A fully automatic logging signal depth calibration method

Abstract

The present invention provides a fully automatic logging signal depth calibration method, which belongs to the field of depth calibration of logging signals. The method includes acquiring logging data and performing preprocessing; performing segmentation preprocessing on the preprocessed data; based on the segmentation results, using the segment mask mechanism to perform segment matching processing on the technical A segment and the completion logging segment to obtain a valid segment set; based on the valid segment set, using the DTW-based segment correlation evaluation and the overall seven-point method to process the segment matching results to obtain continuous and non-overlapping segments; performing depth adjustment on the continuous and non-overlapping segments to complete the fully automatic depth calibration of the logging signal. The present invention solves the problems of ignoring physical coverage and scale differences in logging segment matching, which easily leads to invalid correlation calculations, fails to utilize the dynamic characteristics of time series, has limited matching accuracy, insufficient utilization of segment matching results, lacks a globally unified segment analysis mechanism, has a rough depth alignment method, cannot eliminate subtle offsets between data, and insufficient consideration of boundary processing and cumulative error control.

Description

Technical Field

[0001] The invention belongs to the technical field of depth calibration of well logging signals, and in particular relates to a fully automatic depth calibration method for well logging signals. Background Art

[0002] Depth calibration of well logging signals is an essential step in oil and gas exploration data processing. In actual logging, multiple trips or logging with different tools often result in inconsistent depths. For example, depth drift occurs across wells or logging methods (LWD, wireline logging). Factors such as wireline elastic elongation, wellbore conditions (friction, slip), and measurement hysteresis can all lead to depth errors in the recorded log curves. Without depth calibration, logging curves from different wells will not align at the same depth, leading to confusion in the correspondence between stratigraphic features and severely impacting the accuracy of geological horizon comparison and reservoir interpretation.

[0003] With the advancement of deep learning technology, a method for automatic depth matching of well log curves based on a one-dimensional convolutional neural network (1DCNN) has been proposed. This method uses a CNN model to learn the depth shift relationship between the reference well log (EWL) and the curve to be registered (LWD) within a certain window segment, and directly regresses and outputs the optimal alignment offset. The overall network structure consists of three convolutional layers and two fully connected layers, and combines batch normalization and dropout to prevent overfitting. During training, labeled samples are constructed by introducing artificially simulated offsets in the spliced ​​curves, enabling model fitting and evaluation of gamma log types.

[0004] This method has improved the automation level of deep matching to a certain extent and demonstrated strong feature extraction and pattern recognition capabilities. However, analysis still reveals the following limitations in engineering applications:

[0005] The source of training samples is not realistic enough: the model is trained with artificially generated simulated displacement samples, which fail to fully cover the non-rigid deformations present in actual well logging, such as local tension, compression, and other complex displacement patterns;

[0006] Insufficient validation samples: The model has only been tested on two wells (16 / 1-9 and 16 / 1-21S), and its generalizability to more complex geological conditions has yet to be verified.

[0007] Manual segmentation and window processing is still required: the network input is a sliding window segment of fixed length, and pre-processing still requires manual or rule-based segmentation, failing to achieve a complete closed loop from full-well data to misalignment prediction;

[0008] No modeling mechanism for stretch / squeeze effects: This method only regresses an overall translation value and cannot identify and correct segment deformations. Summary of the Invention

[0009] In view of the above-mentioned deficiencies in the prior art, the present invention provides a fully automatic well logging signal depth calibration method to solve the above-mentioned problems.

[0010] In order to achieve the above objectives, the present invention adopts a technical solution: a fully automatic well logging signal depth calibration method, comprising the following steps:

[0011] Acquire well logging data and perform preprocessing;

[0012] The pre-processed logging data is subjected to segmented logging curve pre-processing, wherein the original logging curve includes a technical A curve and a completion logging curve, and both the technical A curve and the completion logging curve are curves formed by gamma data measured on the same well;

[0013] Based on the segmentation results, the segment mask mechanism is used to perform segment matching processing on the technology A segment and the completion logging segment to obtain a valid segment set, where the technology A segment is the segment corresponding to the technology A curve, and the completion logging segment is the segment corresponding to the completion logging curve;

[0014] According to the valid segment set, segment correlation evaluation based on dynamic time warping (DTW) and the overall seven-point method are used to process the segment matching results to obtain continuous and non-overlapping segments.

[0015] Depth adjustment is performed on continuous and non-overlapping sections to complete fully automatic depth calibration of logging signals.

[0016] Furthermore, the pre-processed logging data is subjected to logging curve segmentation pre-processing, which is specifically as follows:

[0017] Based on the pre-processed logging data, the original logging curve is , perform the first local minimum detection on the original logging curve and obtain the minimum depth set;

[0018] Construct characteristic curves and arrange the minimum depth set in ascending order according to depth to generate a discrete sequence, which is then used as a new logging curve. ;

[0019] New logging curve Perform a second local minimum detection to determine the segmentation point;

[0020] Based on the determined segmentation points, depth segmentation points and segments are generated to complete the segmentation preprocessing of the logging curve.

[0021] Furthermore, the expression of the segment generated by segment preprocessing of the logging curve is as follows:

[0022] ;

[0023] ;

[0024] ;

[0025] ;

[0026] ;

[0027] ;

[0028] in, Indicates the j Split depth segments, Indicates the j The segmentation depth, Equal to the original depth coordinate of the local minimum , Indicates the +1 split depth, Equal to the original depth coordinate of the local minimum represents the segment of the second local minimum segmentation, Indicates the first The depth coordinates of the samples, Corresponding to the position of the second local minimum, Represents the index set of the second local minimum, Representing a collection In ascending order of depth Segment sampling index, Represents a discrete curve In the Gamma logging values ​​at samples, and Represents The gamma logging values ​​of two adjacent sampling points are used to determine Is it a local minimum? represents the number of samples, Represents the discrete curve corresponding to the first local minimum index set, Indicates Gamma logging values ​​in depth coordinates, Represents a discrete curve The universal sampling subscript of Indicates Gamma logging values ​​in depth coordinates, Indicates the first depth coordinates, represents the first local minimum depth set, Indicates depth The gamma log value at represents the depth variable, Represents the original discrete data sampling subscript, N Represents the total number of samples of the original discrete data, Represents the original discrete data The depth coordinates of the sampling points, and Respectively represent The gamma log values ​​of two adjacent samples are used for local minimum detection. Represents the value at the kth position in a discrete sequence.

