curve (solid line) of the multivariate regression model, the corrected neutron porosity curve (solid line), the forward neutron porosity curve (solid line) of the multivariate regression model, the corrected neutron porosity curve (solid line), the forward density curve (solid line) of the multivariate regression model, and the corrected density curve (solid line). By comparing with the measured curves (dashed lines) in each curve channel, the forward modeling and correction data of the non-expanded layer section are highly consistent with the measured data, ensuring the effectiveness of the model prediction; the forward modeling and correction data of the expanded layer section effectively overcome the curve anomaly caused by distortion, ensuring the predictability of the model prediction.

[0092] (3) Process the deep learning sample data of each high-quality well and integrate them to generate a high-quality well curve library

[0093] The processing of the deep learning sample data of each high-quality well includes: removing non-geological factor data differences in the deep learning sample data through multi-well consistency processing using a profile matching technology; and normalizing the logging curves of each high-quality well.

[0094] For deep learning applications, the consistency and normalization of samples and target data are the prerequisites for deep learning model training and application. For the logging curve environmental correction scenario, the purpose of consistency processing is to remove the differences between wells caused by non-geological factors, and the purpose of normalization processing is to process the dynamic range of different curves to be relatively uniform (such as 0 to 1). In this patent, the consistency processing sampling pattern matching technology of multi-well logging data has the following effects: Figure 5 As shown. The normalization process uses a processing method based on the maximum and minimum linear mapping:

[0095]

[0096] Among them, x is the measured value, x min is the measured minimum value, x max is the maximum value measured, x Nor The result after processing.

[0097] (4) Training and validating the deep learning model with high-quality well curve location samples

[0098] The training of the deep learning model of well logging curves mainly includes the construction of the deep learning model structure and the determination of machine learning parameters. Figure 6As shown in the figure, in this step, the machine learning model structure used includes input layer, LSTM layer (long short-term memory), fully connected layer, pooling layer and output layer. The machine learning model parameters are solved using the Adam (Adaptive moment estimation) method. Figure 7 to Figure 9 This method targets Figure 4 The acoustic time difference curve, neutron porosity curve and density curve training examples of the sample well. The dotted line in the second column on the right side of the figure is the original curve after consistency processing, the dotted line in the first column on the right side is the multivariate regression model correction curve after consistency processing, and the solid line curves in the second and first columns on the right side are deep learning prediction curves. It can be seen from the figure that the solid line predicted by the deep learning model is highly consistent with the target curve after correction by the multivariate regression model, which verifies the predictive and generalization ability of the model.

[0099] (5) Input the logging data of low-quality wells into the deep learning model for logging curve prediction

[0100] The logging curve deep learning model established in the first five steps is applied to the same Figure 2a Examples of low to medium quality wells. Figure 10 to Figure 12 The following are examples of the correction of the acoustic time difference curve, neutron porosity curve and density curve of a low-quality well. The dotted line and solid line in the first column on the right side of the figure are the original curve after consistency processing and the curve after deep learning correction, respectively.

[0101] (6) Correct the data of low-quality wells based on the prediction results

[0102] For the problem of long well section curve distortion caused by severe diameter expansion, the deep learning correction results can effectively correct the problematic curve and restore the true response of the formation.

[0103] The correction of the data of the low-quality well according to the prediction result is specifically: replacing the distorted curve of the low-quality well with the prediction curve.

[0104] like Fig.13 As shown, the present invention also proposes a deep learning curve environment correction system based on multivariate regression samples, and the system includes a classification module, a sample generation module, a data processing module, a model verification module, a prediction module and a correction module.

[0105] The classification module is used to classify the studied wells into high-quality wells and low-quality wells according to the wellbore curves and the rock physical trends.

[0106] The sample generation module is used to process the data of high-quality wells and construct deep learning sample data.

[0107] The data processing module is used to process the deep learning sample data of each high-quality well and integrate them to generate a high-quality well curve library.

[0108] The model verification module is used to train and verify the deep learning model with high-quality well curve location samples.

[0109] The prediction module is used to input the logging data of low-quality wells into the deep learning model to predict the logging curve.

[0110] The correction module is used to correct the data of low-quality wells according to the prediction results.

[0111] Those skilled in the art should understand that although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible to modify the technical solutions described in the aforementioned embodiments, or to make equivalent replacements for some of the technical features therein; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A deep learning curve environment correction method based on multivariate regression samples, characterized in that: The method comprises: The study wells were divided into high-quality wells and low-quality wells based on the wellbore curves and rock physical trends; Process the data of high-quality wells and construct deep learning sample data; Process the deep learning sample data of each high-quality well and integrate them to generate a high-quality well curve library; The deep learning model is trained and validated with high-quality well curve location samples; The logging data of low-quality wells are input into the deep learning model for logging curve prediction; The data of low-quality wells are corrected based on the prediction results.

