Coal measure strata well logging curve correction method and device

Abstract

This invention discloses a method for correcting well logging curves in coal-bearing strata. The method includes: inputting training datasets into at least one machine learning algorithm to establish at least one mapping relationship between a reference curve and a target curve; determining the optimal mapping relationship from the at least one mapping relationship based on a test dataset; inputting data from the enlarged-diameter section of the reference curve into the optimal mapping relationship to map and obtain data from the predicted enlarged-diameter section of the target curve; and using the data from the predicted enlarged-diameter section of the target curve to correct the measured data from the enlarged-diameter section of the target curve. This method solves the problem of varying degrees of distortion and deterioration in well logging curve quality caused by wellbore collapse and severe enlargement in coal-bearing strata, effectively improving the quality of affected well logging curves, enhancing the matching degree between well logging curves and formation information, providing a reliable data foundation for comprehensive seismic geological research, and having positive significance for comprehensive geological research of coal-bearing strata.

Description

Technical Field

[0001] This invention relates to the field of geological exploration technology, and in particular to a method and apparatus for correcting well logging curves in coal-bearing strata. Background Technology

[0002] Natural gas generated from coal, carbonaceous mudstone and shale, and dark mudstone and shale in coal-bearing strata possesses enormous resource potential, including various resource types such as coalbed methane, shale gas, and tight gas. With the continuous deepening of oil and gas exploration theories and the continuous improvement of development technologies, coal-bearing strata exploration has become a very important exploration field with huge exploration potential.

[0003] As exploration and development of coal-bearing strata continue to advance, the target strata are gradually thinning and their heterogeneity is increasing. The integrated application of well logging and seismic data is a crucial supporting technology for coal-bearing strata exploration. However, due to the low mechanical strength of coal-bearing sections, wellbore collapse and severe enlargement are prone to occur during drilling. The unevenness of the wellbore wall leads to varying degrees of distortion in acoustic and density logging curves, resulting in poor logging curve quality. Since acoustic and density logging curves play a vital role in well-seismic depth calibration, comprehensive logging analysis, forward modeling of reservoir seismic reflection characteristics, reservoir inversion, and rock physics modeling, they seriously affect comprehensive seismic geological research work such as reservoir prediction and well logging evaluation in coal-bearing strata.

[0004] For well logging curve calibration, common existing technologies include the following approaches: Standard layer statistical analysis calibration method. This method first selects a standard layer in the non-enlarged well section, and then uses curves such as natural gamma, resistivity, and neutron porosity, which are less affected by wellbore conditions, to fit the target curve within the standard layer. The target curve is calibrated through optimal fitting. However, the calibration result of this method is significantly affected by the selected standard layer. Another example is the rock physics modeling curve calibration method. This method establishes a rock physics model of the target layer by adjusting parameters such as skeleton density and volume content, making the model data highly correlated with the measured data. The target curve is then calibrated using the model. However, this method requires repeated trials, is time-consuming and labor-intensive, has high requirements for basic data, is cumbersome, and has a long cycle, making it inconvenient to implement. In general, existing well logging curve calibration methods mainly calibrate a single curve, limiting their application scope. Furthermore, there is currently no well logging curve calibration method specifically for wellbore collapse sections in coal-bearing formations. Summary of the Invention

[0005] In view of the above problems, the present invention is proposed to provide a method and apparatus for correcting logging curves in coal-bearing formations to overcome or at least partially solve the above problems.

[0006] In a first aspect, embodiments of the present invention provide a method for correcting logging curves in coal-bearing formations, comprising:

[0007] Compare the mean logging data of the un-enlarged well section of at least one logging curve with the mean logging data of the enlarged well section.

[0008] Based on the comparison results, at least one reference curve and target curve in the logging curves are determined;

[0009] The data corresponding to the unexpanded well section of the reference curve and the target curve are divided into a training dataset and a test dataset.

[0010] The training dataset is input into at least one machine learning algorithm to establish at least one mapping relationship between the reference curve and the target curve;

[0011] The at least one mapping relationship is verified based on the test dataset, and the optimal mapping relationship is determined from the at least one mapping relationship;

[0012] The data of the enlarged well section of the reference curve is input into the optimal mapping relationship to obtain the data of the enlarged well section of the predicted target curve.

[0013] The data of the enlarged well section of the predicted target curve are corrected using the data of the enlarged well section.

[0014] In one embodiment, the mean logging data corresponding to the un-enlarged well section of at least one logging curve is compared with the mean logging data corresponding to the enlarged well section; based on the comparison result, a reference curve and a target curve are determined among the at least one logging curves, including:

[0015] The average data of at least one logging curve corresponding to the unexpanded well section of at least one logging curve are statistically analyzed, and the average data of at least one logging curve corresponding to the expanded well section of the at least one logging curve are statistically analyzed, so as to obtain the average data of the unexpanded well section logging curve and the average data of the expanded well section logging curve corresponding to any one of the at least one logging curves.

[0016] The difference rate is calculated sequentially for the mean data of the un-expanded well section logging curve and the mean data of the expanded well section logging curve corresponding to any one of the at least one logging curves. If the difference rate is greater than the preset standard rate of the logging curve, the logging curve is determined as the target curve. If the difference rate is less than or equal to the preset standard rate of the logging curve, the logging curve is determined as the reference curve.

[0017] In one embodiment, the data corresponding to the unexpanded well section of the reference curve and the target curve are divided into a training dataset and a test dataset, including:

[0018] The thickness values ​​of the unexpanded well section of the reference curve and the target curve are statistically analyzed, as well as the logging curve response values ​​of the reference curve and the target curve.

[0019] The logging curve response values ​​include: acoustic value, density value, natural gamma value, resistivity value, and compensated neutron value;

[0020] Based on the thickness value and the natural gamma value in the logging curve response value, and according to a preset division ratio, the data corresponding to the unexpanded well section of the reference curve and the target curve are divided into a training dataset and a test dataset.

