Method for improving seismic attribute description reservoir precision

Abstract

A method for improving seismic attribute description reservoir precision relates to the technical field of oil-gas exploration and oil reservoir evaluation, and comprises the following steps: matching a stratigraphic sequence above a well with a time-frequency analysis result of a seismic trace beside the well, determining a corresponding relationship between the well and a seismic large-scale stratigraphic unit, and establishing a time-depth relationship of a research area by synthesizing seismic records; based on the optimized well-seismic calibration result, multiple seismic attributes are extracted along the target geologic horizon; performing correlation analysis on the reservoir attribute at the well point and the corresponding seismic attribute, and selecting the seismic attribute with the maximum correlation coefficient as the optimal prediction parameter of the reservoir attribute; if the decision coefficient is smaller than a preset value, classifying the reservoir attributes by adopting a support vector machine technology, and re-executing each type of reservoir attributes until the decision coefficients of all classifications are larger than the preset value or the sample size is insufficient and regression cannot be carried out; and utilizing the optimized regression formula and the seismic attribute data to calculate the plane distribution of the reservoir attributes of the well-free area, and generating a reservoir attribute plane graph.

Description

Technical Field

[0001] The present invention focuses on the technical field of oil and gas exploration and reservoir evaluation, aiming to provide an innovative method that can significantly improve the accuracy of seismic attributes in describing reservoirs, providing strong support for the efficient development of this field. Background Art

[0002] In the key links of oil and gas exploration and reservoir evaluation, accurately describing reservoir properties is the core element for determining the distribution range of oil and gas reservoirs, accurately assessing reserves, and formulating scientific exploitation plans. Currently, using seismic attributes to describe reservoirs is a widely used technical means in the industry, but traditional methods have many bottlenecks. On the one hand, during the process of determining the formation correspondence between wells and seismic data, the accuracy is insufficient, which directly leads to a significant reduction in the reliability of subsequent seismic attribute extraction; on the other hand, when screening seismic attributes for predicting reservoir properties, there is a lack of a scientific and systematic screening mechanism, resulting in relatively large errors in prediction results. In addition, for complex reservoirs, the existing classification and targeted attribute prediction methods have poor effects. These problems seriously restrict the application accuracy of seismic attributes in reservoir description, and thus have a negative impact on the efficiency and benefits of oil and gas exploration and development. Summary of the Invention

[0003] Object of the Invention The present invention is committed to overcoming the problems in the prior art such as inaccurate well-seismic correspondence, unreasonable screening of seismic attributes, and poor classification and prediction effects for complex reservoirs, and provides a method that can greatly improve the accuracy of seismic attributes in describing reservoirs. Through this method, the distribution range and thickness of reservoirs can be predicted more accurately, providing stable and reliable data support for oil and gas exploration and reservoir evaluation, and helping the industry develop efficiently.

[0004] Technical Solution A method for improving the accuracy of reservoir description using seismic attributes, comprising the following steps: Step (1): Match the stratigraphic sequence in the well with the time-frequency analysis results of the seismic trace beside the well to determine the corresponding relationship between the well and the large-scale seismic stratigraphic unit, and establish the time-depth relationship of the study area through synthetic seismic records; Step (2): Based on the optimized well-seismic calibration results, extract various seismic attributes along the target geological horizon, including the inverted impedance data. Inversion is the core technology connecting seismic response and underground physical parameters, and its essence is to solve the "black box" problem through mathematical optimization. In practical applications, it is necessary to comprehensively consider geological prior knowledge, data quality and algorithm adaptability to select the inversion strategy to achieve reliable prediction of reservoir parameters; Step (3): Conduct a correlation analysis on the reservoir attributes at the well points and the corresponding seismic attributes, and select the seismic attribute with the largest correlation coefficient as the best prediction parameter for the reservoir attributes; Step (4): If the determination coefficient in Step (3) is less than 0.7, use the support vector machine technology to classify the reservoir attributes, and re-execute Step (3) for each class of reservoir attributes until the determination coefficients of all classifications are greater than 0.7 or the sample size is insufficient for regression; Step (5): Use the optimized regression formula and seismic attribute data to calculate the planar distribution of reservoir attributes in the well-free area and generate a reservoir attribute plan view.

[0005] Preferably, in Step (1), the time-frequency analysis includes the matching analysis of the stratigraphic sequence and the seismic trace time series, and corrects the mapping accuracy of the time-depth relationship through synthetic seismic records.

[0006] Preferably, in Step (3), the reservoir attributes include reservoir thickness, and the seismic attributes include at least one of amplitude, frequency, phase and impedance.

