Method and device for determining corrosion rate of pipe of ultra-deep well

By using preset regression model and multi-submodule working together in ultra-deep wells, the problem of low corrosion rate prediction accuracy of ultra-deep well pipes is solved, and more accurate corrosion rate prediction and stronger generalization capabilities are achieved, ensuring that Safe and efficient operation of oil and gas wells.

CN119069049APending Publication Date: 2024-12-03HAINAN BRANCH OF CHINA NATIONAL OFFSHORE OIL (CHINA) CO LTD +1

Patent Information

Application Number
CN202411279703.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-12
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

The existing technology has failed to effectively solve the corrosion problems of downhole equipment caused by high pressure, high temperature and high salinity in ultra-deep well environments, and lacks accurate corrosion rate prediction methods, which affects the safe and efficient operation of oil and gas wells.

Method used

The material characteristic data of ultra-deep well pipes is processed using a preset regression model. Through the coordinated work of multiple submodules and the weighted parameter weighting merge, the accurate corrosion rate is predicted, which improves the accuracy of corrosion rate prediction of low alloy pipes.

Benefits of technology

It improves the accuracy and generalization ability of corrosion rate prediction of ultra-deep well pipes, enhances the reliability and stability of corrosion conditions, and ensures the safe and efficient operation of oil and gas wells.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119069049A_ABST
    Figure CN119069049A_ABST
Patent Text Reader

Abstract

The invention provides a method and device for determining the corrosion rate of an ultra-deep well pipe. Acquiring material characteristic data of the to-be-detected ultra-deep well pipe; acquiring material characteristic data of a to-be-detected target ultra-deep well pipe; respectively processing the material characteristic data of the target ultra-deep well pipe by using a plurality of sub-modules included in a first module in a preset regression model to obtain a plurality of initial corrosion rate predicted values respectively corresponding to the plurality of sub-modules; determining a first corrosion rate predicted value according to the plurality of initial corrosion rate predicted values and the weight parameters of the corresponding sub-modules; processing the first corrosion rate predicted value by using a second module in the preset regression model to obtain a corresponding target corrosion rate predicted value; wherein the target corrosion rate predicted value is used for determining the corrosion condition of the target ultra-deep well pipe. And the precision of predicting the corrosion rate of the ultra-deep well low-alloy pipe is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of natural environment corrosion of metal materials, and in particular to a method and device for determining the corrosion rate of ultra-deep well pipes. Background Art

[0002] With the growing global demand for oil and gas resources, ultra-deep well drilling has become an important means to achieve deep oil and gas resource exploration and development. As oil and gas exploration and development advance to deeper levels, environmental factors such as high pressure, high temperature, and high salinity in ultra-deep well environments have made the corrosion problem of downhole equipment more serious. Therefore, accurately predicting the corrosion behavior of ultra-deep wells is of great significance to ensure the safe and efficient operation of oil and gas wells.

[0003] To address the above problems, no effective solution has been proposed yet. Summary of the invention

[0004] The present specification provides a method and device for determining the corrosion rate of ultra-deep well pipes, which uses a preset regression model to process the material characteristic data of the target ultra-deep well pipes to obtain the corresponding target corrosion rate prediction value, thereby improving the accuracy of the corrosion rate prediction of ultra-deep well low-alloy pipes.

[0005] This specification provides a method for determining the corrosion rate of ultra-deep well tubulars, including:

[0006] Acquire material characteristic data of target ultra-deep well pipes to be inspected;

[0007] The material characteristic data of the target ultra-deep well tubular are processed respectively by using the multiple submodules included in the first module of the preset regression model to obtain multiple initial corrosion rate prediction values ​​corresponding to the multiple submodules respectively; wherein the preset regression model is obtained by pre-training a model constructed based on a machine learning algorithm using historical material characteristic data;

[0008] Determine a first corrosion rate prediction value according to the multiple initial corrosion rate prediction values ​​and weight parameters of corresponding submodules; wherein the weight parameters are determined in advance by processing the historical material characteristic data using the preset regression model;

[0009] The first corrosion rate prediction value is processed by the second module in the preset regression model to obtain a corresponding target corrosion rate prediction value; wherein the target corrosion rate prediction value is used to determine the corrosion condition of the target ultra-deep well tubular.

[0010] In one embodiment, the method further includes: processing the material characteristic data of the target ultra-deep well tubular material by using the plurality of submodules included in the first module of the preset regression model to obtain a plurality of initial corrosion rate prediction values ​​respectively corresponding to the plurality of submodules;

[0011] Acquire first historical material characteristic data of a sample ultra-deep well pipe and a corresponding actual corrosion rate of the sample ultra-deep well pipe as first sample data;

[0012] According to the preset disturbance data, the first historical material characteristic data of the sample ultra-deep well pipe is subjected to data enhancement processing to obtain the second historical material characteristic data of the sample ultra-deep well pipe, and according to the actual corrosion rate of the sample ultra-deep well pipe, the actual corrosion rate of the sample ultra-deep well pipe after the disturbance is determined, and the second historical material characteristic data of the sample ultra-deep well pipe and the actual corrosion rate after the disturbance are combined as the second sample data;

[0013] Combining the first sample data and the second sample data to construct joint sample data;

[0014] The initial regression model is trained using the joint sample data to obtain the preset regression model that meets the requirements.

[0015] In one embodiment, the method of using the joint sample data to train the initial regression model to obtain the preset regression model that meets the requirements includes:

[0016] dividing the joint sample data into first material characteristic data and second material characteristic data;

[0017] Using a plurality of submodules included in the first module in the initial regression module to process the first material characteristic data respectively, to obtain a second corrosion rate prediction value corresponding to each of the submodules;

[0018] Determine whether the submodule converges according to the second corrosion rate prediction value and the actual corrosion rate, and if the submodule does not converge, continue to train the submodule according to the first material characteristic data until the submodule converges to obtain the trained submodule;

[0019] Determining a determination coefficient corresponding to each of the submodules according to an actual corrosion rate corresponding to the first material characteristic data and a second corrosion rate prediction value;

[0020] Determine a weight parameter corresponding to each submodule according to a determination coefficient corresponding to each submodule;

[0021] According to the trained submodules, the second material characteristic data is processed respectively to obtain a third corrosion rate prediction value corresponding to each submodule;

[0022] Determine a fourth corrosion rate prediction value according to the third corrosion rate prediction value and the weight parameter corresponding to each of the submodules;

[0023] Processing the fourth corrosion rate prediction value using the second module of the regression model to obtain a fifth corrosion rate prediction value;

[0024] Calculating a loss value according to the fifth corrosion rate prediction value, the actual corrosion rate and a preset loss function, and detecting whether the regression model converges according to the loss value;

[0025] When it is determined that the initial regression model has not converged, the initial regression model is continuously trained according to the joint sample data until the initial regression model converges to obtain the preset regression model.

[0026] In one embodiment, determining the determination coefficient corresponding to each of the submodules according to the actual corrosion rate corresponding to the first material characteristic data and the second corrosion rate prediction value includes:

[0027] The determination coefficient corresponding to each of the submodules is determined according to the actual corrosion rate corresponding to the first material characteristic data, the second corrosion rate prediction value, and the average of the actual corrosion rate corresponding to the first material characteristic data.

