A method for improving crack prediction accuracy in RS-FD numerical analysis using the Bayes algorithm

Abstract

This application discloses a method for improving the accuracy of fracture prediction using RS-FD numerical analysis with the Bayes algorithm. First, the fracture response locations of various conventional logging curves are determined using RS-FD numerical analysis. Second, for single wells containing core and imaging logging data, the fracture development locations are statistically analyzed. Combined with the RS-FD numerical analysis results, the fracture identification rate of each conventional logging curve is determined to establish the prior probability and likelihood of the Bayesian calculation. Then, according to the Bayesian calculation rule, the Bayesian posterior probability of each conventional logging curve is calculated. Based on the posterior probability, the fracture sensitivity curve is optimized to construct a comprehensive fracture evaluation index. The accuracy of fracture identification is verified, and the fracture development locations of the single well are then predicted. This method provides a practical solution for professionals researching single-well fracture prediction using conventional logging.

Description

Technical Field

[0001] This application relates to the field of geological engineering technology, and in particular to a method for improving the accuracy of fracture prediction in RS-FD numerical analysis using the Bayes algorithm. Background Technology

[0002] With the growth of global energy demand, carbonate oil and gas reservoirs have gradually become important targets for oil and gas exploration and development. Carbonate oil and gas reserves account for more than half of the world's total oil and gas reserves, and the quantity and quality of fractures in carbonate reservoirs are crucial conditions for whether carbonate rocks can become high-quality reservoirs and efficiently accumulate oil and gas. Currently, research on carbonate reservoir fractures mainly includes two methods: surface (outcrops) and subsurface (cores, thin sections, imaging, and seismic data). Surface (outcrop) carbonate fracture research has many advantages such as being intuitive, convenient, and repeatable; however, it also suffers from the phenomenon that original fractures have been damaged or altered due to various factors such as weathering, erosion, temperature differences, and human activities. Therefore, research on subsurface primary reservoir fractures often only has theoretical guiding significance. Subsurface (cores, thin sections, imaging logging, and seismic data) carbonate fracture research preserves the original state of natural fractures to the greatest extent possible, but it also has disadvantages such as high data costs and limited data availability. Therefore, utilizing limited subsurface data to study and predict primary reservoir fractures is an important direction for carbonate reservoir evaluation. In current oil and gas exploration and development, although conventional logging lacks the advantage of directly observing and describing core samples, it is deployed in every well, providing abundant data. Therefore, combining the intuitive fracture information from core, thin section, and imaging logging with the response characteristics of conventional logging at fracture development sites to predict fracture development locations in single wells without core, thin section, or imaging logging is of paramount importance for clarifying the spatial distribution patterns of subsurface fractures in carbonate rocks, determining the distribution patterns of subsurface fracture networks, and optimizing well network deployment.

[0003] Currently, there are numerous methods for fracture identification and characterization using conventional logging, both domestically and internationally. Among them, the R / S-FD method, which combines fractal statistics theory with time series analysis and then uses a finite difference algorithm, can achieve efficient fracture prediction. However, in actual fracture research, it has been found that this method has two shortcomings in its use: (1) Due to the significant differences in the properties of different conventional logging curves, their response capabilities to fractures with different underground properties are significantly different. Using a single conventional logging curve for R / S-FD calculation to predict fractures in a single well has strong limitations and is prone to misjudgment or omission; (2) The method of performing R / S-FD calculation on multiple conventional logging curves, selecting the optimal fracture sensitivity curve, and then multiplying the R / S-FD results of multiple fracture sensitivity curves to ensure the maximum fracture prediction efficiency, although it maximizes the fracture prediction efficiency, often involves a great deal of subjectivity in selecting the optimal fracture sensitivity curve. Practical users tend to choose curves with similar prediction results for multiplication calculation to ensure the number of fracture predictions, but this defeats the purpose of curve selection and results in a strong trace of human control in the final prediction result. Summary of the Invention

[0004] This application provides a method for improving the accuracy of fracture prediction in RS-FD numerical analysis using the Bayes algorithm, aiming to solve the problems of strong limitations and easy misjudgment or omission in the R / S-FD calculation of the single logging curve in the existing technology.

[0005] A method for improving crack prediction accuracy in RS-FD numerical analysis using the Bayes algorithm, the method comprising:

[0006] S1: Processing of conventional logging curves of single wells in the target formation, RS-FD numerical analysis and calculation, establishment of a rectangular coordinate system, and identification of fracture response locations of each curve;

[0007] S2: Based on core and imaging logging data, statistically analyze the fracture development locations in single wells, clarify the fracture identification rate of each curve, and determine the prior probability, likelihood, and total probability calculated by Bayesian methods.

[0008] S3: Based on the Bayesian computational rule, calculate the Bayesian posterior probability of each curve, select the optimal fracture sensitivity curve, construct the comprehensive fracture evaluation index I, verify the accuracy of fracture identification, and predict the fracture development location of the single well in the well to be predicted.

[0009] In the above scheme, step S1 may optionally include:

[0010] S11: Identify the target layer for fracture prediction in a single well and process the conventional logging curves of each well in the target layer;

[0011] The data contained in all conventional logging curves outside the target formation are removed. The processing of conventional logging curves within the target formation includes: unifying the units of all types of logging curves from all single wells in the study area; checking the data distribution of each conventional logging curve in the target formation; directly assigning a value of 0 to abnormal data, including: 0, -999.25, -9999.00, and blank values; accurately determining the reasonable range of data distribution for each conventional logging curve based on the lithological distribution of the target formation; judging the data of conventional logging curves that exceed the reasonable range, and if they are abnormal data, directly assigning a value of 0.

[0012] S12: Based on the RS-FD numerical analysis method, calculate each conventional logging curve in the target layer, establish a rectangular coordinate system, and identify the fracture response location of each curve;

[0013] Based on the RS-FD numerical analysis method, the calculations were performed on each conventional logging curve in the target layer, and the fracture response locations of different conventional logging curves were identified.

[0014]

[0015] Where, n is the total number of data points in the target layer of each conventional logging curve; u is the number of scales that increase sequentially from 0 to n in the target layer of each conventional logging curve starting from the endpoint; Z(i) and Z(j) represent the selected logging curve data; i and j represent the variable of the number of sample points; R(n) represents the range; S(n) represents the standard deviation of the entire process sequence. The second derivative of the R / S value is represented by h, which is the calculation step size.