[0029] Furthermore, the section matching process of the technology A section and the completion logging section is specifically as follows:

[0030] Assume that the technical section A is The completion logging section is ,in, Indicates the first The starting depth of each segment, Indicates the first The end depth of the segment, Indicates the number of completion logging data m The starting depth of each segment, Indicates the number of completion logging data m The end depth of each segment;

[0031] Perform overlapping mask filtering on the Technology A section and the completion logging section:

[0032] ;

[0033] in, Indicates the segment overlap length, if , then remember , this segment does not participate in the relevance evaluation, Indicates the segment mask mark, mask mark 0 means discard, mask mark 1 means retain;

[0034] After filtering the results with overlapping masks, the length consistency mask processing is performed on the technical A section and the completion logging section, and the resulting valid section pair set is:

[0035] ;

[0036] ;

[0037] ;

[0038] in, Indicates the first m Section and completion logging W n The length difference ratio of the segment on the depth axis is When , this segment does not participate in the relevance evaluation, Segment representing technology A m length, Indicates that the completion logging is in the section n length.

[0039] Furthermore, the expression of the valid segment pair set is as follows:

[0040] ;

[0041] in, Represents a set of valid segment pairs, Represents a segment pair.

[0042] Furthermore, the segment matching result is processed as follows:

[0043] Based on the valid segment pair set, interpolation preprocessing is performed on each segment data to obtain two data sequences with the same length. ,in, Representation Technology A The gamma sequence after segment interpolation, Indicates completion logging W n Gamma sequence after segment mapping to the sampling grid of technique A;

[0044] Normalize the two acquired data sequences with the same length;

[0045] Based on the normalized processing results, the dynamic time warping (DTW) distance between the technical A section and the completion logging section is calculated. ;

[0046] Based on the dynamic time warping (DTW) distance between the technical A section and the completion logging section , define the dimensionless similarity coefficient ;

[0047] All calculated dimensionless similarity coefficients Sort and select the first three segment pairs with the highest similarity:

[0048] ;

[0049] in, represents the set of the top three most similar segment pairs, Indicates the selected Technical A section, Indicates the selected The W section of the completion log is i Indicates the sampling number within the segment, 、 and They represent the similarity coefficients corresponding to the top three strongest segment pairs;

[0050] Assume that the overall depth range of the original logging data is ,in, and They represent the lower and upper bounds of the overall depth range of the original well logging data;

[0051] Summarize the boundary points of the first three segment pairs with the highest similarity and add the upper and lower limits of the overall depth to obtain the segmentation depth set :

[0052] ;

[0053] in, and Respectively represent A technology The starting and ending depths of the segment;

[0054] Segmentation depth set Sort by depth from smallest to largest to divide the original well logging data into seven continuous and non-overlapping segments:

[0055] ;

[0056] ;

[0057] in, Indicates the first consecutive and non-overlapping depth intervals, Indicates the first k consecutive and non-overlapping depth intervals, and Respectively represent the first The starting and ending points of consecutive and non-overlapping depth intervals, k Indicates the segment number, and Respectively represent the first k The starting and ending points of consecutive, non-overlapping depth intervals.

[0058] Furthermore, the expression of the dimensionless similarity coefficient is as follows:

[0059] ;

[0060] ;

[0061] ;

[0062] ;

[0063] in, represents the dimensionless similarity coefficient, Indicates segment The dynamic time warping DTW distance of the pair, the input is the two normalized sequences and , represents the dynamic time warping DTW distance function, express The gamma sequence obtained after normalization is express The gamma sequence obtained after normalization is Representation Technology A m The original gamma sequence of the segment, and Respectively The mean and standard deviation of the gamma series, Indicates completion logging W n The original amplitude sequence after the segment is mapped to the sampling grid of technique A, and Respectively The mean and standard deviation of the gamma series.

[0064] Furthermore, the depth adjustment of the continuous and non-overlapping segments is specifically performed as follows:

[0065] The seven consecutive and non-overlapping segments are adjusted using a depth coarse adjustment method based on the seven-point method;

[0066] The fine depth correction method of well logging data based on area feature matching is used to correct the results of rough depth adjustment to complete the fully automatic depth correction of the logging signal.

[0067] Furthermore, the depth rough adjustment method based on the seven-division method is used to adjust seven consecutive and non-overlapping segments, which is specifically as follows:

[0068] Based on seven consecutive and non-overlapping segments, the default depth arrangement of each segment is from shallow to deep, and the first segment of the technology A segment is Make a depth alignment adjustment to move the first segment of the Tech A segment to the The maximum depth of the well is aligned to the maximum depth of the completion logging section:

[0069] ;

[0070] in, Indicates the first paragraph The depth value after the upper boundary is aligned, Indicates the original depth value of the first segment among the seven segments. Indicates the first paragraph of technology A The maximum depth sample point of the internal measurement data, Indicates the first section of completion logging The maximum depth sample point of the internal measurement data, seven segments are seven consecutive and non-overlapping segments;

[0071] The end of the technical A section Make a depth alignment adjustment to move the end of the Tech A segment The maximum depth of the well is aligned to the minimum depth of the completion logging section:

[0072] ;

[0073] in, Indicates the tail segment The depth value after the lower boundary is aligned, Indicates the end of the completion logging section The minimum depth sample point of the internal measurement data, Indicates the end of the technical A section Minimum depth sample point of internal measurement data;

[0074] In the Technology A section, the section scale normalization process is performed on the middle sections except the first and last sections to ensure that the length of the Technology A depth section is consistent with the completion logging section;

[0075] The depth position balance processing is performed on the middle segment after the segment scale normalization processing to align the starting point of the technology A segment and complete the rough depth alignment of each segment of the technology A data.