[0021] Both the training and testing datasets contain the reference curve and the target curve, as well as the well logging response values ​​corresponding to the reference curve and the target curve.

[0022] In one embodiment, the training dataset is input into at least one machine learning algorithm to establish at least one mapping relationship between the reference curve and the target curve, including:

[0023] The at least one machine learning algorithm includes a multivariate regression algorithm and a random forest algorithm;

[0024] The training dataset is input into the multivariate regression algorithm and the random forest algorithm respectively, and a first mapping relationship between the reference curve and the target curve corresponding to the multivariate regression algorithm and a second mapping relationship between the reference curve and the target curve corresponding to the random forest algorithm are established respectively.

[0025] In one embodiment, validating the at least one mapping relationship based on the test dataset and determining the optimal mapping relationship from the at least one mapping relationship includes:

[0026] The reference curves in the test dataset are input into the first mapping relationship and the second mapping relationship respectively. The first prediction target curve is obtained through the first mapping relationship, and the second prediction target curve is obtained through the second mapping relationship.

[0027] The first and second predicted target curves are compared with the target curves in the test dataset to determine their degree of fit. The mapping relationship corresponding to the predicted target curve with a higher degree of fit is determined as the optimal mapping relationship.

[0028] In one embodiment, correcting the data of the enlarged well section of the predicted target curve using data from the enlarged well section includes:

[0029] Using the data of the enlarged well section of the predicted target curve, the data of the enlarged well section of the target curve is replaced to complete the correction of the data of the enlarged well section of the target curve.

[0030] In one embodiment, the unenlarged well section and the enlarged well section are determined in the following manner:

[0031] Based on the logging curve response values ​​corresponding to the logging curves of the coal seams in the study area, and the value range of the coal seam development section in the study area, the coal seam development section in the study area is determined.

[0032] The coal seam development section includes at least one depth point. The absolute value of the difference between the wellbore curve value corresponding to each depth point and the drill bit diameter is taken to obtain the absolute value result corresponding to each depth point. If the absolute value result is less than or equal to the preset wellbore enlargement standard threshold, the depth point corresponding to the absolute value result is a non-enlargement point; otherwise, it is an enlargement point.

[0033] If the formation thickness corresponding to consecutive enlargement points is greater than a preset formation thickness threshold, then the formation segment corresponding to the consecutive enlargement points is an enlargement well segment.

[0034] If the formation thickness corresponding to consecutive non-expansion points is greater than a preset formation thickness threshold, then the formation segment corresponding to the consecutive non-expansion points is an unexpansioned well segment.

[0035] In one embodiment, after correcting the data of the enlarged well section of the target curve, the method further includes:

[0036] The reliability of the correction results is determined by cross-plotting or synthetic recording methods.

[0037] Secondly, embodiments of the present invention provide a well logging curve correction device for coal-bearing formations, comprising:

[0038] The comparison module is used to compare the average logging curve data corresponding to the un-enlarged well section of at least one logging curve with the average logging curve data corresponding to the enlarged well section.

[0039] The first determining module is used to determine, based on the comparison results, at least one reference curve and target curve among the logging curves;

[0040] The partitioning module is used to partition the data corresponding to the unexpanded well section of the reference curve and the target curve into a training dataset and a test dataset.

[0041] A module is established to input the training dataset into at least one machine learning algorithm to establish at least one mapping relationship between the reference curve and the target curve;

[0042] The second determining module is used to verify the at least one mapping relationship based on the test dataset and determine the optimal mapping relationship from the at least one mapping relationship;

[0043] The mapping module is used to input the data of the enlarged well section of the reference curve into the optimal mapping relationship, and map the data of the enlarged well section of the predicted target curve.

[0044] A correction module is used to correct the data of the enlarged well section of the predicted target curve using the data of the enlarged well section.

[0045] Thirdly, embodiments of the present invention provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the aforementioned method for correcting logging curves in coal-bearing formations.

[0046] Fourthly, embodiments of the present invention provide a computer program product, the computer program product including a computer program, which, when executed by a processor, implements the well logging curve correction method for coal-bearing strata as described above.

[0047] The beneficial effects of the above-described technical solutions provided in the embodiments of the present invention include at least the following:

[0048] The well logging curve correction method for coal-bearing formations provided in this invention is based on the actual well logging curve response values ​​of the unexpanded and expanded sections of the coal seam. By comparing the well logging curves of the unexpanded and expanded sections, the method statistically analyzes the mean data of the unexpanded and expanded well logging curves corresponding to any one of at least one well logging curves. It then calculates the difference rate based on the mean data of the unexpanded and expanded well logging curves for any one type of well logging curve. By comparing the difference rate with a preset standard rate for that type of well logging curve, it determines whether the well logging curve is a reference curve or a target curve. This method of determining reference and target curves accurately identifies well logging curves less affected by expansion as reference curves and those more affected by expansion as target curves requiring correction. Then, using a training dataset containing reference and target curves, at least one mapping relationship between the reference and target curves is established through at least one machine learning algorithm. Finally, a test dataset containing reference and target curves is used... This method determines the optimal mapping relationship from at least one mapping relationship. By applying multiple machine learning algorithms, various mapping relationships between reference curves and target curves can be established. This allows for comparison of the mapping effects of multiple mapping relationships, enabling a more accurate determination of the optimal mapping relationship. The optimal mapping relationship is then used to map the data corresponding to the enlarged well section of the predicted target curve, thereby correcting the measured data of the enlarged well section of the target curve. This method solves the problem of varying degrees of distortion in acoustic and density logging curves and the deterioration of logging curve quality caused by wellbore collapse and severe enlargement in coal-bearing sections of coal-bearing strata. It finely corrects acoustic and density logging curves in wellbore collapse sections of coal-bearing strata, enabling the correction of multiple logging curves. This effectively improves the quality of logging curves in coal-bearing sections of coal-bearing strata affected by enlargement and wellbore collapse, and enhances the matching degree between logging curves and formation information. This provides a reliable data foundation for comprehensive seismic geological research such as seismic interpretation and reservoir prediction, and has positive significance for comprehensive geological research of coal-bearing strata.

