A braided river reservoir configuration characterization method based on natural gamma inversion

Abstract

The present application relates to the technical field of oilfield development geological reservoir identification, and particularly relates to a braided river reservoir configuration characterization method based on natural gamma inversion. The present application obtains and processes well-seismic data of a study area, extracts and optimizes seismic attributes beside a well, completes well-seismic data alignment, normalization and resampling, constructs well-seismic joint training samples, and trains a natural gamma curve prediction model; the model is applied to a three-dimensional seismic attribute volume, a natural gamma inversion volume is obtained, and reservoir configuration interpretation is carried out in combination with a braided river reservoir configuration mode. The present application can improve the accuracy of identification of internal configuration units of braided river thick sand bodies under the condition of few wells, and solve the problem that internal configuration units of thick sand bodies between wells are difficult to depict.

Description

Technical Field

[0001] This invention relates to the field of geological reservoir identification technology for oilfield development, and in particular to a method for characterizing braided river reservoir configuration based on natural gamma inversion. Background Technology

[0002] Braided rivers, due to frequent channel migration and the vertical superposition of multiple channel phases, often form thick, banded composite sand bodies. The internal structure of active channels and mid-channel sand bodies is exceptionally complex, making the detailed characterization of braided river reservoir configurations based on a single gene far more challenging than that of other sedimentary reservoirs such as meandering rivers. For a long time, detailed characterization of braided river reservoir configurations has been conducted through methods such as field surveys of typical outcrops, modern sedimentary observations, flume simulation experiments, sedimentary numerical simulations, and detailed well network analysis, yielding numerous research results. However, these methods are difficult to apply to various offshore and onshore oil and gas fields with few or low-well-control development. Therefore, fully utilizing seismic data to achieve detailed characterization of braided river reservoir configurations has become the most realistic and crucial technical approach.

[0003] In recent years, the application of single seismic attribute analysis and seismic inversion techniques for the fine characterization of braided river sand bodies and the identification of interlayers has achieved some results, but it is still difficult to achieve a fine depiction of reservoir configuration units at the single sand body level within the reservoir area. Machine learning methods such as CatBoost, XGBoost, and random forest regression have also made some attempts in reservoir prediction by establishing a mapping relationship between seismic attributes and well logging curves. However, under conditions of few wells, the training sample is limited, the model is prone to overfitting, and it is difficult to fully maintain the continuous variation trend, peak and valley positions, local morphology, and statistical distribution characteristics of the natural gamma curve. At the same time, these methods cannot make full use of the response differences of seismic frequency-division attributes at different frequency band scales, and they also lack a sequence prediction mechanism that takes into account both amplitude fitting and curve trend constraints. Therefore, the predicted results still have a low degree of agreement with the geological characteristics of braided river reservoirs.

[0004] Therefore, in order to address the technical challenge of insufficient accuracy in identifying the internal configuration of thick sand bodies under conditions of few wells, this invention proposes a braided river reservoir configuration characterization method based on natural gamma inversion, which improves the reliability of identifying the internal configuration units of thick sand bodies formed in sedimentary environments such as braided rivers under conditions of few wells. Summary of the Invention

[0005] This invention provides a braided river reservoir configuration characterization method based on natural gamma inversion, which is used to solve the problem that the complex multi-stage superposition, lateral migration and vertical cutting of braided river reservoir configuration units under conditions with few wells makes it difficult for traditional seismic attribute analysis, conventional inversion and traditional machine learning to accurately identify the internal structure of thick sand bodies.

[0006] To achieve the above, the technical solution of the present invention includes the following steps:

[0007] S1. Obtain natural gamma logging curves and frequency-division seismic attribute data for the study area;

[0008] S2. Perform time-depth conversion on the natural gamma logging curves, and perform well-seismic calibration, data cleaning, and outlier processing;

[0009] S3. Extract and select attributes from the seismic attribute data along the seismic traces beside the well to obtain the seismic attributes corresponding to the natural gamma logging curves.