[0076] Furthermore, the method for fine depth correction of well logging data based on area feature matching is used to correct the rough depth adjustment result, which is specifically as follows:

[0077] Based on the depth rough adjustment results, in the technical A section and the corresponding section of the completion logging Select a reference point from , where Indicates the seventh section of the completion logging. i The original depth value of each segment, Indicates the seventh section of Technology A i The original depth value of each segment;

[0078] With each reference point as the center, select a local segment of fixed length and calculate the area characteristics of the completion logging data and technical A data in this segment:

[0079] ;

[0080] ;

[0081] in, Indicates Technology A segment The area enclosed by the gamma logging value and depth, Indicates the completion logging section The area enclosed by the gamma logging value and depth, Indicates the half-width of the window, Indicates the depth of the reference point of technology A. Indicates the depth of the completion logging reference point. Indicates the completion logging at depth coordinate Gamma logging value at ;

[0082] For the current section, with a step length of 1m, starting from the current section, calculate the area enclosed by the technology A data and the depth axis in the current section and normalize it to unit depth;

[0083] Based on the area characteristics, the segment that best matches the normalized segment is found in the completion logging data, and the area of ​​the best matching segment is calculated and normalized.

[0084] Calculate the error function based on the normalized area of ​​the best matching segment and search for a value that minimizes the error within a preset range;

[0085] Based on the search results, the current technology A section is mapped to the completion logging section, and the aligned depth position is updated according to the mapping results;

[0086] Continue processing the next 1m segment, complete processing for all segments, and complete the correction of the depth coarse adjustment result.

[0087] Beneficial effects of the present invention:

[0088] (1) Introducing a segment masking mechanism to effectively avoid invalid matching calculations. This invention performs a dual screening of the degree of overlap and length consistency between segments to ensure that the segments involved in similarity evaluation have physical and structural significance. Experimental measurements show that this screening strategy can eliminate approximately 30% of invalid segment pairs that have no physical significance, significantly improving computational efficiency and accuracy.

[0089] (2) The dynamic time warping (DTW) algorithm improves the sensitivity and stability of dynamic matching. Compared with traditional Euclidean distance or Pearson correlation coefficient, the dynamic time warping (DTW) algorithm can better handle nonlinear deformations, and the average similarity assessment accuracy is improved by approximately 12–18%. It has shown greater adaptability and robustness in various logging curve tests, and is particularly suitable for situations where the curve is oscillating or has severe offset.

[0090] (3) The seven-part structural division takes into account both local focus and overall continuity. This paper combines key high-correlation segments with the overall depth structure to establish a "3+4" segmentation system (3 high-similarity segments + 4 background segments), providing a unified reference for subsequent segment weighted modeling and feature extraction, effectively reducing the risk of local overfitting.

[0091] (4) The dual-layer depth correction strategy effectively improves alignment accuracy. This paper proposes a two-stage depth alignment method of "coarse correction + fine correction". In the coarse correction stage, the depth scale is unified, and in the fine correction stage, an integral area matching mechanism is introduced to achieve sub-meter depth fine-tuning. Experimental verification shows that the standard deviation of the alignment error is reduced from the original approximately 1.2 meters to below 0.35 meters, and the overall alignment accuracy is improved by approximately 70%.

[0092] (5) Good versatility and automatic adaptability. The automatic segmentation, dynamic time warping (DTW) matching, and integral depth calibration methods used in the present invention all rely on specific curve types and can be extended to various logging curves such as density, gamma, neutron, and acoustic waves. They are adaptable to data with different sampling frequencies and noise levels and have good portability and automatic processing capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0093] Figure 1 Flow chart of the method of the present invention. DETAILED DESCRIPTION

[0094] The specific embodiments of the present invention are described below to facilitate understanding of the present invention by those skilled in the art. However, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the appended claims, these changes are obvious, and all inventions and creations utilizing the concepts of the present invention are protected.

[0095] Example

[0096] In this embodiment, technology A refers to instantaneous high-power pulse source emission technology.

[0097] like Figure 1 As shown, the present invention provides a fully automatic well logging signal depth calibration method, and its implementation method is as follows:

[0098] S1. Obtain logging data and perform preprocessing;

[0099] In this embodiment, the preprocessing performed on the acquired well logs includes smoothing.

[0100] S2. Perform segmented logging curve preprocessing on the preprocessed logging data. The implementation method is as follows:

[0101] Based on the pre-processed logging data, the original logging curve is , perform local minimum detection on the original logging curve and obtain the minimum depth set, where the original logging curve Including Technology A Curve and completion logs , Technology A Curve and completion logs All curves are formed by gamma data measured in the same well;

[0102] Construct characteristic curves and sort the minimum depth sets in ascending order of depth , generate a discrete sequence and use the discrete sequence as a new logging curve ;

[0103] New logging curve Perform local minimum detection and determine the segmentation point;

[0104] Based on the determined segmentation points, depth segmentation points and segments are generated to complete the segmentation preprocessing of the logging curve.