[0028] In one embodiment, determining the determination coefficient corresponding to each of the submodules according to the actual corrosion rate corresponding to the first material characteristic data, the second corrosion rate prediction value, and the average of the actual corrosion rate corresponding to the first material characteristic data includes:

[0029] According to the following formula, the determination coefficient corresponding to each submodule is determined:

[0030]

[0031] Wherein, r is the determination coefficient corresponding to the i-th submodule, D is the first material characteristic data, x i is the i-th characteristic data of the first material, y i is the second corrosion rate prediction value corresponding to the i-th characteristic data of the first material, y i * is the actual corrosion rate corresponding to the i-th characteristic data of the first material, is the average of the actual corrosion rates corresponding to the first material characteristic data.

[0032] In one embodiment, the submodule is a module constructed based on an XGBoost regression algorithm, a random forest regression algorithm, a support vector regression algorithm, and a gradient boosting decision tree regression algorithm.

[0033] In one embodiment, the method further comprises:

[0034] Acquiring characteristic data of a third material;

[0035] Using a plurality of submodules included in the first module in the preset regression module to process the third material characteristic data respectively, to obtain a sixth corrosion rate prediction value corresponding to each of the submodules;

[0036] Determine a seventh corrosion rate prediction value according to the sixth corrosion rate prediction value corresponding to each of the submodules and a weight parameter of the corresponding submodule;

[0037] The seventh corrosion rate prediction value is processed by the second module of the preset regression module to obtain an eighth corrosion rate prediction value, and the preset regression module is tested according to the eighth corrosion rate prediction value.

[0038] This specification provides a device for determining the corrosion rate of ultra-deep well pipes, comprising:

[0039] A data acquisition module, used to acquire material characteristic data of the target ultra-deep well tubular to be inspected;

[0040] An initial corrosion determination module is used to process the material characteristic data of the target ultra-deep well tubular material respectively using the multiple submodules included in the first module of the preset regression model to obtain multiple initial corrosion rate prediction values ​​corresponding to the multiple submodules respectively; wherein the preset regression model is obtained by pre-training a model constructed based on a machine learning algorithm using historical material characteristic data;

[0041] A predicted corrosion determination module, configured to determine a first corrosion rate prediction value according to the plurality of initial corrosion rate prediction values ​​and weight parameters of corresponding submodules; wherein the weight parameters are determined in advance by processing the historical material characteristic data using the preset regression model;

[0042] A target corrosion determination module is used to process the first corrosion rate prediction value using the second module in the preset regression model to obtain a corresponding target corrosion rate prediction value; wherein the target corrosion rate prediction value is used to determine the corrosion condition of the target ultra-deep well tubular.

[0043] This specification also provides an electronic device, including a processor and a memory for storing instructions executable by the processor, wherein when the processor executes the instructions, a method for determining the corrosion rate of ultra-deep well tubulars is implemented.

[0044] The present specification also provides a computer-readable storage medium having computer instructions stored thereon, and when the instructions are executed, a method for determining the corrosion rate of ultra-deep well tubulars is implemented.

[0045] A method for determining the corrosion rate of ultra-deep well tubulars provided in this specification is provided, which comprises obtaining material characteristic data of a target ultra-deep well tubular to be detected; processing the material characteristic data of the target ultra-deep well tubular respectively using a plurality of submodules included in a first module in a preset regression model, and obtaining a plurality of initial corrosion rate prediction values ​​corresponding to the plurality of submodules respectively; wherein the preset regression model is obtained by pre-training a model constructed based on a machine learning algorithm using historical material characteristic data; determining a first corrosion rate prediction value according to the plurality of initial corrosion rate prediction values ​​and weight parameters of corresponding submodules; wherein the weight parameters are determined in advance by processing the historical material characteristic data using the preset regression model; processing the first corrosion rate prediction value using a second module in the preset regression model, and obtaining a corresponding target corrosion rate prediction value; wherein the target corrosion rate prediction value is used to determine the corrosion condition of the target ultra-deep well tubular. In this way, firstly, by using multiple different submodules of the first module in the preset regression model, the material characteristic data of the target ultra-deep well pipe can be processed respectively, so as to effectively process different modes in the material characteristic data of the target ultra-deep well pipe, and extract different data features contained in the material characteristic data of the target ultra-deep well pipe, which weakens the part of the single module with poor prediction effect to a certain extent, and has higher accuracy and stronger generalization ability than the single module prediction method and the traditional stacking integrated prediction method. Secondly, since the weight parameter is determined in advance by processing the historical material characteristic data using the preset regression model, the initial corrosion rate prediction value obtained by each submodule and the weight parameter of the corresponding submodule are used to determine the first corrosion rate prediction value, which can improve the accuracy of the corrosion rate prediction. Finally, the first corrosion rate prediction value is processed by the second module to obtain the target corrosion rate prediction value, which can improve the accuracy of the corrosion rate prediction of ultra-deep well low alloy pipes. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the embodiments of this specification, the drawings required for use in the embodiments will be briefly introduced below. The drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0047] Figure 1 It is a flow chart of a method for determining the corrosion rate of ultra-deep well pipes provided by an embodiment of this specification;

[0048] Figure 2 is a schematic diagram of the integrated algorithm regression model structure provided by an embodiment of this specification;

[0049] Figure 3It is a schematic diagram of the structure of an electronic device provided by an embodiment of this specification;

[0050] Figure 4 It is a schematic diagram of the structure of a device for determining the corrosion rate of ultra-deep well pipes provided by an embodiment of the present specification;

[0051] Figure 5 It is a general flow chart for determining the corrosion rate of ultra-deep well tubulars provided by an embodiment of this specification;

[0052] Figure 6 This is a comparison chart between predicted samples and actual samples provided by an embodiment of this specification. DETAILED DESCRIPTION

[0053] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.

[0054] Ultra-deep wells usually refer to wells with a depth exceeding the preset depth range (such as 6,000 meters to 9,000 meters). Their depth is far deeper than conventional oil and gas wells. With the growing global demand for oil and gas resources, ultra-deep well drilling has become an important means to achieve deep oil and gas resource exploration and development. As oil and gas exploration and development advance to deeper levels, environmental factors such as high pressure, high temperature, and high salinity in ultra-deep well environments have made the corrosion problem of downhole equipment more serious. Therefore, accurately predicting the corrosion behavior of ultra-deep wells is of great significance to ensure the safe and efficient operation of oil and gas wells.

[0055] In view of the root cause of the above problems, this specification considers inputting the material characteristic data of the target ultra-deep well pipe to be detected into each submodule in the first module of the regression model to obtain the corresponding initial corrosion rate prediction value, and then weighted merging the initial corrosion rate prediction value with the weight parameters corresponding to each submodule in the first module to determine the first corrosion rate prediction value, and inputting the first corrosion rate prediction value into the second module of the regression model for processing to obtain the corresponding target corrosion rate prediction value, and then determining the corrosion condition of the target ultra-deep well pipe according to the target corrosion rate prediction value. Thereby improving the accuracy of corrosion rate prediction of ultra-deep well low alloy pipes.

[0056] See also Figure 1 As shown, the embodiment of this specification provides a method for determining the corrosion rate of ultra-deep well pipes, wherein the method is specifically applied to the server side. When implemented specifically, the method may include the following contents:

[0057] S101: Acquire material characteristic data of target ultra-deep well tubulars to be inspected.

