Stratigraphic identification method and model training method and device

Abstract

A stratum identification method and a model training method and device are disclosed. The target well is stratified based on an automatic stratification technology to obtain a stratification result of the stratum interface of the target well. The accuracy of the stratum interface represented by the stratification result is evaluated according to the real stratification information of the standard well, the well logging curve of the standard well, and the well logging curve of the target well. The stratum interface is selected from the stratification result based on the evaluation result to obtain the final stratification result of the stratum interface of the target well. Thus, the accuracy of the final stratification result of the stratum interface of the target well can be improved.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of stratum division, and in particular to a stratum identification method and a model training method and device. BACKGROUND

[0002] Stratum division is a basic and key link in well logging data processing and interpretation, and is also a prerequisite for correctly selecting parameters and a key to fluid identification in well logging data processing. The determined stratum has a relatively stable lithological combination, which is of great significance to the evaluation of subsequent vertical generation, reservoir and caprock assemblage and lateral reservoir distribution characteristics.

[0003] At present, well logging curve stratum division is divided into two methods of manual division and automatic layering.

[0004] Manual stratum division is generally divided by manually finding the positions of features (half amplitude points, peak values, etc.) reflecting layer interfaces on well logging curves. However, the features of some layer interfaces on well logging curves are not obvious, which is prone to errors. From the actual manual processing results, the division results of experienced interpreters who are familiar with the geological conditions of the block are more accurate and efficient. For example, in Changqing, it takes about 20 minutes to complete the division of one well, but for interpreters who are not familiar with the geological conditions and lack of experience, it is prone to errors and low efficiency, and it takes about 40 minutes.

[0005] Automatic layering is born to solve the shortcomings of manual layering. The main idea is to set layering standards and use computers to layer well logging curves. Over the years, many scholars at home and abroad have used different methods to study automatic layering of well logging curves from mathematical statistics, well logging curve patterns to signal processing technology, artificial intelligence technology, and have achieved good results. The methods for automatic stratum correlation at home and abroad include dynamic matching algorithm, wavelet transform adaptive algorithm, Walsh transform algorithm, grey correlation method, clustering analysis method, etc. These methods have their own advantages and disadvantages, and also have their own application limitations.

[0006] Therefore, a more effective stratum division scheme is needed. SUMMARY

[0007] One of the technical problems to be solved by the present disclosure is to provide a more effective stratum division scheme.

[0008] According to a first aspect of the present disclosure, a stratum identification method is provided, comprising: performing stratum division on a target well based on an automatic layering technology to obtain a division result of a stratum interface of the target well; evaluating the accuracy of the stratum interface represented by the division result according to real stratum division information of a standard well, well logging curves of the standard well and well logging curves of the target well; and selecting a stratum interface from the division result based on the evaluation result to obtain a final division result of the stratum interface of the target well.

[0009] Optionally, the real formation division information comprises a plurality of formation intervals, and the step of evaluating the accuracy of the formation interfaces comprises: constructing formation interface pairs based on the division results of the formation interfaces of the target well, each formation interface pair comprising two formation interfaces; for each formation interval of the standard well, selecting formation interface pairs whose thicknesses are less than a first predetermined threshold from the thickness of the formation interval; and scoring the formation interface pairs according to the similarity between the first log curve of the target well within the thickness represented by the selected formation interface pairs and the second log curve of the standard well within the thickness represented by the formation interval, the score being positively correlated with the similarity.

[0010] Optionally, the method further comprises: determining a predetermined number of formation levels, each formation level having a corresponding depth range, and the depth ranges corresponding to different formation levels having no overlapping parts; and determining the formation level to which each formation interval of the standard well belongs, wherein the step of selecting, for each formation interval of the standard well, formation interface pairs whose thicknesses are less than a first predetermined threshold from the thickness of the formation interval comprises: for each formation interval of the standard well, selecting, within the depth range corresponding to the formation level to which the formation interval belongs, formation interface pairs whose thicknesses are less than a first predetermined threshold from the thickness of the formation interval.

[0011] Optionally, the step of selecting formation interfaces from the division results based on the evaluation results to obtain the final division results of the formation interfaces of the target well comprises: for each formation interval of the standard well, selecting, from the selected formation interface pairs, the formation interface pair with the highest score as the formation division result of the target well corresponding to the formation interval.

[0012] Optionally, the step of scoring the formation interface pairs comprises: calculating a first feature value of the first log curve within the thickness represented by the selected formation interface pairs and a second feature value of the second log curve of the standard well within the thickness represented by the formation interval; and using a pre-trained scoring model to score the formation interface pairs by taking the difference between the first feature value and the second feature value as the model input.

[0013] Optionally, the step of calculating the first eigenvalue of the first logging curve in the thickness represented by the selected set of formation interfaces and the second eigenvalue of the second logging curve in the thickness represented by the formation interval of the standard well comprises: dividing the thickness interval represented by the selected set of formation interfaces into a plurality of first sub-intervals, and calculating the eigenvalue of the curve part of the first logging curve in each first sub-interval represented by the selected set of formation interfaces as the first eigenvalue; dividing the formation interval into a plurality of second sub-intervals, and calculating the eigenvalue of the curve part of the second logging curve in each second sub-interval represented by the thickness of the formation interval of the standard well as the second eigenvalue.

[0014] Optionally, the automatic layering technology comprises: activity layering method; and / or dynamic matching algorithm; and / or wavelet transform method.

[0015] Optionally, the step of stratifying the target well based on the automatic layering technology comprises: determining an activity function of the logging curve of the target well, the activity function being used to represent the activity at different positions on the logging curve, the activity being used to reflect the dynamic properties of the logging curve; and determining the positions where the activity belongs to local maximum values as the formation interfaces according to the activity function.

[0016] Optionally, the method further comprises: removing negative values in the logging curve; and / or normalizing the logging curve; and / or performing smoothing filtering processing on the logging curve.

[0017] According to a second aspect of the present disclosure, a model training method for scoring the accuracy of stratification results is provided, comprising: obtaining training data, the training data comprising real stratification information of a standard well, stratification results of formation interfaces of at least one comparison well determined based on an automatic layering technology, a logging curve of the standard well, and a logging curve of the comparison well, the real stratification information comprising a plurality of formation intervals; constructing a training sample set based on the training data, the training sample set comprising at least one training sample, each training sample being composed of one formation interval of the standard well and one set of formation interfaces of the comparison well, the set of formation interfaces having a thickness difference from the formation interval lower than a first predetermined threshold, the set of formation interfaces being composed of two formation interfaces, the features of the training sample being used to represent the similarity between a third logging curve of the comparison well in the thickness represented by the set of formation interfaces and a second logging curve of the standard well in the thickness represented by the formation interval, and the label of the training sample being used to represent whether the stratification information represented by the set of formation interfaces is accurate; and training a scoring model for scoring the accuracy of the set of formation interfaces based on the training sample set.

[0018] Optionally, before constructing the training sample set based on the training data, the method further comprises: removing negative values in the logging curve; and / or normalizing the logging curve; and / or performing smoothing filtering processing on the logging curve.

[0019] Optionally, the automatic layering technique comprises: an activity layering method; and / or a dynamic matching algorithm; and / or a wavelet transform method.

[0020] Optionally, the step of obtaining the training data comprises: determining an activity function of the well logging curve of the contrast well, the activity function being used to represent activities at different positions on the well logging curve, the activities being used to reflect dynamic properties of the well logging curve; and determining, according to the activity function, positions at which the activities belong to local maximum values as the stratigraphic interfaces of the contrast well, to obtain a stratigraphic interface division result of the contrast well.

[0021] Optionally, before the activity function of the well logging curve of the contrast well is determined, the method further comprises: adjusting a parameter that affects an effect of the activity function, to improve a recall rate of the activity function, the recall rate being positively correlated with a number of times that a real stratigraphic interface of the contrast well corresponds to a stratigraphic interface determined based on the activity function.

[0022] Optionally, the step of constructing the training sample set based on the training data comprises: calculating a third feature value of a third well logging curve of the contrast well within a thickness represented by the stratigraphic interface pair and a second feature value of a second well logging curve of the standard well within a thickness represented by the stratigraphic interval; and taking a difference between the third feature value and the second feature value as a feature of the training sample.

[0023] Optionally, the step of calculating the third feature value of the third well logging curve of the contrast well within the thickness represented by the stratigraphic interface pair and the second feature value of the second well logging curve of the standard well within the thickness represented by the stratigraphic interval comprises: dividing the thickness interval represented by the stratigraphic interface pair into a plurality of third subintervals, and calculating feature values of curve portions of the third well logging curve of the contrast well within each third subinterval within the thickness represented by the stratigraphic interface pair as the third feature value; and dividing the stratigraphic interval into a plurality of second subintervals, and calculating feature values of curve portions of the second well logging curve of the standard well within each second subinterval within the thickness represented by the stratigraphic interval as the second feature value.

[0024] Optionally, the step of constructing the training sample set based on the training data comprises: determining a label of the training sample according to whether a difference between stratigraphic division information represented by the stratigraphic interface pair and real stratigraphic division information of the contrast well is greater than a second predetermined threshold.

[0025] Optionally, the method further comprises: constructing a verification sample set based on the training data, the verification sample set comprising at least one verification sample, each verification sample being composed of one formation interval of the standard well and one formation interface pair of the contrast well, the thickness of the formation interface pair being less than a first predetermined threshold, the formation interface pair being composed of two formation interfaces, the feature of the verification sample being used to represent the similarity between the third logging curve of the contrast well within the thickness represented by the formation interface pair and the second logging curve of the standard well within the thickness represented by the formation interval; and scoring the formation interface pair in the verification sample of the contrast well constructed based on one formation interval of the standard well using the trained scoring model, and evaluating the accuracy of the scoring model according to the difference between the formation division information represented by the formation interface pair with the highest score and the real formation division information of the contrast well.

[0026] According to a third aspect of the present disclosure, a formation identification device is provided, comprising: a division module configured to divide a target well based on an automatic layering technology to obtain a division result of formation interfaces of the target well; an evaluation module configured to evaluate the accuracy of the formation interfaces represented by the division result according to real formation division information of a standard well, logging curves of the standard well and logging curves of the target well; and a selection module configured to select formation interfaces from the division result based on the evaluation result to obtain a final division result of the formation interfaces of the target well.