[0049] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings.

[0050] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0051] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0052] Figure 1 This is a flowchart of the well logging curve correction method for coal-bearing strata in an embodiment of the present invention;

[0053] Figure 2 This is a schematic diagram of the well logging response characteristics of coal seams and the division of coal seam development sections in coal-bearing strata in an embodiment of the present invention;

[0054] Figure 3 This is a schematic diagram illustrating the division of the unexpanded and expanded well sections and the differences in curves before and after correction in an embodiment of the present invention;

[0055] Figure 4 This is a cross-plot of neutron-density before and after correction in an embodiment of the present invention;

[0056] Figure 5 This is a comparison diagram of the synthesized records before and after correction in an embodiment of the present invention;

[0057] Figure 6 This is a structural block diagram of the well logging curve correction device for coal-bearing strata in an embodiment of the present invention. Detailed Implementation

[0058] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0059] To address the issues of varying degrees of distortion and deterioration in logging curve quality caused by wellbore collapse and severe enlargement in coal-bearing sections of coal-bearing formations, this invention provides a method for correcting logging curves in coal-bearing formations. (Refer to...) Figure 1 As shown, the method includes the following steps:

[0060] S11. Compare the average logging data of the unexpanded well section of at least one logging curve with the average logging data of the expanded well section.

[0061] S12. Based on the comparison results, determine at least one reference curve and target curve among the logging curves;

[0062] S13. Divide the data corresponding to the unexpanded well sections of the reference curve and the target curve into training dataset and test dataset;

[0063] S14. Input the training dataset into at least one machine learning algorithm to establish at least one mapping relationship between the reference curve and the target curve;

[0064] S15. Verify at least one mapping relationship based on the test dataset, and determine the optimal mapping relationship from at least one mapping relationship;

[0065] S16. Input the data of the enlarged well section of the reference curve into the optimal mapping relationship to obtain the data of the enlarged well section of the predicted target curve.

[0066] S17. Use the data of the enlarged well section of the predicted target curve to correct the data of the enlarged well section of the target curve.

[0067] The aforementioned un-enlarged and enlarged well sections can be determined, for example, in the following ways:

[0068] Based on the well logging response values ​​corresponding to the coal seams in the study area, and the value range of the coal seam development section in the study area, the coal seam development section in the study area is determined. Each coal seam development section includes at least one depth point. The absolute value of the difference between the wellbore diameter curve value corresponding to each depth point and the drill bit diameter is taken. If the obtained absolute value is less than or equal to a preset wellbore enlargement standard threshold, the depth point corresponding to that absolute value is a non-enlargement point; otherwise, it is an enlargement point. If the formation thickness corresponding to consecutive enlargement points is greater than a preset formation thickness threshold, the formation segment corresponding to the consecutive enlargement points is an enlargement well section. If the formation thickness corresponding to consecutive non-enlargement points is greater than a preset formation thickness threshold, the formation segment corresponding to the consecutive non-enlargement points is an unenlarged well section.

[0069] The well logging curve correction method for coal-bearing formations provided in this invention first determines a reference curve and a target curve from at least one well logging curve in the coal-bearing formation. The at least one well logging curve may include, for example, acoustic, natural gamma, resistivity, compensated neutron, and density well logging curves. The reference curve is the well logging curve least affected by borehole enlargement, while the target curve is the well logging curve most affected by borehole enlargement. The target curve is the well logging curve that needs correction. Both the reference curve and the target curve include data corresponding to the unenlarged well section and the enlarged well section. Then, the data corresponding to the unenlarged well section of the reference curve and the target curve are divided into training data. The training dataset and the test dataset are used to establish at least one mapping relationship between the reference curve and the target curve by inputting the well logging curve data from the training dataset into at least one machine learning algorithm. The established at least one mapping relationship is then verified based on the test dataset. The optimal mapping relationship is determined from the at least one mapping relationship based on the verification results. The data of the enlarged well section of the reference curve can then be input into the optimal mapping relationship. The data of the enlarged well section of the predicted target curve can be obtained through mapping the data of the enlarged well section of the actual target curve using the data of the enlarged well section of the predicted target curve.

[0070] The well logging curve correction method for coal-bearing strata provided in this invention solves the problems of varying degrees of distortion in acoustic and density logging curves and deterioration in logging curve quality caused by wellbore collapse and severe enlargement in coal-bearing sections of coal-bearing strata. It precisely corrects acoustic and density logging curves in wellbore collapse sections of coal-bearing strata, enabling the correction of multiple logging curves. This effectively improves the quality of logging curves in coal-bearing sections of coal-bearing strata affected by enlargement and wellbore collapse, and enhances the matching degree between logging curves and formation information. Therefore, it provides a reliable data foundation for comprehensive seismic geological research such as seismic interpretation and reservoir prediction, and has positive significance for the comprehensive geological research of coal-bearing strata.

[0071] Furthermore, in steps S11 and S12 above, at least one reference curve and target curve in the well logging curve are determined in the following manner:

[0072] The average data of at least one logging curve corresponding to the unexpanded well section of at least one logging curve are statistically analyzed, and the average data of at least one logging curve corresponding to the expanded well section of the at least one logging curve are statistically analyzed, so as to obtain the average data of the unexpanded well section logging curve and the average data of the expanded well section logging curve corresponding to any one of the at least one logging curves.