[0105] In this embodiment, the original logging curve is ,in, d Indicates depth in meters. The automatic segmentation process involves the following four steps:

[0106] 1. The first local minimum detection to obtain the minimum depth set and its amplitude :

[0107] ;

[0108] 2. Construct a characteristic curve and arrange the minimum depth set in ascending order of depth to generate a discrete sequence:

[0109] , ;

[0110] and consider it as a new well log ;

[0111] 3. Second local minimum detection (determine the segmentation point):

[0112] ;

[0113] 4. Generate depth segmentation points and segments:

[0114] , ;

[0115] in, Indicates the j Split depth segments, Indicates the j The segmentation depth, Equal to the original depth coordinate of the local minimum , Indicates the +1 split depth, Equal to the original depth coordinate of the local minimum represents the segment of the second local minimum segmentation, Indicates the first The depth coordinates of the samples, Corresponding to the position of the second local minimum, Represents the index set of the second local minimum, Representing a collection In ascending order of depth Segment sampling index, Represents a discrete curve In the Gamma logging values ​​at samples, and Represents The gamma logging values ​​of two adjacent sampling points are used to determine Is it a local minimum? represents the number of samples, Represents the discrete curve corresponding to the first local minimum index set, Indicates Gamma logging values ​​in depth coordinates, Represents a discrete curve The universal sampling subscript of Indicates Gamma logging values ​​in depth coordinates, Indicates the first depth coordinates, represents the first local minimum depth set, Indicates depth The gamma log value at represents the depth variable, Represents the original discrete data sampling subscript, N Represents the total number of samples of the original discrete data, Represents the original discrete data The depth coordinates of the sampling points, and Respectively represent The gamma log values ​​of two adjacent samples are used for local minimum detection. Represents the value at the kth position in a discrete sequence.

[0116] In summary, The original logging curve Divide into several non-overlapping segments ,in, Indicates the depth segmentation point.

[0117] In this embodiment, in order to maintain the consistency of multi-source data, the above four steps are synchronously applied to the technology A curve. and completion logs , ensuring that both are split at the same depth Align the top of the image to lay a unified segmentation foundation for subsequent fusion analysis.

[0118] S3. Based on the segmentation results, the segment mask mechanism is used to perform segment matching processing on the technology A segment and the completion logging segment to obtain a valid segment set, where the technology A segment is the segment corresponding to the technology A curve, and the completion logging segment is the segment corresponding to the completion logging curve. The implementation method is as follows:

[0119] Assume that the technical section A is The completion logging section is ,in, Indicates the technology A data The starting depth of each segment, Indicates the technology A data The end depth of the segment, Indicates the number of completion logging data m The starting depth of each segment, Indicates the number of completion logging data m The end depth of each segment;

[0120] Perform overlapping mask filtering on the technical A section and the completion logging section;

[0121] After filtering the results with overlapping masks, the length consistency mask processing is performed on the technical A section and the completion logging section to obtain a set of valid section pairs.

[0122] In this embodiment, it is assumed that the technical section A is The completion logging section is ,in, Indicates the technology A data The starting depth of each segment, Indicates the technology A data The end depth of the segment, Indicates the number of completion logging data m The starting depth of each segment, Indicates the number of completion logging data m The end depth of each segment is filtered using two levels of masking:

[0123] 1. Overlap mask: Perform overlapping mask filtering on the technical A section and the completion logging section:

[0124] ;

[0125] in, Indicates the segment overlap length, the first m Section and completion logging W n The actual overlapping thickness of the segments on the depth axis, if , then remember , this segment does not participate in the relevance evaluation, Indicates the segment mask mark, mask mark 0 means discard, mask mark 1 means retain.

[0126] 2. Length consistency mask: After filtering the results with overlapping masks, length consistency mask processing is performed on the technical A section and the completion logging section, resulting in a set of valid section pairs:

[0127] ;

[0128] ;

[0129] ;

[0130] in, Indicates the first m Section and completion logging W n The length difference ratio of the segment on the depth axis is When , this segment does not participate in the relevance evaluation, Segment representing technology A m length, Indicates that the completion logging is in the section n length.

[0131] Ultimately, only when When the segment Only then enter the correlation evaluation process between technology A and completion logging.

[0132] S4. Based on the valid segment set, segment matching results are processed using segment correlation evaluation based on dynamic time warping (DTW) and the overall seven-point method to obtain continuous and non-overlapping segments, which are specifically:

[0133] Based on the valid segment pair set, interpolation preprocessing is performed on each segment data to obtain two data sequences with the same length. ,in, Representation Technology A The gamma sequence after segment interpolation, Indicates completion logging W n Gamma sequence after segment mapping to the sampling grid of technique A;

[0134] Normalize the two acquired data sequences with the same length;

[0135] Based on the normalized processing results, the dynamic time warping (DTW) distance between the technical A section and the completion logging section is calculated. ;

[0136] Based on the dynamic time warping (DTW) distance between the technical A section and the completion logging section , define the dimensionless similarity coefficient ;

[0137] All calculated dimensionless similarity coefficients Sort and select the first three segment pairs with the highest similarity;

[0138] Assume that the overall depth range of the original logging data is ,in, and They represent the lower and upper bounds of the overall depth range of the original well logging data;

[0139] Summarize the boundary points of the first three segments with the highest similarity and add the upper and lower limits of the overall depth of the data to obtain the segmentation depth set ;

[0140] Segmentation depth set The original well logging data are sorted in ascending order of depth to divide them into seven continuous and non-overlapping segments.

[0141] In this embodiment, the dynamic time warping (DTW) is summarized as follows: for two time series of arbitrary lengths, , Dynamic Time Warping (DTW) finds the best alignment path for the two by nonlinearly scaling the time axis. , , ,in, and Represents the tail element (last sample value) of the two time series, Represents time series The number of sampling points.

[0142] First, construct the local cost matrix (Usually take the absolute difference or squared difference , and then use dynamic programming to calculate the cumulative cost :

[0143] ;

[0144] in, and All represent the common sampling values ​​of the sequence, represents the local distance function, represents the local cost matrix element, 、 Represents and Respectively represent the calculation of cumulative cost The cumulative cost of the left, top, and left-top neighbors referenced when represents the cumulative cost at the starting point, and , represents the local cost of the starting point, .

[0145] Cumulative cost of the optimal path After path length normalization, it is defined as the dynamic time warping DTW distance:

[0146] ;

[0147] in, Represents dynamic time warping distance: the cumulative cost of the aligned paths is calculated based on the optimal matching path length K The normalized result is K Indicates the optimal matching path length.