[0027] Optionally, the real formation division information comprises a plurality of formation intervals, and the evaluation module comprises: a construction module configured to construct formation interface pairs based on the division result of the formation interfaces of the target well, each formation interface pair being composed of two formation interfaces; an interface pair selection module configured to select, for each formation interval of the standard well, a formation interface pair whose thickness is less than a first predetermined threshold from the thickness of the formation interval; and a scoring module configured to score the formation interface pair according to the similarity between the first logging curve of the target well within the thickness represented by the selected formation interface pair and the second logging curve of the standard well within the thickness represented by the formation interval, the score being positively correlated with the similarity.

[0028] Optionally, the device further comprises: a first determination module configured to determine a predetermined number of formation levels, each formation level having a corresponding depth range, and the depth ranges corresponding to different formation levels having no overlapping parts; and a second determination module configured to determine the formation level to which each formation interval of the standard well belongs, and the interface pair selection module is configured to select, for each formation interval of the standard well, a formation interface pair whose thickness is less than a first predetermined threshold from the thickness of the formation interval within the depth range corresponding to the formation level to which the formation interval belongs.

[0029] Optionally, the selecting module selects, for each stratigraphic interval of the standard well, a stratigraphic interface pair with the highest score from the selected stratigraphic interface pairs as the stratigraphic division result corresponding to the stratigraphic interval of the target well.

[0030] Optionally, the scoring module comprises: a calculating module configured to calculate a first feature value of the first logging curve within the thickness represented by the selected stratigraphic interface pair and a second feature value of the second logging curve within the thickness represented by the stratigraphic interval of the standard well; and a predicting module configured to use the pre-trained scoring model to score the stratigraphic interface pair by taking the difference between the first feature value and the second feature value as the model input.

[0031] Optionally, the calculating module comprises: a first calculating module configured to divide the thickness interval represented by the selected stratigraphic interface pair into a plurality of first sub-intervals, and calculate the feature values of the curve portions of the first logging curve within each first sub-interval as the first feature values; and a second calculating module configured to divide the stratigraphic interval into a plurality of second sub-intervals, and calculate the feature values of the curve portions of the second logging curve within each second sub-interval as the second feature values.

[0032] Optionally, the automatic stratigraphic division technology comprises: an activity stratigraphic division method; and / or a dynamic matching algorithm; and / or a wavelet transform method.

[0033] Optionally, the dividing module comprises: an activity function determining module configured to determine an activity function of the logging curve of the target well, the activity function being configured to represent the activity at different positions on the logging curve, the activity being configured to reflect the dynamic property of the logging curve; and a stratigraphic interface determining module configured to determine the positions at which the activity is a local maximum as the stratigraphic interfaces according to the activity function.

[0034] Optionally, the apparatus further comprises: a removing module configured to remove negative values in the logging curve; and / or a normalization processing module configured to perform normalization processing on the logging curve; and / or a filtering processing module configured to perform smoothing filtering processing on the logging curve.

[0035] According to a fourth aspect of the present disclosure, a model training apparatus for scoring accuracy of stratigraphic division results is provided, comprising: an obtaining module configured to obtain training data, the training data comprising real stratigraphic division information of a standard well, stratigraphic interface division results of at least one contrast well determined based on an automatic stratigraphic division technology, well logging curves of the standard well, and well logging curves of the contrast well, the real stratigraphic division information comprising a plurality of stratigraphic intervals; a constructing module configured to construct a training sample set based on the training data, the training sample set comprising at least one training sample, each training sample being composed of one stratigraphic interval of the standard well and one stratigraphic interface pair of the contrast well, the thickness of the stratigraphic interval and the stratigraphic interface pair being less than a first predetermined threshold, the stratigraphic interface pair being composed of two stratigraphic interfaces, the feature of the training sample being configured to represent the similarity between a third well logging curve of the contrast well within the thickness represented by the stratigraphic interface pair and a second well logging curve of the standard well within the thickness represented by the stratigraphic interval, and the label of the training sample being configured to represent whether the stratigraphic division information represented by the stratigraphic interface pair is accurate; and a training module configured to train a scoring model for scoring accuracy of the stratigraphic interface pair based on the training sample set.

[0036] Optionally, the apparatus further comprises: a removing module configured to remove negative values in the well logging curves; and / or a normalization processing module configured to perform normalization processing on the well logging curves; and / or a filtering processing module configured to perform smoothing filtering processing on the well logging curves.

[0037] Optionally, the automatic stratigraphic division technology comprises: an activity stratigraphic division method; and / or a dynamic matching algorithm; and / or a wavelet transform method.

[0038] Optionally, the obtaining module comprises: an activity function determining module configured to determine an activity function of the well logging curves of the contrast well, the activity function being configured to represent activities at different positions on the well logging curves, the activities being configured to reflect dynamic properties of the well logging curves; and a stratigraphic interface determining module configured to determine, according to the activity function, positions at which the activities are local maximum values as the stratigraphic interfaces of the contrast well, to obtain the stratigraphic interface division results of the contrast well.

[0039] Optionally, the apparatus further comprises: a parameter adjusting module configured to adjust parameters affecting effects of the activity function before the activity function determining module determines the activity function of the well logging curves of the contrast well, to improve a recall rate of the activity function, the recall rate being positively correlated with a number of times that the real stratigraphic interfaces of the contrast well correspond to the stratigraphic interfaces determined based on the activity function.

[0040] Optionally, the constructing module comprises: a calculating module, configured to calculate a third feature value of a third logging curve of the comparison well in a thickness represented by the group of formation interfaces, and a second feature value of a second logging curve of the standard well in a thickness represented by the formation interval; and a sample feature determining module, configured to take a difference between the third feature value and the second feature value as a feature of the training sample.

[0041] Optionally, the calculating module comprises: a third calculating module, configured to divide the thickness interval represented by the group of formation interfaces into a plurality of third sub-intervals, and calculate a feature value of a curve portion of the third logging curve of the comparison well in each third sub-interval as the third feature value; and a fourth calculating module, configured to divide the formation interval into a plurality of second sub-intervals, and calculate a feature value of a curve portion of the second logging curve of the standard well in each second sub-interval as the second feature value.

[0042] Optionally, the constructing module comprises: a sample label determining module, configured to determine a label of the training sample according to whether a difference between the formation division information represented by the group of formation interfaces and the real formation division information of the comparison well is greater than a second predetermined threshold.

[0043] Optionally, the constructing module further constructs a verification sample set based on the training data, the verification sample set comprises at least one verification sample, each verification sample is composed of one formation interval of the standard well and one group of formation interfaces of the comparison well, the thickness of the group of formation interfaces is less than the thickness of the formation interval by less than a first predetermined threshold, the group of formation interfaces is composed of two formation interfaces, and a feature of the verification sample is used to represent a similarity between the third logging curve of the comparison well in a thickness represented by the group of formation interfaces and the second logging curve of the standard well in a thickness represented by the formation interval. The device can further comprise: a model evaluating module, configured to use the trained scoring model to score the group of formation interfaces in the verification sample of the same comparison well constructed based on one formation interval of the standard well, and evaluate an accuracy of the scoring model according to a difference between a formation represented by the group of formation interfaces with the highest score and real formation division information of the comparison well.

[0044] According to a fifth aspect of the present disclosure, a computing device is provided, comprising: a processor; and a memory having executable code stored thereon, which, when executed by the processor, causes the processor to perform the method according to the first aspect.

[0045] According to a sixth aspect of the present disclosure, a non-transitory machine-readable storage medium is provided, having executable code stored thereon, which, when executed by a processor of an electronic device, causes the processor to perform the method according to the first aspect.

[0046] Thus, the present disclosure evaluates the accuracy of the formation interface represented by the division result of the target well determined based on the automatic layering technology according to the real formation division information of the standard well, the well logging curve of the standard well and the well logging curve of the target well, and selects the formation interface from the division result based on the evaluation result, so that the accuracy of the division result of the formation interface of the target well is improved. BRIEF DESCRIPTION OF DRAWINGS

[0047] The above and other objects, features and advantages of the present disclosure will become more apparent from the following detailed description when taken in conjunction with the accompanying drawings in which like reference characters refer to like parts throughout the figures, and in which:

[0048] Figure 1 A schematic flowchart of a formation identification method according to one embodiment of the present disclosure is shown.

[0049] Figure 2 A schematic flowchart of a model training method according to one embodiment of the present disclosure is shown.

[0050] Figure 3 A schematic diagram of formation interfaces extracted based on the activity layering method is shown.

[0051] Figure 4 A schematic diagram showing that the thickness interval represented by a formation interval of a standard well and a formation interface pair of a contrast well, respectively, is divided into a plurality of sub-intervals is shown.

[0052] Figure 5 A schematic diagram showing the calculation method of the error between the formation represented by the formation interface pair and the real formation is shown.

[0053] Figure 6 A structural block diagram of a formation identification apparatus according to one embodiment of the present disclosure is shown.

[0054] Figure 7 A structural block diagram of a model training apparatus according to one embodiment of the present disclosure is shown.

[0055] Figure 8 A structural block diagram of a computing device according to one embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0056] Preferred embodiments of the present disclosure will be described herein below with reference to the accompanying drawings. While preferred embodiments of the present disclosure are shown in the drawings, it is understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that the present disclosure will be thorough and complete, and will fully convey the scope of the present disclosure to those skilled in the art.

[0057] First, the advantages and disadvantages of some typical automatic layering technologies in the prior art are briefly summarized.

[0058] The automatic layering technology can include but is not limited to activity layering method, dynamic matching algorithm, wavelet transform method.

[0059] The activity layering method (also known as activity function method) is to calculate the dispersion of points by using the square of the curve value and the average value. Because it is simple and convenient to calculate, it has been one of the methods frequently used by people. However, because it is generally analyzed for a logging curve, the information is lacking and not representative, and the information is easily affected by noise, which often leads to too many errors in layering interfaces and affects the effect of layering.