[0073] The difference rate is calculated sequentially for the mean data of the unexpanded well section and the mean data of the expanded well section corresponding to any one of the at least one logging curves. If the difference rate is greater than the preset standard rate of the logging curve, the logging curve is determined as the target curve. If the difference rate is less than or equal to the preset standard rate of the logging curve, the logging curve is determined as the reference curve.

[0074] In this embodiment of the invention, the reference curve and the target curve are determined by sequentially calculating the difference rate of the average data of the un-expanded well section and the average data of the expanded well section corresponding to any one of the at least one logging curves. First, the average data of at least one logging curve corresponding to the un-expanded well section and the average data of at least one logging curve corresponding to the expanded well section are statistically analyzed. After statistical analysis, the average data of the un-expanded well section and the average data of the expanded well section corresponding to any one logging curve can be obtained. Then, the difference rate of the reference curve and the target curve is calculated by sequentially calculating the difference rate of the reference curve and the target curve. The difference rate is calculated between the average data of the un-expanded well logging section and the average data of the expanded well logging section corresponding to a certain logging curve. If the difference rate is greater than the preset standard rate of the logging curve, it indicates that the logging curve is relatively sensitive to expansion and needs to be corrected. In this case, the logging curve is determined as the target curve. If the difference rate is less than or equal to the preset standard rate of the logging curve, it indicates that the logging curve is not sensitive to expansion. In this case, the logging curve is determined as the reference curve. The above steps of calculating the difference rate are performed sequentially on at least one logging curve to determine the reference curve and the target curve.

[0075] Furthermore, in step S13 above, the data corresponding to the unexpanded well sections of the reference curve and the target curve are divided into a training dataset and a test dataset in the following manner:

[0076] The thickness values ​​of the unexpanded well sections of the reference curve and the target curve are statistically analyzed, as well as the logging response values ​​of the reference curve and the target curve. The logging response values ​​may include, for example, sonic values, density values, natural gamma values, resistivity values, and compensated neutron values.

[0077] Based on the thickness value and the natural gamma value in the logging curve response value, and according to the preset division ratio, the data corresponding to the unexpanded well section of the reference curve and the target curve are divided into training dataset and test dataset. Both the training dataset and the test dataset contain the reference curve and the target curve, as well as the logging curve response value corresponding to the reference curve and the target curve.

[0078] In the above steps, the thickness of the unexpanded well sections of the reference curve and target curve, as well as the logging response value of each reference curve and target curve, are statistically analyzed. Then, the unexpanded well sections of the reference curve and target curve are sorted first by thickness, and second by natural gamma value, to obtain the sorted data of the unexpanded well sections of the reference curve and target curve. Then, according to a preset division ratio (e.g., the division ratio of the test dataset is 10%, and the division ratio of the training dataset is 90%), the test dataset is selected from the data of the reference curve and target curve with similar thickness or natural gamma value of the unexpanded well section, and the remaining data of the reference curve and target curve with the unexpanded well section are used as the training dataset.

[0079] Furthermore, in step S14 above, at least one mapping relationship between the reference curve and the target curve is established in the following manner:

[0080] The training dataset is input into at least one machine learning algorithm to establish at least one mapping relationship between the reference curve and the target curve;

[0081] At least one machine learning algorithm may be used, such as multiple regression algorithm and random forest algorithm. Of course, the embodiments of the present invention are not limited to the above two machine learning algorithms. Any algorithm that can establish a mapping relationship between the reference curve and the target curve is acceptable. The following uses the specific implementation of these two algorithms as examples for illustration.

[0082] By employing various machine learning algorithms, multiple mapping relationships between reference curves and target curves can be established. This allows for comparison of the mapping effects of different relationships, selection of the optimal mapping relationship, and ultimately, improvement in the accuracy of target curve prediction.

[0083] The training dataset is input into the multivariate regression algorithm and the random forest algorithm respectively, and the first mapping relationship between the reference curve and the target curve corresponding to the multivariate regression algorithm and the second mapping relationship between the reference curve and the target curve corresponding to the random forest algorithm are established respectively.

[0084] After establishing at least one mapping relationship between the reference curve and the target curve through the above steps, the optimal mapping relationship is determined from at least one mapping relationship in the following manner:

[0085] The reference curves in the test dataset are input into the first mapping relationship and the second mapping relationship respectively. The first prediction target curve is obtained by mapping through the first mapping relationship, and the second prediction target curve is obtained by mapping through the second mapping relationship.

[0086] The obtained first and second predicted target curves are compared with the measured target curves in the test dataset to determine the mapping relationship corresponding to the predicted target curve with higher matching degree.

[0087] Furthermore, after determining the optimal mapping relationship through the above steps, the data of the enlarged well section of the measured target curve are corrected using the optimal mapping relationship in the following manner:

[0088] The data of the enlarged well section of the reference curve is input into the optimal mapping relationship to obtain the data of the enlarged well section of the predicted target curve.

[0089] The data of the enlarged well section of the predicted target curve is used to replace the data of the enlarged well section of the measured target curve, thus completing the correction of the enlarged well section data of the measured target curve.

[0090] The well logging curve correction method for coal-bearing formations provided in this embodiment of the invention, after correcting the data of the enlarged diameter section of the measured target curve, can also perform the following steps:

[0091] The reliability of the above correction results is determined by cross-plotting or synthetic recording methods.

[0092] The reliability analysis of the correction results of the well logging curve correction method for coal-bearing strata provided in this embodiment of the invention can be achieved, for example, through cross-plotting or synthetic recording methods. First, cross-plots are made using the well logging curves before and after correction. By comparing the cross-plots of the well logging curves before and after correction, it can be seen that the corrected curve is more consistent with geological laws. Then, synthetic seismic records are made using the well logging curves before and after correction. Comparative analysis of the synthetic records shows that the calibration results of the synthetic records of the corrected curve are significantly improved, especially in the enlarged well section, where the correlation coefficient is higher, thereby verifying the reliability of the above correction method.

