Transverse logging series acoustic curve reconstruction method

Abstract

The invention relates to the technical field of oilfield development logging, and discloses a transverse logging series acoustic curve reconstruction method which comprises the following steps: S1, normalizing all curves except a target curve; s2, analyzing according to the correlation between the normalized curve and a target reconstruction curve; s3, on the basis of the selected model curve, establishing a feature sample library, and performing training by adopting a multivariate judgment-based graph group clustering model; s4, setting a model logging facies minimum value, a model logging facies maximum value and the number of selectable models, and adjusting and optimizing the models in real time; and S5, after the prediction curve is generated, the curve is processed through smooth filtering work. According to the method provided by the invention, related features can be extracted from other logging curves through a graph group clustering technology, and then a missing acoustic curve is reconstructed. Therefore, the problem that the reservoir physical property parameters cannot be calculated in early development logging is effectively solved, and reliable data support is provided for subsequent oil reservoir evaluation and development decision.

Description

Technical Field

[0001] The present invention relates to the technical field of oilfield development logging, and specifically provides a method for reconstructing acoustic curves of a lateral logging series. Background Art

[0002] In the early stage of oilfield development, the electric logging curves are mostly electrode system curves, which have good guiding significance for determining reservoir thickness and perforation construction in the initial development stage. However, physical property parameter curves such as acoustic travel time are missing, and key reservoir parameters such as porosity and permeability cannot be calculated, affecting reserve calculation and subsequent development adjustment.

[0003] In the acoustic curve reconstruction technology, the core idea is mainly the effective identification and construction of reservoir geophysical characteristics. Generally starting from reservoir properties, the "four-property" relationship between lithology, physical properties, oil-bearing property and electrical property is studied. By applying certain mathematical model means and methods, geophysical information reflecting reservoir characteristics is reasonably extracted and characterized, and the "pseudo-acoustic" curve reflecting reservoir characteristics is reconstructed. Currently, there are two common types of curve reconstruction technologies. One is mathematical statistics regression or empirical formula. Mathematical statistics regression mainly requires determining the correlation between the reconstructed curve and other logging curves such as density and gamma, and geological information, and selecting curves with better correlation for multi-dimensional information statistical regression. Empirical formulas include the Archie formula, etc., which obtain acoustic curve information through the calculation of physical property parameters. The limitation of these methods is that too much emphasis is placed on the setting of characteristic curves and formula parameters during the reconstruction process, and the reconstructed curve is overly dependent on the characteristic curve, resulting in distortion of the reconstructed curve. The other is the curve reconstruction technology based on mathematical methods such as wavelet transform and curve forward and inverse correction. The basic principle is that after wavelet transform of various standardized logging data, their detailed characteristics are mapped to their respective tower structures according to the size of the resolution. In principle of information fusion at the same resolution, feature selection of regions is carried out in different frequency bands of the corresponding layer. Usually, the reconstruction is based on the acoustic curve. This method has high requirements for the window length setting. If the window length is small, the signal is too short, resulting in inaccurate frequency analysis. If the window length is large, the time domain is not fine enough, and the time resolution is low, affecting the accuracy of the reconstructed curve. How to establish a high-precision acoustic curve reconstruction method to meet the development geological requirements has become a major problem for logging technicians.

[0004] Therefore, the present invention specifically provides a method for reconstructing acoustic curves of a lateral logging series to solve the deficiencies of the prior art. Summary of the Invention

[0005] Aiming at the deficiencies of the prior art, the present invention provides a method for reconstructing acoustic curves of a lateral logging series, which solves the problems that the existing acoustic curve reconstruction technology overly relies on the setting of characteristic parameters and the reconstructed curve is distorted.

[0006] To achieve the above object, the present invention is realized through the following technical solutions: A method for reconstructing a lateral logging series acoustic wave curve, comprising the following steps:

[0007] S1. By normalizing all curves except the target curve, the well logging curves are standardized to the range of [0, 1];

[0008] S2. According to the correlation analysis between the normalized curve and the target reconstruction curve, select the curve with a higher correlation coefficient as the model curve for the subsequent reconstruction process;

[0009] S3. Based on the selected model curve, establish a feature sample library and train it using a multivariate decision-based graphical group clustering model to achieve the reconstruction of the acoustic wave curve;

[0010] S4. Set the minimum value, maximum value of the model logging facies and the number of optional models, and adjust and optimize the model in real time;

[0011] S5. After generating the prediction curve, process the curve through smoothing filtering, and use the core well analysis data to test and optimize the calibration model and results to ensure that the reconstructed acoustic wave curve conforms to the formation characteristics;

[0012] S6. On the basis of S1-S5, form an acoustic wave curve reconstruction method and software module based on graphical group clustering.