[0010] Seismic attribute selection is based on the analysis of the response and correlation characteristics between the natural gamma logging curve and the candidate seismic attributes. Each candidate seismic attribute sequence is aligned with the natural gamma logging curve under a unified time sampling grid. The similarity between the shape of the natural gamma logging curve and the well-side seismic attribute curve is identified. The Pearson correlation coefficient is used to calculate the degree of correlation between the natural gamma logging curve and the candidate seismic attributes, thereby selecting the seismic attribute that reflects the natural gamma trend.

[0011] S4. Align the natural gamma logging curves processed in S2 with the seismic attributes obtained in S3, perform normalization and resampling, and ensure that the two are consistent under the same time sampling grid. Construct a well-seismic joint training sample consisting of seismic attribute sequences and natural gamma curve sequences in the manner of independent sliding window for each well.

[0012] S5. Using the seismic attribute sequence as input and the natural gamma logging curve sequence as output, a natural gamma curve prediction model is trained based on the joint well-seismic training samples; the specific steps are as follows:

[0013] (1) Based on residual dilated temporal convolutional network, including input feature mapping layer, multiple dilated one-dimensional convolutional residual modules and output regression layer; input feature mapping layer maps seismic attribute sequence to high-dimensional temporal feature space; residual module uses one-dimensional convolution with different dilation rate to extract local response features at different time scales, and stably transmits shallow and deep features through residual connection; output regression layer maps temporal features to natural gamma logging curve prediction values ​​at each time sampling point;

[0014] (2) For the natural gamma curve prediction model, a weighted trend composite loss function of natural gamma response value, consisting of natural gamma numerical regression loss and first-order difference trend loss, is adopted to constrain the magnitude error of the predicted value and the changing trend of adjacent sampling points, respectively. A weighting coefficient of natural gamma response value is introduced to improve the prediction accuracy of the natural gamma response segment. The formula is as follows:

[0015]

[0016]

[0017]

[0018]

[0019] Among them, L GR L represents the weighted trend composite loss function of the natural gamma response value; v L represents the natural gamma numerical regression loss; d This represents the first-order difference trend loss; y represents the model-predicted natural gamma value; y represents the true natural gamma value; ŷ t and y t Represent the predicted and actual values ​​corresponding to the t-th time sampling point, respectively; w represents the weighting coefficient of the natural gamma response value; ŷ t+1 and y t+1 denoted as the predicted value and the actual value corresponding to the (t+1)th time sampling point, respectively; η represents the response value weight adjustment coefficient; α and β represent the weight coefficients of the natural gamma numerical regression loss and the first difference trend loss, respectively; mean represents the average error over all sampling points or adjacent sampling points;

[0020] (3) During the model training process, the parameters of the residual dilation time convolutional network are updated using training well samples, and the inter-well generalization ability of the model is evaluated using test well samples, so that the natural gamma curve prediction results are consistent with the real well logging curves in terms of amplitude, peak and valley positions and continuous change trends.

[0021] S6. The natural gamma curve prediction model obtained in S5 is used to obtain the 3D seismic attribute volume of the study area to obtain the seismic inversion natural gamma volume. The reservoir configuration unit is then interpreted by combining the braided river reservoir configuration model with the well point configuration unit logging response chart. Specifically, the 3D seismic attribute volume of the entire area used for prediction is read according to seismic traces, and multi-attribute time-series features are constructed according to the normalization method used in the training phase. Along the time sampling direction of each seismic trace, feature sequences are extracted using a sliding window and input into the trained natural gamma curve intelligent prediction model to obtain the natural gamma prediction corresponding to each window. The results were analyzed; for the same sampling point in the overlapping area of ​​adjacent windows, a continuous natural gamma curve was generated by averaging multiple predicted values; finally, the prediction results were denormalized, and the natural gamma inversion body of the whole area was written according to the spatial position of the original 3D seismic body; based on the geological model, configuration unit and configuration style of the braided river reservoir configuration in the study area, the configuration units of the core bar, active channel, abandoned channel and floodplain reservoir were identified, and the well point configuration unit logging interpretation results were used for verification, and then the inter-well configuration units and the contact style of configuration units were characterized.