[0093] The well logging curve correction method for coal-bearing formations provided in this invention is based on the actual well logging curve response values ​​of the unexpanded and expanded sections of the coal seam. By comparing the well logging curves of the unexpanded and expanded sections, the method statistically analyzes the mean data of the unexpanded and expanded well logging curves corresponding to any one of at least one well logging curves. It then calculates the difference rate based on the mean data of the unexpanded and expanded well logging curves for any one type of well logging curve. By comparing the difference rate with a preset standard rate for that type of well logging curve, it determines whether the well logging curve is a reference curve or a target curve. This method of determining reference and target curves accurately identifies well logging curves less affected by expansion as reference curves and those more affected by expansion as target curves requiring correction. Then, using a training dataset containing reference and target curves, at least one mapping relationship between the reference and target curves is established through at least one machine learning algorithm. Finally, a test dataset containing reference and target curves is used... This method determines the optimal mapping relationship from at least one mapping relationship. By applying multiple machine learning algorithms, various mapping relationships between reference curves and target curves can be established. This allows for comparison of the mapping effects of multiple mapping relationships, enabling a more accurate determination of the optimal mapping relationship. The optimal mapping relationship is then used to map the data corresponding to the enlarged well section of the predicted target curve, thereby correcting the measured data of the enlarged well section of the target curve. This method solves the problem of varying degrees of distortion in acoustic and density logging curves and the deterioration of logging curve quality caused by wellbore collapse and severe enlargement in coal-bearing sections of coal-bearing strata. It finely corrects acoustic and density logging curves in wellbore collapse sections of coal-bearing strata, enabling the correction of multiple logging curves. This effectively improves the quality of logging curves in coal-bearing sections of coal-bearing strata affected by enlargement and wellbore collapse, and enhances the matching degree between logging curves and formation information. This provides a reliable data foundation for comprehensive seismic geological research such as seismic interpretation and reservoir prediction, and has positive significance for comprehensive geological research of coal-bearing strata.

[0094] The following detailed description of the well logging curve correction method for coal-bearing strata of the present invention will be provided using a specific example, taking well logging in a coal-bearing strata in a certain exploration area as an example.

[0095] First, based on the logging response characteristics of the coal seam, the coal seam development sections of the coal-bearing strata are divided in the target area. The mineral composition and physical properties of the coal seam determine its unique response characteristics on the logging curve, generally exhibiting the characteristics of "three highs and two lows," namely, high neutron porosity (e.g., compensated neutron porosity > 40), high acoustic transit time (e.g., acoustic transit time > 360), high resistivity (e.g., resistivity > 20), low natural gamma (e.g., natural gamma < 90), and low density (e.g., density < 1.95). Based on the logging response characteristics of coal seams, coal seam development sections are divided. In this embodiment, through comprehensive analysis of the acoustic (DT), natural gamma ray (GR), resistivity (RT), compensated neutron log (CNL), and density (DEN) logging curves within the work area, combined with the logging response characteristics of coal seams, sections meeting the conditions of "GR<90, DEN<1.95, RT>20, DT>360, CNL>40" are classified as coal seams. Figure 2 As shown, a total of 26 coal seam development segments were identified.

[0096] Then, through comparative analysis of the caliper curve and drill bit diameter, the un-expanded and expanded sections of the coal seam development segment are determined. Specifically, the standard threshold C for determining caliper expansion and the formation thickness threshold T for continuous caliper expansion are first determined. Then, at a certain depth point, the absolute value of the difference between the caliper curve value and the drill bit diameter is taken. If this value is less than or equal to the standard threshold C for caliper expansion, then that depth point is an un-expanded point; if the value is greater than the standard threshold C, then it is an expanded point. Based on the expanded and un-expanded points of the coal seam development segment, if the formation thickness for continuous caliper expansion is greater than the given formation thickness threshold T, then that segment is an expanded section; if the formation thickness for continuous non-expanded caliper expansion is greater than the given formation thickness threshold T, then that segment is an un-expanded section.

[0097] Reference Figure 3 As shown, with a selected drill bit diameter of 250mm, a C value of 5mm, and a T value of 1m, the well sections from 2868m to 2875m are identified as unexpanded well sections, based on the principle that the formation thickness of continuously unexpanded wells is greater than T. The well sections from 2860.5m to 2865.1m and from 2934.2m to 2939.3m are identified as expanded well sections, based on the principle that the formation thickness of continuously expanded wells is greater than T.

[0098] The average values ​​of various logging curves in the coal seam development section are statistically analyzed for both the enlarged and unenlarged sections. If the difference rate between the average value of the enlarged and unenlarged sections of a certain logging curve is above a specific standard N, it indicates that the curve is sensitive to enlargement and needs correction; this curve is considered the target curve. If the difference rate is below the specific standard N, it indicates that the curve is not sensitive to enlargement and can be used as a reference curve. Following this method, the above analysis is performed on each curve to determine the reference and target curves.

[0099] In this embodiment, taking the density curve as an example, the standard rate N is first determined to be 5%, and then the average density of the unexpanded coal seam section is statistically analyzed to be 1.607 g / cm³. 3 The average density of the enlarged coal seam section is 1.502 g / cm³. 3 The difference rate between the two is 6.5%, which is greater than the standard rate of 5%, indicating that the density curve is sensitive to diameter expansion and needs to be corrected. This is the target curve. Using the same method, the difference rates of the acoustic curve, natural gamma curve, resistivity curve and compensated neutron curve are all determined to be less than N, and these are determined to be the reference curves.

[0100] In this embodiment, the logging response values ​​of acoustic waves, density, natural gamma, resistivity, and compensated neutrons in the unexpanded coal seam section can be read, for example, using logging software. That is, the logging curves are loaded into the logging software, and according to the pre-divided unexpanded well section, the logging response values ​​of the reference curve and the target curve at different depth points are read at a certain sampling interval.