[0013] Preferably, in the step S1, the standard formula for normalizing the curve is:

[0014]

[0015] where b is the normalized curve, a is the model curve, min is the reasonable minimum value of the curve, and max is the reasonable maximum value of the curve.

[0016] Preferably, the model curves are SP curve, RMN curve, RMG curve, R025 curve, R045 curve. The reasonable minimum and maximum values of the SP curve are respectively distributed at the extreme points of large sections of sandstone and the extreme points of large sections of mudstone. The reasonable minimum and maximum values of the RMN curve, RMG curve, R025 curve, and R045 curve are respectively distributed at the extreme points of large sections of mudstone and the extreme points of large sections of sandstone.

[0017] Preferably, in the graphical group clustering model, the optimal number of partitions is determined by calculating the KRI factor value, which is used to calculate the distance between each sample and other samples.

[0018] Preferably, the calculation formula of the KRI factor is KRI(X) = NI(X) * M(X, Y) * D(X, Y), where Let \(m\) represent the adjacent exponent, \(m\) represent the remaining samples except \(X\), and \(\alpha\) be a fixed parameter; \(M(X, Y)\) represents the \(m\)th sample of \(X\); \(D(X, Y)\) represents the distance between \(X\) and \(Y\) samples.

[0019] Preferably, the smoothing filtering operation uses a band - pass filter, wavelet transform, or Kalman filter to reduce noise and improve the accuracy of the reconstructed acoustic wave curve.

[0020] Preferably, the correlation analysis is to obtain the model curve by calculating the Pearson correlation coefficient.

[0021] Preferably, the normalization process includes amplitude normalization and depth normalization of well logging curves.

[0022] The present invention provides a method for reconstructing acoustic wave curves of lateral logging series. It has the following beneficial effects:

[0023] 1. In the early stage of oilfield development, due to possible technical limitations or equipment problems, acoustic wave curves may be missing from well logging data. The method proposed by the present invention can extract relevant features from other well logging curves through graph - based clustering technology, and then reconstruct the missing acoustic wave curves. This effectively solves the problem of being unable to calculate reservoir physical property parameters in early - stage development well logging, providing reliable data support for subsequent reservoir evaluation and development decision - making.

[0024] 2. The method of the present invention takes graph - based clustering as the core, performs clustering analysis on other well logging curves, and reconstructs the acoustic wave curve according to the clustering results. The reconstructed acoustic wave curve can more accurately reflect the acoustic wave propagation characteristics of the formation, thereby improving the reliability of acoustic wave time difference values. This is of great significance for evaluating physical property parameters such as reservoir porosity and permeability, providing an accurate reference basis for reservoir development.

[0025] 3. The method of the present invention verifies the accuracy and reliability of the reconstructed acoustic wave curve through the inspection of sealed coring data. Sealed coring data is usually the reference standard for geological engineers to evaluate reservoir physical property parameters. By comparing and analyzing with sealed coring data, the effectiveness of the acoustic wave curve reconstruction method based on graph - based clustering proposed by the present invention can be verified, and further improve the evaluation accuracy of reservoir physical property parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 is the flow schematic diagram of the present invention;

[0027] Figure 2 is the normalization schematic diagram of the SP curve, RMN curve, RMG curve, R025 curve, and R045 curve of the present invention;

[0028] Figure 3The graphical group clustering model of the present invention determines the optimal partition cluster number tree diagram through the KRI method;

[0029] Figure 4 It is a comprehensive comparison diagram of the acoustic travel time in Well X of the present invention, the reconstructed acoustic travel time, and the porosity analyzed with the core. Detailed implementation manners

[0030] Next, in combination with the accompanying drawings of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts shall fall within the protection scope of the present invention.