[0022] Compared with existing technologies, the beneficial effects of this invention are as follows: First, it uses residual dilation time convolutional networks to establish a natural gamma curve sequence prediction model and improves prediction accuracy and curve continuity through high-value weighted trend composite loss; Second, it constructs a three-dimensional data volume of natural gamma inversion, which, together with the natural gamma features of braided river reservoir configuration units, interprets reservoir configuration units and reservoir structure patterns, solving the problem of difficulty in characterizing the internal configuration units of thick sand bodies between wells. Attached Figure Description

[0023] Figure 1 This is a schematic flowchart of a braided river reservoir configuration characterization method based on natural gamma inversion applicable to conditions with few wells, according to the present invention.

[0024] Figure 2 This is a schematic diagram of the natural gamma curve in the time domain according to an embodiment of the present invention;

[0025] Figure 3 This is a heatmap of the Pearson correlation coefficients between different seismic attribute volumes and natural gamma curves in an embodiment of the present invention.

[0026] Figure 4 This is a schematic diagram of the natural gamma curve prediction model according to an embodiment of the present invention;

[0027] Figure 5 This is a schematic diagram of the natural gamma curve prediction results according to an embodiment of the present invention;

[0028] Figure 6 This is a schematic diagram of the natural gamma response of a braided river reservoir configuration unit according to an embodiment of the present invention;

[0029] Figure 7 This is a schematic diagram of the braided river reservoir configuration unit combination style according to an embodiment of the present invention;

[0030] Figure 8 This is a schematic diagram illustrating the reservoir configuration of the natural gamma prediction body according to an embodiment of the present invention. Detailed Implementation

[0031] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0032] Taking the X oilfield in the Bohai Sea area of ​​the study region as an example, the scheme of this application will be described in conjunction with the attached figures. Figure 1 As shown, the specific implementation steps of this invention are as follows:

[0033] S1. Obtain natural gamma-ray logging curves and frequency-division seismic attribute data for the study area. Natural gamma-ray curves characterize the continuous variations in different lithologies and clay content through curve morphology and combination features. Compared with other logging curves, they are more advantageous in identifying braided river reservoir configuration units such as core bars, active channels, abandoned channels, and floodplains.

[0034] Natural gamma curve data and various seismic attribute data were obtained from 10 wells in the study area. Specifically, the seismic attributes included: amplitude envelope, 90-Hz amplitude envelope, instantaneous frequency, 90-Hz instantaneous frequency, and root mean square amplitude.

[0035] S2. Using the wellbore time-depth relationship, the depth-domain natural gamma curve obtained in S1 is transformed to the seismic time domain. Figure 2 As shown in the figure, the correspondence between well logging response and seismic reflection phase axis is established through well-seismic calibration, thereby ensuring the consistency between natural gamma curve and seismic data in vertical position and geological strata. At the same time, the impact of noise and acquisition error on curve characteristics is reduced through data cleaning and outlier processing. Specifically, firstly, a depth-time conversion relationship is established based on time-depth calibration data, and the natural gamma curve is resampled from the depth domain to the seismic time domain; then, a synthetic seismic record is produced by combining seismic wavelets, and well-seismic calibration is completed through stratum alignment, time shift correction, etc.; finally, missing values, abrupt values, and outliers exceeding the reasonable range in the natural gamma curve are identified and corrected.

[0036] S3. Import the seismic attributes obtained in S1 into the time-domain grid in Petrel, aligning it with the natural gamma ray logging curve. Then, extract the well-side seismic attribute data in the time domain and optimize the seismic attribute selection based on the response and correlation characteristics between the natural gamma ray logging curve and the candidate seismic attributes. Figure 3 As shown. Specifically, the correlation between the natural gamma logging curve and each candidate seismic attribute is calculated using the Pearson correlation coefficient, as shown in the formula:

[0037]

[0038] Where, ρ G,Aj Represents the relationship between the natural gamma logging curve G and the j-th candidate seismic attribute A. j The Pearson correlation coefficient between them, g i Let represent the natural gamma value at the i-th time sampling point, and n be the total number of time sampling points participating in the correlation analysis, i.e., the sample size; a ijThis represents the j-th candidate seismic attribute value corresponding to the i-th time sampling point. and ā j Let G and A represent respectively. j The mean.