[0058] S102: Using the multiple submodules included in the first module in the preset regression model to respectively process the material characteristic data of the target ultra-deep well tubular, and obtain multiple initial corrosion rate prediction values ​​corresponding to the multiple submodules respectively; wherein the preset regression model is obtained by pre-training a model constructed based on a machine learning algorithm using historical material characteristic data.

[0059] S103: Determine a first corrosion rate prediction value according to the multiple initial corrosion rate prediction values ​​and weight parameters of corresponding submodules; wherein the weight parameters are pre-determined by processing the historical material characteristic data using the preset regression model.

[0060] S104: Processing the first corrosion rate prediction value using the second module in the preset regression model to obtain a corresponding target corrosion rate prediction value; wherein the target corrosion rate prediction value is used to determine the corrosion condition of the target ultra-deep well tubular.

[0061] The ultra-deep well pipe may be a low alloy pipe; and the material characteristic data may be obtained by merging corrosion data sets of ultra-deep well pipe materials.

[0062] The material characteristic data of the target ultra-deep well pipe may include material type (such as high-strength alloy steel, composite materials, super alloys and other materials with corrosion resistance, wear resistance, high temperature resistance and fatigue resistance), structural design (such as multi-layer structure, special cross-sectional shape), environmental influence (such as temperature, water content, H2S partial pressure, CO2 partial pressure), ion concentration influence (such as Cl- concentration, Mg2+Ca2+ ion concentration, SO42- ion concentration) and other characteristic data.

[0063] Specifically, for example, the material type may be low alloy steel P110 and P110S, martensitic stainless steel S13Cr, duplex stainless steel 2507, nickel-based alloy 2535, and the like.

[0064] In addition, different material characteristic data can be selected according to the depth corresponding to the target ultra-deep well pipe. For example, when the depth corresponding to the target ultra-deep well pipe is in the shallow stage (such as 0-1000 meters), since the pressure mainly borne by the pipe in the shallow stage is relatively small, the strength and corrosion resistance of the material are the key. At this time, the shape design of the pipe can be relatively simple, and the standard circular cross-section can meet the requirements. Therefore, when the depth corresponding to the target ultra-deep well pipe is in the shallow stage, strength and corrosion resistance can be selected as material characteristic data.

[0065] When the depth corresponding to the target ultra-deep well pipe is in the middle stage (such as 1000-3000 meters), the pipe not only needs to withstand higher formation pressure and temperature, but also needs to have better fatigue resistance and wear resistance. At this time, the toughness and ductility of the material become particularly important, so a more complex shape design can be used to disperse the stress, such as reinforcing ribs or spiral structures. Therefore, when the depth corresponding to the target ultra-deep well pipe is in the middle stage, strength, corrosion resistance, and structural design can be selected as material characteristic data.

[0066] When the depth corresponding to the target ultra-deep well pipe is in the deep stage (such as above 3,000 meters), the pipe will face extremely high formation pressure and temperature. At this time, the high-temperature creep resistance and corrosion resistance of the material become the focus of the design. In order to improve durability, alloy materials or composite materials can be used, and the characteristics of compression resistance and bending resistance need to be considered in the shape design. For example, a multi-layer structure pipe can be used, with high-strength materials inside and corrosion-resistant materials outside to ensure the stability and long life of the pipe in extreme environments. That is, when the depth corresponding to the target ultra-deep well pipe is in the deep stage, strength, corrosion resistance, structural design, and material type can be selected as material characteristic data.

[0067] In addition, there are many different methods for selecting the material characteristic data of the target ultra-deep well tubular. Different material characteristic data can be selected according to different actual application scenarios, and the embodiments of this specification do not specifically limit this.

[0068] In some embodiments, the weight parameter may be determined by processing the historical material characteristic data using the preset regression model in the following manner, including:

[0069] Processing the historical material characteristic data using the preset regression model to obtain a corrosion rate prediction value corresponding to the historical material characteristic data;

[0070] A weight parameter is determined according to the corrosion rate prediction value corresponding to the historical material characteristic data and the actual corrosion rate corresponding to the historical material characteristic data.

[0071] In some embodiments, the second module of the above-mentioned preset regression model can be constructed by a linear regression algorithm, and the input of the second module is a first corrosion rate prediction value obtained by preliminarily predicting the material characteristic data of the target ultra-deep well pipe through the first module of the preset regression module, and then the first corrosion rate prediction value is processed by the second module to obtain the target corrosion rate prediction value.

[0072] Therefore, the second module of the preset regression model can effectively capture the complex relationships and nonlinear features that the first module fails to fully model, and can uniformly manage and optimize the first module of the preset regression model, thereby improving the generalization ability of the preset regression model. Furthermore, the second module can also improve the stability and reliability of the final prediction results of the regression model.

[0073] Based on the above embodiments, firstly, by using multiple different submodules of the first module in the preset regression model, the material characteristic data of the target ultra-deep well pipe can be processed respectively, so as to effectively process different modes in the material characteristic data of the target ultra-deep well pipe, and extract different data features contained in the material characteristic data of the target ultra-deep well pipe, which weakens the part of the single module with poor prediction effect to a certain extent, and has higher accuracy and stronger generalization ability than the single module prediction method and the traditional stacking integrated prediction method. Secondly, since the weight parameter is determined in advance by processing the historical material characteristic data using the preset regression model, the initial corrosion rate prediction value obtained by each submodule and the weight parameter of the corresponding submodule are used to determine the first corrosion rate prediction value, which can improve the accuracy of the corrosion rate prediction. Finally, the first corrosion rate prediction value is processed by the second module to obtain the target corrosion rate prediction value, which can improve the accuracy of the corrosion rate prediction of ultra-deep well low alloy pipes.

[0074] In some embodiments, before the material characteristic data of the target ultra-deep well tubular are processed by using the multiple submodules included in the first module of the preset regression model to obtain multiple initial corrosion rate prediction values ​​corresponding to the multiple submodules, the method may further include the following contents when implemented:

[0075] S1: Acquire first historical material characteristic data of a sample ultra-deep well tubular material and the corresponding actual corrosion rate of the sample ultra-deep well tubular material as first sample data.

[0076] S2: According to the preset disturbance data, the first historical material characteristic data of the sample ultra-deep well tubular is subjected to data enhancement processing to obtain the second historical material characteristic data of the sample ultra-deep well tubular, and according to the actual corrosion rate of the sample ultra-deep well tubular, the actual corrosion rate of the sample ultra-deep well tubular after the disturbance is determined, and the second historical material characteristic data of the sample ultra-deep well tubular and the actual corrosion rate after the disturbance are combined as the second sample data.

[0077] S3: Combine the first sample data and the second sample data to construct joint sample data.

[0078] S4: Using the joint sample data, train the initial regression model to obtain the preset regression model that meets the requirements.

[0079] In some embodiments, according to the preset disturbance data, data enhancement processing is performed on the first historical material characteristic data of the sample ultra-deep well tubular material to obtain the second historical material characteristic data of the sample ultra-deep well tubular material. When specifically implemented, it may include:

[0080] Preset disturbance data is added to the first historical material characteristic data of the sample ultra-deep well tubular material to obtain second historical material characteristic data of the sample ultra-deep well tubular material.