[0101] In this embodiment, for example, the logging curves can be loaded into Powerlog software. Based on the pre-divided unexpanded well sections, the logging curve response values ​​of the reference curve and the target curve at each depth sampling point of each unexpanded well section are read sequentially at a sampling interval of 0.05m. As shown in Table 1, Table 1 shows the logging curve response values ​​of each logging curve in the unexpanded well sections of different coal seams.

[0102] Table 1

[0103]

[0104]

[0105] The data corresponding to the unexpanded well sections of the reference and target curves were divided into training and test datasets. The training dataset was used to establish the mapping relationship between the reference and target curves during curve calibration, while the test dataset was used to verify the reliability of the mapping relationship between the target and reference curves. The thickness of the unexpanded well sections and the logging response values ​​for each reference and target curve were statistically analyzed. Then, the sections were sorted first by thickness, and secondly by natural gamma value. Test datasets were selected from well sections with similar thickness or natural gamma values, representing 10% of the total number of unexpanded well sections. The remaining unexpanded well sections served as the training dataset for calculating the mapping relationship between the target and reference curves.

[0106] In this embodiment, 26 unexpanded coal seam development segments were divided into coal-bearing strata in a certain block. The unexpanded coal seam development segments were sorted by "thickness" as the primary keyword and "natural gamma" as the secondary keyword. The results are shown in Table 2. At a ratio of approximately 10%, one unexpanded well segment of about 5m, 2m and 1m was selected as the test dataset. The well segments with serial numbers 2, 8 and 16 in Table 2 are the selected test datasets, and the rest are used as training datasets.

[0107] Table 2

[0108]

[0109]

[0110] This study establishes a mapping relationship between the target curve and the reference curve on the training dataset using two different algorithms: multivariate regression and random forest regression. The test dataset is then used to verify the accuracy of the predictions from each algorithm and to find the optimal mapping relationship. Specifically, firstly, both multivariate regression and random forest regression are applied to the training dataset to establish two different mapping relationships between the reference and target curves. Then, using these two different mapping relationships, the reference curve in the test dataset is mapped to the target curve, resulting in two different predicted target curves. Finally, the accuracy of these two predicted target curves is compared with the original target curve (the measured target curve) in the test dataset. The mapping relationship corresponding to the predicted target curve with the highest accuracy is determined as the optimal mapping relationship.

[0111] In this embodiment, the target curve is a density curve, with a sampling interval of 0.05m. One depth sampling point corresponds to one sample. There are a total of 820 samples in 23 unexpanded well sections in the training set. Each sample has 4 features: DT, CNL, GR and RT. DEN is the observed value.

[0112] First, the multiple linear regression method is applied to construct the following model:

[0113] DEN=a+b*DT+c*CNL+d*GR+e*lg(RT) (1)

[0114] In the above formula, DEN is the density logging value, DT is the sonic logging value, CNL is the compensated neutron logging value, GR is the natural gamma logging value, RT is the resistivity logging value, and a, b, c, d, and e are the multiple linear regression fitting constants. Substituting the 820 sample values ​​from 23 unexpanded well sections in the training set into this formula, the following expression for the multiple regression mapping relationship is obtained:

[0115] DEN=1.414+0.0005*DT-0.0043*CNL+0.0043*GR-0.081*lg(RT) (2)

[0116] Substituting the DT, CNL, GR, and RT values ​​of 170 samples from three unexpanded well sections in the test set into the mapping equation, the predicted density value DEN' is obtained. The matching rate w of the mapping is then calculated using the following formula:

[0117]

[0118] In the above formula, N is the total number of test sample points, which is 170, i is the i-th sample, and DEN′ i Let DEN be the predicted density value for the i-th sample. i The actual logging curve value for the i-th sample is given, and the matching rate of this mapping prediction is 90.1%.

[0119] The following section uses the random forest regression method to establish the mapping relationship between the target curve and the reference curve. The random forest regression algorithm is a machine learning algorithm based on the idea of ​​ensemble learning. It randomly samples data to form multiple different decision trees, and then combines the results of each decision tree using a certain combination strategy (usually taking the average of the predictions from all decision trees) to obtain the final prediction result of the random forest. Its main steps are:

[0120] ① Using the Bagging method, randomly sample from the original sample set to construct k sample subsets, namely θ1, θ2, ..., θ k In this embodiment, a subset of 350 samples was constructed from 820 samples in 23 unexpanded well sections of the training set.

[0121] ② Using the random subspace method, m features are randomly selected from all feature attributes, and node splitting is performed. The optimal splitting method among these m features is selected to construct a single regression decision tree. In this embodiment, each sample has 4 attribute features: DT, CNL, GR and RT. Three attribute features are randomly selected (m=3) to construct a single regression decision tree.

[0122] ③ Repeat steps ① and ② above to build multiple regression decision trees, i.e., {H(x,θ1),H(x,θ2),…,H(x,θ)} k The maximum depth r of the trees is selected to maximize the growth of each tree and form a forest. The selection of the maximum depth r of the trees has a significant impact on the prediction results. In this embodiment, through repeated experiments, the maximum depth of the trees is selected as 10.

[0123] ④ The final prediction result is the average of the prediction results from all regression decision trees. That is:

[0124]

[0125] In the above formula: This represents the final prediction result of the random forest; h(x,θ) i H(x,θ) is the decision number for the i-th tree. i The prediction results are given by x, where x is the independent variable and k is the number of decision trees.

[0126] The three parameters ultimately required in the random forest regression method are: the number of sample subsets (i.e., decision trees, k), the extracted feature attributes, m, and the maximum tree growth depth, r. These three parameters need to be optimized through repeated experiments based on the criterion of minimizing the average error. In this embodiment, k = 350, m = 3, and r = 10. Applying these parameters, a mapping relationship between the reference curve and the target curve was established through random forest regression on 820 samples.