[0007] Preferably, in Step (4), the support vector machine technology is used to classify reservoir sand bodies, and the classification basis includes thickness distribution characteristics and seismic response patterns.

[0008] Preferably, in Step (5), the regression formula is a multiple linear regression model, and its input parameters are the optimized seismic attribute combinations.

[0009] Preferably, the method is used for quantitative prediction of the distribution range and thickness of reservoirs in the oil and gas exploration and reservoir evaluation stages.

[0010] Preferably, the determination coefficient is the square of the correlation coefficient.

[0011] Step (1) Exact Matching and Relationship Establishment: With the help of the stratigraphic sequence in the well, depth-time frequency analysis is carried out on the seismic traces beside the well. In this process, the variation characteristics of different frequency components of the seismic traces over time are studied in depth to achieve refined matching between the stratigraphic sequence and the time sequence of the seismic traces, thereby accurately determining the corresponding relationship between the well and the large-scale seismic stratigraphic units. Subsequently, the time-depth relationship of the study area is constructed through synthetic seismic records. Synthetic seismic records are obtained by convolving well logging data with seismic wavelets. By carefully comparing with the actual seismic traces and continuously adjusting relevant parameters, the mapping accuracy of the time-depth relationship is continuously optimized to ensure the accuracy of the basic data.

[0012] Step (2) Multi-attribute Extraction: Based on the optimized well-seismic calibration results, a professional seismic interpretation software is used to comprehensively extract various seismic attributes along the target geological horizon. These attributes not only cover the wave impedance data obtained through seismic inversion technology but also include key attributes such as amplitude, frequency, and phase. Wave impedance data can intuitively reflect the lithological and physical property changes of underground strata, providing an essential basis for subsequent reservoir analysis.

[0013] Step (3) Correlation Analysis and Parameter Selection: In-depth correlation analysis is carried out on reservoir attributes at well points, such as reservoir thickness, and corresponding seismic attributes, including at least one of amplitude, frequency, phase, and wave impedance. Using scientific statistical analysis methods, the correlation coefficient between the two is accurately calculated, and the seismic attribute with the largest correlation coefficient is selected as the best prediction parameter for reservoir attributes, thereby significantly improving the accuracy of reservoir attribute prediction.

[0014] Step (4) Complex Reservoir Classification and Optimization: If the coefficient of determination (the coefficient of determination is the square of the correlation coefficient) in Step (3) is less than 0.7, advanced support vector machine technology is used to classify reservoir attributes. This technology mainly targets reservoir sand bodies and classifies them according to thickness distribution characteristics and seismic response patterns. There are obvious differences in the thickness distribution and seismic section response characteristics of different types of reservoir sand bodies, and support vector machines can accurately identify these differences and classify the reservoirs into different categories. Subsequently, the relevant analysis in Step (3) is re-executed for each class of reservoir attributes until the coefficient of determination of all classifications is greater than 0.7, or regression analysis cannot be performed due to insufficient sample size.

[0015] Step (5) Planar Distribution Prediction and Mapping: Using the optimized regression formula, seismic attribute data is input into a multiple linear regression model. This model takes the optimized combination of seismic attributes as input parameters and realizes accurate prediction of reservoir attributes in well-free areas by establishing an accurate mathematical relationship between seismic attribute data and reservoir attributes. Based on the prediction results, a planar map of reservoir attributes is drawn to visually display the distribution range and thickness of the reservoir.

[0016] Beneficial Effects Precise guarantee of basic data: By establishing precise well-seismic matching and time-depth relationships, the accuracy of seismic attribute extraction has been significantly improved, laying a solid foundation for subsequent reservoir analysis.

[0017] Remarkable improvement in prediction accuracy: Scientific seismic attribute screening methods ensure that the selected attributes are highly correlated with reservoir attributes, effectively improving the prediction accuracy of reservoir attributes.

[0018] Enhanced adaptability to complex reservoirs: Through the classification and processing of complex reservoirs, more accurate predictions can be obtained for different types of reservoirs, significantly improving the adaptability to complex geological conditions.

[0019] Intuitive and reliable decision-making basis: The generated reservoir attribute plane maps can visually display the distribution range and thickness of the reservoir, providing an intuitive and reliable decision-making basis for oil and gas exploration and reservoir evaluation, and effectively promoting the improvement of the efficiency and success rate of oil and gas exploration and development. Description of the Drawings

[0020] Figure 1 It is a flow chart of the method of the present invention. Detailed Implementation Manner

[0021] Implementation Preparation Comprehensively collect basic data such as logging data and seismic data in the study area, and use professional data processing tools to preprocess the data to ensure the accuracy, integrity, and consistency of the data, laying a solid data foundation for subsequent analysis.