[0081] Specifically, the second historical material characteristic data of the sample ultra-deep well pipe can be obtained according to the following random perturbation formula:

[0082]

[0083] in, is the second historical material feature data obtained by enhancing the i-th first historical material feature data, x i is the characteristic data of the first historical material, a x is the preset disturbance data corresponding to the i-th first historical material characteristic data, n x is the random disturbance range corresponding to the preset disturbance data.

[0084] In some embodiments, according to the actual corrosion rate of the sample ultra-deep well tubular, determining the actual corrosion rate of the sample ultra-deep well tubular after disturbance may include:

[0085] Using a preset feature extraction model, feature extraction processing is performed on the first historical material feature data and the second historical material feature data of the ultra-deep well tubular to obtain a first feature vector corresponding to the first historical material feature data and a second feature vector corresponding to the second historical material feature data; wherein the feature extraction model is a feature extraction model constructed based on a preset machine learning algorithm;

[0086] Obtaining a similarity between the second feature vector and each first feature vector, and determining a target feature vector in the first feature vector corresponding to the second feature vector based on the similarity;

[0087] Based on the actual corrosion rate corresponding to the target eigenvector, the actual corrosion rate corresponding to the second eigenvector is determined, and based on the actual corrosion rate corresponding to the first eigenvector and the actual corrosion rate corresponding to the second eigenvector, the actual corrosion rate of the sample ultra-deep well tubular after disturbance is determined.

[0088] For example, assume that the first historical material characteristic data of the sample ultra-deep well pipe includes characteristic data 1, characteristic data 2 and characteristic data 3, and the actual corrosion rate of characteristic data 1 is 50%, the actual corrosion rate of characteristic data 2 is 60%, and the actual corrosion rate of characteristic data 3 is 70%.

[0089] Assuming that the three first historical material feature data are enhanced according to the preset disturbance data, and the obtained second historical material feature data include feature data 4, then, the preset feature extraction model can be used to perform feature extraction processing on the above feature data 1 to feature data 4 respectively to obtain the feature vector corresponding to each feature data, that is, the first feature vector 1 corresponding to feature data 1, the first feature vector 2 corresponding to feature data 2, the first feature vector 3 corresponding to feature data 3, and the second feature vector 1 corresponding to feature data 4.

[0090] The similarity between the second feature vector 1 and each first feature vector can be calculated respectively, and based on the similarity, the target feature vector whose similarity with the second feature vector is greater than a preset similarity threshold is screened out from the first feature vector. Assuming that the target feature vectors are the first feature vector 2 and the first feature vector 3, then the actual corrosion rate corresponding to the feature data 4 can be determined based on the actual corrosion rate corresponding to the first feature vector 2 and the first feature vector 3, such as the mean (or maximum value, minimum value, etc.) of the actual corrosion rate corresponding to the first feature vector 2 and the first feature vector 3 can be determined as the actual corrosion rate corresponding to the feature data 4.

[0091] In this way, the combined sample data may include feature data 1, feature data 2, feature data 3, and feature data 4, as well as the actual corrosion rate corresponding to each feature data.

[0092] In some embodiments, see Figure 2 As shown, the combined sample data is used to train the initial regression model to obtain the preset regression model that meets the requirements. When the method is implemented specifically, the following contents may also be included:

[0093] S1: Divide the joint sample data into first material characteristic data and second material characteristic data.

[0094] S2: using the multiple submodules included in the first module in the initial regression module to process the first material characteristic data respectively, to obtain a second corrosion rate prediction value corresponding to each of the submodules.

[0095] S3: Determine whether the submodule has converged based on the second corrosion rate prediction value and the actual corrosion rate. If the submodule has not converged, continue to train the submodule based on the first material characteristic data until the submodule converges to obtain the trained submodule.

[0096] S4: Determine a determination coefficient corresponding to each of the submodules according to the actual corrosion rate corresponding to the first material characteristic data and the second corrosion rate prediction value.

[0097] S5: Determine a weight parameter corresponding to each of the submodules according to the determination coefficient corresponding to each of the submodules.

[0098] S6: Processing the second material characteristic data respectively according to the trained sub-modules to obtain a third corrosion rate prediction value corresponding to each of the sub-modules.

[0099] S7: Determine a fourth corrosion rate prediction value according to the third corrosion rate prediction value and the weight parameter corresponding to each of the submodules.

[0100] S8: Using the second module of the regression model to process the fourth corrosion rate prediction value to obtain a fifth corrosion rate prediction value.

[0101] S9: Calculate a loss value according to the fifth corrosion rate prediction value, the actual corrosion rate and a preset loss function, and detect whether the regression model converges according to the loss value.

[0102] S10: When it is determined that the initial regression model has not converged, continue to train the initial regression model according to the joint sample data until the initial regression model converges to obtain the preset regression model.

[0103] In some embodiments, the determination coefficient corresponding to each of the submodules is determined according to the actual corrosion rate corresponding to the first material characteristic data and the second corrosion rate prediction value. When the method is specifically implemented, the following contents may also be included:

[0104] S1: Determine a determination coefficient corresponding to each of the submodules according to an actual corrosion rate corresponding to the first material characteristic data, a predicted value of the second corrosion rate, and an average value of the actual corrosion rate corresponding to the first material characteristic data.

[0105] According to the following formula, the determination coefficient corresponding to each submodule is determined:

[0106]

[0107] Wherein, r is the determination coefficient corresponding to the i-th submodule, D is the first material characteristic data, xi is the i-th characteristic data of the first material, y i is the second corrosion rate prediction value corresponding to the i-th characteristic data of the first material, y i * is the actual corrosion rate corresponding to the i-th characteristic data of the first material, is the average of the actual corrosion rates corresponding to the first material characteristic data.

[0108] In some embodiments, the submodule is a module constructed based on an XGBoost regression algorithm, a random forest regression algorithm, a support vector regression algorithm, and a gradient boosting decision tree regression algorithm.

[0109] Since different sub-modules are built based on different regression algorithms, they can flexibly cope with data sets of different sizes. For example, on small data sets, sub-modules built based on the support vector regression algorithm and sub-modules built based on the random forest regression algorithm can effectively prevent the problem of overfitting, while sub-modules built based on the XGBoost regression algorithm and sub-modules built based on the gradient boosting decision tree algorithm can process large data sets through powerful feature selection and incremental learning capabilities to improve the generalization ability of the model.

[0110] In addition to the above-mentioned regression algorithm, since the material characteristic data may contain data of multiple different data types (such as text data, audio data, video data, image data, etc.), the sub-module can also be constructed based on a variety of machine learning algorithms for processing different data types to process data of different data types contained in the material characteristic data.

[0111] In addition, these submodules can collaborate with each other in an integrated learning manner to process different patterns in the data and perform feature extraction. The output of the first module is determined by weighted averaging and other methods, thereby improving the prediction accuracy and generalization ability of the overall regression model. This integration strategy utilizes the advantages and characteristics of different algorithms, reduces the bias and variance risks of a single algorithm, improves the prediction accuracy of the regression model, and enhances the robustness and stability of the regression model when processing complex regression tasks.

[0112] In some embodiments, the method may further include the following when implemented:

[0113] S1: Acquire characteristic data of a third material.