[0039] By combining the similarity of the natural gamma logging curves and the seismic attribute curves in the well-side channels, the sensitivity evaluation and redundancy identification of candidate seismic attributes are carried out. Seismic attributes with strong correlation to the response of the natural gamma logging curves and low mutual redundancy are retained to obtain the preferred seismic attribute set for subsequent well-seismic joint sample construction. Finally, the following seismic attributes are selected: first derivative (Seismic Attribute D1), second derivative (Seismic Attribute D2), 90 Hz original seismic data (90-Hz Original Seismic), 90 Hz seismic amplitude volume (90-Hz Seismic Amplitude Volume), seismic amplitude volume (Seismic Amplitude Volume), and 90 Hz relative acoustic impedance (90-Hz Relative Acoustic Impedance).

[0040] S4. Align the natural gamma logging curves and optimized seismic attributes obtained from S2 and S3, perform normalization and resampling, and ensure consistency between the two under a unified time sampling grid. Construct a well-seismic joint training sample consisting of seismic attribute sequences and natural gamma curve sequences using a single-well independent sliding window method, specifically including:

[0041] ① Linear interpolation resampling was performed on the natural gamma logging curves and seismic attribute sequences of each well at a TWT sampling interval of 0.25 ms. Training and test wells were distinguished by well name. Test wells were not involved in model training or normalized parameter fitting. Normalized parameters were determined only from training well data. To reduce the impact of outliers and noise on data scaling and model training, improve training stability and robustness, and preserve the positive and negative variation characteristics of seismic attributes, the natural gamma logging curves were normalized to the 0-1 interval according to the hundredth and ninety-ninth percentiles of the training wells, and the seismic attributes were normalized to the -1 to 1 interval according to the hundredth and ninety-ninth percentiles of the training wells. This was done to avoid information leakage from test wells and reduce the impact of outliers on model training.

[0042] ② The data obtained in step ① is segmented into well-seismic joint sample pairs by using a single well as the unit. The input data is a multidimensional seismic attribute sequence within each time window, and the output label is a natural gamma curve sequence aligned with its time. The segmentation process is performed independently within each well, without cross-well splicing, to avoid misinterpreting discontinuous sequences from different wells as the same continuous curve. The preferred sequence length of the sliding window is 16 time sampling points, and the preferred time sampling interval is 0.25 ms, but the invention is not limited to these.

[0043] S5. Construct and train a natural gamma curve prediction model. In this embodiment, the prediction model uses a residual temporal convolutional network. The purpose is to effectively extract the longitudinal variation features of well logging curves and seismic attributes at different scales through dilated convolution, and to capture long-range sequence dependencies. At the same time, residual connections are used to alleviate the gradient vanishing and performance degradation problems in deep network training, retain the original feature information, and improve the training stability of the model. Figure 4 As shown, the seismic attribute sequence within a sliding window is used as input, and the natural gamma curve sequence within the same sliding window is used as the output target. This is used to learn the nonlinear mapping relationship between the seismic attribute sequence and the natural gamma curve sequence, specifically including:

[0044] ① In this embodiment, the sequence window length for both input and output is set to 16, the input feature dimension is 7, including relative time features and filtered seismic attribute features, and the output is a normalized natural gamma curve sequence. The natural gamma curve prediction model includes an input feature mapping layer, three dilated one-dimensional convolutional residual modules, and an output regression layer. The input mapping layer uses a one-dimensional convolution with a kernel size of 1 to map the 7-dimensional input features to 64-dimensional temporal features. The three dilated one-dimensional convolutional residual modules each have 64 channels, a kernel size of 5, and dilation rates of 1, 2, and 4 respectively, with corresponding padding sizes of 2, 4, and 8 respectively. Each residual temporal convolutional module internally includes a one-dimensional dilated convolution, a LeakyReLU activation function, and a Dropout layer. The module input and convolution output are summed to maintain stable transfer of shallow features to deep features. The output regression layer uses a one-dimensional convolution with a kernel size of 1 to map the 64-dimensional temporal features to a 1-dimensional natural gamma prediction value.

