Shale platinum box artificial intelligence identification method based on element logging

Abstract

This invention discloses an artificial intelligence identification method for shale platinum box formations based on elemental logging, comprising the following steps: Step S1, extracting elemental curves from horizontal well logging while drilling; Step S2, designing the structure of a quantitative identification method for shale platinum box formations; Step S3, designing the application flow of the quantitative identification method for shale platinum box formations. This invention's artificial intelligence identification method for shale platinum box formations based on elemental logging can automate the identification of shale sub-layers, improve the objectivity and reliability of the identification results, and enhance the resolution of shale sub-layer identification.

Description

Technical Field

[0001] This invention relates to the field of geological steering technology for horizontal well drilling, and in particular to an artificial intelligence identification method for shale platinum box based on elemental logging. Background Technology

[0002] Shale oil and gas, as an important unconventional oil and gas resource, has relatively thin high-quality sub-layers. Therefore, during horizontal well drilling of shale oil and gas, it is necessary to determine the relative position of the current well bottom and the high-quality shale sub-layers in real time in order to adjust the well trajectory in a timely manner and improve the drilling rate of high-quality shale oil and gas reservoirs. In engineering, high-quality shale sub-layers are defined as platinum boxes, and during horizontal well drilling, it is determined in real time whether the current well bottom is inside the platinum box. Therefore, the identification of shale platinum boxes can be transformed into identifying the shale sub-layer to which the well bottom belongs. Currently, the main method for identifying the shale sub-layer to which the well bottom belongs is: on-site engineers analyze the characteristics of each shale sub-layer on the elemental logging curves of adjacent completed horizontal and vertical wells, and qualitatively or quantitatively establish the elemental characteristic patterns of each shale sub-layer, which serve as the identification basis for the current horizontal well drilling.

[0003] It is evident that current technology primarily relies on manual analysis. However, manual analysis struggles to simultaneously capture the changing characteristics of all element curves. Furthermore, due to measurement errors and improper on-site worker procedures, element content curves often contain noise, making manual analysis susceptible to its influence. On the other hand, manual analysis has low resolution, making it difficult to identify thinner layers from elemental logging curves, and even more difficult to identify shale sub-layers at each sampling point.

[0004] In summary, the current method of identifying shale sub-layers using elemental logging is subjective, with low reliability and resolution of the identification results, and a low degree of automation in the identification process. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the existing technology and provide an artificial intelligence identification method for shale platinum box formations based on elemental logging. This invention uses a deep neural network as the core for shale sub-layer identification and a hierarchical progressive calculation method for accurately locating shale sub-layers, designing an artificial intelligence identification method for shale platinum box formations based on elemental logging. This method automates the shale sub-layer identification process, improves the objectivity and reliability of the identification results, and enhances the resolution of shale sub-layer identification.

[0006] The solution of the present invention is as follows:

[0007] An AI-based method for identifying platinum shale enclosures based on elemental logging includes the following steps:

[0008] Step S1: Extract the logging-while-drilling element curves of the horizontal well;

[0009] Step S2: Design the structure of a quantitative identification method for shale platinum boxes;

[0010] Step S3: Design the application process of the quantitative identification method for shale platinum boxes.

[0011] Specifically, step S2 includes the following sub-steps:

[0012] Step S21: Design the structure of the three-factor orthogonal method for quantitative identification of shale platinum boxes;

[0013] Step S22: Evaluate the generalization ability of the shale platinum box quantitative identification algorithm for vertical wells to horizontal wells;

[0014] Step S23: Optimize the structure of the quantitative identification method for shale platinum boxes.

[0015] Specifically, the structure in step S21 is a neural network structure, including a framework, a number of modules, and activation functions;

[0016] The framework includes convolutional neural networks, recurrent neural networks, or perceptron neural networks;

[0017] The number of modules includes 2, 3, or 4;

[0018] The activation function includes ReLU, LeakyReLU, or Tanh.