[0114] S2: using the plurality of submodules included in the first module in the preset regression module to process the third material characteristic data respectively, and obtaining a sixth corrosion rate prediction value corresponding to each of the submodules.

[0115] S3: Determine the seventh corrosion rate prediction value according to the sixth corrosion rate prediction value corresponding to each of the submodules and the weight parameters of the corresponding submodules.

[0116] S4: using the second module of the preset regression module to process the seventh corrosion rate prediction value to obtain an eighth corrosion rate prediction value, and testing the preset regression module according to the eighth corrosion rate prediction value.

[0117] As can be seen from the above, an ultra-deep well tubular corrosion rate determination method provided in an embodiment of the present specification obtains material characteristic data of a target ultra-deep well tubular to be detected; uses a plurality of submodules included in a first module in a preset regression model to respectively process the material characteristic data of the target ultra-deep well tubular, and obtains a plurality of initial corrosion rate prediction values ​​corresponding to the plurality of submodules; wherein the preset regression model is obtained by pre-training a model constructed based on a machine learning algorithm using historical material characteristic data; a first corrosion rate prediction value is determined according to the plurality of initial corrosion rate prediction values ​​and weight parameters of corresponding submodules; wherein the weight parameters are determined in advance by processing the historical material characteristic data using the preset regression model; the first corrosion rate prediction value is processed by the second module in the preset regression model to obtain a corresponding target corrosion rate prediction value; wherein the target corrosion rate prediction value is used to determine the corrosion condition of the target ultra-deep well tubular. In this way, firstly, by using multiple different submodules of the first module in the preset regression model, the material characteristic data of the target ultra-deep well pipe can be processed respectively, so as to effectively process different modes in the material characteristic data of the target ultra-deep well pipe, and extract different data features contained in the material characteristic data of the target ultra-deep well pipe, which weakens the part of the single module with poor prediction effect to a certain extent, and has higher accuracy and stronger generalization ability than the single module prediction method and the traditional stacking integrated prediction method. Secondly, since the weight parameter is determined in advance by processing the historical material characteristic data using the preset regression model, the initial corrosion rate prediction value obtained by each submodule and the weight parameter of the corresponding submodule are used to determine the first corrosion rate prediction value, which can improve the accuracy of the corrosion rate prediction. Finally, the first corrosion rate prediction value is processed by the second module to obtain the target corrosion rate prediction value, which can improve the accuracy of the corrosion rate prediction of ultra-deep well low alloy pipes.

[0118] See also Figure 3 As shown, the embodiment of this specification also provides a specific electronic device, wherein the electronic device includes a network communication port 301, a processor 302 and a memory 303, and the above structures are connected through internal cables so that each structure can perform specific data interaction.

[0119] The network communication port 301 can be used to obtain material characteristic data of the target ultra-deep well tubular to be detected.

[0120] The processor 302 can be specifically used to use multiple submodules included in the first module of the preset regression model to process the material characteristic data of the target ultra-deep well tubular respectively, and obtain multiple initial corrosion rate prediction values ​​corresponding to the multiple submodules respectively; wherein the preset regression model is obtained by pre-training a model constructed based on a machine learning algorithm using historical material characteristic data; a first corrosion rate prediction value is determined according to the multiple initial corrosion rate prediction values ​​and weight parameters of corresponding submodules; wherein the weight parameters are determined in advance by processing the historical material characteristic data using the preset regression model; the first corrosion rate prediction value is processed using the second module in the preset regression model to obtain a corresponding target corrosion rate prediction value; wherein the target corrosion rate prediction value is used to determine the corrosion condition of the target ultra-deep well tubular.

[0121] The memory 303 may be specifically used to store corresponding instruction programs.

[0122] Based on the above method, the relevant structural performance of the electronic equipment can be effectively utilized, the data processing speed of the electronic equipment can be improved, and the method for determining the corrosion rate of ultra-deep well pipes can be efficiently realized.

[0123] In this embodiment, the network communication port 301 can be a virtual port that is bound to different communication protocols so that different data can be sent or received. For example, the network communication port can be a port responsible for web data communication, a port responsible for FTP data communication, or a port responsible for email data communication. In addition, the network communication port can also be a physical communication interface or communication chip. For example, it can be a wireless mobile network communication chip, such as GSM, CDMA, etc.; it can also be a Wifi chip; it can also be a Bluetooth chip.

[0124] In this embodiment, the processor 302 may be implemented in any appropriate manner. For example, the processor may take the form of a microprocessor or processor and a computer-readable medium storing a computer-readable program code (such as software or firmware) executable by the (micro)processor, a logic gate, a switch, an application specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller, etc. This specification does not limit this.

[0125] In this embodiment, the memory 303 may include multiple levels. In a digital system, anything that can store binary data can be a memory; in an integrated circuit, a circuit with a storage function but no physical form is also called a memory, such as RAM, FIFO, etc.; in a system, a storage device with a physical form is also called a memory, such as a memory stick, TF card, etc.

[0126] The embodiment of the present specification also provides a computer-readable storage medium based on the above-mentioned method for determining the corrosion rate of ultra-deep well tubulars, wherein the computer-readable storage medium stores computer program instructions, and when the computer program instructions are executed, the following are implemented: obtaining material characteristic data of the target ultra-deep well tubular to be detected; using multiple submodules contained in the first module in the preset regression model to process the material characteristic data of the target ultra-deep well tubular respectively, and obtaining multiple initial corrosion rate prediction values ​​corresponding to the multiple submodules respectively; wherein the preset regression model is obtained by pre-training a model constructed based on a machine learning algorithm using historical material characteristic data; determining a first corrosion rate prediction value according to the multiple initial corrosion rate prediction values ​​and weight parameters of the corresponding submodules; wherein the weight parameters are determined in advance by processing the historical material characteristic data using the preset regression model; processing the first corrosion rate prediction value using the second module in the preset regression model to obtain a corresponding target corrosion rate prediction value; wherein the target corrosion rate prediction value is used to determine the corrosion condition of the target ultra-deep well tubular.

[0127] In this embodiment, the storage medium includes, but is not limited to, a random access memory (RAM), a read-only memory (ROM), a cache, a hard disk (HDD), or a memory card. The memory may be used to store computer program instructions. The network communication unit may be an interface for network connection communication set in accordance with the standard specified by the communication protocol.

[0128] In this embodiment, the functions and effects specifically implemented by the program instructions stored in the computer-readable storage medium can be explained in comparison with other implementations and will not be described in detail here.

[0129] See also Figure 4 At the software level, the embodiment of this specification also provides a device for determining the corrosion rate of ultra-deep well pipes, which may specifically include the following structural modules:

[0130] The data acquisition module 401 is used to acquire material characteristic data of the target ultra-deep well tubular to be detected;

[0131] The initial corrosion determination module 402 is used to process the material characteristic data of the target ultra-deep well tubular material respectively using the multiple submodules included in the first module of the preset regression model to obtain multiple initial corrosion rate prediction values ​​corresponding to the multiple submodules respectively; wherein the preset regression model is obtained by pre-training a model constructed based on a machine learning algorithm using historical material characteristic data;

[0132] The predicted corrosion determination module 403 is used to determine a first corrosion rate prediction value according to the multiple initial corrosion rate prediction values ​​and the weight parameters of the corresponding submodules; wherein the weight parameters are determined in advance by processing the historical material characteristic data using the preset regression model;

[0133] The target corrosion determination module 404 is used to process the first corrosion rate prediction value using the second module in the preset regression model to obtain a corresponding target corrosion rate prediction value; wherein the target corrosion rate prediction value is used to determine the corrosion condition of the target ultra-deep well tubular.