[0016] The method for determining the fracture response location of different conventional logging curves is as follows: After calculating the R / S ratio of each conventional logging curve and obtaining the logarithm of the corresponding number of scales u, a rectangular coordinate system is established with lg(R / S) as the ordinate and lg(u) as the abscissa. A scatter plot of the relationship between the two is then plotted. The locations of abrupt slope changes in the scatter plot are the fracture response points. A rectangular coordinate system is then established with the second derivative of the R / S value as the ordinate and lg(u) as the abscissa. A bar chart of the relationship between the two is then plotted. Abnormally high values ​​in the bar chart are the fracture sensitive points. The two charts are then merged along the X-axis to create a double Y-axis chart. The matching between the fracture response points and fracture sensitive points is observed to determine the fracture response location of each conventional logging curve.

[0017] In the above scheme, step S2 may optionally include:

[0018] S21: Conduct statistical analysis of the fracture development locations in the core samples from the target formation, and identify the fracture development zones for each formation.

[0019] S22: Conduct statistical analysis of fracture development locations in single wells with imaging logging deployed, and identify the fracture development zones of each well.

[0020] S23: Match the fracture development intervals obtained from the core layer and imaging logging layer with the fracture response locations identified by each conventional logging curve after calculation based on the RS-FD numerical analysis method, and clarify the fracture identification rate of each conventional logging curve;

[0021] S24: Based on the fracture identification rate of each conventional logging curve, determine the prior probability, likelihood, and total probability of each conventional logging curve calculated by Bayes.

[0022] In the above scheme, optional step S21 includes: when conducting statistics on the development of fractures in a single well in the core layer of the target layer, before core observation, determining the top depth, bottom depth and core recovery rate of the core section, and then resetting all the cores in sequence, measuring the top and bottom depths corresponding to the fracture development locations of the cores with developed fractures, so as to determine the fracture length and development location.

[0023] Step S22 includes: statistical analysis of fracture development locations in the target formation of imaging logging; observation of dynamic and static images from imaging logging; statistical analysis of natural fractures with sine and cosine curve characteristics; and exclusion of induced fractures with a goose-like arrangement caused by drilling tools in the longitudinal direction. The statistical analysis of natural fractures clarifies the top and bottom depths of their development locations, indicating the longitudinal influence range of the fractures.

[0024] In the above scheme, optionally, step S23 includes: if imaging logging is deployed in the core well, the fractures identified by imaging logging are compared and merged with the fractures identified by the core well, and all fractures are completely identified in the target layer; then, the fracture response locations identified by conventional logging curves calculated based on RS-FD numerical analysis method are matched, and the fracture identification rate of each conventional logging curve is calculated.

[0025] If no imaging logging is deployed in the core well, the fracture information obtained by statistical analysis of the core layer is matched with the fracture response locations identified by each curve after calculation based on the RS-FD numerical analysis method, and the fracture identification rate of each conventional logging curve is calculated.

[0026] If a single well does not undergo coring and only imaging logging is deployed, the fracture information obtained through imaging logging layer statistics is matched with the fracture response locations identified by each curve after calculation based on the RS-FD numerical analysis method, and the fracture identification rate of each conventional logging curve is calculated.

[0027] When matching the fracture information obtained from core and imaging logging with the fracture response locations identified by the curves calculated by the RS-FD numerical analysis method, it is clear that the fracture information obtained from core and imaging logging is the accurate fracture development location and fracture length information of a single fracture in the vertical direction of a single well, which belongs to a certain range. If the fracture response location identified by the curves calculated by the RS-FD numerical analysis method is within the range of the accurate fracture development location obtained from core and imaging logging, then a fracture is considered to have been identified.

[0028] In the above scheme, step S24 optionally includes:

[0029] When performing Bayesian calculations using various conventional well logging curves, the prior probability... This should be the probability of randomly selecting any conventional logging curve and performing Bayesian calculations. The prior probability is calculated based on the number of conventional logging curves in the actual situation, and the formula is:

[0030]

[0031] Its likelihood This should be the fracture identification rate (P) of any randomly selected conventional logging curve. k The formula is:

[0032]

[0033] Its total probability It should be the sum of the probabilities that each conventional logging curve is randomly selected for subsequent fracture prediction in actual practice, and its formula is:

[0034]

[0035] Where n is the total number of conventional logging curves; i is the type of each conventional logging curve.

[0036] In the above scheme, step S3 may optionally include:

[0037] S31: Based on Bayesian computation rules, perform Bayesian posterior probability analysis on each conventional well logging curve. Calculation;

[0038] S32: Based on the Bayesian posterior probability of each conventional logging curve obtained by calculation, the fracture sensitivity curve is selected first;

[0039] S33: Construct a comprehensive crack evaluation index I based on the selected crack sensitivity curve as the basic parameter;

[0040] S34: By re-matching the fracture development intervals obtained from core layer and imaging logging layer statistics, the accuracy of the constructed comprehensive fracture evaluation index is verified, and the single-well fracture development location of the well to be predicted is completed.

[0041] In the above scheme, optionally, step S31 includes: after calculating the prior probability, likelihood, and total probability of each conventional logging curve using the Bayesian algorithm, calculating the Bayesian posterior probability of each conventional logging curve according to the Bayesian calculation rule. Its formula is:

[0042]

[0043] Step S32 includes: The principle for selecting the optimal fracture sensitivity curve is to compare the posterior probability and prior probability calculated using the Bayesian algorithm for each conventional logging curve. Curves with a posterior probability exceeding the prior probability are considered fracture sensitive, while those with a posterior probability less than the prior probability are considered not to have fracture sensitivity. The posterior probabilities calculated using the Bayesian algorithm for each conventional logging curve are sorted from largest to smallest. Curves with a higher posterior probability than the prior probability have stronger fracture sensitivity and better fracture prediction capabilities. Conversely, conventional logging curves with a posterior probability greater than the prior probability but closer to the prior probability have weaker fracture sensitivity and poorer fracture prediction capabilities. A comprehensive evaluation is performed by comparing the fracture sensitivity of each conventional logging curve on different wells. If the sensitivity of the same curve differs significantly across different wells, that curve is directly discarded.