[0148] The time complexity of this algorithm is generally , effectively reduced to ,in, Indicates the time complexity level, the number of operations in the worst case is Proportional, Represents time series 、 length, Indicates asymptotic complexity, placeholder represents the independent variable, represents the width of the Sakoe–Chiba or Itakura restraint belt, Represents the larger of the two sequence lengths.

[0149] In this embodiment, the dynamic time warping (DTW) correlation based on the mask segment is calculated as follows:

[0150] Based on the valid segment pair set, interpolation preprocessing is performed on each segment data to obtain two data sequences with the same length. , taking the depth sequence of technology A data as the benchmark, the completion logging data sequence is interpolated to the corresponding depth point of technology A data by linear interpolation to obtain two data sequences with the same length, where, Representation Technology A The gamma sequence after segment interpolation, Indicates completion logging W n Gamma sequence after segment mapping to the Technique A sampling grid.

[0151] Then the two data sequences with the same length are normalized, such as using z-score normalization:

[0152] , ;

[0153] Then calculate the dynamic time warping (DTW) distance between the technical A section and the completion logging section:

[0154] ;

[0155] And define the corresponding dimensionless similarity coefficient as:

[0156] ;

[0157] in, represents the dimensionless similarity coefficient, Indicates segment The dynamic time warping DTW distance of the pair, the input is the two normalized sequences and , represents the dynamic time warping DTW distance function, express The gamma sequence obtained after normalization is express The gamma sequence obtained after normalization is Representation Technology A m The original gamma sequence of the segment, and Respectively The mean and standard deviation of the gamma series, Indicates completion logging W n The original amplitude sequence after the segment is mapped to the sampling grid of technique A, and Respectively The mean and standard deviation of the gamma series.

[0158] In this embodiment, the three segment pairs with the highest correlation are selected as follows:

[0159] All the calculated dimensionless similarity coefficients Sort and select the top three segment pairs with the highest similarity:

[0160] ;

[0161] Specifically expressed as:

[0162] ;

[0163] in, represents the set of the top three most similar segment pairs, Indicates the selected Technical A section, Indicates the selected The W section of the completion log is i Indicates the sampling number within the segment, 、 and They represent the similarity coefficients of the top three strongest segment pairs respectively.

[0164] In this embodiment, the overall seven-point method is as follows:

[0165] The overall depth range of the original logging data is ,in, and They represent the lower and upper bounds of the overall depth range of the original well logging data;

[0166] Summarize the boundary points of the three segment pairs selected above and add the depth upper and lower limits of the entire data:

[0167] ;

[0168] in, and Respectively represent A technology The starting and ending depths of the segment;

[0169] Then split the depth set Sort by depth from smallest to largest:

[0170] ;

[0171] Indicates the lower bound of the overall depth range of the original logging data, Indicates the upper bound of the overall depth range of the original logging data, Indicates that the set is sorted in ascending order The elements, .

[0172] In this way, the original logging data is divided into seven continuous and non-overlapping segments:

[0173] ;

[0174] ;

[0175] in, Indicates the first consecutive and non-overlapping depth intervals, Indicates the first k consecutive and non-overlapping depth intervals, and Respectively represent the first The starting and ending points of consecutive and non-overlapping depth intervals, k Indicates the segment number, and Respectively represent the first k The starting and ending points of consecutive, non-overlapping depth intervals.

[0176] The three highly correlated segments were selected as feature segments, and the remaining four segments were considered background segments. This segmentation strategy not only highlights the key segments with the highest correlation, but also takes into account the continuity and integrity of the overall data, laying a good structural foundation for subsequent data mining and interpretation.

[0177] S5. Adjust the depth of continuous and non-overlapping sections to complete the fully automatic depth calibration of the logging signal. The implementation method is as follows:

[0178] The seven consecutive and non-overlapping segments are adjusted using a depth coarse adjustment method based on the seven-point method. The implementation method is as follows:

[0179] Based on seven consecutive and non-overlapping segments, the default depth arrangement of each segment is from shallow to deep, and the first segment of the technology A segment is Make a depth alignment adjustment to move the first segment of the Tech A segment to the The maximum depth of the well is aligned to the maximum depth of the completion log;

[0180] The end of the technical A section Make a depth alignment adjustment to move the end of the Tech A segment The maximum depth of the well is aligned to the minimum depth of the completion log;

[0181] In the Technology A section, the section scale normalization process is performed on the middle sections except the first and last sections to ensure that the length of the Technology A depth section is consistent with the completion logging section;

[0182] The depth position balance processing is performed on the middle segment after the segment scale normalization processing to align the starting point of the technology A segment and complete the rough depth alignment of each segment of the technology A data.

[0183] The fine depth correction method for well logging data based on area feature matching is used to correct the rough depth adjustment results and complete the fully automatic depth correction of the logging signal. The implementation method is as follows:

[0184] Based on the depth rough adjustment results, in the technical A section and the corresponding section of the completion logging Select a reference point from , where Indicates the seventh section of the completion logging. i The original depth value of each segment, Indicates the seventh section of Technology A i The original depth value of each segment;

[0185] With each reference point as the center, select a local segment of fixed length and calculate the area characteristics of the completion logging data and technical A data in this segment:

[0186] ;

[0187] ;

[0188] in, Indicates Technology A segment The area enclosed by the gamma logging value and depth, Indicates the completion logging section The area enclosed by the gamma logging value and depth, Indicates the half-width of the window, Indicates the depth of the reference point of technology A. Indicates the depth of the completion logging reference point. Indicates the completion logging at depth coordinate Gamma logging value at ;

[0189] For the current section, with a step length of 1m, starting from the current section, calculate the area enclosed by the technology A data and the depth axis in the current section and normalize it to unit depth;

[0190] Based on the area characteristics, the segment that best matches the normalized segment is found in the completion logging data, and the area of ​​the best matching segment is calculated and normalized.