[0060] The dynamic matching algorithm sets multiple parameters to realize automatic stratum correlation, but in the process of calculating the distance between two strata, the initial setting of the weight will become an important link in the calculation, and the setting of the weight is mostly determined according to the experience value, which means that the dynamic pattern matching algorithm cannot adapt to various complex underground conditions. However, the underground conditions are complex and each region has different characteristics, which puts a high test on the setting of the weight. Therefore, the adaptability of the dynamic pattern matching algorithm is not very strong, and it can only be used in one region. Different regions cannot use the same set of weight values for calculation, otherwise, it will be difficult to match the comparison results with the manual comparison results.

[0061] The wavelet transform processing method (i.e. wavelet transform method) has advantages and disadvantages, and almost all of them face the problem of removing false strata to improve the layering accuracy. The traditional processing method is to adjust the layering results by setting multiple thresholds, but the parameter adjustment process is often complicated and difficult to obtain ideal results. In processing logging curves with relatively uniform geology, the wavelet transform method can indeed better filter out stratum interfaces and achieve good layering effect. However, for relatively complex geological features, the wavelet transform method will encounter obstacles in finding stratum interfaces, and even how to set the threshold value will result in a certain number of false strata or real stratum interfaces being removed together with false strata. The selection of wavelet function is mainly based on experience, and there is no unified method and standard, which is difficult to operate.

[0062] In view of the shortcomings of the existing automatic layering technology (especially the activity layering method), the present disclosure proposes that on the basis of determining the division result of the stratum interval of the target well based on the automatic layering technology (such as the activity layering method), the accuracy of the division result is further evaluated with the real stratum division information of the standard well to remove the stratum interfaces with inaccurate division and improve the accuracy of the final division result.

[0063] The target well refers to a well to be subjected to formation division, and the standard well refers to a well of which real formation division information is known. The standard well can refer to a well located in the same geographical area (e.g., separated by no more than a predetermined distance) as the target well, or a well of which geological conditions are the same as or similar to those of the target well. That is, the subsurface layering structure of the standard well is similar to that of the target well.

[0064] Thus, the accuracy of the formation division result of the target well determined based on the automatic layering technology can be evaluated with the aid of the real formation division information of the standard well, so that the formation division result of the target well can be screened or filtered according to the evaluation result, so that the accuracy of the final formation division result of the target well can be guaranteed.

[0065] Details involved in the present disclosure will be further described below in conjunction with the accompanying drawings.

[0066] Figure 1 A schematic flowchart of a formation identification method according to one embodiment of the present disclosure is shown. Figure 1 The method shown can be implemented in software by a computer program, and can also be executed by a specially configured computing device Figure 1 The method shown.

[0067] Referring to Figure 1 In step S110, formation division is performed on the target well based on an automatic layering technology, to obtain a division result of the formation interface of the target well.

[0068] The automatic layering technology can adopt one or more of, but is not limited to, activity layering method, dynamic matching algorithm, and wavelet transform method. Considering that the activity layering method is simple to calculate and the obtained formation interface is relatively comprehensive, it can cover the real formation interface as much as possible, and the present disclosure can filter the formation interface determined based on the automatic layering technology by comparison, so the activity layering method can be preferred for formation division of the target well. The activity layering method is a mature technology in the art, and one exemplary process of formation division of the target well based on the activity layering method is as follows.

[0069] The logging curve has mobility, in order to represent the dynamic property of the logging curve, the activity function of the logging curve of the target well can be determined (i.e., defined) first, the activity function is used to represent the activity at different positions on the logging curve, and the activity is used to reflect the dynamic property of the logging curve. As an example, the activity function can be expressed as,

[0070]

[0071]

[0072]

[0073]

[0074] wherein E(d) represents the activity of the logging curve at depth d, n is a positive integer related to the window length in calculating the activity, the window length is 2n+1, and x(i) represents the value of the original logging curve; represents the average value of the logging curve in the upper and lower n ranges at d; R(d) represents the sum of squares of the logging values of the logging curve in the upper and lower n ranges at d; and T(d) represents the sum of the logging values of the logging curve in the upper and lower n ranges at d.

[0075] After the activity function is determined, the position of the activity belonging to the local maximum can be taken as the formation interface according to the activity function, that is, the depth corresponding to each maximum value in the activity function can be taken as the formation interface of the target well, and thus the division result of the formation interface of the target well can be obtained. The formation interface refers to the interface between two adjacent formations.

[0076] The activity function in the embodiment can refer to an activity function after adjustment of parameters affecting the effect of the activity function. The parameter adjustment process can be referred to the related description of the model training part below.

[0077] In step S120, the accuracy of the formation interface represented by the division result is evaluated according to the real formation division information of the standard well, the logging curve of the standard well, and the logging curve of the target well.

[0078] A single well can be selected as the standard well by an expert in advance, and the formation boundary is divided to obtain the real formation division information that can reflect the real formation division result of the standard well.

[0079] The real formation division information can include a plurality of formation intervals. There is no overlapping part between the plurality of formation intervals, that is, each formation interval is composed of two adjacent formation interfaces of the standard well.

[0080] The logging curve is a curve formed during logging. There are many types of logging curves. Common logging curves include, but are not limited to, natural gamma (GR) logging curve, acoustic time difference (AC) logging curve, natural potential (SP) logging curve, shallow detection induction (RLLD) logging curve, deep resistivity (RT) logging curve, caliper (CAL) logging curve, and acoustic amplitude (CBL) logging curve. The logging curve in the embodiment can include one or more of the above.

[0081] Before using the logging curve, the logging curve can be preprocessed first to ensure the availability and use effect of the logging curve. The preprocessing performed on the logging curve can include, but is not limited to, removing negative values of the curve, normalization, and smoothing filtering.

[0082] 1. Removing negative values of the curve

[0083] Since the current well logging curve data has extremely small negative values ​​due to the measuring instrument, these negative values ​​are usually judged to be invalid. Therefore, for at least part (e.g., all) of the well logging curves with negative values, the negative values ​​in the well logging curves can be removed.

[0084] 2. Normalization

[0085] Since different logging curves have different scale dimensions and the logging curves of different wells may have different value ranges, in order to ensure that the algorithm can fairly compare the logging curves of the standard well and the comparison well, the logging curve of each well can be normalized to [0, 1] in advance. Suppose the logging curve sample sequence of a single well is {x i}={x1,x2,…,x n}, for each data x in the sequence i Normalize it and get:

[0086] Among them, max l≤k≤n {x k} is the maximum amplitude of the logging curve for this well section, min l≤k≤n {x k} is the minimum amplitude of the logging curve for this well section.

[0087] 3. Smoothing Filter

[0088] Random data variation noise in the logging curve may cause statistical fluctuations in the logging curve that are unrelated to the formation, so it is necessary to filter out these interfering noises.

[0089] The mean filter method can be used to smooth the logging curve. The calculation process of the mean filter is as follows.

[0090] Let S ab Represents a one-dimensional array with a center at (a,0) and a length of 2b. The calculation process of the mean filter is to calculate the mean value of s ab The average value of the disturbed data g(a,0) in the field within the well logging curve is used to restore the original data at any point (a,0) in the well logging curve. The value is to construct s with this point as the center point ab , use s ab The average value of the logging values ​​within the field in the array is calculated, that is:

[0091]

[0092] After performing the above processing on the logging curve, the accuracy of the stratigraphic interface represented by the division result can be evaluated based on the actual stratigraphic division information of the standard well, the processed logging curve of the standard well, and the processed logging curve of the target well.

[0093] The well log curves corresponding to the stratum interval of the same or similar thickness (i.e. the thickness difference is not more than a predetermined threshold, such as 20 meters) between the standard well and the target well should have certain similarity. The stratum interval of the standard well (referred to as the standard stratum interval) is known and accurate, and any two stratum boundaries among all stratum boundaries determined based on the automatic layering technology can constitute an accuracy unknown (i.e. to be evaluated) stratum interval (referred to as the target stratum interval) of the target well, and the well logs of the standard well and the target well can be obtained through logging technology.

[0094] Therefore, for a standard stratum interval of the standard well, a target stratum interval with a thickness difference less than a first predetermined threshold (such as ± 20 meters) from the thickness of the standard stratum interval can be selected from all target stratum intervals of the target well, and the two stratum boundaries constituting the target stratum interval are scored according to the similarity between the well log curves of the target stratum interval and the well log curves of the standard stratum interval. Wherein, the score is positively correlated with the similarity, the higher the score, the more similar the well log curves, and the higher the possibility that the two stratum boundaries constituting the target stratum interval are consistent with the true stratum boundaries.

[0095] As an example, all possible stratum boundary pairs can be constructed based on the division results of the stratum boundaries of the target well, each stratum boundary pair is composed of any two stratum boundaries, and each stratum boundary pair constitutes an accuracy to be evaluated stratum interval (i.e. target stratum interval) of the target well; then for each stratum interval of the standard well (i.e. standard stratum interval), a stratum boundary pair with a thickness difference less than a first predetermined threshold from the thickness of the stratum interval can be selected; and the stratum boundary pair is scored according to the similarity between the well log curves of the target well within the thickness represented by the selected stratum boundary pair (for the sake of distinction, it can be referred to as the first well log curve) and the well log curves of the standard well within the thickness represented by the stratum interval (for the sake of distinction, it can be referred to as the second well log curve).

[0096] The difference in thickness referred to in the present disclosure can refer to the absolute thickness difference of the stratum deviating from the depth, or the thickness difference between strata within the same depth range, i.e. the relative thickness difference. Preferably, it can refer to the thickness difference between strata within the same depth range.

[0097] For example, assuming that the standard formation intervals of the standard well are represented as {…, 1200-1250m, …, 1850-1900m, …}, and the target formation intervals of the target well include 1210-1260m and 1870-1905m, then according to the absolute thickness difference, for the standard formation interval of 1200-1250m of the standard well with a thickness of 50m, both 1210-1260m and 1870-1905m of the target well can be compared with the logging curves of the standard formation interval of 1200-1250m to determine whether the division of the formation interfaces corresponding to the target formation intervals of 1210-1260m and 1870-1905m is accurate.