[0044] In the above scheme, optional step S33 includes: establishing a comprehensive crack evaluation index I based on the actual preferred conditions, the formula of which is:

[0045]

[0046] Among them, I Z , is the product of the R / S-FD calculation results of the optimized fracture sensitivity curves; n, is the total number of conventional logging curves under the actual optimized conditions;

[0047] By applying normalization, the data is processed to facilitate data analysis. The formula is as follows:

[0048]

[0049] Among them, I Zn , represents each value of the product of the optimized crack sensitivity curve R / S-FD calculation results; n, represents the total number of the optimized crack sensitivity curve summation results;

[0050] A direct coordinate system is established with each value of the comprehensive crack evaluation index I as the vertical axis and the values ​​corresponding to the order of each value as the horizontal axis. A cumulative curve is then created. The upper and lower ends of the crack development area are shown as a sudden increase in the Y value on the cumulative curve. After exceeding the crack development area, the Y value change quickly returns to the normal growth rate. In addition, the middle of the crack will show a local maximum value of Y value on the cumulative curve. The crack development area on the comprehensive crack evaluation index I curve is thus determined.

[0051] In the above scheme, optionally, step S34 includes:

[0052] The accuracy of the current fracture comprehensive evaluation index I curve in predicting fractures in a single well is verified by rematching the fracture development locations predicted on the fracture comprehensive evaluation index I curve with the fracture development segments obtained from the core layer and imaging logging layer. When the accuracy of single-well fracture prediction reaches the prediction standard of the study area, the fracture comprehensive evaluation index I curve is used to carry out single-well fracture prediction. If it does not reach the prediction standard of the study area, the curve with the lowest posterior probability in the fracture sensitivity curve is removed, and the fracture comprehensive evaluation index I curve is reconstructed.

[0053] When the constructed fracture comprehensive evaluation index I curve has reached the prediction standard of the study area, several conventional logging curves that are the same as the constructed fracture comprehensive evaluation index I curve in the conventional logging curves of the well to be predicted are selected in a targeted manner to calculate R / S-FD. After calculation, the fracture comprehensive evaluation index I curve of the well to be predicted is directly constructed by applying the calculation formula of fracture comprehensive evaluation index I, so as to carry out the prediction of the fracture development location of the well to be predicted.

[0054] Compared with the prior art, this application has at least the following beneficial effects:

[0055] This application, based on further analysis and research of existing technical problems, recognizes the limitations of existing R / S-FD calculations using a single logging curve, which are prone to misjudgment or omission. By introducing a Bayesian algorithm to optimize the fracture prediction process using conventional logging data, this application significantly improves the accuracy of fracture prediction in carbonate reservoirs based on the RS-FD method, solving common problems of misjudgment, omission, and subjective selection in existing technologies. Existing R / S-FD methods suffer from significant differences in logging curve responses and strong subjectivity in selecting fracture sensitivity curves, limiting the accuracy of fracture prediction. The solution proposed in this application prioritizes processing each conventional logging curve in the target interval, clarifying the response characteristics of different logging curves to fractures, thus effectively avoiding misjudgment when using a single logging curve. By introducing prior probability and likelihood through the Bayesian algorithm, combined with fracture statistical analysis of high-quality data such as core and imaging logging, the fracture identification rate of each logging curve can be theoretically calculated accurately, reducing the impact of subjective selection on the results. The calculation of Bayesian posterior probability can objectively optimize fracture sensitivity curves, further construct a comprehensive fracture evaluation index, and achieve accurate prediction of fracture development locations in the well to be predicted. Furthermore, by combining well history, production data, and leakage data, the accuracy of fracture prediction can be more comprehensively verified, ensuring the reliability of the prediction results. Overall, this approach not only improves the accuracy and efficiency of fracture prediction but also reduces human intervention, enhancing the scientific rigor and objectivity of fracture identification. It provides more precise technical support for oil and gas exploration and development in carbonate reservoirs, solving the problems caused by limitations in logging data and subjective factors in traditional methods. This allows for effective prediction and study of fracture development locations and distribution patterns even with limited conventional logging data, thus promoting further development in carbonate reservoir fracture research.

[0056] This method clarifies the fracture response locations of each curve from the perspective of RS-FD numerical calculation. Then, for single wells containing core and imaging logging data, it statistically analyzes the fracture development locations and, combined with RS-FD numerical analysis results, clarifies the fracture identification rate of each curve to determine the prior probability and likelihood of Bayesian calculation. Next, based on Bayesian calculation rules, it calculates the Bayesian posterior probability of each curve. Based on the posterior probability, it selects the optimal fracture sensitivity curve, constructs a comprehensive fracture evaluation index, and verifies the fracture identification accuracy. Finally, based on the comprehensive fracture evaluation index, it predicts the fracture development locations of the well to be predicted, and, by combining well history, single-well production, venting and leakage data, and fracture identification accuracy, accurately predicts the fracture development locations of the single well. This method provides a practical and feasible solution for professionals researching single-well fracture prediction using conventional logging. Attached Figure Description

[0057] Figure 1A flowchart illustrating a method for improving crack prediction accuracy in RS-FD numerical analysis using the Bayes algorithm, provided in one embodiment of this application;

[0058] Figure 2 This application provides an embodiment of a fracture response location identification map for eight conventional logging curves.

[0059] Figure 3 An analysis chart of the fracture identification rate of eight conventional logging curves provided in one embodiment of this application;

[0060] Figure 4 A statistical graph showing the prior probability, likelihood, and posterior probability of eight conventional well logging curves calculated by Bayes for one embodiment of this application;

[0061] Figure 5 This is a crack comprehensive evaluation index accuracy discrimination chart provided in one embodiment of this application. Detailed Implementation

[0062] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0063] In one embodiment, such as Figure 1 As shown, a method for improving crack prediction accuracy in RS-FD numerical analysis using the Bayes algorithm is provided, including the following steps:

[0064] S1: Processing of conventional logging curves of single wells in the target formation, RS-FD numerical analysis and calculation, establishment of a rectangular coordinate system, and identification of fracture response locations of each curve.

[0065] S2: Based on core and imaging logging data, the locations of fracture development in single wells are statistically analyzed, and the fracture identification rate (P) of each curve is determined. k Determine the prior probabilities of Bayesian computation. Likelihood and total probability .

[0066] S3: Based on the Bayesian computational rule, calculate the Bayesian posterior probability of each curve, select the optimal fracture sensitivity curve, construct the comprehensive fracture evaluation index I, verify the accuracy of fracture identification, and predict the fracture development location of the single well in the well to be predicted.

[0067] In this embodiment, step S1 includes:

[0068] S11: Identify the target layer for fracture prediction in a single well and process the conventional logging curves of each well in the target layer.