[0134] In some embodiments, before the above-mentioned initial corrosion determination module 402, during specific implementation, first historical material characteristic data of the sample ultra-deep well tubular and the corresponding actual corrosion rate of the sample ultra-deep well tubular are obtained as first sample data; according to preset disturbance data, the first historical material characteristic data of the sample ultra-deep well tubular is subjected to data enhancement processing to obtain second historical material characteristic data of the sample ultra-deep well tubular, and according to the actual corrosion rate of the sample ultra-deep well tubular, the actual corrosion rate of the sample ultra-deep well tubular after disturbance is determined, and the second historical material characteristic data of the sample ultra-deep well tubular and the actual corrosion rate after disturbance are combined as second sample data; the first sample data and the second sample data are combined to construct joint sample data; a regression model determination module is used to train an initial regression model using the joint sample data to obtain the preset regression model that meets the requirements.

[0135] In some embodiments, when the above-mentioned regression model determination module is specifically implemented, the joint sample data is divided into first material characteristic data and second material characteristic data; the first material characteristic data are processed respectively using multiple submodules included in the first module in the initial regression module to obtain a second corrosion rate prediction value corresponding to each submodule; according to the second corrosion rate prediction value and the actual corrosion rate, it is determined whether the submodule has converged. If the submodule has not converged, the submodule is continuously trained according to the first material characteristic data until the submodule converges to obtain the trained submodule; a determination coefficient determination module is used to determine the determination coefficient corresponding to each submodule according to the actual corrosion rate corresponding to the first material characteristic data and the second corrosion rate prediction value; according to each submodule The determination coefficient corresponding to the block is used to determine the weight parameter corresponding to each of the sub-modules; the second material characteristic data is processed respectively according to the trained sub-modules to obtain the third corrosion rate prediction value corresponding to each of the sub-modules; the fourth corrosion rate prediction value is determined according to the third corrosion rate prediction value and the weight parameter corresponding to each of the sub-modules; the fourth corrosion rate prediction value is processed by the second module of the regression model to obtain the fifth corrosion rate prediction value; the loss value is calculated according to the fifth corrosion rate prediction value, the actual corrosion rate and the preset loss function, and the regression model is detected whether it converges according to the loss value; when it is determined that the initial regression model has not converged, the initial regression model is continuously trained according to the joint sample data until the initial regression model converges to obtain the preset regression model.

[0136] In some embodiments, when the above-mentioned determination coefficient determination module is implemented, the determination coefficient corresponding to each sub-module is determined according to the actual corrosion rate corresponding to the first material characteristic data, the second corrosion rate prediction value and the average of the actual corrosion rate corresponding to the first material characteristic data.

[0137] Specifically, the determination coefficient corresponding to each submodule is determined according to the following formula:

[0138]

[0139] Wherein, r is the determination coefficient corresponding to the i-th submodule, D is the first material characteristic data, x i is the i-th characteristic data of the first material, y i is the second corrosion rate prediction value corresponding to the i-th characteristic data of the first material, y i * is the actual corrosion rate corresponding to the i-th characteristic data of the first material, is the average of the actual corrosion rates corresponding to the first material characteristic data.

[0140] In some embodiments, the submodule is a module constructed based on an XGBoost regression algorithm, a random forest regression algorithm, a support vector regression algorithm, and a gradient boosting decision tree regression algorithm.

[0141] In some embodiments, the apparatus further comprises:

[0142] A data acquisition module, used to acquire characteristic data of a third material;

[0143] A corrosion rate determination module, used to process the third material characteristic data respectively using a plurality of submodules included in the first module in the preset regression module to obtain a sixth corrosion rate prediction value corresponding to each of the submodules;

[0144] A weight merging module, used to determine the seventh corrosion rate prediction value according to the sixth corrosion rate prediction value corresponding to each of the submodules and the weight parameters of the corresponding submodules;

[0145] The test verification module is used to process the seventh corrosion rate prediction value by using the second module of the preset regression module to obtain the eighth corrosion rate prediction value, and test the preset regression module according to the eighth corrosion rate prediction value.

[0146] It should be noted that the units, devices or modules described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. For the convenience of description, the above devices are described separately by functions divided into various modules. Of course, when implementing this specification, the functions of each module can be implemented in the same or more software and / or hardware, or the modules that implement the same function can be implemented by a combination of multiple sub-modules or sub-units. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0147] As can be seen from the above, based on the ultra-deep well pipe corrosion rate determination device provided by the embodiment of this specification, first, using multiple different sub-modules of the first module in the preset regression model, the material characteristic data of the target ultra-deep well pipe can be processed respectively, so as to effectively process different modes in the material characteristic data of the target ultra-deep well pipe, and extract different data features contained in the material characteristic data of the target ultra-deep well pipe, which weakens the part of the single module with poor prediction effect to a certain extent, and has higher accuracy and stronger generalization ability than the single module prediction method and the traditional stacked integrated prediction method. Secondly, since the weight parameter is determined in advance by processing the historical material characteristic data using the preset regression model, the initial corrosion rate prediction value obtained by each sub-module and the weight parameter of the corresponding sub-module are used to determine the first corrosion rate prediction value, which can improve the accuracy of the corrosion rate prediction. Finally, the first corrosion rate prediction value is processed by the second module to obtain the target corrosion rate prediction value, which can improve the accuracy of the corrosion rate prediction of ultra-deep well low alloy pipes.

[0148] In a specific scenario example, a method, device and equipment for determining the corrosion rate of ultra-deep well pipes provided in this specification can be applied to process the material characteristic data of the target ultra-deep well pipes using a preset regression model to obtain the corresponding target corrosion rate prediction value, which solves the problem of low prediction accuracy of traditional models and improves the accuracy of corrosion rate prediction of ultra-deep well low-alloy pipes. The specific implementation process may include the following contents, see Figure 5 shown.

[0149] S1: Data merging: The data sets of different materials obtained from the ultra-deep well low alloy pipe corrosion experiment are merged to form a corrosion data set containing environmental factors, ion concentration, material type and corrosion rate. The corrosion rate in the corrosion data set is determined as the actual corrosion rate of the sample ultra-deep well pipe, and the remaining characteristic data is determined as the first historical material characteristic data of the sample ultra-deep well pipe.

[0150] S2: Data processing: For the first historical material characteristic data of the sample ultra-deep well pipe, the material type is converted into a numerical variable using the unique hot encoding method. For example, assuming that the first historical material characteristic data corresponds to K types of ultra-deep well low alloy pipe materials, then the material parameter can be set to K bits, the i-th bit of the i-th material is 1, and the remainder is 0. The processed results are shown in Table 1.

[0151] Table 1

[0152] Material Type No. 1 No. 2 …… kth position P110S 1 0 0 0 P110 0 1 0 0 …… 0 0 0 0 Kth material 0 0 0 1

[0153] The processed first historical material characteristic data is standardized, and the expression is:

[0154]

[0155] Among them, x i is the standardized characteristic data of the first historical material, x′ i is the i-th characteristic data of the first historical material, μ i is the mean value of the data sequence of the i-th characteristic data of the first historical material, σ i is the standard deviation of the data sequence of the i-th characteristic data of the first historical material.