[0098] Considering that when the depth difference is too large, the logging curves of the formation intervals with the same or similar thicknesses may have certain differences, which cannot guarantee the accuracy of the evaluation based on the curve similarity, the thickness difference in the present disclosure can preferably refer to the thickness difference between the formations in the same depth range (i.e., the depth difference is not large, such as less than a third predetermined threshold). Still taking the above example, for the standard formation interval of 1200-1250m of the standard well with a thickness of 50m, only the logging curves of the target formation interval (i.e., 1210-1260m) of the target well in the same depth range (both around 1200m) as the standard formation interval of 1200-1250m of the standard well can be compared with the logging curves of the standard formation interval of 1200-1250m to determine whether the division of the formation interface corresponding to the target formation interval of 1210-1260m is accurate. As for the accuracy evaluation of the formation interface corresponding to the target formation interval of 1850-1900m of the target well, a standard formation interval with a small depth difference from the target formation interval of 1850-1900m and a thickness difference less than the first predetermined threshold, such as the standard formation interval of 1830-1870m, can be used for comparison.

[0099] That is, for each formation interval of the standard well, when selecting the group of formation interfaces for comparison with the formation interval, the group of formation interfaces with a small difference in depth (such as a depth difference less than a third predetermined threshold) and a thickness difference less than the first predetermined threshold from the thickness of the formation interval can be selected.

[0100] As an example, the present disclosure can determine a predetermined number of formation levels, and determine the formation level to which each formation interval of the standard well belongs according to the real formation division information of the standard well. Each formation level has a corresponding depth range, and the depth ranges corresponding to different formation levels do not overlap.

[0101] The stratigraphic level can represent a basic unit of the stratum. The larger the stratigraphic level, the smaller the thickness, and the shorter the age represented. The number of stratigraphic levels (i.e., the predetermined number) can be set according to actual conditions. When the predetermined number is 1, it means that all stratigraphic intervals of the standard well belong to the same stratigraphic level. When the predetermined number is 2, it means that all stratigraphic intervals of the standard well are divided into two stratigraphic levels corresponding to different depth ranges. It can be seen that the number of stratigraphic levels is positively correlated with the division precision, that is, the more stratigraphic levels, the more precise the division. When setting the number of stratigraphic levels and the depth range corresponding to each stratigraphic level, the number of stratigraphic levels and the depth range can be set based on the well depth. That is, the predetermined number of stratigraphic levels can be set according to the well depth, and the depth range corresponding to each stratigraphic level is determined. For example, the well depth can be divided into 5 stratigraphic levels, and each stratigraphic level corresponds to one fifth of the well depth.

[0102] Therefore, when selecting a group of stratigraphic interfaces (i.e., target stratigraphic intervals) for comparison (i.e., similarity between comparison logging curves) with each stratigraphic interval (i.e., standard stratigraphic interval) of the standard well, a group of stratigraphic interfaces with a thickness difference less than a first predetermined threshold value from the thickness of the stratigraphic interval can be selected within the depth range corresponding to the stratigraphic level to which the stratigraphic interval belongs.

[0103] That is, in units of stratigraphic levels, for each stratigraphic interval of the standard well within each stratigraphic level, a group of stratigraphic interfaces with a thickness difference less than a first predetermined threshold value from the thickness of the stratigraphic interval can be selected from a group of stratigraphic interfaces of the comparison well that also belong to the stratigraphic level.

[0104] In step S130, the stratigraphic interfaces are selected from the division results based on the evaluation results to obtain the final division results of the stratigraphic interfaces of the target well.

[0105] The evaluation results can reflect the accuracy of the stratigraphic interfaces represented by the stratigraphic division results obtained based on the automatic stratigraphic division technology. Therefore, based on the evaluation results, stratigraphic interfaces with high evaluation can be selected from the division results, that is, stratigraphic interfaces with low evaluation are filtered, to improve the accuracy of the final division results of the stratigraphic interfaces of the target well.

[0106] Taking the above-mentioned scoring of the group of stratigraphic interfaces as an example, all groups of stratigraphic interfaces can be sorted according to the scores, and then the stratigraphic interfaces corresponding to the predetermined number of groups of stratigraphic interfaces with high ranking (and scores greater than a fourth predetermined threshold value) are selected as the final division results of the stratigraphic interfaces of the target well.

[0107] As an example, in the case that the formation structure of the standard well and the target well is similar, for each formation interval of the standard well, the formation interface pair group with the highest score can be selected from the selected formation interface pair groups for comparison with the formation interval as the formation division result corresponding to the formation interval of the target well.

[0108] The operation of scoring the formation interface pair group according to the similarity between the logging curves mentioned above can be realized by calculating the similarity between the logging curves, and assigning the score of the formation interface pair group according to the similarity calculation result. The automatic scoring of the formation interface pair group can also be realized by using a pre-trained model.

[0109] Taking the automatic scoring of the formation interface pair group by using the pre-trained model as an example, the training process of the model will be described first. Figure 2 A schematic flow chart of a model training method for scoring the accuracy of the formation division result according to an embodiment of the present disclosure is shown.

[0110] As shown in Figure 2 In step S210, training data is obtained.

[0111] The training data includes the real formation division information of the standard well, the division result of the formation interface of at least one contrast well determined based on the automatic layering technology, the logging curves of the standard well and the logging curves of the contrast well, and the real formation division information includes a plurality of formation intervals.

[0112] The automatic layering technology can adopt one or more of, but is not limited to, the activity layering method, the dynamic matching algorithm and the wavelet transform method. Preferably, the contrast well is divided into layers based on the activity layering method. The process of dividing the contrast well into layers based on the activity layering method can be referred to the related description above.

[0113] It should be noted that before determining the activity function of the logging curves of the contrast well, the present disclosure can also adjust the parameters affecting the effect of the activity function to improve the recall rate of the activity function, the recall rate is positively correlated with the number of times the real formation interface of the contrast well corresponds to the formation interface determined based on the activity function, that is, the more the number of times, the greater the recall rate. By adjusting the parameters affecting the effect of the activity function based on the recall rate, the formation interface determined based on the adjusted activity function can cover the real formation interface as much as possible.

[0114] The parameters affecting the effect of the activity function can include, but are not limited to, the smoothing filter window length, the activity function window length, and the extraction activity curve local maximum function parameter (scipy.signal.find_peaks). The values of the parameters affecting the effect of the activity function can be determined based on the principle of maximizing the recall rate, and the activity function is determined by using the parameters.

[0115] As an example, to find the optimal parameters, we can define two loss functions to measure the effectiveness of each parameter set. The calculation logic of these two loss functions is as follows: The recall rate is calculated by counting the number of correspondences between the true formation boundaries and the boundaries extracted by the activity function; the precision rate is calculated by counting the number of correspondences between the boundaries extracted by the activity function and the true boundaries. The optimal parameters can be determined by maximizing both recall and precision, and the activity function is determined based on these optimal parameters.

[0116] In step S220, a training sample set is constructed based on the training data.

[0117] Before executing step S220 to construct a training sample set based on the training data, the well logging curves may be preprocessed to ensure their usability and effectiveness. Preprocessing operations performed on the well logging curves may include, but are not limited to, removing negative values, normalizing, and smoothing filtering. For details on the implementation of preprocessing operations, please refer to the relevant description above.

[0118] The training sample set includes at least one training sample, each training sample consists of a stratigraphic interface pair of a standard well and a comparison well whose thickness difference with the stratigraphic interval is lower than a first predetermined threshold, and the stratigraphic interface pair consists of two stratigraphic interfaces.

[0119] by Figure 3 Taking the stratigraphic interface extracted based on the activity stratification method as shown in the figure as an example, all possible stratigraphic interface pairs can be constructed as follows: (1,2), (1,3), (1,4), …, (3,4), (3,5), (3,6), …, (22,23), (22,24), (23,24). For a stratigraphic interval of a standard well, stratigraphic interface pairs with a thickness of ±20 meters of the stratigraphic interval of the standard well are selected from all pairs for comparison to form a training sample.

[0120] When selecting a stratigraphic interface pair group for comparison with the stratigraphic interval of a standard well so that the selected stratigraphic interface pair group and the stratigraphic interval constitute a training sample, a stratigraphic interface pair group can be selected whose depth difference with the stratigraphic interval of the standard well is not much (such as the depth difference is less than the third predetermined threshold value) and whose thickness difference with the thickness of the stratigraphic interval is less than the first predetermined threshold value, and constitute a training sample with the stratigraphic interval of the standard well.

[0121] As an example, a predetermined number of formation levels can be determined, and according to the real formation division information of the standard well, the formation level to which the formation interval of the standard well belongs is determined, each formation level has a corresponding depth range, and the depth ranges corresponding to different formation levels do not overlap. For each formation interval of the standard well, a group of formation interfaces can be selected in the depth range corresponding to the formation level to which the formation interval belongs, and the thickness of the formation interval is less than a first predetermined threshold. The formation interval of the standard well forms a training sample.

[0122] The feature of the training sample is used to represent the similarity between the logging curve (for the sake of distinction, it can be referred to as the third logging curve) of the contrast well in the thickness represented by the group of formation interfaces and the logging curve (for the sake of distinction, it can be referred to as the second logging curve) of the standard well in the thickness represented by the formation interval.

[0123] When constructing the feature of the training sample, a third feature value of the third logging curve and a second feature value of the second logging curve can be calculated, and the difference between the third feature value and the second feature value is taken as the feature of the training sample.

[0124] The third feature value and the second feature value can refer to a numerical value used to reflect the curve shape feature of the logging curve. The third feature value and the second feature value can be calculated by using an algorithm suitable for feature extraction of time series waveform data. The algorithm that can be used can include but is not limited to at least one of DTW / MJC, a statistical feature algorithm, a Shapelet algorithm, a ROCKET algorithm, and an InceptionTime algorithm. DTW is the abbreviation of Dynamic Time Warping; MJC is the abbreviation of Minimum Jump Cost; the Shapelet algorithm, the ROCKET algorithm, and the InceptionTime algorithm are all algorithms suitable for feature extraction of time series waveform data. The statistical feature algorithm can refer to statistical features such as mean, mean square error, variance, and standard deviation of the logging curve.