[0031] Embodiment:

[0032] Please refer to the attached Figure 1 - attached Figure 4 The embodiments of the present invention provide a method for reconstructing a lateral logging series acoustic wave curve, including the following steps:

[0033] S1. By normalizing all curves except the target curve, the well logging curves are standardized to the range of [0, 1]. Through the normalization process, the numerical ranges of different well logging curves can be mapped to the same range, which is convenient for subsequent correlation analysis and model training;

[0034] S2. According to the correlation analysis between the normalized curve and the target reconstruction curve, select the curve with a higher correlation coefficient as the model curve for the subsequent reconstruction process. Selecting the curve with a higher correlation coefficient can improve the accuracy of the reconstruction result;

[0035] S3. Based on the selected model curve, establish a feature sample library and train it using a multivariate decision-based graphical group clustering model to achieve the reconstruction of the acoustic wave curve. By establishing a feature sample library and training using a graphical group clustering model, the acoustic wave curve patterns corresponding to different formation features can be learned. During the reconstruction process, the trained model is used for prediction to obtain the reconstructed acoustic wave curve;

[0036] S4. Set the minimum value, maximum value, and optional model quantity of the model logging facies, and adjust and optimize the model in real time. Setting the minimum and maximum values of the model logging facies can limit the range of the model to ensure that the generated acoustic wave curve conforms to the actual situation. According to needs, the quantity of the optional model can also be set to control the diversity of the reconstruction result. During the reconstruction process, the model can be adjusted and optimized according to the real-time situation to improve the reconstruction effect;

[0037] S5. After generating the prediction curve, the curve is processed through smoothing filtering, and the calibration model and results are verified and optimized using coring well analysis data to ensure that the reconstructed acoustic curve conforms to the formation characteristics. Smoothing filtering can remove noise and mutation points, making the reconstructed curve smoother and more continuous. By comparing and calibrating with the coring well analysis data, the accuracy of the reconstruction model can be verified, and the reconstruction results can be further optimized to better conform to the actual formation characteristics;

[0038] S6. Based on S1 - S5, an acoustic curve reconstruction method and software module based on graphic group clustering are formed, which can provide an efficient and accurate acoustic curve reconstruction solution.

[0039] This method uses techniques such as normalization processing, correlation analysis, and graphic group clustering training. It can reconstruct acoustic curves that conform to the formation characteristics based on the existing logging curve data. Through this method, the understanding of the formation structure and lithology in oil and gas exploration and development can be improved, providing useful support for the development and management of oil and gas resources.

[0040] In step S1, the standard formula for normalizing the curve is:

[0041]

[0042] where b is the normalized coring, a is the model curve, min is the reasonable minimum value of the curve, and max is the reasonable maximum value of the curve.

[0043] The model curves are SP curve, RMN curve, RMG curve, R025 curve, and R045 curve.

[0044] In the graphic group clustering model, the optimal number of partitions is determined by calculating the KRI factor value, which is used to calculate the distance between each sample and other samples. By using the calculated KRI factor value to determine the optimal number of partitions, the problem of subjectively selecting the number of partitions can be avoided, improving the reliability and consistency of the graphic group clustering model. The calculation formula of the KRI factor is KRI(X) = NI(X) * M(X, Y) * D(X, Y), where represents the adjacent index, m represents the remaining samples except X, α is a fixed parameter; M(X, Y) is the m-th sample of X; D(X, Y) is the distance between samples X and Y.

[0045] The smoothing filtering work adopts a band - pass filter, wavelet transform, or Kalman filter to reduce noise and improve the accuracy of the reconstructed acoustic curve.

[0046] Band - pass filter: A band - pass filter can select signals within a specific frequency range for filtering and filter out noise in other frequency ranges. In the smoothing filtering operation, the band - pass filter can select an appropriate frequency range to remove noise signals unrelated to the acoustic wave curve while retaining the frequency components related to formation characteristics, thereby reducing noise and improving the accuracy of the reconstructed curve.

[0047] Wavelet transform: Wavelet transform is a time - frequency analysis method that can decompose a signal into sub - bands of different frequencies and filter and reconstruct each sub - band. In the smoothing filtering operation, wavelet transform can be used to decompose the acoustic wave curve into sub - bands of different frequencies, and then each sub - band is filtered by selecting an appropriate threshold to remove noise signals. Finally, the filtered sub - bands are reconstructed to obtain a smoothed acoustic wave curve.

[0048] Kalman filter: The Kalman filter is a recursive filtering algorithm that can perform state estimation and filtering based on the system model and observation data. In the smoothing filtering operation, the Kalman filter can be used to filter the acoustic wave curve. By balancing the prior information and observation data, the influence of noise is reduced and the accuracy of the reconstructed curve is improved. The Kalman filter usually requires defining the system model and measurement model and adjusting parameters according to the actual situation.

[0049] Correlation analysis is to obtain the model curve by calculating the Pearson correlation coefficient.

[0050] The Pearson correlation coefficient is a commonly used statistic for measuring the strength and direction of the linear relationship between two variables. Its value ranges from - 1 to 1, where - 1 indicates a perfect negative correlation, 0 indicates no correlation, and 1 indicates a perfect positive correlation.