[0069] S12: Based on the RS-FD numerical analysis method, calculate each conventional logging curve in the target layer, establish a rectangular coordinate system, and identify the fracture response location of each curve.

[0070] In this embodiment, step S11 includes: For oil exploration and development, conventional logging typically covers a large vertical range. Therefore, the first step in single-well fracture prediction is to identify the target layer and remove data from conventional logging curves outside the target layer to avoid increased workload or missed predictions due to unclear target layers. After identifying the target layer, the conventional logging curves within that layer need to be processed. First, the units of all types of logging curves from all single wells in the study area need to be standardized. Standardized units are a prerequisite for subsequent comparison of fracture prediction results. Second, based on standardized units, the data distribution of each conventional logging curve within the target layer is checked. Abnormal data such as 0, -999.25, -9999.00, and blank values ​​can be directly assigned a value of 0 for easier subsequent operations.

[0071] In this embodiment, the numerical variations of conventional logging curves vary considerably depending on the lithology of the subsurface strata and the degree of fracture development within the same lithology. For example, in carbonate formations, AC (acoustic transit time) logging is typically 48-65 μs / ft, DEN (density) logging is mostly 2.63-2.72 g / cm³, CNL (neutron) logging is mainly between 1%-5%, GR (natural gamma) logging is mostly less than 30 API, fractured reservoirs are 10-70 API, and RL (resistivity logging) logging is mainly distributed between 5-1000 Ω·m. The difference in fracture size affects the variation in resistivity. Therefore, the reasonable range of data distribution for each conventional logging curve should be accurately determined based on the lithological distribution of the target stratum. Conventional logging curves exceeding the reasonable range should be evaluated; if they are abnormal data, they can be directly assigned a value of 0.

[0072] In this embodiment, step S12 includes: calculating the conventional logging curves of each layer based on the RS-FD numerical analysis method formula, and identifying the fracture response locations of different conventional logging curves.

[0073]

[0074]

[0075] Where, n is the total number of data points in the target layer of each conventional logging curve; u is the number of scales that increase sequentially from 0 to n in the target layer of each conventional logging curve starting from the endpoint; Z(i) and Z(j) represent the selected logging curve data; i and j represent the variable of the number of sample points; R(n) represents the range; S(n) represents the standard deviation of the entire process sequence. represents the second derivative of the R / S value; h is the calculation step size.

[0076] In this embodiment, the method for determining the fracture response location of different conventional logging curves is as follows: First, by calculating the logarithm of the R / S ratio of each conventional logging curve and the corresponding number of scales u, a rectangular coordinate system is established with lg(R / S) as the ordinate and lg(u) as the abscissa, and a scatter plot of the relationship between the two is drawn. The locations of abrupt slope changes in the scatter plot are the fracture response points. Second, a rectangular coordinate system can be established with the second derivative of the R / S value as the ordinate and lg(u) as the abscissa, and a bar chart of the relationship between the two is drawn. The abnormally high values ​​in the bar chart are the fracture sensitive points. Finally, the two charts are merged with the same x-axis to create a double y-axis chart. By observing the matching between the fracture response points and the fracture sensitive points, the fracture response location of each conventional logging curve can be determined.

[0077] In this embodiment, step S2 includes: S21: statistically analyzing the fracture development locations of single wells in the core layer of the target formation, and identifying the fracture development zones.

[0078] S22: Conduct statistical analysis of fracture development locations in single wells with imaging logging deployed, and identify the fracture development zones of each well.

[0079] S23: Match the fracture development intervals obtained from core and imaging logging with the fracture response locations identified by each conventional logging curve calculated using the RS-FD numerical analysis method to determine the fracture identification rate (P) of each conventional logging curve. k ).

[0080] S24: Based on the fracture identification rate of each conventional logging curve, determine the prior probability calculated by Bayes for each conventional logging curve. Likelihood and total probability .

[0081] In this embodiment, step S21 includes: when conducting statistical analysis of the fracture development locations of a single well in the target layer core segment of a coring well, before core observation, it is necessary to first determine the top depth, bottom depth, and core recovery rate of the coring well segment, and then reposition all the cores in sequence, and measure the top and bottom depths corresponding to the fracture development locations of the cores with developed fractures to determine the fracture length and development location, providing accurate basic data for subsequent analysis of the fracture response locations identified by the RS-FD numerical analysis method.

[0082] In this embodiment, step S22 includes: statistically analyzing the fracture development locations in the target formation of the imaging logging, primarily based on observations of dynamic and static images from the imaging logging. Here, only natural fractures exhibiting sine and cosine curve characteristics need to be counted, while induced fractures with a goose-like arrangement caused by drilling tools in the vertical direction need to be excluded. The statistical analysis of natural fractures only needs to determine the top and bottom depths of their development locations to represent the longitudinal influence range of the fractures, and to provide accurate basic data for subsequent analysis of the fracture response locations identified by the RS-FD numerical analysis method.

[0083] In this embodiment, step S23 includes: Core sampling in the target layer is mostly targeted partial coring, and cannot achieve full coverage of the target layer; it is extremely rare to find complete coring of the entire target layer. Therefore, fracture statistics at the core layer level are mostly highly discrete, with limited and concentrated fracture data. Imaging logging, on the other hand, is mostly deployed continuously in a single well. In this case, if imaging logging is deployed in the core sampling well, the fractures identified by the imaging logging can be compared and merged with those identified by the core sampling well, allowing for complete identification of all fractures in the target layer. Then, this is matched with the fracture response locations identified by each conventional logging curve calculated using the RS-FD numerical analysis method to calculate the fracture identification rate (P) of each conventional logging curve. k If the core well does not have imaging logging deployed, the fracture identification rate (P) of each conventional logging curve can only be calculated by matching the fracture information obtained from the core layer statistics with the fracture response locations identified by each curve calculated based on the RS-FD numerical analysis method. k If a single well does not undergo coring and only imaging logging is deployed, the fracture information obtained from the imaging logging layers can only be matched with the fracture response locations identified by each curve after calculation using the RS-FD numerical analysis method to calculate the fracture identification rate (P) of each conventional logging curve. kDue to significant differences in the deployment of imaging logging and coring wells across different study areas, some wells in the target layer have both imaging logging and coring capabilities, while others only have coring or imaging logging in the target layer. Therefore, it is necessary to analyze the fracture identification rate based on the actual data. However, it is essential to ensure that fracture identification rates are calculated for all wells in the study area that have both imaging logging and coring capabilities in the target layer, as well as those that only have coring or imaging logging in the target layer.