[0045] ② For the prediction model established in step ①, a high-value weighted trend composite loss function is used to constrain the model's prediction results. The high-value weighted trend composite loss function includes natural gamma numerical regression loss and first-order difference trend loss. The natural gamma numerical regression loss is used to constrain the magnitude error between the predicted and true natural gamma values, while the first-order difference trend loss is used to constrain the consistency of the trend between the predicted and true curves at adjacent time sampling points. To improve the model's fitting ability to the high natural gamma response segment, a high-value segment weighting coefficient is introduced into the natural gamma numerical regression loss, making the model pay more attention to the prediction accuracy of the high natural gamma segment during training. Specifically, the coefficient α of the natural gamma numerical regression loss is set to 1, the coefficient β of the first-order difference trend loss is set to 0.5, and the high-value segment weighting coefficient η is set to 2.5, as shown in the following formula:

[0046]

[0047]

[0048]

[0049]

[0050] Among them, L GR L represents the weighted trend composite loss function of the natural gamma response value; v L represents the natural gamma numerical regression loss; d This represents the first-order difference trend loss; y represents the model-predicted natural gamma value; y represents the true natural gamma value; ŷ t and y t Let ŷ represent the predicted value and the actual value corresponding to the t-th time sampling point, respectively; t+1 and y t+1 denoted as the predicted value and the actual value corresponding to the (t+1)th time sampling point, respectively; w represents the weighting coefficient of the natural gamma response value; η represents the response value weight adjustment coefficient; α and β represent the weighting coefficients of the natural gamma numerical regression loss and the first-order difference trend loss, respectively; mean represents the average error over all sampling points or adjacent sampling points.

[0051] ③ During model training, the well-seismic joint training samples constructed in step S4 were used as the training set. During training, the network parameters were updated using training well samples, and the model's inter-well generalization ability was evaluated using test well samples. The predicted curve was made consistent with the overall trend of the real natural gamma curve, showing good synchronicity in the rapid initial drop, the local sudden increase in the middle, and the subsequent fluctuation range. This indicates that the model can effectively capture the overall change pattern and main response characteristics of the natural gamma curve, with an accuracy of 71.6%. The results are as follows: Figure 5 As shown.

[0052] S6. The prediction model obtained in S5 is used to obtain the three-dimensional seismic attribute volume of the study area, and the seismic inversion natural gamma body is obtained. This is then combined with the braided river reservoir configuration model to interpret the reservoir configuration units, specifically including:

[0053] ① The three-dimensional seismic attribute volume of the whole region used for prediction is read according to the seismic traces, and multi-attribute temporal features are constructed according to the normalization method of the training stage; along the time sampling direction of each seismic trace, the feature sequence is extracted by sliding window and input into the trained natural gamma curve intelligent prediction model to obtain the natural gamma prediction results corresponding to each window; for the same sampling point in the overlapping area of ​​adjacent windows, the average fusion method of multiple prediction values ​​is used to generate a continuous natural gamma curve; finally, the prediction results are denormalized, and the natural gamma inversion volume of the whole region is written according to the spatial position of the original three-dimensional seismic volume.

[0054] ②For example Figure 6 and Figure 7 As shown, based on the natural gamma response charts of the braided river reservoir configuration geological model, configuration units, and configuration patterns in the study area, configuration units were identified in the natural gamma inversion obtained in step ①. Reservoir configuration units such as core bars, active channels, abandoned channels, and floodplains were identified, and verified against well point configuration unit logging interpretation results. This completed the characterization of inter-well configuration units and their contact patterns. Figure 8 As shown, this solves the problem of the difficulty in accurately depicting the internal structural units of thick sand bodies between wells.

[0055] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the protection scope of the claims and specification of the present invention.