[0019] Specifically, step S22 uses the sensitivity element curves of the pilot well corresponding to the adjacent completed horizontal well and the identification results of the shale platinum box to verify the accuracy of the predicted data, i.e. the generalization performance of the method, as the evaluation index to select structural design schemes with excellent performance.

[0020] Preferably, the high-performance structural design is a recurrent neural network with three recurrent layers, and the activation function is Tanh or LeakyReLU.

[0021] Specifically, the optimization of step S23 includes introducing the Dropout method to achieve positive feedback.

[0022] Specifically, step S3 includes the following sub-steps:

[0023] Step S31: Design a hierarchical and progressive calculation route for the quantitative identification method of shale platinum boxes;

[0024] Step S32: Design a three-stage quantitative identification method to predict the route, consisting of "pilot well - landing section - horizontal section".

[0025] Specifically, step S31 involves predicting the position of the platinum box at the next level based on the prediction results of the previous level, thus forming a progressive calculation method.

[0026] Specifically, step S32 is as follows:

[0027] For pilot wells, after drilling is completed, detailed platinum box identification results are obtained by logging curves. The optimized identification method is used to train multiple identification method models with different subdivision levels according to the hierarchical progressive calculation route and identification accuracy requirements.

[0028] For the landing segment, the location of the platinum box is predicted using multiple identification method models trained through the pilot well. The prediction results are then used together with the identification results from the pilot well to train multiple identification method models at different subdivision levels, thereby identifying the platinum box in the horizontal segment.

[0029] The beneficial effects of this invention are:

[0030] The invention presents an artificial intelligence identification method for shale platinum box based on element logging, which can automate the identification of shale sub-layers, improve the objectivity and reliability of the identification results, and enhance the identification resolution of shale sub-layers. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0032] Figure 1 This is a technical roadmap of the present invention;

[0033] Figure 2 This is a schematic diagram of the sensitive element curves of the horizontal well corresponding to the pilot well in the present invention;

[0034] Figure 3 This is a structural diagram of the method for identifying shale platinum box structures in horizontal wells, as described in this invention.

[0035] Figure 4 This is the hierarchical and progressive calculation roadmap of the present invention;

[0036] Figure 5 This is a prediction roadmap for the three-stage quantitative identification method of the present invention;

[0037] Figure 6 This is a diagram showing the identification results of the horizontal well platinum box in this invention. Detailed Implementation

[0038] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0039] To provide a clearer understanding of the technical features, objectives, and beneficial effects of this invention, the following detailed description of the technical solution is provided. Obviously, the described embodiments are only a portion of the embodiments of this invention, not all of them, and should not be construed as limiting the scope of implementation of this invention. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort are within the protection scope of this invention.

[0040] like Figure 1 As shown, this invention designs an artificial intelligence identification method for shale platinum box formations based on elemental logging. The method first analyzes the sensitivity of all elemental logging curves to changes in shale sub-layers, extracting sensitive elemental curves. Second, it designs a deep neural network structure using a three-factor orthogonal method, considering layer type, number of layers, and number of layer nodes. Based on the deep neural networks formed by different structural design schemes, it learns the elemental curves and platinum box distribution characteristics of vertical wells, and evaluates the performance of different design schemes by assessing their generalization ability to horizontal wells, thereby selecting the optimal structural design scheme. The network performance is then optimized by combining different deep neural network functional modules. Finally, a hierarchical, progressive approach is adopted to design the computational route of the identification algorithm. Based on the actual horizontal well drilling process, a three-stage prediction route for the quantitative identification algorithm of shale platinum boxes—"pilot well – horizontal well landing section – horizontal well horizontal section"—is designed. This quantitative identification method for shale platinum boxes automates shale sub-layer identification, improves the objectivity and reliability of the identification results, and enhances the resolution of shale sub-layer identification.

[0041] 1. Extraction of element curves from logging while drilling in horizontal wells

[0042] Based on the identification results of shale platinum box formations in adjacent completed horizontal wells, the sensitivity of each element curve to the identification of shale platinum box formations was analyzed. Sensitive elements were selected as the data basis for identification, and the effectiveness of the sensitive elements was verified using the pilot well corresponding to the horizontal well to be drilled.