[0125] In the case where the third feature value and the second feature value both include feature values of multiple dimensions, the difference between the third feature value and the second feature value can refer to the result of mutual subtraction between feature values of the same dimension.

[0126] In order to better extract the local shape feature of the logging curve, the formation interval can also be divided into a plurality of second subintervals, the feature value of the curve part of the second logging curve in each second subinterval is calculated as the second feature value, and the thickness interval represented by the group of formation interfaces is divided into a plurality of third subintervals, and the feature value of the curve part of the third logging curve in each third subinterval is calculated as the third feature value.

[0127] Thus, the second eigenvalue may include multiple eigenvalues ​​of the second sub-interval, the third eigenvalue may include multiple eigenvalues ​​of the third sub-interval, and the difference between the third eigenvalue and the second eigenvalue may refer to the difference between the eigenvalues ​​of the third sub-interval and the second sub-interval corresponding to each other in thickness.

[0128] like Figure 4 As shown, one interval of the standard well NP204 and the comparison well NP203 can be decomposed into 6 small intervals respectively. The characteristics of the logging curves in each small interval are calculated by the above-mentioned calculation method of the curve morphological characteristics, and then the difference between the characteristics of the logging curves in the corresponding small intervals is calculated to characterize the similarity between the logging curves in the small intervals.

[0129] The labels of the training samples are used to indicate whether the stratigraphic division information represented by the stratigraphic interface pair is accurate.

[0130] The label of the training sample can be determined based on whether the difference between the stratigraphic division information represented by the stratigraphic interface pair and the actual stratigraphic division information of the comparison well is greater than a second predetermined threshold. The stratigraphic division information represented by the stratigraphic interface pair can refer to the stratum formed by the two stratigraphic interfaces that constitute the stratigraphic interface pair. The stratigraphic interface in the stratigraphic interface pair that is closer to the ground can be considered as the top depth of the formed stratum, the stratigraphic interface in the stratigraphic interface pair that is farther from the ground can be considered as the bottom depth of the formed stratum, and the thickness between the top depth and the bottom depth can be considered as the layer thickness of the formed stratum.

[0131] After discussions with experts, we can evaluate the difference between the stratigraphic division information represented by the stratigraphic interface pair and the actual stratigraphic division information based on three dimensions: top depth error, bottom depth error, and layer thickness error. This difference may include but is not limited to one or more of the following: top depth error, bottom depth error, and layer thickness error.

[0132] like Figure 5 As shown, the top depth error can refer to the absolute value of the difference between the top depth of the true formation boundary (i.e., the actual formation boundary shown in the figure) and the top depth of the formation boundary represented by the formation interface pair (i.e., the model formation boundary shown in the figure). The bottom depth error can refer to the absolute value of the difference between the bottom depth of the true formation boundary and the bottom depth of the formation interface pair. The layer thickness error can refer to the absolute value of the difference between the layer thickness of the true formation boundary and the layer thickness of the formation interface pair. The layer thickness error can be equal to the top depth error + the bottom depth error.

[0133] For the top depth, the bottom depth, and the layer thickness, the error value that can be tolerated in each dimension can be set to 5%-10% of the actual layer thickness, referred to as error_margin. Taking 10% as an example, the calculation logic of error_margin is as follows: error_margin = 0.10 x actual layer thickness. The total error value that can be tolerated is calculated by error_margin and can be referred to as total_allowed_error. The specific calculation logic of total_allowed_error is as follows: total_allowed_error = error_margin x 3. Finally, a total error value referred to as total_error is calculated. The calculation logic is as follows: total_error = top depth error + bottom depth error + layer thickness error. If total_error <= total_allowed_error, the label is 1, otherwise it is 0.

[0134] In step S230, a scoring model for scoring the accuracy of the pair of formation interfaces is trained based on the training sample set.

[0135] The scoring model can be used to predict the accuracy of the pair of formation interfaces, and output a score value reflecting the accuracy of the pair of formation interfaces. The score value can be a probability value between 0 and 1, and the probability value is positively correlated with the accuracy.

[0136] The scoring model can be a binary classification model, specifically but not limited to a GBDT model. The scoring model can output a probability value of two logging curves being of the same type. This probability value can be understood as a curve similarity, and this probability value can be used as a score of the pair of formation interfaces.

[0137] The present disclosure can also construct a validation sample set based on the training data. The validation sample set can include at least one validation sample. Each validation sample is composed of a formation interval of a standard well and a pair of formation interfaces of a comparison well having a thickness difference less than a first predetermined threshold from the formation interval. The pair of formation interfaces is composed of two formation interfaces. The features of the validation sample are used to represent the similarity between a third logging curve of the comparison well within the thickness represented by the pair of formation interfaces and a second logging curve of the standard well within the thickness represented by the formation interval.

[0138] For a formation interval of a standard well, the trained scoring model can be used to score the pair of formation interfaces in the validation sample of the corresponding comparison well constructed based on the formation interval. The accuracy of the scoring model is evaluated according to the difference between the formation division information represented by the pair of formation interfaces with the highest score and the real formation division information.

[0139] The stratum interface pair set with the highest score can be regarded as the division result of the stratum corresponding to the stratum interval of the standard well of the contrast well. If the difference between the stratum division information represented by the stratum interface pair set with the highest score and the real stratum division information is greater than the maximum error that can be tolerated (i.e., a second predetermined threshold, such as 0.1), it is considered that the scoring model is inaccurate. Thus, the accuracy of the scoring model can be tested based on the verification sample set. If the accuracy of the scoring model does not meet the standard (such as the accuracy being lower than 60%), the scoring model can be continuously trained until the accuracy of the scoring model meets the standard.

[0140] So far, the training process of the scoring model has been described. Figures 2 to 5 The training process of the scoring model has been described.

[0141] Returning to the description of step S120 of the above method, Figure 1 In the process of scoring the stratum interface pair set using the trained scoring model, the first feature value of the first logging curve in the thickness represented by the selected stratum interface pair set and the second feature value of the second logging curve in the thickness represented by the standard well in the stratum interval can be calculated, and the difference between the first feature value and the second feature value is used as the model input to score the stratum interface pair set using the pre-trained scoring model. In the process of calculating the first feature value of the first logging curve in the thickness represented by the selected stratum interface pair set and the second feature value of the second logging curve in the thickness represented by the standard well in the stratum interval, the thickness interval represented by the selected stratum interface pair set can be divided into a plurality of first subintervals, the feature values of the curve portions of the first logging curve in each first subinterval in the thickness represented by the selected stratum interface pair set are calculated as the first feature value, and the stratum interval is divided into a plurality of second subintervals, and the feature values of the curve portions of the second logging curve in each second subinterval in the thickness represented by the standard well in the stratum interval are calculated as the second feature value.

[0142] Particular application examples

[0143] In order to make the purposes, technical solutions and advantages of the present disclosure clearer, the technical solutions and processes of the present disclosure will be described in detail below in combination with the processing flow of actual well data in a block.

[0144] 1) Select 114 well data of a block, randomly select 60 wells as a training set, and 54 wells as a test set. The selected logging curves are gamma (GR), acoustic (AC), natural potential (SP), and shallow detection induction (RLLD).

[0145] First, all data is cleaned and preprocessed to remove invalid values (i.e., negative values) in the well data, and extreme value normalization and smoothing filtering are performed on the curves.

[0146] 2) Select NP204 as the standard well, and the other 60 wells in the training set as the comparison wells.

[0147] 3) Automatically stratify the 60 comparison wells in the training set by the activity function to obtain the stratigraphic interface division information of each well.

[0148] 4) Group the 60 comparison wells in the training set.

[0149] 5) Adjust the activity function by the actual stratigraphic division information of the NP204 standard well and the grouping of the 60 comparison wells. For the parameters that affect the effect of the activity function, the following groups of parameters are tested, and the recall rates are calculated:

[0150] a. Smooth filtering window length: [10, 30, 50, 70, 100];

[0151] b. Activity function window length: 30 equal length values between 0 and 202 are tested, i.e. 30 values with fixed difference are tested in the range of 0-202;

[0152] c. Extract local maximum value function parameters of activity curve (scipy.signal.find_peaks).

[0153] i. Promninences: 15 equal length values between 0 and 0.2 are tested;

[0154] ii. Heights: 15 equal length values between 0 and 0.2 are tested;

[0155] iii. Distances: 0, 5, 10…50 are tested respectively, i.e. one value every 5 between 0 and 52 is tested.

[0156] The current highest recall rate parameter group is:

[0157] · Smooth filtering window length: 30

[0158] · Activity function window length: 30

[0159] · Prominence: 0.0

[0160] · Height: 0.2

[0161] · Distance: 45.0

[0162] 6) For the contrast well pair group extracted by the activity function of the well, and the standard well label, the interval of the standard well and the contrast well is respectively divided into 6 small intervals, and the local curve shape feature is constructed. Through the above-constructed features, the feature values extracted from the standard well formation can be obtained, and the feature values extracted from the contrast well are subtracted to obtain the absolute difference value. The contrast of each layer of the standard well and the contrast well can be understood as a sample.

[0163] 7) For the above sample, the top depth error, low depth error and layer thickness error are calculated, and the label of tolerable error value of 0 and 1 is input into GBDT for binary classification learning. GBDT will finally output two curves as the probability value of the same class. This probability value is understood as the similarity of the curves.

[0164] 8) Similarly, the activity function of the well in the 54 test set is determined, and the pair group and the feature are constructed. Finally, the GBDT model trained is used to score the pair group belonging to each layer, and the pair group with the highest score is selected as the prediction result of each layer, and the test set accuracy is calculated. The current tolerable error value is 0.15, and the accuracy is 61%.

[0165] Compared with the existing automatic layering technology, the present disclosure has at least the following advantages.

[0166] The activity function method is generally analyzed for a logging curve, and the information is lacking and not representative. The present disclosure constructs a pair group by the activity function method and takes it as a sample of the training model. The constructed sample extracts curve shape features through multiple logging curves, which is more informative and representative.