[0127] The DT, CNL, GR and RT values ​​of 170 samples from 3 unexpanded well sections in the test set were applied to the mapping relationship to obtain the predicted density value DEN'. The matching rate w of the mapping relationship was calculated according to formula (3). The matching rate of this mapping relationship prediction was 96.5%.

[0128] Therefore, the random forest regression method with the highest consistency is selected as the optimal algorithm, and the mapping relationship obtained by it is the optimal mapping relationship.

[0129] The obtained optimal mapping relationship is used to predict the data of the expanded coal seam section of the measured target curve. In the expanded section, the measured target curve value of that section is replaced with the predicted target curve value.

[0130] In this embodiment, during the enlarged well section, the values ​​of the enlarged well section of the reference curve are first read. Then, using the preferred random forest regression algorithm and optimal mapping relationship, the enlarged well section data of the reference curve is mapped to the target curve using this optimal mapping relationship, i.e., the mapping of the predicted target curve is performed. Finally, the enlarged well section data of the predicted target curve is used to replace the measured enlarged well section data of the target curve, completing the correction of the target curve of the coal seam enlarged well section. Figure 3 As shown, Figure 3 The diagram shows the relative difference between the original density curve and the corrected density curve.

[0131] The reliability of the curve correction method is analyzed below using cross-plotting and synthetic recording methods.

[0132] In this embodiment, firstly, cross-plots are plotted between the density curves before and after correction and the neutron curve, respectively, and then referred to... Figure 4 As shown, Figure 4 The middle part includes the upper half Figure 4 (a) and the lower half Figure 4 (b), Figure 4 (a) is the neutron-density cross plot before correction. Figure 4 (b) is the corrected neutron-density cross plot. Figure 4 The horizontal axis represents the neutron logging curve, and the vertical axis represents the density logging curve. A comparison of the cross-plots of the density curves before and after correction reveals the results. Figure 4 (a) The correlation between the neutron-density curves is not obvious. Figure 4 (b) The neutron-density curves show a clear positive correlation, which is more consistent with the geological patterns of the study area, indicating that the curve correction is highly reliable.

[0133] Next, synthetic seismic records were created by combining the density curves before and after correction with the acoustic curves, referring to... Figure 5 As shown, Figure 5 The middle includes the left half Figure 5 (a) and the right half Figure 5 (b), Figure 5 (a) is the depth calibration map of the synthesized record before correction. Figure 5 (b) is the time-depth calibration plot of the synthesized record after correction. The comparative analysis of the synthesized records shows that the calibration results of the synthesized record after curve correction are significantly improved. Figure 5 (a) indicates that in the enlarged well section, the waveforms of the synthetic seismic record before curve correction differ significantly from those of the seismic trace, resulting in a low degree of agreement. Figure 5 (b) indicates that the corrected synthetic seismic record has a high waveform match with the seismic trace and a higher correlation coefficient, which further verifies the reliability of the correction method.

[0134] The well logging curve correction method for coal-bearing strata provided by this invention effectively improves the quality of well logging curves in coal-bearing sections affected by wellbore enlargement and wellbore collapse, enhancing the representation of formation information by the well logging curves. This provides a solid data foundation for comprehensive seismic geological research, including well-seismic depth calibration and reservoir prediction. The well logging curve correction method for coal-bearing strata provided by this invention is easy to implement, highly reliable, and can correct multiple well logging curves, making it of great significance for comprehensive geological research in oil and gas exploration of coal-bearing strata.

[0135] Based on the same inventive concept, this invention also provides a logging curve correction device for coal-bearing formations. Since the principle of the problem solved by this device is similar to the aforementioned logging curve correction method for coal-bearing formations, the implementation of this device can refer to the implementation of the aforementioned method, and the repeated parts will not be described again.

[0136] This invention provides a well logging curve correction device for coal-bearing formations, with reference to... Figure 6 As shown, it includes:

[0137] Comparison module 61 is used to compare the average logging curve data corresponding to the un-enlarged well section of at least one logging curve with the average logging curve data corresponding to the enlarged well section.

[0138] The first determining module 62 is used to determine, based on the comparison results, at least one reference curve and target curve among the logging curves;

[0139] The partitioning module 63 is used to partition the data corresponding to the unexpanded well section of the reference curve and the target curve into a training dataset and a test dataset.

[0140] Module 64 is configured to input the training dataset into at least one machine learning algorithm to establish at least one mapping relationship between the reference curve and the target curve;

[0141] The second determining module 65 is used to verify the at least one mapping relationship based on the test dataset and determine the optimal mapping relationship from the at least one mapping relationship;

[0142] Mapping module 66 is used to input the data of the enlarged well section of the reference curve into the optimal mapping relationship, and map the data of the enlarged well section of the predicted target curve.

[0143] The correction module 67 is used to correct the data of the enlarged well section of the target curve using the data of the enlarged well section of the predicted target curve.

[0144] This invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned method for correcting logging curves in coal-bearing formations.

[0145] This invention provides a computer program product, which includes a computer program that, when executed by a processor, implements the aforementioned method for correcting logging curves in coal-bearing formations.