[0043] The case study horizontal well is located in a shale gas block in Luzhou, southern Sichuan. By comparing the element curves of adjacent completed horizontal wells with the identification results of the platinum box, nine relatively sensitive element curves were identified: aluminum, silicon, calcium, thorium, uranium, magnesium, nickel, zinc, and manganese. Figure 2 The variation characteristics of these nine sensitive element curves on the pilot well corresponding to the horizontal well to be drilled are shown. Obviously, these curves have obvious variation characteristics at the interface of each sub-layer, and also have obvious numerical differences within the sub-layer. Therefore, it is feasible to use these nine sensitive element curves to carry out the identification of shale platinum box during the horizontal well drilling process.

[0044] 2. Structural Design of Quantitative Identification Algorithm for Shale Platinum Boxes

[0045] The algorithm structure was designed from three aspects: framework, number of modules, and activation function using the three-factor orthogonal method. The verification route of the structure scheme was determined. The generalization of vertical well to horizontal well shale platinum box identification was used as the evaluation criterion. The performance of each structure design scheme was compared and the structure design scheme with the best generalization performance was selected. The functional modules were optimized by integrating the algorithm performance to form a deep learning shale platinum box identification algorithm.

[0046] ① Algorithm structure design for quantitative identification of shale platinum box using the three-factor orthogonal method

[0047] Starting from three levels of factors—the framework used in the algorithm, the number of modules corresponding to the framework, and the activation function—the structure of the algorithm is designed to identify the algorithm. Based on the degree of influence of each factor on the algorithm's performance, the design scheme verification route is determined.

[0048] Factors affecting neural network performance include: framework, number of modules, number of nodes in each module, activation function, optimizer, training parameters, and data preprocessing methods. Among these, framework, number of modules, and activation function are selected as the three most important factors for structural design. For the neural network algorithm framework, three main types are considered: convolutional neural networks, recurrent neural networks, and perceptron neural networks. For the number of modules, three levels are considered: 2, 3, and 4. For activation functions, ReLU, LeakyReLU, and Tanh are considered. Directly comparing twenty-seven structural design schemes with these three factors and three levels significantly increases the overall time cost; therefore, orthogonal design is used to formulate structural design schemes. Based on the importance of their impact on neural network performance, the performance of the three basic frameworks is first compared. After determining the framework, the impact of the number of modules on neural network performance is further compared. Finally, with the framework and number of modules clearly defined, the activation function type is determined.

[0049] ② Evaluation of the generalization ability of the shale platinum box quantitative identification algorithm for vertical wells to horizontal wells

[0050] Using the sensitive element curves of pilot wells corresponding to adjacent completed horizontal wells and the identification results of shale platinum box as training data, and the sensitive element curves and identification results of completed horizontal wells as prediction data, the identification algorithm is trained based on the training data. The verification accuracy of the prediction data, i.e. the generalization performance of the algorithm, is used as the evaluation index to select structural design schemes with excellent performance.

[0051] As shown in Table 1, to compare the performance of the neural network under each scheme, the data of pilot well elements corresponding to nearby completed wells and the platinum box identification results were used as training data to examine the generalization ability of different schemes to horizontal wells, which was used as a screening criterion. The verification results for the structural scheme designed in ① are shown in Table 1. It can be seen that when a recurrent neural network is used as the basic framework of the network, and three recurrent layers are constructed, with the activation function of each recurrent layer being either Tanh or LeakyReLU, the performance of the identification algorithm reaches its maximum. Although the performance of the neural network using LeakyReLU is slightly lower than that using Tanh, the difference is small, and if the average of the maximum values ​​is taken multiple times, the two are comparable.