[0167] The adaptability of the dynamic pattern matching algorithm is not very strong, and it can only be used in one region. Different regions cannot use the same set of weights for calculation, otherwise it will be difficult to match the contrast results with the manual contrast results. The modeling scheme of the present disclosure divides each curve into small segments and constructs the difference of each segment, in order to better extract local and global shape features. This method can make up for the poor adaptability of the dynamic pattern matching algorithm.

[0168] The wavelet transform method for finding the formation interface will encounter obstacles, and even how to set the threshold value will cause a certain number of false formations or real formation interfaces to be removed together with false formations. Through the scoring of each pair group by the model, the model will give a higher score to the real formation pair group, so that it will be ranked in the front during sorting, and the false formation will be ranked in the back. The scoring of the model can be understood as a filtering mechanism.

[0169] The formation identification method of the present disclosure can also be implemented as a formation identification device. Figure 6A structural block diagram of a formation identification apparatus is shown according to an example embodiment of the present disclosure. The functional units of the formation identification apparatus can be implemented by hardware, software or a combination of hardware and software to realize the principles of the present disclosure. Those skilled in the art can understand that, Figure 6 The described functional units can be combined or divided into sub-units to realize the principles of the above-mentioned invention. Therefore, the description herein can support any possible combination, or division, or further limitation of the functional units described herein.

[0170] The following briefly describes the functional units that the formation identification apparatus can have and the operations that the functional units can perform. For the details involved therein, please refer to the relevant description above, which will not be repeated here.

[0171] Referring to Figure 6 , the formation identification apparatus 600 includes a division module 610, an evaluation module 620 and a selection module 630.

[0172] The division module 610 is configured to divide the target well based on an automatic layering technique to obtain a division result of the formation interface of the target well. The evaluation module 620 is configured to evaluate the accuracy of the formation interface represented by the division result according to the real formation division information of the standard well, the logging curve of the standard well and the logging curve of the target well. The selection module 630 is configured to select the formation interface from the division result based on the evaluation result to obtain the final division result of the formation interface of the target well.

[0173] The real formation division information can include a plurality of formation intervals, and the evaluation module 620 can include a construction module, an interface pair selection module and a scoring module. The construction module is configured to construct a formation interface pair based on the division result of the formation interface of the target well, the formation interface pair consisting of two formation interfaces. The interface pair selection module is configured to select, for each formation interval of the standard well, a formation interface pair whose thickness difference from the thickness of the formation interval is less than a first predetermined threshold. The scoring module is configured to score the formation interface pair according to the similarity between the first logging curve of the target well within the thickness represented by the selected formation interface pair and the second logging curve of the standard well within the thickness represented by the formation interval, the score being positively correlated with the similarity.

[0174] The formation identification apparatus 600 can further include a first determination module and a second determination module. The first determination module is configured to determine a predetermined number of formation levels, each formation level having a corresponding depth range, and the corresponding depth ranges of different formation levels have no overlapping parts. The second determination module is configured to determine a formation level to which a formation interval of the standard well belongs. The interface pair selection module can select, for each formation interval of the standard well, a formation interface pair group from the selected formation interface pair group, the formation interface pair group having a thickness difference with the thickness of the formation interval less than a first predetermined threshold in the depth range corresponding to the formation level to which the formation interval belongs.

[0175] The selection module 630 can select, for each formation interval of the standard well, a formation interface pair group with the highest score from the selected formation interface pair group as the formation division result corresponding to the formation interval of the target well.

[0176] The scoring module can include a calculation module and a prediction module. The calculation module is configured to calculate a first feature value of the first logging curve in the thickness represented by the selected formation interface pair group and a second feature value of the second logging curve in the thickness represented by the standard well in the formation interval. The prediction module is configured to use the pre-trained scoring model to score the formation interface pair group by taking the difference between the first feature value and the second feature value as the model input.

[0177] The calculation module can include a first calculation module and a second calculation module. The first calculation module is configured to divide the thickness interval represented by the selected formation interface pair group into a plurality of first sub-intervals, and calculate the feature values of the curve portions of the first logging curve in each first sub-interval within the thickness represented by the selected formation interface pair group as the first feature values. The second calculation module is configured to divide the formation interval into a plurality of second sub-intervals, and calculate the feature values of the curve portions of the second logging curve in each second sub-interval within the thickness represented by the standard well in the formation interval as the second feature values.

[0178] The division module 610 can include an activity function determination module and a formation interface determination module. The activity function determination module is configured to determine an activity function of the logging curve of the target well, the activity function being used to represent the activity at different positions on the logging curve, the activity being used to reflect the dynamic properties of the logging curve. The formation interface determination module is configured to determine, according to the activity function, the positions where the activity belongs to local maximum values as the formation interfaces.

[0179] The formation identification apparatus 600 can further include at least one of a removal module, a normalization processing module, and a filtering processing module. The removal module is configured to remove negative values in the logging curve. The normalization processing module is configured to perform normalization processing on the logging curve. The filtering processing module is configured to perform smoothing filtering processing on the logging curve.

[0180] The model training method of the present disclosure can also be implemented as a model training apparatus.Figure 7 A structural block diagram of a model training apparatus according to an example embodiment of the present disclosure is shown. The functional units of the model training apparatus can be implemented by hardware, software or a combination of hardware and software implementing the principles of the present disclosure. Those skilled in the art can understand that, Figure 7 The described functional units can be combined or divided into sub-units to implement the principles of the above-described application. Therefore, the description herein can support any possible combination, or division, or further limitation of the functional units described herein.

[0181] The functional units that the model training apparatus can have and the operations that each functional unit can perform are briefly described below, and the detailed parts involved therein can be referred to the related description above, which will not be repeated here.

[0182] Referring to Figure 7 The model training apparatus 700 includes an acquisition module 710, a construction module 720 and a training module 730.

[0183] The acquisition module 710 is configured to acquire training data, the training data including real formation division information of a standard well, a division result of a formation interface of at least one contrast well determined based on an automatic layering technology, a well logging curve of the standard well and a well logging curve of the contrast well, the real formation division information including a plurality of formation intervals.

[0184] The construction module 720 is configured to construct a training sample set based on the training data, the training sample set including at least one training sample, each training sample being composed of one formation interval of the standard well and one formation interface pair of the contrast well having a thickness difference lower than a first predetermined threshold value from the formation interval, the formation interface pair being composed of two formation interfaces, a feature of the training sample being configured to represent a similarity between a third well logging curve of the contrast well within a thickness represented by the formation interface pair and a second well logging curve of the standard well within a thickness represented by the formation interval, and a label of the training sample being configured to represent whether the formation division information represented by the formation interface pair is accurate.

[0185] The training module 730 is configured to train a scoring model for scoring the accuracy of the formation interface pair based on the training sample set.

[0186] The model training apparatus 700 can further include at least one of a removal module, a normalization processing module and a filtering processing module. Before the construction module 720 constructs the training sample set based on the training data, the removal module can be used to remove negative values in the well logging curve; and / or the normalization processing module can be used to normalize the well logging curve; and / or the filtering processing module can be used to perform smoothing filtering processing on the well logging curve.

[0187] The acquisition module 710 can include an activity function determination module and a formation interface determination module. The activity function determination module is configured to determine an activity function of the logging curve of the contrast well, the activity function being used to represent the activity at different positions on the logging curve, and the activity being used to reflect the dynamic property of the logging curve. The formation interface determination module is configured to, according to the activity function, take the position where the activity is a local maximum as the formation interface of the contrast well, to obtain the division result of the formation interface of the contrast well.

[0188] The model training apparatus 700 can further include a parameter adjustment module. The parameter adjustment module is configured to, before the activity function determination module determines the activity function of the logging curve of the contrast well, adjust a parameter that affects the effect of the activity function, to improve the recall rate of the activity function, the recall rate being positively correlated with the number of times that the real formation interface of the contrast well corresponds to the formation interface determined based on the activity function.

[0189] The construction module 720 can include a calculation module and a sample feature determination module. The calculation module is configured to calculate a third feature value of the third logging curve of the contrast well within the thickness represented by the group of formation interfaces, and a second feature value of the second logging curve of the standard well within the thickness represented by the formation interval. The sample feature determination module is configured to take the difference between the third feature value and the second feature value as the feature of the training sample.

[0190] The calculation module can include a third calculation module and a fourth calculation module. The third calculation module is configured to divide the thickness interval represented by the group of formation interfaces into a plurality of third sub-intervals, and calculate the feature value of the curve part of the third logging curve of the contrast well within each third sub-interval as the third feature value. The fourth calculation module is configured to divide the formation interval into a plurality of second sub-intervals, and calculate the feature value of the curve part of the second logging curve of the standard well within each second sub-interval as the second feature value.

[0191] The construction module 720 can include a sample label determination module. The sample label determination module is configured to determine the label of the training sample according to whether the difference between the formation division information represented by the group of formation interfaces and the real formation division information of the contrast well is greater than a second predetermined threshold.

[0192] The construction module 720 can also construct a verification sample set based on the training data, the verification sample set including at least one verification sample, each verification sample being composed of one formation interval of a standard well and one formation interface pair of a contrast well, the formation interface pair being composed of two formation interfaces, the difference between the thicknesses of the formation interval and the formation interface pair being less than a first predetermined threshold, the characteristics of the verification sample being used to represent the similarity between the third logging curve of the contrast well within the thickness represented by the formation interface pair and the second logging curve of the standard well within the thickness represented by the formation interval. The model training apparatus 700 can further include a model evaluation module. The model evaluation module is configured to score, for one formation interval of a standard well, the formation interface pair in the verification sample of the corresponding contrast well constructed based on the formation interval using the trained scoring model, and evaluate the accuracy of the scoring model according to the difference between the formation represented by the formation interface pair with the highest score and the real formation division information of the contrast well.

[0193] Figure 8 A structural schematic diagram of a computing device according to an embodiment of the present disclosure is shown, which can be used to implement the above-mentioned formation identification method or model training method.

[0194] Referring to Figure 8 The computing device 800 includes a memory 810 and a processor 820.