[0084] In this embodiment, when matching the fracture information obtained from the core layer and imaging logging layer with the fracture response locations identified by each curve calculated by the RS-FD numerical analysis method, it is necessary to clarify that the fracture information obtained from the core layer and imaging logging layer represents the accurate fracture development location and fracture length information of a single fracture in the vertical direction of a single well, which falls within a certain range. Therefore, as long as the fracture response locations identified by each curve calculated by the RS-FD numerical analysis method are within the range of the accurate fracture development locations obtained from the core layer and imaging logging layer, it can be considered that a fracture has been identified.

[0085] In this embodiment, step S24 includes: when performing Bayesian calculations using each conventional logging curve, its prior probability This should be the probability of randomly selecting any conventional logging curve and performing Bayesian calculations. Therefore, the prior probability needs to be calculated based on the number of conventional logging curves in the actual situation. The formula is as follows:

[0086]

[0087] Its likelihood This should be the fracture identification rate (P) of any randomly selected conventional logging curve. k The formula is:

[0088]

[0089] Its total probability It should be the sum of the probabilities that each conventional logging curve is randomly selected for subsequent fracture prediction in actual practice, and its formula is:

[0090]

[0091] Where n is the total number of conventional logging curves; i is the type of each conventional logging curve.

[0092] In this embodiment, step S3 includes:

[0093] S31: Based on Bayesian computation rules, perform Bayesian posterior probability analysis on each conventional well logging curve. The calculation.

[0094] S32: Based on the Bayesian posterior probabilities of each conventional logging curve obtained from the calculation, the fracture sensitivity curve is selected.

[0095] S33: Construct a comprehensive crack evaluation index I based on the selected crack sensitivity curve as the basic parameter.

[0096] S34: By re-matching the fracture development intervals obtained from core layer and imaging logging layer statistics, the accuracy of the constructed comprehensive fracture evaluation index is verified, and the single-well fracture development location of the well to be predicted is completed.

[0097] In this embodiment, step S31 includes: calculating the prior probability of each conventional logging curve using a Bayesian algorithm. Likelihood and total probability Then, the Bayesian posterior probability of each conventional logging curve can be calculated according to the Bayesian computational rule. Its formula is:

[0098]

[0099] In this embodiment, step S32 includes the following principle for selecting the fracture sensitivity curve: First, compare the posterior probability and prior probability calculated using the Bayesian algorithm for each conventional logging curve. Curves with a posterior probability value exceeding the prior probability are considered fracture sensitive curves, while those with a posterior probability value less than the prior probability are considered to lack fracture sensitivity. Second, sort the posterior probabilities calculated using the Bayesian algorithm for each conventional logging curve from largest to smallest. Conventional logging curves with a posterior probability greater than the prior probability and a higher posterior probability value have stronger fracture sensitivity and a stronger fracture prediction ability. Conversely, conventional logging curves with a posterior probability greater than the prior probability and a posterior probability value closer to the prior probability have weaker fracture sensitivity and a poorer fracture prediction ability. Finally, since we need to perform RS-FD numerical analysis and Bayesian calculations on single wells that have both imaging logging and coring at the target layer, and single wells that only have coring or imaging logging at the target layer, it is necessary to comprehensively compare the fracture sensitivity of each conventional logging curve on different wells for a comprehensive evaluation. Normally, the sensitivity of different conventional logging curves to fractures is not significantly different across different wells. If the same curve shows a large difference in sensitivity across different wells, that curve can be directly discarded.

[0100] In this embodiment, step S33 includes: establishing a comprehensive crack evaluation index I based on actual preferred conditions, the formula of which is:

[0101]

[0102] Among them, IZ , is the product of the R / S-FD calculation results of the optimized fracture sensitivity curves; n is the total number of conventional logging curves under the actual optimized conditions.

[0103] Since the product of the optimized crack sensitivity curves after Bayesian calculation significantly enhances the numerical dispersion, it is necessary to process the data using a normalization method to facilitate data analysis. The formula is as follows:

[0104]

[0105] Among them, I Zn , represents each value of the product of the optimized crack sensitivity curve R / S-FD calculation results; n, represents the total number of the optimized crack sensitivity curve summation results.

[0106] In this embodiment, by multiplying the calculated results of the optimized crack sensitivity curves R / S-FD, the accuracy of crack prediction can be further improved, reducing the occurrence of misjudgments or multiple judgments in crack prediction by a single crack sensitivity curve. Multiple curves jointly constrain the crack prediction results to improve accuracy. At this point, a direct coordinate system is established with each value of the crack comprehensive evaluation index I as the ordinate and the values ​​corresponding to the order of each value as the abscissa, creating a cumulative curve. The upper and lower endpoints of the crack development area are represented by a sudden increase in the Y-value on the cumulative curve. After exceeding the crack development area, the Y-value change rapidly returns to a normal growth rate, and the middle of the crack exhibits a local maximum Y-value on the cumulative curve. This method can quickly determine the crack development area on the crack comprehensive evaluation index I curve.

[0107] In this embodiment, step S34 includes: re-matching the fracture development locations predicted on the fracture comprehensive evaluation index I curve with the fracture development segments statistically obtained from core and imaging logging layers to verify the accuracy of the currently constructed fracture comprehensive evaluation index I curve in predicting fractures in a single well. Once the single-well fracture prediction accuracy reaches the prediction standard for the study area, the fracture comprehensive evaluation index I curve can be used for single-well fracture prediction. If it does not reach the prediction standard for the study area, the curve with the lowest posterior probability in the fracture sensitivity curve can be removed, and the fracture comprehensive evaluation index I curve can be reconstructed to improve the single-well fracture prediction accuracy.

[0108] In this embodiment, the fracture prediction standard for the study area should be based on the actual requirements of the area. It is necessary to first discuss and determine the fracture prediction standard suitable for the area with experts and engineers in the study area before evaluating the compliance of the single-well fracture prediction accuracy.

[0109] In this embodiment, step S34 includes: when the constructed fracture comprehensive evaluation index I curve has reached the prediction standard of the study area, several conventional logging curves that are the same as the constructed fracture comprehensive evaluation index I curve in the conventional logging curves of the well to be predicted can be selected in a targeted manner to calculate R / S-FD. After calculation, the fracture comprehensive evaluation index I curve of the well to be predicted is directly constructed by applying the calculation formula of fracture comprehensive evaluation index I, so as to carry out the prediction of the fracture development location of the single well to be predicted.