[0052] Table 1 Performance Comparison of Horizontal Well Structure Design Schemes in Case Studies

[0053] Algorithm framework Number of modules loss function accuracy Convolutional Neural Networks 3 ReLU 80.32％ Recurrent Neural Networks 3 ReLU 84.79％ Multilayer perceptron 3 ReLU 84.04％ Recurrent Neural Networks 2 ReLU 84.57％ Recurrent Neural Networks 4 ReLU 83.11％ Recurrent Neural Networks 3 LeakyReLU 85.45％ Recurrent Neural Networks 3 Tanh 84.91％

[0054] ③ Optimization of the quantitative identification algorithm structure for shale platinum boxes

[0055] By comparing the improvement effect of commonly used generalization performance optimization functional modules on the algorithm, appropriate functional modules are embedded into the algorithm to form the final shale platinum box identification algorithm.

[0056] Building upon step ②, this study compares the performance changes of the algorithm when introducing adjustment modules such as Dropout, residual connections, and inter-layer normalization into the neural network. This allows for the selection of adjustment modules that provide positive feedback to the algorithm's performance, resulting in the final shale platinum box identification algorithm. The comparative results show that the Dropout module slightly alters the identification algorithm's performance, while the other modules have no significant impact on performance. The final identification algorithm structure is as follows: Figure 3 As shown.

[0057] 3. Application Flow Design of Quantitative Identification Algorithm for Shale Platinum Boxes

[0058] Based on the specified subdivision level, the platinum box is subdivided level by level. For each subdivision level, a separate identification algorithm is trained to perform prediction, forming a hierarchical and progressive identification algorithm calculation route. According to the on-site horizontal well drilling procedure, hierarchical and progressive identification algorithm calculations are carried out for the pilot well, the horizontal well landing section, and the horizontal section of the horizontal well, forming a three-stage shale platinum box prediction.

[0059] ① Design of a hierarchical and progressive calculation route for the quantitative identification algorithm of shale platinum boxes

[0060] Based on actual engineering needs, the subdivision levels of shale platinum boxes are determined, and a separate identification algorithm is trained for each subdivision level. During prediction, the corresponding identification algorithm is used to calculate the bottom platinum box position at the current subdivision level, from coarse to fine, in order to obtain a more precise bottom platinum box position.

[0061] In actual horizontal well drilling, determining the current bottom hole position typically requires precision down to the upper, middle, and lower positions within a specific layer. For complex well sections, the accuracy requirements are even higher. Therefore, a progressively subdivided prediction process was designed. One identification algorithm predicts the current bottom hole layer, which can be considered a first-level subdivision of the platinum box. Based on this subdivision, another identification algorithm predicts the upper, middle, and lower positions within that subdivision (a second-level subdivision of the platinum box). This process continues, using multiple identification algorithm models to progressively subdivide the platinum box. Based on the prediction results of the previous level, the position of the next level of platinum box is predicted, forming a progressive algorithmic calculation route. Figure 4 As shown.

[0062] ② Quantitative identification algorithm for route prediction design in the three stages of "pilot well - landing section - horizontal section"

[0063] A hierarchical and progressive identification algorithm training was conducted for pilot wells. The trained algorithm was used to predict the landing section of the horizontal well, and the prediction results were added to the training set. The identification algorithm was further trained hierarchically, and the trained algorithm was used to predict the platinum box of the horizontal section of the horizontal well.

[0064] Before drilling a horizontal well, a pilot well is usually drilled as a basis for evaluating the platinum box range, high-quality reservoir distribution characteristics, and small-layer permeability performance of the horizontal well. During the drilling of the horizontal well, the well trajectory usually goes through the build-up section, landing section, and horizontal section. The build-up section is far from the layers covered by the platinum box, so platinum box prediction work is usually only carried out for the landing section and the horizontal section.

[0065] For pilot wells, detailed platinum box identification results can be obtained through logging curves after drilling completion. The optimized identification algorithm is then used according to… Figure 4 The process and recognition accuracy requirements shown illustrate the step-by-step training of multiple recognition algorithm models at different subdivision levels; for example... Figure 5 As shown, for the landing section of the horizontal well, multiple identification algorithm models trained through the pilot well are used to predict the location of the platinum box. The prediction results and the identification results of the pilot well are used to train multiple identification algorithm models at different subdivision levels, thereby identifying the platinum box in the horizontal section.