[0156] S3: Data partitioning: Figure 5 As shown, an initial data set can be constructed based on the third material characteristic data and the first historical material characteristic data of the sample ultra-deep well tubular, a training data set can be constructed based on the first historical material characteristic data, and a test data set can be constructed based on the third material characteristic data.

[0157] In addition, the data volume ratio of the third material characteristic data and the first historical material characteristic data of the sample ultra-deep well tubular may be a preset ratio (eg, 1:9).

[0158] S4: Data enhancement: According to the preset disturbance data, the first historical material characteristic data of the sample ultra-deep well tubular is subjected to data enhancement processing to obtain the second historical material characteristic data of the sample ultra-deep well tubular, and according to the actual corrosion rate of the sample ultra-deep well tubular, the actual corrosion rate of the sample ultra-deep well tubular after the disturbance is determined, and the second historical material characteristic data of the sample ultra-deep well tubular and the actual corrosion rate after the disturbance are combined as the second sample data; the first sample data and the second sample data are combined to construct the joint sample data.

[0159] S5: using the plurality of submodules included in the first module in the preset regression model to process the joint sample data respectively, and obtaining a plurality of second corrosion rate prediction values ​​respectively corresponding to the plurality of submodules.

[0160] In some embodiments, the multiple sub-modules are modules constructed based on the XGBoost regression algorithm, the random forest regression algorithm, the support vector regression algorithm, and the gradient boosting decision tree regression algorithm.

[0161] Among them, the XGBoost regression algorithm can handle nonlinear relationships and prevent overfitting due to its parallel computing and regularization capabilities, but parameter tuning is relatively complex; the random forest regression algorithm improves prediction accuracy through integration, has good noise resistance and generalization performance, but may produce overfitting; the support vector regression algorithm performs regression prediction by finding the optimal hyperplane in high-dimensional space, and is good at solving nonlinear and high-dimensional problems, but is slightly disadvantaged when processing large amounts of data; the gradient boosting decision tree regression algorithm optimizes the prediction results by iteratively fitting the residuals, and can capture complex relationships in the data, but the computational complexity is relatively high. Therefore, combining different submodules together as the first module of the regression model can combine the advantages of the algorithms corresponding to each submodule and overcome the disadvantages of the algorithms corresponding to each submodule.

[0162] In some embodiments, the first module can be trained using a cross-validation method with a preset number of folds. Taking 10-fold cross-validation as an example, the joint sample data can be divided into 10 folds, of which 9 folds are used as the first material characteristic data and 1 fold is used as the second material characteristic data. The multiple submodules in the first module of the preset regression model can be trained 10 times based on the first material characteristic data obtained by the division.

[0163] In some embodiments, Figure 2 As shown, a new training data set can be constructed based on the fourth corrosion rate prediction value, and a new test data set can be constructed based on the seventh corrosion rate prediction value.

[0164] S6: Determine a determination coefficient corresponding to each of the submodules according to the second corrosion rate prediction value obtained by processing the first material characteristic data by the multiple submodules, and the actual corrosion rate corresponding to the first material characteristic data.

[0165] S7: Determine the overall determination coefficient of each submodule according to the determination coefficient corresponding to each submodule;

[0166] Specifically, for example, the overall determination coefficient of each submodule can be calculated according to the following formula:

[0167]

[0168] Among them, R i is the overall determination coefficient of the ith submodule, r n is the determination coefficient of the ith submodule in the nth fold, and K is the number of cross-validation folds.

[0169] S8: Further, the weight parameter of each submodule may be determined according to the following formula:

[0170]

[0171] Among them, W i is the weight parameter of the ith submodule, R i is the overall determination coefficient of the ith submodule, N is the number of submodules, R j The overall determination coefficient of the jth submodule.

[0172] S9: According to the trained sub-modules, the second material characteristic data is processed respectively to obtain the third corrosion rate prediction value corresponding to each of the sub-modules, and the fourth corrosion rate prediction value is determined according to the third corrosion rate prediction value corresponding to each of the sub-modules and the weight parameter.

[0173] Among them, during the 10 training processes of each sub-module, the trained sub-module is used to predict the characteristic data of the second material, and the results are averaged to obtain the averaged third corrosion rate prediction value, and the fourth corrosion rate prediction value is determined based on the averaged third corrosion rate prediction value and the weight parameter.

[0174] S10: Utilize the second module in the preset regression model to process the fourth corrosion rate prediction value to obtain a corresponding fifth corrosion rate prediction value.

[0175] S11: Use mean absolute error and coefficient of determination to determine the prediction accuracy of the preset regression model.

[0176] Specifically, the mean absolute error can be obtained according to the following formula:

[0177]

[0178] Among them, MAE is the mean absolute error.

[0179] Specifically, the gradient descent algorithm can be used for learning, and its formula is as follows:

[0180]

[0181] Among them, w i+1 is the model parameter after the i-th iteration, w i is the model parameter before the i-th iteration, α is the learning rate, y (j) is the actual corrosion rate of the jth sample, is the i-th material characteristic data of the j-th sample.

[0182] Furthermore, the loss between the fifth corrosion rate prediction value and the actual corrosion rate can be calculated, and the prediction loss of the data set is minimized by iteratively updating the parameters. At this time, the corresponding model parameters are the required model parameters, and the formula is as follows:

[0183]

[0184] S12: Use grid search (GridSearchCV) hyperparameter search to adjust the parameters of the preset regression model.

[0185] Among them, the above-mentioned grid search (GridSearchCV) is a tool for model hyperparameter optimization. Through cross-validation, it performs an exhaustive search on the specified parameter grid to find the parameter combination that performs best on the data set.

[0186] Specifically, for the submodules in the first module of the preset regression model (such as submodules constructed based on the XGBoost regression algorithm, the random forest regression algorithm, and the gradient boosting decision tree regression algorithm, respectively), the grid search (GridSearchCV) hyperparameter search can be used to tune the two parameters of the number of decision trees and the maximum depth, as shown in Table 2.

[0187] Table 2

[0188] Submodules Number of decision trees Maximum Depth Submodule 1 built based on XGBoost regression algorithm 300 10 Submodule 2 based on random forest regression algorithm 100 10 Submodule 3 based on gradient boosting decision tree regression algorithm 100 4

[0189] For the submodule in the first module of the preset regression model (such as the submodule built based on the support vector regression algorithm), the grid search (GridSearchCV) hyperparameter search is used to tune the two parameters of the function type and the penalty parameter, as shown in Table 3.

[0190] Table 3

[0191] Submodules Kernel function type Penalty Parameters Submodules built based on support vector regression algorithm poly 0.1

[0192] S13: The eighth corrosion rate value obtained by processing the third material characteristic data using the preset regression model can be seen Figure 6 As shown, the value of the mean absolute error is 0.0198 and the value of the determination coefficient is 0.9423.