2. The deep learning curve environment correction method based on multivariate regression samples according to claim 1 is characterized in that: The data of the high-quality well includes high-quality layer segment data and distorted layer segment data.

3. The deep learning curve environment correction method based on multivariate regression samples according to claim 2 is characterized in that: The processing of the data of high-quality wells includes: Perform multiple regression fitting on high-quality interval data of high-quality wells; Correct and optimize the distorted layer data of high-quality wells.

4. The deep learning curve environment correction method based on multivariate regression samples according to claim 3 is characterized in that: Correction and optimization of distorted layer data of high-quality wells, including: sonic time difference data correction, neutron porosity data correction and density data correction.

5. The deep learning curve environment correction method based on multivariate regression samples according to claim 4 is characterized in that: The acoustic wave time difference data correction comprises the following steps: Using high-quality layer data, a multivariate regression model for the acoustic time difference curve is established; The acoustic time difference curve is synthesized by using the multivariate regression model of the acoustic time difference curve to correct the distorted layer interval acoustic time difference data.

6. The deep learning curve environment correction method based on multivariate regression samples according to claim 4 is characterized in that: The neutron porosity data correction comprises the following steps: Using high-quality interval data, a multivariate regression model for neutron porosity curves is established; The neutron porosity curve is synthesized by using the multivariate regression model of the neutron porosity curve to correct the distorted neutron porosity data.

7. The deep learning curve environment correction method based on multivariate regression samples according to claim 4 is characterized in that: The density data correction comprises the following steps: Using high-quality layer data, a multivariate regression model for density curves is established; The multi-regression model of density curve is used to synthesize density curve and correct the distorted layer density data.

8. The deep learning curve environment correction method based on multivariate regression samples according to claim 1 is characterized in that: The processing of the deep learning sample data of each high-quality well includes: The non-geological data differences in deep learning sample data are removed through multi-well consistency processing using pattern matching technology; The logging curves of each high-quality well are normalized.

9. The deep learning curve environment correction method based on multivariate regression samples according to claim 1 is characterized in that: The deep learning model is trained and verified with high-quality well curve location samples, including: building the deep learning model structure and determining the machine learning parameters.

10. The deep learning curve environment correction method based on multivariate regression samples according to claim 1, characterized in that: The correction of the data of the low-quality well according to the prediction result is specifically: replacing the distorted curve of the low-quality well with the prediction curve.

11. A deep learning curve environment correction system based on multivariate regression samples, characterized in that: The system includes a classification module, a sample generation module, a data processing module, a model verification module, a prediction module and a correction module; Wherein, the classification module is used to classify the studied wells into high-quality wells and low-quality wells according to the wellbore curve and the rock physical trend; The sample generation module is used to process the data of high-quality wells and construct deep learning sample data; The data processing module is used to process the deep learning sample data of each high-quality well and integrate and generate a high-quality well curve library; The model verification module is used to train and verify the deep learning model with high-quality well curve storage location samples; The prediction module is used to input the logging data of low-quality wells into the deep learning model to predict the logging curve; The correction module is used to correct the data of low-quality wells according to the prediction results.

12. The deep learning curve environment correction system based on multivariate regression samples according to claim 11, characterized in that: The data of the high-quality well includes high-quality layer segment data and distorted layer segment data.

13. The deep learning curve environment correction system based on multivariate regression samples according to claim 12, characterized in that: The sample generation module is used to process the data of high-quality wells, including: Perform multiple regression fitting on high-quality interval data of high-quality wells; Correct and optimize the distorted layer data of high-quality wells.

14. The deep learning curve environment correction system based on multivariate regression samples according to claim 13, characterized in that: Correction and optimization of distorted layer data of high-quality wells, including: sonic time difference data correction, neutron porosity data correction and density data correction.

15. The deep learning curve environment correction system based on multivariate regression samples according to claim 11, characterized in that: The data processing module is used to process the deep learning sample data of each high-quality well, including: The non-geological data differences in deep learning sample data are removed through multi-well consistency processing using pattern matching technology; The logging curves of each high-quality well are normalized.

16. The deep learning curve environment correction system based on multivariate regression samples according to claim 11, characterized in that: The model verification module is used to train and verify the deep learning model with high-quality well curve storage location samples, including: building a deep learning model structure and determining machine learning parameters.

17. The deep learning curve environment correction system based on multivariate regression samples according to claim 11, characterized in that: The correction module is used to correct the data of the low-quality well according to the prediction result, specifically: replace the distorted curve of the low-quality well with the prediction curve.

Citation Information

Cited By

  • Shaft core and logging curve data verification homing method and system based on DPSO-ANN evaluation

    CN121211411A

  • A Method and System for Verification and Relocation of Wellbore Core and Logging Data Based on DPSO-ANN Evaluation

    CN121211411B