[0191] Calculate the error function based on the normalized area of ​​the best matching segment and search for a value that minimizes the error within a preset range;

[0192] Based on the search results, the current technology A data segment is mapped to the completion logging data segment, and the aligned depth position is updated according to the mapping results;

[0193] Continue processing the next 1m segment, complete processing for all segments, and complete the correction of the depth coarse adjustment result.

[0194] In this embodiment, the depth rough adjustment method based on the seven-point method is as follows:

[0195] After the aforementioned seven-part overall processing, the completion logging data and the Technical A data are divided into seven depth segments:

[0196] The completion logging sections are: ;Technical A section is: ;

[0197] In order to unify the depth benchmark and facilitate subsequent detailed analysis, a rough adjustment is made to the depth range of technology A. The specific steps are as follows:

[0198] 1. First section depth ( Segment) Adjustment: By default, the depth of each segment is arranged from shallow to deep. First, the first segment of the technical A segment To make a depth alignment adjustment:

[0199] ;

[0200] in, Indicates the first paragraph The depth value after the upper boundary is aligned, Indicates the original depth value of the first segment among the seven segments (before alignment). Indicates the first paragraph of technology A The maximum depth sample point of the internal measurement data, Indicates the first section of well logging The maximum depth sample point of the internal measurement data, the seven segments are seven continuous and non-overlapping segments.

[0201] That is, the maximum depth of the first section of Technology A is aligned to the maximum depth of the completion logging.

[0202] 2. Adjustment of the final depth (b7 segment): Then perform similar operations on the final b7 segment of the Technical A data. The adjustment formula is:

[0203] ;

[0204] in, Indicates the tail segment The depth value after the lower boundary is aligned, Indicates the end of the completion technology The minimum depth sample point of the internal measurement data, Indicates the end of technology A The minimum depth sample point of the internal measurement data.

[0205] That is, the minimum depth of the end section of Technology A is aligned to the minimum depth of the completion logging.

[0206] 3. Depth adjustment of other intermediate segments (taking segment b2 as an example): For segments other than the first and last segments, taking segment b2 as an example, the depth scale and position are adjusted through the following two steps:

[0207] Step 1: Normalize the segment scale (stretch or compress) to ensure that the length of the depth segment is consistent with the completion logging segment:

[0208] ;

[0209] in, Indicates segment The maximum and minimum depths, and Indicates segment The maximum and minimum depths.

[0210] Step 2: Shift the depth position to align the segment start point:

[0211] ;

[0212] in, Indicates the aligned , Indicates segment The maximum depth of Indicates segment The maximum depth of

[0213] Through the above steps, the rough depth alignment of each segment of the technology A data is completed, the depth benchmark for subsequent analysis is unified, and the consistency of each data segment in the depth scale is ensured.

[0214] In this embodiment, the previous step achieves completion data through linear depth adjustment. and Technology A Data However, due to the slight differences in the data sources, there is still some misalignment between the two sets of data after the rough depth calibration. To further improve the matching accuracy, it is necessary to introduce a fine depth calibration method to make more precise adjustments to the depth. The fine depth calibration adopts a strategy based on area feature matching, and calculates the area feature of the curve through numerical integration (shape integral method trapz), and further matches the corresponding paragraphs. and The following is a detailed description of the specific steps for fine depth calibration:

[0215] 1. Reference point selection: First, select the corresponding section of each team. Select a reliable reference point. The reference point should be located inside the segment, be representative and easy to identify in the two sets of data. The selection of the reference point provides a starting point for the subsequent area matching, so that there is a clear anchor point for fine-tuning. Segment Selection ,for Segment Selection , select other sections uniformly .

[0216] 2. Area feature construction: Centered at each reference point, select a local segment of a certain length (e.g., ±2m) and calculate the area feature of the completion data and Technical A data within that segment. The area is defined as the integral of the GR curve over depth, using trapezoidal integration (trapz) as an approximation.

[0217] 3. Best matching point search:

[0218] By segment For example, Segment is , Segment is In order to achieve more accurate depth alignment, the local area matching method is used to The specific steps are as follows:

[0219] First, with a step length of 1m, start Start and process in sequence Segment. For each 1m segment, that is, segment , do the following:

[0220] Step 1: Calculate the area enclosed by the technical A data and the depth axis within the segment , and normalized to unit depth:

[0221] ;

[0222] in, express Normalized to its length, Indicates segment The area enclosed by the gamma logging value (GR) and the depth is calculated using the trapezoidal integral approximation.

[0223] Step 2: Completion logging data In the search, find the The best matching segment. Specifically in the range Select any , calculate the corresponding segment The area of ​​​​and normalized:

[0224] ;

[0225] in, Indicates the end depth of the segment, the value range is , express The normalized value, Represents completion logging data from arrive area.

[0226] Step 3: Calculate the error function , and in Search within the range to minimize the error :

[0227] ;

[0228] in, Represents completion logging data exist and x The result after area normalization is: express and x The result after area normalization is: express z The optimal solution, x express x axis, z It means selecting any value in the range [a1+0.95,a1+1.05).

[0229] Step 4: Set the current technology A number segment Mapping to completion log segments by linear stretching (or compression) , the stretching formula is:

[0230] ;

[0231] ;

[0232] in, Indicates the new depth value after adjustment. Indicates original The depth value within the segment.

[0233] Step 5: Update the depth position after alignment:

[0234] ;

[0235] ;

[0236] in, Represents completion logging data A new starting point, Represents completion logging data The new starting point is the next point to be processed.

[0237] Move the new reference position to , preparing for the subsequent matching of 1-meter segments.

[0238] Step 6: Continue to process the next 1m segment, i.e. Repeat the above steps until the whole Segment Processing completed.