[0146] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0147] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0148] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0149] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0150] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0151] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for correcting well logging curves in coal-bearing formations, characterized in that, include: Compare the mean logging data of the un-enlarged well section of at least one logging curve with the mean logging data of the enlarged well section. Based on the comparison results, at least one reference curve and target curve in the logging curves are determined; The thickness values ​​of the unexpanded well section of the reference curve and the target curve are statistically analyzed, as well as the logging curve response values ​​of the reference curve and the target curve. The logging curve response values ​​include: acoustic value, density value, natural gamma value, resistivity value, and compensated neutron value; Based on the thickness value and the natural gamma value in the logging curve response value, and according to a preset division ratio, the data corresponding to the unexpanded well section of the reference curve and the target curve are divided into a training dataset and a test dataset; both the training dataset and the test dataset contain the reference curve and the target curve, as well as the logging curve response values ​​corresponding to the reference curve and the target curve. The training dataset is input into the multivariate regression algorithm and the random forest algorithm respectively, and a first mapping relationship between the reference curve and the target curve corresponding to the multivariate regression algorithm and a second mapping relationship between the reference curve and the target curve corresponding to the random forest algorithm are established respectively. The reference curves in the test dataset are input into the first mapping relationship and the second mapping relationship respectively. The first prediction target curve is obtained through the first mapping relationship, and the second prediction target curve is obtained through the second mapping relationship. The first and second predicted target curves are compared with the target curves in the test dataset to determine the mapping relationship corresponding to the predicted target curve with higher matching degree, and the mapping relationship corresponding to the predicted target curve with higher matching degree is determined as the optimal mapping relationship. The data of the enlarged well section of the reference curve is input into the optimal mapping relationship to obtain the data of the enlarged well section of the predicted target curve. The data of the enlarged well section of the predicted target curve are corrected using the data of the enlarged well section.

2. The method as described in claim 1, characterized in that, Compare the mean logging data of the un-enlarged well section of at least one logging curve with the mean logging data of the enlarged well section. Based on the comparison results, at least one reference curve and target curve in the well logging curves are determined, including: The average data of at least one logging curve corresponding to the unexpanded well section of at least one logging curve are statistically analyzed, and the average data of at least one logging curve corresponding to the expanded well section of the at least one logging curve are statistically analyzed, so as to obtain the average data of the unexpanded well section logging curve and the average data of the expanded well section logging curve corresponding to any one of the at least one logging curves. The difference rate is calculated sequentially for the mean data of the un-expanded well section logging curve and the mean data of the expanded well section logging curve corresponding to any one of the at least one logging curves. If the difference rate is greater than the preset standard rate of the logging curve, the logging curve is determined as the target curve. If the difference rate is less than or equal to the preset standard rate of the logging curve, the logging curve is determined as the reference curve.

3. The method as described in claim 1, characterized in that, Using data from the enlarged well section of the predicted target curve, the data from the enlarged well section of the target curve are corrected, including: Using the data of the enlarged well section of the predicted target curve, the data of the enlarged well section of the target curve is replaced to complete the correction of the data of the enlarged well section of the target curve.

4. The method according to any one of claims 1-3, characterized in that, The unexpanded well section and the expanded well section are determined in the following manner: Based on the logging curve response values ​​corresponding to the logging curves of the coal seams in the study area, and the value range of the coal seam development section in the study area, the coal seam development section in the study area is determined. The coal seam development section includes at least one depth point. The absolute value of the difference between the wellbore curve value corresponding to each depth point and the drill bit diameter is taken to obtain the absolute value result corresponding to each depth point. If the absolute value result is less than or equal to the preset wellbore enlargement standard threshold, the depth point corresponding to the absolute value result is a non-enlargement point; otherwise, it is an enlargement point. If the formation thickness corresponding to consecutive enlargement points is greater than a preset formation thickness threshold, then the formation segment corresponding to the consecutive enlargement points is an enlargement well segment. If the formation thickness corresponding to consecutive non-expansion points is greater than a preset formation thickness threshold, then the formation segment corresponding to the consecutive non-expansion points is an unexpansioned well segment.

5. The method as described in claim 1, characterized in that, After correcting the data of the enlarged well section of the target curve, the method further includes: The reliability of the correction results is determined by cross-plotting or synthetic recording methods.

6. A well logging curve correction device for coal-bearing formations, characterized in that, include: The comparison module is used to compare the average logging curve data corresponding to the un-enlarged well section of at least one logging curve with the average logging curve data corresponding to the enlarged well section. The first determining module is used to determine, based on the comparison results, at least one reference curve and target curve among the logging curves; The partitioning module is used to partition the data corresponding to the unexpanded well section of the reference curve and the target curve into a training dataset and a test dataset. A module is established to statistically analyze the thickness values ​​of the unexpanded well section of the reference curve and the target curve, as well as the logging curve response values ​​of the reference curve and the target curve. The well logging curve response values ​​include: sonic value, density value, natural gamma value, resistivity value, and compensated neutron value. Based on the thickness value and the natural gamma value in the well logging curve response values, and according to a preset division ratio, the data corresponding to the unexpanded well sections of the reference curve and the target curve are divided into a training dataset and a test dataset. Both the training dataset and the test dataset contain the reference curve and the target curve, as well as the well logging curve response values ​​corresponding to the reference curve and the target curve. The training dataset is input into a multivariate regression algorithm and a random forest algorithm, respectively, to establish a first mapping relationship between the reference curve and the target curve corresponding to the multivariate regression algorithm, and a second mapping relationship between the reference curve and the target curve corresponding to the random forest algorithm. The second determining module is used to input the reference curves in the test dataset into the first mapping relationship and the second mapping relationship respectively, obtain a first predicted target curve through the first mapping relationship, and obtain a second predicted target curve through the second mapping relationship; compare the first predicted target curve and the second predicted target curve with the target curves in the test dataset respectively, and determine the mapping relationship corresponding to the predicted target curve with the higher matching degree as the optimal mapping relationship; The mapping module is used to input the data of the enlarged well section of the reference curve into the optimal mapping relationship, and map the data of the enlarged well section of the predicted target curve. A correction module is used to correct the data of the enlarged well section of the predicted target curve using the data of the enlarged well section.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the well logging curve correction method for coal-bearing formations as described in any one of claims 1 to 5.

8. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the well logging curve correction method for coal-bearing formations as described in any one of claims 1 to 5.

Images