[0051] In correlation analysis, the model curves can be regarded as one variable, and then the Pearson correlation coefficient between them is calculated. The specific steps are as follows:

[0052] a. Collect the model curve data to be subjected to correlation analysis;

[0053] b. For each pair of model curves, calculate their Pearson correlation coefficient. The calculation of the Pearson correlation coefficient involves the values of each model curve at the same position;

[0054] c. According to the value of the Pearson correlation coefficient, the correlation between the model curves can be judged. If the correlation coefficient is close to 1, it means that the two model curves have a strong positive correlation; if the correlation coefficient is close to - 1, it means that the two model curves have a strong negative correlation; if the correlation coefficient is close to 0, it means that there is no obvious linear relationship between the two model curves.

[0055] By calculating the Pearson correlation coefficient, it can help analyze the correlation between model curves. This is necessary for understanding the associations between models, discovering the relationships between variables, and evaluating the accuracy and reliability of models.

[0056] Normalization processing includes amplitude normalization and depth normalization of well logging curves.

[0057] Amplitude normalization: In well logging curves, different curves may have different amplitude ranges, which can affect subsequent data processing and analysis. To eliminate amplitude differences, amplitude normalization can be performed on well logging curves. Common amplitude normalization methods include maximum-minimum normalization, standardization, and mean normalization, etc. These methods can scale the amplitude values of well logging curves to a specific range, such as [0, 1] or [-1, 1], for better comparison and analysis of the amplitude variations between different curves.

[0058] Depth normalization: Well logging curves usually use depth as the abscissa, and different wells may have different well depth ranges. To unify the well depth ranges of different wells, depth normalization can be performed on well logging curves. A common method of depth normalization is to map the well depth to the range of [0, 1] according to a certain ratio, or convert the well depth to relative depth (such as percentage).

[0059] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for reconstructing acoustic wave curves of a lateral logging series, characterized in that, it includes the following steps: S1. By normalizing all curves except the target curve, the well logging curves are standardized to the range of [0, 1]; S2. According to the correlation analysis between the normalized curves and the target reconstructed curve, select the curve with a higher correlation coefficient as the model curve for the subsequent reconstruction process; S3. Based on the selected model curve, establish a characteristic sample library and train it using a multivariate decision-based graphical group clustering model to achieve the reconstruction of acoustic wave curves; S4. Set the minimum value, maximum value of the model logging facies and the number of optional models, and adjust and optimize the model in real time; S5. After generating the prediction curve, process the curve through smoothing filtering, and use the core well analysis data to test and optimize the calibration model and results to ensure that the reconstructed acoustic wave curve conforms to the formation characteristics; S6. Based on S1 - S5, form an acoustic wave curve reconstruction method and software module based on graphical group clustering.

2. A method for reconstructing acoustic wave curves of a lateral logging series according to claim 1, characterized in that, in the step S1, the standard formula for normalizing the curve is: where b is the normalized curve, a is the model curve, min is the reasonable minimum value of the curve, and max is the reasonable maximum value of the curve.

3. A method for reconstructing acoustic wave curves of a lateral logging series according to claim 1, characterized in that, the model curve is SP curve, RMN curve, RMG curve, R025 curve, R045 curve.

4. A method for reconstructing acoustic wave curves of a lateral logging series according to claim 1, characterized in that, in the graphical group clustering model, the optimal number of partitions is determined by calculating the KRI factor value, which is used to calculate the distance between each sample and other samples.

5. A method for reconstructing acoustic wave curves of a lateral logging series according to claim 4, characterized in that, The KRI factor calculation formula is KRI(X) = NI(X) * M(X,Y) * D(X,Y), where represents the adjacent index, m represents the remaining samples except X, α is a fixed parameter; M(X,Y) is the m-th sample of X; D(X,Y) is the distance between the X and Y samples.

6. A method for reconstructing acoustic wave curves of a lateral logging series according to claim 1, characterized in that, the smoothing filtering work adopts a band-pass filter, wavelet transform or Kalman filter to reduce noise and improve the accuracy of the reconstructed acoustic wave curve.

7. A method for reconstructing acoustic wave curves of a lateral logging series according to claim 1, characterized in that, the correlation analysis obtains the model curve by calculating the Pearson correlation coefficient.

8. A method for reconstructing acoustic wave curves of a lateral logging series according to claim 1, characterized in that, the normalization process includes amplitude normalization and depth normalization of the well logging curves.

Images