[0195] The processor 820 can be a multi-core processor, or can include multiple processors. In some embodiments, the processor 820 can include a general-purpose main processor and one or more special-purpose coprocessors, such as a graphics processing unit (GPU), a digital signal processor (DSP), and the like. In some embodiments, the processor 820 can be implemented using a customized circuit, such as an application specific integrated circuit (ASIC) or a field programmable gate array (FPGA).

[0196] The memory 810 can include various types of storage units, such as a system memory, a read-only memory (ROM), and a permanent storage device. Among them, the ROM can store static data or instructions required by the processor 820 or other modules of the computer. The permanent storage device can be a read-write storage device. The permanent storage device can be a non-volatile storage device that does not lose stored instructions and data even after the computer is powered off. In some embodiments, the permanent storage device uses a mass storage device (such as a magnetic or optical disk, a flash memory) as a permanent storage device. In some other embodiments, the permanent storage device can be a removable storage device (such as a floppy disk, an optical drive). The system memory can be a read-write storage device or a volatile read-write storage device, such as a dynamic random access memory. The system memory can store some or all of the instructions and data required by the processor during runtime. In addition, the memory 810 can include a combination of any computer readable storage media, including various types of semiconductor storage chips (DRAM, SRAM, SDRAM, flash memory, programmable read-only memory), magnetic disks and / or optical disks. In some embodiments, the memory 810 can include a read and / or write removable storage device, such as a compact disc (CD), a read-only digital versatile disc (such as DVD-ROM, double-layer DVD-ROM), a read-only Blu-ray disc, an ultra-density optical disc, a flash memory card (such as an SD card, a min SD card, a Micro-SD card, etc.), a magnetic floppy disk, etc. The computer readable storage medium does not include a carrier wave and a transient electronic signal transmitted through wireless or wired transmission.

[0197] The memory 810 stores executable code, which, when processed by the processor 820, can cause the processor 820 to perform the formation identification method or the model training method described above.

[0198] The formation identification method or the model training method according to the present disclosure, the formation identification device, the model training device, and the computing device have been described in detail above with reference to the accompanying drawings.

[0199] In addition, the method according to the present disclosure can also be implemented as a computer program or a computer program product, which includes computer program code instructions for performing each step defined in the above method of the present disclosure.

[0200] Alternatively, the present disclosure can also be implemented as a non-transitory machine readable storage medium (or computer readable storage medium, or machine readable storage medium) having stored executable code (or computer program, or computer instruction code) thereon, which, when executed by a processor of an electronic device (or computing device, server, etc.), causes the processor to perform each step of the above method according to the present disclosure.

[0201] Those skilled in the art will further appreciate that the various illustrative logical blocks, modules, circuits, and algorithm steps described in connection with the disclosure herein can be implemented as electronic hardware, computer software, or combinations of both.

[0202] The flow diagrams and the block diagrams in the drawings are presented to illustrate the architecture, functionality, and operation of possible implementations of systems and methods in accordance with the present disclosure. In this regard, each block in the flow diagrams and the block diagrams can represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that in some alternative implementations, the functions noted in the block can occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and / or flow diagrams, and combinations thereof, can be implemented by special purpose hardware-based systems that perform the specified functions or operations, or combinations of special purpose hardware and

[0203] Embodiments of the present disclosure have been described above, and the description is intended to be illustrative of the embodiments and not exhaustive, and is not limited to the embodiments disclosed. Numerous modifications and adaptations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The selection of the terms to be used in the description is intended to best describe the principles of the embodiments, practical application, or improvement to the art, or to enable other ordinary skilled in the art to understand the embodiments disclosed herein.

Claims

1. A formation identification method, comprising: Performing stratigraphic division on the target well based on the automatic stratification technology to obtain a division result of the stratigraphic interface of the target well; Evaluating the accuracy of the stratigraphic interfaces represented by the division results based on the actual stratigraphic division information of the standard well, the well logging curve of the standard well, and the well logging curve of the target well; as well as Selecting a formation interface from the division results based on the evaluation result to obtain a final division result of the formation interface of the target well; The actual stratigraphic division information includes multiple stratigraphic intervals. The step of evaluating the accuracy of the stratigraphic interface based on the actual stratigraphic division information of the standard well, the well logging curve of the standard well, and the well logging curve of the target well includes: constructing a formation interface pair group based on the division result of the formation interface of the target well, wherein the formation interface pair group consists of two formation interfaces; For each stratigraphic interval of the standard well, selecting a pair of stratigraphic interfaces whose thickness differs from the thickness of the stratigraphic interval by less than a first predetermined threshold; Scoring the stratigraphic interface pair according to the similarity between the first well logging curve of the target well within the thickness represented by the selected stratigraphic interface pair and the second well logging curve of the standard well within the thickness represented by the stratigraphic interval, wherein the score is positively correlated with the similarity; The step of scoring the formation interface pair comprises: Calculating a first characteristic value of a first well logging curve within the thickness represented by the selected pair of formation interfaces and a second characteristic value of a second well logging curve of the standard well within the thickness represented by the formation interval; The difference between the first eigenvalue and the second eigenvalue is used as a model input, and a pre-trained scoring model is used to score the group of formation interfaces.

2. The method according to claim 1, further comprising: Determining a predetermined number of stratigraphic levels, each stratigraphic level having a corresponding depth range, and depth ranges corresponding to different stratigraphic levels having no overlapping portions; Determine the stratigraphic level to which the stratigraphic interval of the standard well belongs, Among them, for each stratigraphic interval of the standard well, the step of selecting a stratigraphic interface pair whose thickness difference from the thickness of the stratigraphic interval is less than a first predetermined threshold includes: for each stratigraphic interval of the standard well, selecting a stratigraphic interface pair whose thickness difference from the thickness of the stratigraphic interval is less than a first predetermined threshold within the depth range corresponding to the stratigraphic level to which the stratigraphic interval belongs.

3. The method according to claim 1, wherein The step of selecting a formation interface from the division results based on the evaluation result to obtain a final division result of the formation interface of the target well includes: For each stratigraphic interval of the standard well, a stratigraphic interface pair group with the highest score is selected from the selected stratigraphic interface pair groups as the stratigraphic division result corresponding to the target well and the stratigraphic interval.

4. The method according to claim 1, wherein The step of calculating a first characteristic value of a first well logging curve within the thickness represented by the selected pair of formation interfaces and a second characteristic value of a second well logging curve of the standard well within the thickness represented by the formation interval comprises: Dividing the thickness interval represented by the selected stratigraphic interface pair into a plurality of first subintervals, and calculating a characteristic value of a curve portion of the first well logging curve within the thickness represented by the selected stratigraphic interface pair in each of the first subintervals as a first characteristic value; The formation interval is divided into a plurality of second subintervals, and characteristic values ​​of curve portions of the second well logging curve of the standard well within the thickness represented by the formation interval within each second subinterval are calculated as second characteristic values.

5. The method according to claim 1, wherein The automatic tiering technology includes: Activity stratification; and / or Dynamic matching algorithms; and / or Wavelet transform method.

6. The method according to claim 1, wherein The steps for dividing the target well into strata based on the automatic stratification technology include: Determining an activity function of a well logging curve of the target well, wherein the activity function is used to characterize the activity at different positions on the well logging curve, and the activity is used to reflect the dynamic properties of the well logging curve; According to the activity function, the position where the activity is at a local maximum is taken as the formation interface.

7. The method according to claim 1, further comprising: removing negative values ​​from the well logging curve; and / or performing normalization processing on the well logging curve; and / or The well logging curve is smoothed and filtered.

8. A model training method for scoring the accuracy of stratigraphic division results, comprising: Acquiring training data, the training data including real stratigraphic division information of a standard well, division results of stratigraphic interfaces of at least one comparison well determined based on automatic stratification technology, well logging curves of the standard well, and well logging curves of the comparison well; Constructing a training sample set based on the training data, the training sample set including at least one training sample, each training sample consisting of a stratigraphic interface pair of a stratigraphic interval of the standard well and a stratigraphic interface pair of the comparison well, the difference in thickness between the stratigraphic interval and the stratigraphic interval being less than a first predetermined threshold, the stratigraphic interface pair consisting of two stratigraphic interfaces, the features of the training samples being used to characterize the similarity between a third well logging curve of the comparison well within the thickness represented by the stratigraphic interface pair and a second well logging curve of the standard well within the thickness represented by the stratigraphic interval, and the labels of the training samples being used to characterize whether stratigraphic division information represented by the stratigraphic interface pair is accurate; The step of constructing a training sample set based on the training data includes: calculating a third eigenvalue of a third well logging curve of the comparison well within the thickness represented by the formation interface pair and a second eigenvalue of a second well logging curve of the standard well within the thickness represented by the formation interval; and using the difference between the third eigenvalue and the second eigenvalue as a feature of the training sample; A scoring model for scoring the accuracy of the set of bed interface pairs is trained based on the training sample set.

9. The method according to claim 8, wherein Before constructing a training sample set based on the training data, the method further includes: removing negative values ​​from the well log; and / or performing normalization processing on the well logging curve; and / or The logging curve is smoothed and filtered.

10. The method according to claim 8, wherein The automatic tiering technology includes: Activity stratification; and / or Dynamic matching algorithms; and / or Wavelet transform method.

11. The method according to claim 8, wherein The steps to obtain training data include: Determining an activity function of the well logging curve of the comparison well, wherein the activity function is used to characterize the activity at different positions on the well logging curve, and the activity is used to reflect the dynamic properties of the well logging curve; According to the activity function, the position where the activity belongs to the local maximum value is used as the formation interface of the comparison well to obtain the division result of the formation interface of the comparison well.

12. The method according to claim 11, wherein Before determining the activity function of the well logging curve of the comparison well, the method further includes: Parameters affecting the effect of the activity function are adjusted to improve the recall rate of the activity function, wherein the recall rate is positively correlated with the number of times the actual formation interface of the comparison well corresponds to the formation interface determined based on the activity function.