[0022] Example 1: Conventional sandstone reservoir 1. Execute step (1): Select the well-known time-frequency analysis software Geopsy to conduct time-frequency analysis on the seismic traces beside the wells in Block A of a certain basin. Utilize the powerful signal processing function of the software to obtain high-resolution frequency-time information. Carefully compare the detailed stratigraphic sequence on the well with the analysis results one by one, and use standard stratigraphic correlation techniques to successfully determine their accurate corresponding relationship. Then, synthesize seismic records using logging data, and finally optimize to obtain a high-precision time-depth relationship by adjusting parameters such as the frequency and phase of the seismic wavelet multiple times.

[0023] 2. Execute step (2): Based on the optimized well-seismic calibration results, use the commonly used seismic interpretation software Landmark in the industry to accurately extract various seismic attributes including impedance data along the target geological horizon. Utilize the built-in data quality control module of the software to strictly evaluate and screen the extracted attributes, and eliminate abnormal data to ensure data reliability.

[0024] 3. Execute step (3): The system organizes the reservoir thickness and other attribute data at the well point, associates and integrates them with the corresponding seismic attribute data, and constructs a complete data set. The Pearson correlation coefficient analysis method is used to accurately calculate the correlation coefficient between the two, and finally determine that the wave impedance is the best prediction parameter for reservoir thickness.

[0025] 4. Execute step (4): After calculation, the determination coefficient is greater than 0.7, so there is no need to perform support vector machine classification and go directly to the next step.

[0026] 5. Execute step (5): Input the seismic attribute data into the multivariate linear regression model and use the professional data analysis software R to perform model calculations. Based on the calculation results, the drawing software Surfer is used to draw a reservoir attribute plane map to clearly display the reservoir distribution.

[0027] Example 2: Carbonate reservoir 1. Execute step (1): Use the advanced time-frequency analysis software SeisLab to perform deep time-frequency analysis on the wellside seismic traces in Block B of a certain sea area to obtain rich frequency-time details. Through professional stratigraphic comparison methods, carefully match the well strata with the analysis results to determine the corresponding relationship. Use well logging data to synthesize seismic records, repeatedly adjust the convolution operation parameters, and optimize the time-depth relationship.

[0028] 2. Execute step (2): Based on the optimized well-seismic calibration results, use Petrel seismic interpretation software to comprehensively extract various seismic attributes along the target geological layer, including wave impedance, amplitude, frequency, etc. A strict data quality control process is used to ensure data reliability.

[0029] 3. Execute step (3): sort out the reservoir porosity, permeability and other attribute data at the well point, and correlate and integrate them with the corresponding seismic attributes. Use the Spearman rank correlation coefficient analysis method to calculate the correlation coefficient and determine that the amplitude is the best prediction parameter for reservoir porosity.

[0030] 4. Execute step (4): Calculate the determination coefficient to be less than 0.7, use the support vector machine algorithm, with the help of Python's Scikit-learn machine learning library, and classify the reservoirs according to the reservoir thickness distribution characteristics and seismic response patterns. Repeat the correlation analysis of step (3) for each type of reservoir until the determination coefficient is greater than 0.7.

[0031] 5. Execute step (5): Input the optimized seismic attribute data into the multivariate linear regression model and use Python's Statsmodels library to perform model calculations. Based on the calculation results, use professional drawing software GMT to draw a reservoir attribute plane map to intuitively display the reservoir distribution.

[0032] Example 3: Complex fault-block reservoir 1. Execute step (1): Use the efficient time-frequency analysis software ProMAX to perform time-frequency analysis on the seismic traces beside the wells in Block C of a certain oilfield to obtain detailed frequency-time information. Apply professional stratigraphic correlation techniques to precisely match the formation above the well with the analysis results and determine the corresponding relationship. Optimize the time-depth relationship by continuously adjusting the parameters of the synthetic seismic record.

[0033] 2. Execute step (2): According to the optimized well-seismic calibration results, use the Jason seismic interpretation software to extract various seismic attributes along the target geological horizon, including wave impedance, phase, etc. Apply strict data quality control techniques to ensure data quality.

[0034] 3. Execute step (3): Organize the attribute data such as oil saturation and reservoir thickness at the well points, and correlate them with the corresponding seismic attributes. Use the canonical correlation analysis method to calculate the correlation coefficient and determine that the phase is the best prediction parameter for reservoir oil saturation.