[0066] After the above three technical steps, the platinum box identification result of the horizontal well can be obtained. The final platinum box identification result of the horizontal well in this case is as follows: Figure 6 As shown, the well has been subdivided into two levels, with the platinum box containing both the first and second sub-layers divided into two parts, A and B.

[0067] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

[0068] It should be noted that, for the sake of simplicity, the aforementioned method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, because according to this application, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and units involved are not necessarily essential to this application.

[0069] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0070] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, ROM, RAM, etc.

[0071] The above description discloses only preferred embodiments of the present invention and should not be construed as limiting the scope of the present invention. Therefore, equivalent variations made in accordance with the claims of the present invention are still within the scope of the present invention.

Claims

1. A method for AI-based identification of shale platinum enclosures based on elemental logging, characterized in that, Includes the following steps: Step S1: Extract the logging-while-drilling element curves of the horizontal well; Step S2, design the structure of a quantitative identification method for shale platinum boxes; step S2 includes the following sub-steps: Step S21: Design the structure of the three-factor orthogonal method for quantitative identification of shale platinum box; Step S22: Evaluate the generalization of the shale platinum box quantitative identification algorithm from vertical wells to horizontal wells; Step S23: Optimize the structure of the shale platinum box quantitative identification method. The structure described in step S21 is a neural network structure; Step S3, design the application process of the quantitative identification method for shale platinum boxes; step S3 includes the following sub-steps: Step S31: Design a hierarchical and progressive calculation route for the quantitative identification method of shale platinum boxes; Step S32: Design a three-stage quantitative identification method to predict the route, consisting of "pilot well - landing section - horizontal section"; Step S31 specifically involves predicting the position of the next level of platinum box based on the prediction results of the previous level of subdivision, forming a progressive calculation method. Specifically, it adopts a hierarchical progressive prediction process of step-by-step subdivision, that is: using an identification algorithm to predict the current bottom layer, and the layer can be regarded as a first-level subdivision of the platinum box. Based on the determination of the layer, another identification algorithm is used to predict the upper, middle and lower positions inside the layer. In this way, multiple identification algorithm models are used to subdivide the platinum box step by step, and the position of the next level of platinum box is predicted based on the prediction results of the previous level of subdivision, forming a progressive algorithm calculation method. Step S32 specifically involves: For pilot wells, after drilling is completed, detailed platinum box identification results are obtained by logging curves. The optimized identification method is used to train multiple identification method models with different subdivision levels according to the hierarchical progressive calculation route and identification accuracy requirements. For the landing segment, the location of the platinum box is predicted using multiple identification method models trained through the pilot well. The prediction results are then used together with the identification results from the pilot well to train multiple identification method models at different subdivision levels, thereby identifying the platinum box in the horizontal segment.

2. The artificial intelligence identification method for shale platinum box based on elemental logging according to claim 1, characterized in that, The neural network structure includes the framework, the number of modules, and the activation function; The framework includes convolutional neural networks, recurrent neural networks, or perceptron neural networks; The number of modules includes 2, 3, or 4; The activation function includes ReLU, LeakyReLU, or Tanh.

3. The artificial intelligence identification method for shale platinum box based on elemental logging according to claim 2, characterized in that, Step S22 uses the sensitivity element curves of the pilot well corresponding to the adjacent completed horizontal well and the identification results of the shale platinum box to verify the accuracy of the predicted data, i.e. the generalization performance of the method, as the evaluation index, and selects a structural design scheme with excellent performance; the structural design scheme with excellent performance is a recurrent neural network with 3 recurrent layers and the activation function is Tanh or LeakyReLU.

4. The artificial intelligence identification method for shale platinum box based on elemental logging according to claim 3, characterized in that, The optimization of step S23 includes introducing the Dropout method to achieve positive feedback.

Images