13. The method according to claim 8, wherein The step of calculating a third characteristic value of the third well logging curve of the comparison well within the thickness represented by the formation interface pair and a second characteristic value of the second well logging curve of the standard well within the thickness represented by the formation interval includes: Dividing the thickness interval represented by the formation interface pair into a plurality of third subintervals, and calculating the characteristic value of the curve portion of the third well logging curve of the comparison well within the thickness represented by the formation interface pair in each of the third subintervals as a third characteristic value; The formation interval is divided into a plurality of second subintervals, and characteristic values ​​of curve portions of the second well logging curve of the standard well within the thickness represented by the formation interval within each second subinterval are calculated as second characteristic values.

14. The method according to claim 8, wherein The step of constructing a training sample set based on the training data includes: The label of the training sample is determined according to whether the difference between the stratigraphic division information represented by the stratigraphic interface pair group and the actual stratigraphic division information of the comparison well is greater than a second predetermined threshold.

15. The method according to claim 8, further comprising: Constructing a validation sample set based on the training data, the validation sample set including at least one validation sample, each validation sample consisting of a stratigraphic interface pair of a standard well and a comparison well, the difference in thickness between the stratigraphic interface pair and the stratigraphic interval being less than a first predetermined threshold, the stratigraphic interface pair consisting of two stratigraphic interfaces, and features of the validation sample being used to characterize similarity between a third well logging curve of the comparison well within the thickness represented by the stratigraphic interface pair and a second well logging curve of the standard well within the thickness represented by the stratigraphic interval; For a stratigraphic interval of the standard well, the trained scoring model is used to score the stratigraphic interface pairs in the verification sample corresponding to the same comparison well constructed based on the stratigraphic interval, and the accuracy of the scoring model is evaluated based on the difference between the stratigraphic division information represented by the stratigraphic interface pair with the highest score and the actual stratigraphic division information of the comparison well.

16. A stratum identification device, comprising: A division module, configured to divide the target well into strata based on an automatic stratification technology, and obtain a division result of the stratum interface of the target well; An evaluation module, configured to evaluate the accuracy of the stratigraphic interface represented by the division result based on the actual stratigraphic division information of the standard well, the well logging curve of the standard well, and the well logging curve of the target well; as well as A selection module, configured to select a formation interface from the division results based on an evaluation result, and obtain a final division result of the formation interface of the target well; The actual stratigraphic division information includes multiple stratigraphic intervals, and the evaluation module includes: A construction module, configured to construct a formation interface pair group based on the division result of the formation interface of the target well, wherein the formation interface pair group consists of two formation interfaces; an interface pair selection module, configured to select, for each stratigraphic interval of the standard well, a stratigraphic interface pair whose thickness difference from the thickness of the stratigraphic interval is less than a first predetermined threshold; a scoring module for scoring the selected stratigraphic interface pair based on similarity between a first well logging curve of the target well within the thickness represented by the selected stratigraphic interface pair and a second well logging curve of the reference well within the thickness represented by the stratigraphic interval, wherein the score is positively correlated with the similarity; The scoring module includes: a calculation module for calculating a first characteristic value of a first well logging curve within a thickness represented by the selected pair of formation interfaces and a second characteristic value of a second well logging curve of the standard well within a thickness represented by the formation interval; A prediction module is configured to use a difference between the first eigenvalue and the second eigenvalue as a model input and to score the formation interface pair using a pre-trained scoring model.

17. The apparatus according to claim 16, further comprising: A first determining module is configured to determine a predetermined number of stratigraphic levels, each stratigraphic level having a corresponding depth range, and depth ranges corresponding to different stratigraphic levels having no overlapping portions; The second determination module is used to determine the stratigraphic level to which the stratigraphic interval of the standard well belongs. The interface pair selection module selects, for each stratigraphic interval of the standard well, stratigraphic interface pairs whose thickness difference from the thickness of the stratigraphic interval is less than a first predetermined threshold within a depth range corresponding to the stratigraphic level to which the stratigraphic interval belongs.

18. The device according to claim 16, wherein The selection module selects, for each stratigraphic interval of the standard well, a stratigraphic interface pair group with the highest score from the selected stratigraphic interface pairs as a stratigraphic division result corresponding to the target well and the stratigraphic interval.

19. The device according to claim 16, wherein The calculation module includes: a first calculation module, configured to divide the thickness interval represented by the selected formation interface pair into a plurality of first subintervals, and calculate a characteristic value of a curve portion of the first well logging curve within the thickness represented by the selected formation interface pair in each of the first subintervals as a first characteristic value; The second calculation module is used to divide the formation interval into multiple second sub-intervals, and calculate the characteristic value of the curve portion of the second logging curve of the standard well within the thickness represented by the formation interval in each second sub-interval as the second characteristic value.

20. The apparatus according to claim 16, wherein The automatic tiering technology includes: Activity stratification; and / or Dynamic matching algorithms; and / or Wavelet transform method.

21. The apparatus according to claim 16, wherein The division module includes: an activity function determination module, configured to determine an activity function of a well logging curve of the target well, wherein the activity function is configured to characterize the activity at different positions on the well logging curve, and the activity is configured to reflect the dynamic properties of the well logging curve; The formation interface determination module is used to determine the position where the activity belongs to the local maximum value as the formation interface according to the activity function.

22. The apparatus of claim 16, further comprising: A removal module, configured to remove negative values ​​from the well logging curve; and / or A normalization processing module, used for performing normalization processing on the well logging curve; and / or The filtering processing module is used to perform smooth filtering on the logging curve.

23. A model training device for scoring the accuracy of stratigraphic division results, comprising: an acquisition module, configured to acquire training data, the training data including real stratigraphic division information of a standard well, division results of stratigraphic interfaces of at least one comparison well determined based on automatic stratification technology, well logging curves of the standard well, and well logging curves of the comparison well; the real stratigraphic division information including multiple stratigraphic intervals; a construction module for constructing a training sample set based on the training data, the training sample set including at least one training sample, each training sample consisting of a stratigraphic interface pair of a stratigraphic interval of the standard well and a stratigraphic interface pair of the comparison well, the difference in thickness between the stratigraphic interval and the stratigraphic interval being less than a first predetermined threshold, the stratigraphic interface pair consisting of two stratigraphic interfaces, the features of the training sample being used to characterize the similarity between a third well logging curve of the comparison well within the thickness represented by the stratigraphic interface pair and a second well logging curve of the standard well within the thickness represented by the stratigraphic interval, and the label of the training sample being used to characterize whether the stratigraphic division information represented by the stratigraphic interface pair is accurate; A training module, configured to train a scoring model for scoring the accuracy of the formation interface pair based on the training sample set; Among them, the construction module includes: a calculation module, used to calculate the third eigenvalue of the third logging curve of the comparison well within the thickness represented by the formation interface pair group and the second eigenvalue of the second logging curve of the standard well within the thickness represented by the formation interval; a sample feature determination module, used to use the difference between the third eigenvalue and the second eigenvalue as the feature of the training sample.

24. The apparatus according to claim 23, further comprising: A removal module, configured to remove negative values ​​from the well logging curve; and / or A normalization processing module, used for performing normalization processing on the well logging curve; and / or The filtering processing module is used to perform smooth filtering on the logging curve.

25. The apparatus according to claim 23, wherein The automatic tiering technology includes: Activity stratification; and / or Dynamic matching algorithms; and / or Wavelet transform method.

26. The device according to claim 23, wherein The acquisition module includes: an activity function determination module, configured to determine an activity function of the well logging curve of the comparison well, wherein the activity function is configured to characterize the activity at different positions on the well logging curve, and the activity is configured to reflect the dynamic properties of the well logging curve; The formation interface determination module is used to take the position where the activity belongs to the local maximum value as the formation interface of the comparison well according to the activity function, so as to obtain the division result of the formation interface of the comparison well.

27. The device according to claim 26, wherein The device also includes: A parameter adjustment module is used to adjust the parameters that affect the effect of the activity function before the activity function determination module determines the activity function of the logging curve of the comparison well, so as to improve the recall rate of the activity function. The recall rate is positively correlated with the number of times the real formation interface of the comparison well corresponds to the formation interface determined based on the activity function.

28. The apparatus according to claim 23, wherein The calculation module includes: a third calculation module, configured to divide the thickness interval represented by the formation interface pair into a plurality of third subintervals, and calculate a characteristic value of a curve portion of the third well logging curve of the comparison well within the thickness represented by the formation interface pair within each of the third subintervals as a third characteristic value; The fourth calculation module is used to divide the formation interval into multiple second sub-intervals, and calculate the characteristic value of the curve portion of the second logging curve of the standard well within the thickness represented by the formation interval in each second sub-interval as the second characteristic value.

29. The apparatus according to claim 23, wherein The construction module includes: The sample label determination module is used to determine the label of the training sample according to whether the difference between the stratigraphic division information represented by the stratigraphic interface pair group and the actual stratigraphic division information of the comparison well is greater than a second predetermined threshold.

30. The apparatus of claim 23, wherein: The construction module further constructs a validation sample set based on the training data, the validation sample set including at least one validation sample, each validation sample consisting of a stratigraphic interface pair of a standard well and a comparison well, the difference in thickness between the stratigraphic interface pair and the stratigraphic interval being less than a first predetermined threshold, the stratigraphic interface pair consisting of two stratigraphic interfaces, the features of the validation sample being used to characterize the similarity between a third well logging curve of the comparison well within the thickness represented by the stratigraphic interface pair and a second well logging curve of the standard well within the thickness represented by the stratigraphic interval, the device further comprising: The model evaluation module is used to score the stratigraphic interface pairs in the verification sample corresponding to the same comparison well constructed based on a stratigraphic interval of the standard well using the trained scoring model, and evaluate the accuracy of the scoring model based on the difference between the stratigraphic division information of the stratigraphic interface pair with the highest score and the actual stratigraphic division information of the comparison well.

31. A computing device comprising: processor; as well as A memory having executable codes stored thereon, which, when executed by the processor, causes the processor to perform the method according to any one of claims 1 to 15.

32. A non-transitory machine-readable storage medium having executable codes stored thereon, which, when executed by a processor of an electronic device, causes the processor to perform the method according to any one of claims 1 to 15.

Images