[0239] In summary, the present invention solves the following problems based on the above design:

[0240] 1. Ignoring physical coverage and scale differences in well logging segment matching can easily lead to invalid correlation calculations. Traditional methods fail to fully consider the actual overlap and length consistency between segments when matching segments, often resulting in many invalid or even misleading segments being included in similarity calculations. This invention introduces a dual-masking mechanism that filters segment overlap and scale consistency separately, retaining only valid segment pairs and significantly improving matching efficiency and accuracy.

[0241] 2. Failure to leverage the dynamic characteristics of time series, limiting matching accuracy. Traditional curve matching often uses simple distance or correlation metrics, which are difficult to adapt to curve shape changes or nonlinear alignment. This paper introduces the Dynamic Time Warping (DTW) algorithm, which achieves high-precision similarity calculation through nonlinear time axis scaling, effectively capturing the dynamic correlation between Technology A and completion logging data.

[0242] 3. Segment matching results are underutilized, and a globally unified segment analysis mechanism is lacking. Existing methods often only perform local analysis on a few highly correlated segments, failing to establish a unified, global standard for deep segmentation, impacting the consistency of subsequent analysis. This invention utilizes a "seven-part holistic approach" to unify the modeling of key and background segments, balancing matching focus with overall segment continuity, thus improving global interpretation capabilities.

[0243] 4. Depth alignment methods are crude and cannot eliminate subtle offsets between data. Traditional depth correction methods are mostly based on simple linear translation or scaling, which cannot handle small misalignments between different data sources caused by sampling errors, tool response differences, etc. This paper proposes a fine-grained depth correction strategy based on area features, which achieves precise fine-tuning through local integral matching, further improving data alignment accuracy.

[0244] 5. Existing methods lack sufficient consideration for boundary processing and cumulative error control, impacting subsequent modeling accuracy. Many automatic alignment methods fail to fully consider the continuity and consistency of segment boundaries, easily leading to dislocation or jumps at the boundaries, and cumulative errors exacerbate with depth. In this method, the first and last segments are aligned using boundary points, and the middle segment uses a combination of scale normalization and position translation to systematically reduce depth error transmission and ensure depth consistency across the entire segment.

Claims

1. A fully automatic logging signal depth calibration method, characterized in that: The following steps are involved: Acquire well logging data and perform preprocessing; The pre-processed logging data is subjected to segmented logging curve pre-processing, wherein the original logging curve includes a technical A curve and a completion logging curve, and both the technical A curve and the completion logging curve are curves formed by gamma data measured on the same well; Based on the segmentation results, the segment mask mechanism is used to perform segment matching processing on the technology A segment and the completion logging segment to obtain a valid segment set, where the technology A segment is the segment corresponding to the technology A curve, and the completion logging segment is the segment corresponding to the completion logging curve; According to the valid segment set, segment correlation evaluation based on dynamic time warping (DTW) and the overall seven-point method are used to process the segment matching results to obtain continuous and non-overlapping segments. The processing of the segment matching results is specifically as follows: Based on the valid segment pair set, interpolation preprocessing is performed on each segment data to obtain two data sequences with the same length. ,in, Representation Technology A The gamma sequence after segment interpolation, Indicates completion logging W n Gamma sequence after segment mapping to the sampling grid of technique A; Normalize the two acquired data sequences with the same length; Based on the normalized processing results, the dynamic time warping (DTW) distance between the technical A section and the completion logging section is calculated. ; Based on the dynamic time warping (DTW) distance between the technical A section and the completion logging section , define the dimensionless similarity coefficient ; All calculated dimensionless similarity coefficients Sort and select the first three segment pairs with the highest similarity: ； in, represents the set of the top three most similar segment pairs, Indicates the selected Technical A section, Indicates the selected The W section of the completion log is i Indicates the sampling number within the segment, 、 and They represent the similarity coefficients corresponding to the top three strongest segment pairs; Assume that the overall depth range of the original logging data is ,in, and They represent the lower and upper bounds of the overall depth range of the original well logging data; Summarize the boundary points of the first three segment pairs with the highest similarity and add the upper and lower limits of the overall depth to obtain the segmentation depth set : ； in, and Respectively represent A technology The starting and ending depths of the segment; Segmentation depth set Sort the data by depth from small to large to divide the original logging data into seven continuous and non-overlapping segments; ； ； in, Indicates the first consecutive and non-overlapping depth intervals, Indicates the first k consecutive and non-overlapping depth intervals, and Respectively represent the first The starting and ending points of consecutive and non-overlapping depth intervals, k Indicates the segment number, and Respectively represent the first k The starting and ending points of consecutive and non-overlapping depth intervals; The expression of the dimensionless similarity coefficient is as follows: ； ； ； ； in, represents the dimensionless similarity coefficient, Indicates segment The dynamic time warping DTW distance of the pair, the input is the two normalized sequences and , represents the dynamic time warping DTW distance function, express The gamma sequence obtained after normalization is express The gamma sequence obtained after normalization is Representation Technology A m The original gamma sequence of the segment, and Respectively The mean and standard deviation of the gamma series, Indicates completion logging W n The original amplitude sequence after the segment is mapped to the sampling grid of technique A, and Respectively Mean and standard deviation of the gamma series; Depth adjustment is performed on continuous and non-overlapping sections to complete fully automatic depth calibration of logging signals.

2. The fully automatic logging signal depth calibration method according to claim 1, characterized in that: The pre-processed logging data is subjected to logging curve segmentation pre-processing, which is specifically as follows: Based on the pre-processed logging data, the original logging curve is , perform the first local minimum detection on the original logging curve and obtain the minimum depth set; Construct characteristic curves and arrange the minimum depth set in ascending order according to depth to generate a discrete sequence, which is then used as a new logging curve. ; New logging curve Perform a second local minimum detection to determine the segmentation point; Based on the determined segmentation points, depth segmentation points and segments are generated to complete the segmentation preprocessing of the logging curve.