Claims

1. A method for characterizing braided river reservoir configuration based on natural gamma inversion, characterized in that, Includes the following steps: S1. Obtain natural gamma logging curves and frequency-division seismic attribute data for the study area; S2. Perform time-depth conversion on the natural gamma logging curves, and perform well-seismic calibration, data cleaning, and outlier processing; S3. Extract and select attributes from the seismic attribute data along the seismic traces beside the well to obtain the seismic attributes corresponding to the natural gamma logging curves. S4. Align the natural gamma logging curves processed in S2 with the seismic attributes obtained in S3, perform normalization and resampling, and ensure that the two are consistent under the same time sampling grid. Construct a well-seismic joint training sample consisting of seismic attribute sequences and natural gamma curve sequences in the manner of independent sliding window for each well. S5. Using the seismic attribute sequence as input and the natural gamma logging curve sequence as output target, a natural gamma curve prediction model is trained based on the well-seismic joint training samples. The natural gamma curve prediction model employs a weighted trend composite loss function of natural gamma response values, consisting of natural gamma numerical regression loss and first-order difference trend loss. This function constrains the magnitude error of the predicted value and the changing trend of adjacent sampling points, respectively. Furthermore, a weighting coefficient for the natural gamma response value is introduced to improve the prediction accuracy of the natural gamma response segment. The formula is as follows: Among them, L GR L represents the weighted trend composite loss function of the natural gamma response value; v L represents the natural gamma numerical regression loss; d This represents the first-order difference trend loss; y represents the model-predicted natural gamma value; y represents the true natural gamma value; ŷ t and y t Let ŷ represent the predicted value and the actual value corresponding to the t-th time sampling point, respectively; t+1 and y t+1 denoted as the predicted value and the actual value corresponding to the (t+1)th time sampling point, respectively; w represents the weighting coefficient of the natural gamma response value; η represents the response value weight adjustment coefficient; α and β represent the weighting coefficients of the natural gamma numerical regression loss and the first difference trend loss, respectively; mean represents the average error over all sampling points or adjacent sampling points; S6. The natural gamma curve prediction model obtained in S5 is used to obtain the three-dimensional seismic attribute volume of the study area to obtain the seismic inversion natural gamma volume. Combined with the braided river reservoir configuration model and the well point configuration unit logging response chart, the reservoir configuration unit is interpreted.

2. The method for characterizing braided river reservoir configuration based on natural gamma inversion according to claim 1, characterized in that, The attribute selection described in S3 is based on the analysis of the response and correlation characteristics between the natural gamma logging curve and the candidate seismic attributes. Each candidate seismic attribute sequence is aligned with the natural gamma logging curve under a unified time sampling grid. The similarity between the shape of the natural gamma logging curve and the well-side seismic attribute curve is identified. The Pearson correlation coefficient is used to calculate the degree of correlation between the natural gamma logging curve and the candidate seismic attributes, thereby selecting the seismic attributes that reflect the natural gamma trend.

3. The method for characterizing braided river reservoir configuration based on natural gamma inversion according to claim 1, characterized in that, The specific method for training the natural gamma curve prediction model described in S5 is as follows: based on a residual dilated temporal convolutional network, including an input feature mapping layer, multiple dilated one-dimensional convolutional residual modules, and an output regression layer; the input feature mapping layer maps the earthquake attribute sequence to a high-dimensional temporal feature space; the residual modules use one-dimensional convolution with different dilation rates to extract local response features at different time scales, and stably transmit shallow and deep features through residual connections; The output regression layer maps the temporal features to the predicted values ​​of the natural gamma logging curves at each time sampling point.

4. The method for characterizing braided river reservoir configuration based on natural gamma inversion according to claim 3, characterized in that, During the training process of the natural gamma curve prediction model, the parameters of the residual dilation time convolutional network are updated using training well samples, and the inter-well generalization ability of the natural gamma curve prediction model is evaluated using test well samples, so that the natural gamma curve prediction results are consistent with the real well logging curves in terms of amplitude, peak and valley positions and continuous variation trends.

5. The method for characterizing braided river reservoir configuration based on natural gamma inversion according to claim 1, characterized in that, The specific approach of S6 is as follows: the three-dimensional seismic attribute volume of the entire region used for prediction is read according to seismic traces, and multi-attribute temporal features are constructed according to the normalization method of the training phase; along the time sampling direction of each seismic trace, the feature sequence is extracted by sliding window and input into the trained natural gamma curve prediction model to obtain the natural gamma prediction results corresponding to each window; for the same sampling point in the overlapping area of ​​adjacent windows, the average fusion method of multiple prediction values ​​is used to generate a continuous natural gamma curve; finally, the prediction results are denormalized, and the natural gamma inversion volume of the entire region is written according to the spatial position of the original three-dimensional seismic volume; based on the geological model of braided river reservoir configuration, configuration units and configuration patterns of the study area, the configuration units of the core bar, active channel, abandoned channel and floodplain reservoir are identified, and the well point configuration unit logging interpretation results are used for verification, and the inter-well configuration units and configuration unit contact patterns are characterized.

Images