[0035] 4. Execute step (4): Since the coefficient of determination is less than 0.7, adopt the support vector machine algorithm. Use the machine learning toolbox of MATLAB to classify the reservoir based on the reservoir thickness distribution characteristics and seismic response patterns. Re-perform the canonical correlation analysis in step (3) for each type of reservoir until the coefficient of determination meets the requirements or the sample size is insufficient for regression.

[0036] 5. Execute step (5): Input the optimized seismic attribute data into the multiple linear regression model and perform model operations using MATLAB software. Use the operation results to draw the reservoir attribute plan view with the professional mapping software ArcGIS to clearly display the reservoir distribution.

[0037] Example 1: Prediction of conventional sandstone reservoir Application scenario: Sandstone reservoir in a continental basin, with reservoir thickness of 5 - 20m and burial depth of 1800 - 2500m, porosity distribution needs to be predicted. Implementation steps: 1. Optimization of well-seismic calibration: Adopt wavelet transform time-frequency analysis to match the seismic traces beside the well, and combine the synthetic seismic record to correct the time-depth relationship, with the error controlled within ±2ms.

[0038] 2. Attribute extraction and screening: Extract amplitude, wave impedance and frequency attributes, and find that the wave impedance has the highest correlation coefficient with porosity (R = 0.82).

[0039] 3. Classification optimization: Since the initial determination coefficient R² = 0.65 (<0.7), support vector machines were used to classify according to sandstone thickness (thick layers > 10m, thin layers ≤ 10m). After classification, the R² values of the two categories were increased to 0.75 and 0.78 respectively.

[0040] 4. Planar prediction: · Using a multiple linear regression model, inputting the combined parameters of wave impedance + amplitude, generating a porosity distribution map, and the error between the prediction result and the verification well is <8%.

[0041] Example 2: Prediction of thin interbed reservoirs Application scenario: Thin interbed reservoirs (single layer thickness 2 - 8m) in coastal areas, and the distribution range of sand bodies needs to be predicted. Technical adjustments: 1. High-precision time-frequency analysis: Using the generalized S transform to improve time-frequency resolution and optimizing the matching accuracy of synthetic seismic records, with the time-depth relationship error <1.5ms.

[0042] 2. Attribute combination optimization: Extracting arc length, skewness, and waveform envelope attributes, and screening out that the skewness attribute has the best correlation with sand body thickness (R = 0.75).

[0043] 3. Dynamic classification strategy: After the first classification, R² = 0.68. Classifying again according to the seismic response mode (strong amplitude / weak amplitude), and finally the R² values of both categories are >0.72.

[0044] 4. Non-isochronous window extraction: Defining non-isochronous windows along the top and bottom of the main lobe of the seismic waveform to reduce the interference of non-reservoir information, and the accuracy of predicting the sand body boundary is improved by 15%.

[0045] Example 3: Classification and prediction of reservoirs in complex structural areas Application scenario: Areas with developed faults, and the reservoirs are affected by tectonic compression to form multiple types of sand bodies (channel sand, beach bar sand). Technical highlights: 1. Multi-attribute fusion analysis: Combining amplitude, phase, and seismic inversion velocity attributes to construct a multi-parameter cross plot to identify the differential responses of channel sand and beach bar sand.

[0046] 2. Support vector machine classification optimization: Classifying according to the thickness distribution (thick layers > 15m, medium-thick layers 5 - 15m, thin layers < 5m) and the seismic waveform distortion degree, and after classification, the R² value is increased to 0.73 - 0.81.

[0047] 3. Regional regression modeling: Establishing regression formulas for different structural units (fault blocks / slopes) respectively, and the input attribute combination includes amplitude + wave impedance + arc length, and the coincidence rate of the planar map is >85%.

[0048] Example 4: Reservoir Prediction in Low Signal-to-Noise Ratio Area Application Scenario: The signal-to-noise ratio of seismic data is low (main frequency < 30 Hz), and the ability to identify thin reservoirs needs to be improved. Adaptive Improvements: 1. Data Preprocessing: Frequency compensation and prestack depth migration techniques are used to improve the data quality, and the main frequency of the target layer is increased to 45 Hz.

[0049] 2. Attribute Enhancement Technique: The instantaneous frequency attribute in the high-frequency band (40 - 60 Hz) is extracted, and the correlation coefficient R with the reservoir thickness is 0.71, which is better than that of the full-frequency band attribute (R = 0.58).