3. The fully automatic logging signal depth calibration method according to claim 2, characterized in that: The expression of the segment generated by segment preprocessing of the logging curve is as follows: ； ； ； ； ； ； in, Indicates the j Split depth segments, Indicates the j The segmentation depth, Equal to the original depth coordinate of the local minimum , Indicates the +1 split depth, Equal to the original depth coordinate of the local minimum represents the segment of the second local minimum segmentation, Indicates the first The depth coordinates of the samples, Corresponding to the position of the second local minimum, Represents the index set of the second local minimum, Representing a collection In ascending order of depth Segment sampling index, Represents a discrete curve In the Gamma logging values ​​at samples, and Represents The gamma logging values ​​of two adjacent sampling points are used to determine Is it a local minimum? represents the number of samples, Represents the discrete curve corresponding to the first local minimum index set, Indicates Gamma logging values ​​in depth coordinates, Represents a discrete curve The universal sampling subscript of Indicates Gamma logging values ​​in depth coordinates, Indicates the first depth coordinates, represents the first local minimum depth set, Indicates depth The gamma log value at represents the depth variable, Represents the original discrete data sampling subscript, N Represents the total number of samples of the original discrete data, Represents the original discrete data The depth coordinates of the sampling points, and Respectively represent The gamma log values ​​of two adjacent samples are used for local minimum detection. Represents the value at the kth position in a discrete sequence.

4. The fully automatic well logging signal depth calibration method according to claim 1, characterized in that: The section matching process for the technical A section and the completion logging section is specifically as follows: Assume that the technical section A is The completion logging section is ,in, Indicates the first The starting depth of each segment, Indicates the first The end depth of the segment, Indicates the number of completion logging data m The starting depth of each segment, Indicates the number of completion logging data m The end depth of each section; the technology A data and completion logging data are the gamma data measured by technology A and completion technology for the same well respectively; Perform overlapping mask filtering on the Technology A section and the completion logging section: ； in, Indicates the segment overlap length, if , then remember , this segment does not participate in the relevance evaluation, Indicates the segment mask mark, mask mark 0 means discard, mask mark 1 means retain; After filtering the results with overlapping masks, the length consistency mask processing is performed on the technical A section and the completion logging section, and the resulting valid section pair set is: ； ； ； in, Indicates the first m Section and completion logging W n The length difference ratio of the segment on the depth axis is When , this segment does not participate in the relevance evaluation, Segment representing technology A m length, Indicates that the completion logging is in the section n length.

5. The fully automatic well logging signal depth calibration method according to claim 4, characterized in that: The expression of the valid segment pair set is as follows: ； in, Represents a set of valid segment pairs, Represents a segment pair.

6. The fully automatic logging signal depth calibration method according to claim 2, characterized in that: The depth adjustment of the continuous and non-overlapping segments is specifically as follows: The seven consecutive and non-overlapping segments are adjusted using a depth coarse adjustment method based on the seven-point method; The fine depth correction method of well logging data based on area feature matching is used to correct the results of rough depth adjustment to complete the fully automatic depth correction of the logging signal.

7. The fully automatic logging signal depth calibration method according to claim 6, characterized in that: The depth rough adjustment method based on the seven-part method is used to adjust seven consecutive and non-overlapping segments, which is specifically as follows: Based on seven consecutive and non-overlapping segments, the default depth arrangement of each segment is from shallow to deep, and the first segment of the technology A segment is Make a depth alignment adjustment to move the first segment of the Tech A segment to the The maximum depth of the well is aligned to the maximum depth of the completion logging section: ； in, Indicates the first paragraph The depth value after the upper boundary is aligned, Indicates the original depth value of the first segment among the seven segments. Indicates the first paragraph of technology A The maximum depth sample point of the internal measurement data, Indicates the first section of completion logging The maximum depth sample point of the internal measurement data, seven segments are seven consecutive and non-overlapping segments; The end of the technical A section Make a depth alignment adjustment to move the end of the Tech A segment The maximum depth of the well is aligned to the minimum depth of the completion logging section: ； in, Indicates the tail segment The depth value after the lower boundary is aligned, Indicates the end of the completion logging section The minimum depth sample point of the internal measurement data, Indicates the end of the technical A section Minimum depth sample point of internal measurement data; In the Technology A section, the section scale normalization process is performed on the middle sections except the first and last sections to ensure that the length of the Technology A depth section is consistent with the completion logging section; The depth position balance processing is performed on the middle segment after the segment scale normalization processing to align the starting point of the technology A segment and complete the rough depth alignment of each segment of the technology A data.

8. The fully automatic well logging signal depth calibration method according to claim 6, characterized in that: The method for fine depth correction of well logging data based on area feature matching is used to correct the result of rough depth adjustment, which is specifically as follows: Based on the depth rough adjustment results, in the technical A section and the corresponding section of the completion logging Select a reference point from , where Indicates the seventh section of the completion logging. i The original depth value of each segment, Indicates the seventh section of Technology A i The original depth value of each segment; With each reference point as the center, select a local segment of fixed length and calculate the area characteristics of the completion logging data and technical A data in this segment: ； ； in, Indicates Technology A segment The area enclosed by the gamma logging value and depth, Indicates the completion logging section The area enclosed by the gamma logging value and depth, Indicates the half-width of the window, Indicates the depth of the reference point of technology A. Indicates the depth of the completion logging reference point. Indicates the completion logging at depth coordinate Gamma logging value at ; For the current section, with a step length of 1m, starting from the current section, calculate the area enclosed by the technology A data and the depth axis in the current section and normalize it to unit depth; Based on the area characteristics, the segment that best matches the normalized segment is found in the completion logging data, and the area of ​​the best matching segment is calculated and normalized. Calculate the error function based on the normalized area of ​​the best matching segment and search for a value that minimizes the error within a preset range; Based on the search results, the current technology A section is mapped to the completion logging section, and the aligned depth position is updated according to the mapping results; Continue processing the next 1m segment, complete processing for all segments, and complete the correction of the depth coarse adjustment result.

Images