[0110] This embodiment proposes a method to improve the accuracy of fracture prediction in RS-FD numerical analysis using the Bayes algorithm. To address the current issues of missed or incorrect fracture predictions when using conventional logging curves as the basis for fracture prediction via RS-FD numerical analysis, and the inherent subjectivity in selecting the optimal fracture sensitivity curve after performing RS-FD calculations on multiple conventional logging curves, this method provides a solution that prioritizes processing each conventional logging curve in the target formation before performing RS-FD numerical analysis. This clarifies the fracture response location of each curve from the perspective of RS-FD numerical calculation. Then, it further processes core and imaging logging data... This method involves statistically analyzing the fracture development locations of individual wells based on data, and combining this with RS-FD numerical analysis results to determine the fracture identification rate of each curve, thereby establishing the prior probability and likelihood of Bayesian calculations. Then, according to the Bayesian calculation rules, the Bayesian posterior probability of each curve is calculated. Based on the posterior probability, the optimal fracture sensitivity curve is selected, a comprehensive fracture evaluation index is constructed, and the fracture identification accuracy is verified. Finally, based on the comprehensive fracture evaluation index, the fracture development locations of the well to be predicted are calculated. Combining well history, single-well production, venting and leakage data, and fracture identification accuracy, the fracture development locations of the single well are accurately predicted. This method provides a practical and feasible solution for professionals researching single-well fracture prediction using conventional logging.

[0111] In one specific embodiment, this example describes a method for predicting the location of fracture development in a single well in a carbonate reservoir by using the Bayes algorithm to improve the accuracy of fracture prediction in RS-FD numerical analysis.

[0112] First, a detailed survey of the distribution of coring wells and the deployment of imaging logging in the study area revealed that only one well in the area possesses both coring data from the target formation and continuous imaging logging data within that formation. The remaining wells are either single-wells with only coring data from the target formation or single-wells with imaging logging deployed within the same formation. In the practical application of the Bayes algorithm to improve the accuracy of fracture prediction using RS-FD numerical analysis, RS-FD numerical analysis should be performed on all these wells to determine the fracture identification rate of different conventional logging curves and to optimize the fracture sensitivity curve using Bayes calculations. This embodiment only uses the single well in the study area that simultaneously possesses both coring data from the target formation and continuous imaging logging data within that formation as an example.

[0113] Through the collection and analysis of conventional logging data in the study area, it was found that each well in the study area simultaneously possessed eight conventional logging curves in the target interval: AC (acoustic transit time), DEN (density), CAL (well diameter), GR (natural gamma), RS (shallow lateral resistivity), RD (deep lateral resistivity), SP (spontaneous potential), and CNL (neutron). Therefore, it is necessary to process these eight conventional logging data in the study area first, removing outliers and standardizing units. Then, RS-FD numerical analysis calculations of the eight conventional logging curves were performed on a single well in this embodiment. After logarithmically calculating the R / S ratio of each curve and the corresponding number of scales u, a rectangular coordinate system was established. Then, the second derivative of the R / S value and the corresponding number of scales u were logarithmically calculated and a rectangular coordinate system was established. Finally, the two graphs were merged with the same x-axis to create a double Y-axis graph. The fracture response points of the R / S calculation results and the fracture sensitive points of the second derivative of the R / S value were observed to determine the fracture response points of each conventional logging data. Analysis of the fracture response locations on the curves revealed that, in the example, the conventional logging curves for a single well in this study identified 9 fracture response locations for RS (shallow lateral resistivity), 8 for AC (acoustic transit time), 8 for DEN (density), and 8 for RD (deep lateral resistivity), 6 for CNL (neutron), 4 for SP (spontaneous potential), and 3 for CAL (caliber), while GR (natural gamma) showed no fracture response locations. Figure 2 As shown.

[0114] Statistical analysis revealed that although the single well in this embodiment contained coring in the target formation, the coring was discontinuous and short, with only 3 fractures identified at the core scale. Imaging logging, on the other hand, was continuous and complete in the target formation, identifying a total of 21 fractures (including 3 at the core scale). However, the imaging logging showed a phenomenon of multiple fractures appearing densely together. In this case, the numerous densely developed fracture zones were combined into a single fracture segment for statistical analysis, resulting in 4 fracture development segments. Two of these segments contained 6 fractures each, one contained 2 fractures, one contained 3 fractures, and the remaining 4 were individual fractures (including 2 at the core scale). Figure 3 As shown. Matching the fracture development intervals statistically obtained from core and imaging logging layers with the fracture response locations identified by conventional logging curves calculated using the RS-FD numerical analysis method reveals the following fracture identification rates: AC (acoustic transit time) 0.75, RD (deep lateral resistivity) 0.75, RS (shallow lateral resistivity) 0.78, CNL (neutron) 0.50, DEN (density) 0.75, and SP (spontaneous potential)... The fracture identification rate for ( ) was 0.50, and for CAL (well caliper) it was 0.33; the likelihoods for AC (acoustic transit time) were 0.093, RD (deep lateral resistivity) were 0.093, RS (shallow lateral resistivity) were 0.098, CNL (neutron) was 0.063, DEN (density) was 0.093, SP (spontaneous potential) was 0.063, and CAL (well caliper) was 0.041; the total probability was 0.565. Figure 4 As shown. After calculating the posterior probability using the prior probability, likelihood, and total probability of each conventional logging curve, the posterior probability and prior probability of each conventional logging curve are compared to optimize the fracture sensitivity curve. In this embodiment, four fracture sensitivity curves are optimized: AC (acoustic transit time) posterior probability is 16.46%, DEN (density) posterior probability is 16.46%, RD (deep lateral resistivity) posterior probability is 16.46%, and RS (shallow lateral resistivity) posterior probability is 17.35%, all of which are greater than the prior probability of 12.50%. The second derivative of the R / S values ​​of the four conventional logging curves is multiplied and then normalized to construct fracture sensitivity curve I. By re-matching with the fracture development intervals statistically obtained from the core layer and imaging logging layers of the well, the accuracy of the constructed fracture comprehensive evaluation index is verified to be 70.00%, which is a high accuracy. Figure 5 As shown. Therefore, the AC (acoustic transit time), DEN (density), RD (deep lateral resistivity), and RS (shallow lateral resistivity) of the well to be predicted are also processed by the second derivative of the R / S value, and a fracture sensitivity curve I is constructed to carry out single-well fracture prediction of the well to be predicted.