[0050] 3. Small Sample Size Processing: When the single-class sample size < 10, Bayesian regression is used to replace the linear model, and the prediction error is reduced by 12%.

[0051] Example 5: Parameter Sensitivity Test Test Purpose: To verify the influence of the coefficient of determination threshold (0.7) and attribute combinations on the prediction results. Test Design: 1. Threshold Comparison: By comparing the thresholds 0.6 / 0.7 / 0.8, it is found that when the threshold is 0.7, the number of classification categories is moderate (3 - 5 categories), and the total prediction error is the smallest (average error 9.2%).

[0052] 2. Attribute Combination Comparison: The prediction error of the single attribute (acoustic impedance) is 15.3%, the error of the dual attribute (acoustic impedance + amplitude) is 10.1%, and the error of the triple attribute (+ arc length) is 8.5%.

[0053] 3. Algorithm Comparison: The classification accuracy of the support vector machine (89%) is better than that of K-means clustering (76%) and random forest (82%).

[0054] Verification of Technical Effects All examples are evaluated by the cross-validation method: The well data is divided into a training group (70%) and a validation group (30%), and the mean absolute error (MAE) and coefficient of determination R² are calculated.

[0055] The MAE of Examples 1 - 4 is 7.2 - 9.8%, and R² is 0.72 - 0.84, which is better than the traditional single-attribute prediction method (MAE > 15%, R² < 0.6).

[0056] The above examples fully support the technical features of Claims 1 - 7 through different scenarios and parameter configurations, reflecting the universality and innovation of the present invention.

[0057] Through the above detailed and scientific implementation steps, the method of the present invention can be effectively implemented, significantly improving the accuracy of reservoir description by seismic attributes and providing strong support for oil and gas exploration and reservoir evaluation.

Claims

1. A method for improving the accuracy of reservoir description using seismic attributes, characterized in that, Including the following steps: Step (1): Use the stratigraphic sequence in the well and the time-frequency analysis results of the seismic trace beside the well for matching to determine the corresponding relationship between the well and the large-scale seismic stratigraphic unit, and establish the time-depth relationship of the study area through synthetic seismic records; Step (2): Based on the optimized well-seismic calibration results, extract various seismic attributes along the target geological horizon; Step (3): Conduct a correlation analysis on the reservoir attributes at the well points and the corresponding seismic attributes, and select the seismic attribute with the largest correlation coefficient as the best prediction parameter for the reservoir attributes; Step (4): If the determination coefficient in Step (3) is less than the predetermined value, use the support vector machine technology to classify the reservoir attributes, and re-execute Step (3) for each class of reservoir attributes until the determination coefficients of all classifications are greater than the predetermined value or the sample size is insufficient for regression; Step (5): Use the optimized regression formula and seismic attribute data to calculate the planar distribution of the reservoir attributes in the well-free area and generate a reservoir attribute plan view.

2. The method for improving the accuracy of reservoir description using seismic attributes according to claim 1, wherein The time-frequency analysis in Step (1) includes the matching analysis of the stratigraphic sequence and the seismic trace time series, and corrects the mapping accuracy of the time-depth relationship through synthetic seismic records.

3. The method for improving the accuracy of reservoir description by seismic attributes according to claim 2, characterized in that The reservoir attributes in Step (3) include reservoir thickness, and the seismic attributes include at least one of amplitude, frequency, phase, and wave impedance.

4. The method for improving the accuracy of reservoir description using seismic attributes according to claim 3, wherein The support vector machine technology in Step (4) is used to classify the reservoir sand bodies, and the classification basis includes the thickness distribution characteristics and seismic response patterns.

5. The method for improving the accuracy of reservoir description using seismic attributes according to claim 4, characterized in that, The regression formula in Step (5) is a multiple linear regression model, and its input parameters are the optimized seismic attribute combinations.

6. The method for improving the accuracy of reservoir description using seismic attributes according to claim 5, characterized in that, The method is used for the quantitative prediction of the distribution range and thickness of reservoirs in the oil and gas exploration and reservoir evaluation stages.

7. The method for improving the accuracy of reservoir description by seismic attributes according to claim 6, characterized in that, The determination coefficient is the square of the correlation coefficient.

8. The method for improving the accuracy of reservoir description by seismic attributes according to claim 7, characterized in that The predetermined value is 0.

7.

9. The method for improving the accuracy of reservoir description using seismic attributes according to claim 8, characterized in that, The various seismic attributes in Step (1) include the wave impedance data obtained by inversion.

Images