[0115] This embodiment describes a method for predicting fracture development locations in single wells using conventional logging in oil and gas exploration and development. It provides a method that first processes conventional logging curves from the target formation and then performs RS-FD numerical analysis to identify the fracture response locations of each curve from an RS-FD numerical perspective. Next, for single wells containing core and imaging logging data, a statistical analysis of fracture development locations is conducted. Combined with the RS-FD numerical analysis results, the fracture identification rate of each curve is determined to establish the prior probability and likelihood of Bayesian calculations. Then, based on Bayesian calculation rules, the Bayesian posterior probability of each curve is calculated. Based on the posterior probability, the fracture sensitivity curve is optimized, a comprehensive fracture evaluation index is constructed, and the fracture identification accuracy is verified. Finally, based on the comprehensive fracture evaluation index, the method for predicting fracture development locations in the well to be predicted is implemented. This patent effectively solves the problems of significant missed or misjudged fractures when using conventional logging curve data as basic data and employing RS-FD numerical analysis for fracture prediction, as well as the problem of strong human subjectivity when optimizing fracture sensitivity curves after performing RS-FD calculations on multiple conventional logging curves.

[0116] This embodiment primarily provides a method for professionals using conventional logging to predict fractures in single wells. It involves prioritizing the processing of conventional logging curves in the target formation before performing RS-FD numerical analysis to identify the fracture response locations of each curve from an RS-FD numerical perspective. Next, for single wells containing core and imaging logging data, a statistical analysis of fracture development locations is conducted. Combined with the RS-FD numerical analysis results, the fracture identification rate of each curve is determined to establish the prior probability and likelihood of Bayesian calculations. Then, based on Bayesian calculation rules, the Bayesian posterior probability of each curve is calculated. Based on the posterior probability, the fracture sensitivity curve is optimized, a comprehensive fracture evaluation index is constructed, and the fracture identification accuracy is verified. Finally, based on the comprehensive fracture evaluation index, a scientific method for predicting the fracture development locations of the well to be predicted is implemented. This method effectively avoids the problem of missed or incorrect fracture predictions that often occur when using RS-FD numerical analysis for fracture prediction based on a single logging curve. It also better addresses the inherent subjectivity in selecting the optimal fracture sensitivity curve after performing RS-FD calculations on multiple conventional logging curves. This provides a practical solution for professionals using conventional logging for single-well fracture prediction in oil and gas geological exploration and development.

[0117] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A method for improving crack prediction accuracy in RS-FD numerical analysis using the Bayes algorithm, characterized in that, The method includes: S1: Processing of conventional logging curves of single wells in the target formation, RS-FD numerical analysis and calculation, establishment of a rectangular coordinate system, and identification of fracture response locations of each curve; S2: Based on core and imaging logging data, statistically analyze the fracture development locations in single wells, clarify the fracture identification rate of each curve, and determine the prior probability, likelihood, and total probability calculated by Bayesian methods. S3: Based on the Bayesian computational rule, calculate the Bayesian posterior probability of each curve, select the optimal fracture sensitivity curve, construct the comprehensive fracture evaluation index I, verify the accuracy of fracture identification, and predict the fracture development location of the single well in the well to be predicted. Step S3 includes: S31: Based on Bayesian computation rules, perform Bayesian posterior probability analysis on each conventional well logging curve. Calculation; S32: Based on the Bayesian posterior probability of each conventional logging curve obtained by calculation, the fracture sensitivity curve is selected first; S33: Construct a comprehensive crack evaluation index I based on the selected crack sensitivity curve as the basic parameter; S34: By rematching the fracture development intervals obtained from core layer and imaging logging layer statistics, the accuracy of the constructed comprehensive fracture evaluation index is verified, and the single-well fracture development location of the well to be predicted is completed. Step S31 includes: after calculating the prior probability, likelihood, and total probability of each conventional logging curve using the Bayesian algorithm, calculating the Bayesian posterior probability of each conventional logging curve according to the Bayesian calculation rule. The formula is: Step S32 includes: The principle for selecting the optimal fracture sensitivity curve is to compare the posterior probability and prior probability calculated using the Bayesian algorithm for each conventional logging curve. Curves with a posterior probability exceeding the prior probability are considered fracture sensitive, while those with a posterior probability less than the prior probability are considered not to have fracture sensitivity. The posterior probabilities calculated using the Bayesian algorithm for each conventional logging curve are sorted from largest to smallest. Curves with a higher posterior probability than the prior probability have stronger fracture sensitivity and a stronger fracture prediction ability. Conversely, conventional logging curves with a posterior probability greater than the prior probability but closer to the prior probability have weaker fracture sensitivity and a poorer fracture prediction ability. A comprehensive evaluation is performed by comparing the fracture sensitivity of each conventional logging curve on different wells. If the sensitivity of the same curve differs significantly across different wells, that curve is directly discarded.

2. The method according to claim 1, characterized in that, Step S1 includes: S11: Identify the target layer for fracture prediction in a single well and process the conventional logging curves of each well in the target layer; The data contained in all conventional logging curves outside the target formation are removed. The processing of conventional logging curves within the target formation includes: unifying the units of all types of logging curves from all single wells in the study area; checking the data distribution of each conventional logging curve in the target formation; directly assigning a value of 0 to abnormal data, including: 0, -999.25, -9999.00, and blank values; accurately determining the reasonable range of data distribution for each conventional logging curve based on the lithological distribution of the target formation; judging the data of conventional logging curves that exceed the reasonable range, and if they are abnormal data, directly assigning a value of 0. S12: Based on the RS-FD numerical analysis method, calculate each conventional logging curve in the target layer, establish a rectangular coordinate system, and identify the fracture response location of each curve; Based on the RS-FD numerical analysis method, the calculations were performed on each conventional logging curve in the target layer, and the fracture response locations of different conventional logging curves were identified. Where, n is the total number of data points in the target layer of each conventional logging curve; u is the number of scales that increase sequentially from 0 to n in the target layer of each conventional logging curve starting from the endpoint; Z(i) and Z(j) represent the selected logging curve data; i and j represent the variable of the number of sample points; R(n) represents the range; S(n) represents the standard deviation of the entire process sequence. The second derivative of the R / S value is represented by h, which is the calculation step size. The method for determining the fracture response location of different conventional logging curves is as follows: After calculating the R / S ratio of each conventional logging curve and obtaining the logarithm of the corresponding number of scales u, a rectangular coordinate system is established with lg(R / S) as the ordinate and lg(u) as the abscissa. A scatter plot of the relationship between the two is drawn for each conventional logging curve. The location of the abrupt change in slope in the scatter plot is the fracture response point. A rectangular coordinate system is established with the second derivative of R / S as the ordinate and lg(u) as the abscissa. A bar chart of the relationship between the two is drawn for each conventional logging curve. The abnormally high values ​​in the bar chart are the fracture sensitive points. The two charts are merged along the X-axis to create a double Y-axis chart. The matching between the fracture response points and the fracture sensitive points is observed to determine the fracture response location of each conventional logging curve.

3. The method according to claim 1, characterized in that, Step S2 includes: S21: Conduct statistical analysis of the fracture development locations in the core samples from the target formation, and identify the fracture development zones for each formation. S22: Conduct statistical analysis of fracture development locations in single wells with imaging logging deployed, and identify the fracture development zones of each well. S23: Match the fracture development intervals obtained from the core layer and imaging logging layer with the fracture response locations identified by each conventional logging curve after calculation based on the RS-FD numerical analysis method, and clarify the fracture identification rate of each conventional logging curve; S24: Based on the fracture identification rate of each conventional logging curve, determine the prior probability, likelihood, and total probability of each conventional logging curve calculated by Bayes.

4. The method according to claim 3, characterized in that Step S21 includes: when conducting statistics on the development of fractures in a single well in the target layer of the core well, before core observation, determining the top depth, bottom depth and core recovery rate of the core well section, and then resetting all the cores in sequence, measuring the top and bottom depths corresponding to the fracture development locations of the cores with developed fractures, in order to determine the fracture length and development range. Step S22 includes: statistical analysis of fracture development locations in the target formation of imaging logging; observation of dynamic and static images from imaging logging; statistical analysis of natural fractures with sine and cosine curve characteristics; and exclusion of induced fractures with a goose-like arrangement caused by drilling tools in the longitudinal direction. The statistical analysis of natural fractures clarifies the top and bottom depths of their development locations, indicating the longitudinal influence range of the fractures.

5. The method according to claim 3, characterized in that, Step S23 includes: if imaging logging is deployed in the core well, the fractures identified by the imaging logging are compared and merged with those identified by the core well to completely identify all fractures in the target layer; then, the fracture response locations identified by each conventional logging curve calculated based on the RS-FD numerical analysis method are matched to calculate the fracture identification rate of each conventional logging curve. If no imaging logging is deployed in the core well, the fracture information obtained by statistical analysis of the core layer is matched with the fracture response locations identified by each curve after calculation based on the RS-FD numerical analysis method, and the fracture identification rate of each conventional logging curve is calculated. If a single well does not have coring and only imaging logging is deployed, the fracture information obtained by imaging logging layer statistics is matched with the fracture response locations identified by each curve after calculation based on RS-FD numerical analysis method, and the fracture identification rate of each conventional logging curve is calculated. When matching the fracture information obtained from core logging and imaging logging with the fracture response locations identified by conventional logging curves calculated using the RS-FD numerical analysis method, it is clear that the fracture information obtained from core logging and imaging logging is the accurate fracture development location and fracture length information of a single fracture in the longitudinal direction of a single well, which belongs to a certain range. If the fracture response locations identified by conventional logging curves calculated using the RS-FD numerical analysis method are within the range of accurate fracture development locations obtained from core logging and imaging logging, then a fracture is considered to have been identified.

6. The method according to claim 3, characterized in that, Step S24 includes: When performing Bayesian calculations using various conventional well logging curves, the prior probability... This should be the probability of randomly selecting any conventional logging curve and performing Bayesian calculations. The prior probability is calculated based on the number of conventional logging curves in the actual situation, and the formula is: Its likelihood It should be the fracture identification rate P of any randomly selected conventional logging curve. k Its formula is: Its total probability It should be the sum of the probabilities that each conventional logging curve is randomly selected for subsequent fracture prediction in actual practice, and its formula is: Where n is the total number of conventional logging curves; i is the type of each conventional logging curve.

7. The method according to claim 1, characterized in that, Step S33 includes: establishing a comprehensive crack evaluation index I based on actual optimization conditions, the formula of which is: Among them, I Z , is the product of the R / S-FD calculation results of the optimized fracture sensitivity curves; n, is the total number of conventional logging curves under the actual optimized conditions; By applying normalization, the data is processed to facilitate data analysis. The formula is as follows: Among them, I Zn , represents each value of the product of the optimized crack sensitivity curve R / S-FD calculation results; n, represents the total number of the optimized crack sensitivity curve summation results; A direct coordinate system is established with each value of the comprehensive crack evaluation index I as the vertical axis and the values ​​corresponding to the order of each value as the horizontal axis. A cumulative curve is then created. The upper and lower ends of the crack development area are shown as a sudden increase in the Y value on the cumulative curve. After exceeding the crack development area, the Y value change quickly returns to the normal growth rate. In addition, the middle of the crack will show a local maximum value of Y value on the cumulative curve. The crack development area on the comprehensive crack evaluation index I curve is thus determined.

8. The method according to claim 1, characterized in that, Step S34 includes: The accuracy of the current fracture comprehensive evaluation index I curve in predicting fractures in a single well is verified by rematching the fracture development locations predicted on the fracture comprehensive evaluation index I curve with the fracture development segments obtained from the core layer and imaging logging layer. When the accuracy of single-well fracture prediction reaches the prediction standard of the study area, the fracture comprehensive evaluation index I curve is used to carry out single-well fracture prediction. If it does not reach the prediction standard of the study area, the curve with the lowest posterior probability in the fracture sensitivity curve is removed, and the fracture comprehensive evaluation index I curve is reconstructed. When the constructed fracture comprehensive evaluation index I curve has reached the prediction standard of the study area, several conventional logging curves that are the same as the constructed fracture comprehensive evaluation index I curve in the conventional logging curves of the well to be predicted are selected in a targeted manner to calculate R / S-FD. After calculation, the fracture comprehensive evaluation index I curve of the well to be predicted is directly constructed by applying the calculation formula of fracture comprehensive evaluation index I, so as to carry out the prediction of the fracture development location of the well to be predicted.

Images