[0193] Although the present specification provides method operation steps as described in the embodiments or flow charts, more or less operation steps may be included based on conventional or non-creative means. The order of steps listed in the embodiments is only one way of executing the order of many steps, and does not represent a unique execution order. When the device or client product in practice is executed, it can be executed in sequence or in parallel according to the method shown in the embodiments or the drawings (for example, a parallel processor or a multi-threaded processing environment, or even a distributed data processing environment). The term "include", "include" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, product or device including a series of elements includes not only those elements, but also includes other elements that are not explicitly listed, or also includes elements inherent to such a process, method, product or device. In the absence of more restrictions, it is not excluded that there are other identical or equivalent elements in the process, method, product or device including the elements. The first, second, etc. words are used to represent the name, and do not represent any particular order.

[0194] Those skilled in the art also know that, in addition to implementing the controller in a purely computer-readable program code, the controller can be made to implement the same function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, such a controller can be considered as a hardware component, and the devices for implementing various functions included therein can also be considered as structures within the hardware component. Or even, the devices for implementing various functions can be considered as both software modules for implementing the method and structures within the hardware component.

[0195] Through the description of the above embodiments, it can be known that those skilled in the art can clearly understand that the present specification can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present specification can essentially be embodied in the form of a software product, which can be stored in a storage medium such as ROM / RAM, a disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a mobile terminal, a server, or a network device, etc.) to execute the methods described in each embodiment of the present specification or some parts of the embodiments.

[0196] Although the present specification is described through embodiments, those skilled in the art will appreciate that there are many modifications and changes to the present specification without departing from the spirit of the present specification, and it is intended that the appended claims include these modifications and changes without departing from the spirit of the present specification.

Claims

1. A method for determining the corrosion rate of ultra-deep well pipes, characterized in that: include: Acquire material characteristic data of target ultra-deep well pipes to be inspected; The material characteristic data of the target ultra-deep well tubular are processed respectively by using the multiple submodules included in the first module of the preset regression model to obtain multiple initial corrosion rate prediction values ​​corresponding to the multiple submodules respectively; wherein the preset regression model is obtained by pre-training a model constructed based on a machine learning algorithm using historical material characteristic data; Determine a first corrosion rate prediction value according to the multiple initial corrosion rate prediction values ​​and weight parameters of corresponding submodules; wherein the weight parameters are determined in advance by processing the historical material characteristic data using the preset regression model; The first corrosion rate prediction value is processed by the second module in the preset regression model to obtain a corresponding target corrosion rate prediction value; wherein the target corrosion rate prediction value is used to determine the corrosion condition of the target ultra-deep well tubular.

2. The method according to claim 1, characterized in that Before the method uses the plurality of submodules included in the first module of the preset regression model to process the material characteristic data of the target ultra-deep well tubular respectively to obtain a plurality of initial corrosion rate prediction values ​​respectively corresponding to the plurality of submodules, it also includes: Acquire first historical material characteristic data of a sample ultra-deep well pipe and a corresponding actual corrosion rate of the sample ultra-deep well pipe as first sample data; According to the preset disturbance data, the first historical material characteristic data of the sample ultra-deep well pipe is subjected to data enhancement processing to obtain the second historical material characteristic data of the sample ultra-deep well pipe, and according to the actual corrosion rate of the sample ultra-deep well pipe, the actual corrosion rate of the sample ultra-deep well pipe after the disturbance is determined, and the second historical material characteristic data of the sample ultra-deep well pipe and the actual corrosion rate after the disturbance are combined as the second sample data; Combining the first sample data and the second sample data to construct joint sample data; The initial regression model is trained using the joint sample data to obtain the preset regression model that meets the requirements.

3. The method according to claim 2, characterized in that The method of using the combined sample data to train the initial regression model to obtain the preset regression model that meets the requirements includes: dividing the joint sample data into first material characteristic data and second material characteristic data; Using a plurality of submodules included in the first module in the initial regression module to process the first material characteristic data respectively, to obtain a second corrosion rate prediction value corresponding to each of the submodules; Determine whether the submodule converges according to the second corrosion rate prediction value and the actual corrosion rate, and if the submodule does not converge, continue to train the submodule according to the first material characteristic data until the submodule converges to obtain the trained submodule; Determining a determination coefficient corresponding to each of the submodules according to an actual corrosion rate corresponding to the first material characteristic data and a second corrosion rate prediction value; Determine a weight parameter corresponding to each submodule according to a determination coefficient corresponding to each submodule; According to the trained submodules, the second material characteristic data is processed respectively to obtain a third corrosion rate prediction value corresponding to each submodule; Determine a fourth corrosion rate prediction value according to the third corrosion rate prediction value and the weight parameter corresponding to each of the submodules; Processing the fourth corrosion rate prediction value using the second module of the regression model to obtain a fifth corrosion rate prediction value; Calculating a loss value according to the fifth corrosion rate prediction value, the actual corrosion rate and a preset loss function, and detecting whether the regression model converges according to the loss value; When it is determined that the initial regression model has not converged, the initial regression model is continuously trained according to the joint sample data until the initial regression model converges to obtain the preset regression model.

4. The method according to claim 3, characterized in that The determining the determination coefficient corresponding to each of the submodules according to the actual corrosion rate corresponding to the first material characteristic data and the second corrosion rate prediction value includes: The determination coefficient corresponding to each of the submodules is determined according to the actual corrosion rate corresponding to the first material characteristic data, the second corrosion rate prediction value, and the average of the actual corrosion rate corresponding to the first material characteristic data.

5. The method according to claim 4, characterized in that Determining the determination coefficient corresponding to each of the submodules according to the actual corrosion rate corresponding to the first material characteristic data, the second corrosion rate prediction value, and the average of the actual corrosion rate corresponding to the first material characteristic data includes: According to the following formula, the determination coefficient corresponding to each submodule is determined: Wherein, r is the determination coefficient corresponding to the submodule, D is the first material characteristic data, x i is the i-th characteristic data of the first material, y i is the second corrosion rate prediction value corresponding to the i-th characteristic data of the first material, y i * is the actual corrosion rate corresponding to the i-th characteristic data of the first material, is the average of the actual corrosion rates corresponding to the first material characteristic data.

6. The method according to claim 5, characterized in that The submodule is a module constructed based on the XGBoost regression algorithm, the random forest regression algorithm, the support vector regression algorithm and the gradient boosting decision tree regression algorithm.

7. The method according to claim 6, characterized in that The method further comprises: Acquiring characteristic data of a third material; Using a plurality of submodules included in the first module in the preset regression module to process the third material characteristic data respectively, to obtain a sixth corrosion rate prediction value corresponding to each of the submodules; Determine a seventh corrosion rate prediction value according to the sixth corrosion rate prediction value corresponding to each of the submodules and a weight parameter of the corresponding submodule; The seventh corrosion rate prediction value is processed by the second module of the preset regression module to obtain an eighth corrosion rate prediction value, and the preset regression module is tested according to the eighth corrosion rate prediction value.

8. An ultra-deep well pipe corrosion rate determination device, characterized in that: include: A data acquisition module, used to acquire material characteristic data of the target ultra-deep well tubular to be inspected; An initial corrosion determination module is used to process the material characteristic data of the target ultra-deep well tubular material respectively using the multiple submodules included in the first module of the preset regression model to obtain multiple initial corrosion rate prediction values ​​corresponding to the multiple submodules respectively; wherein the preset regression model is obtained by pre-training a model constructed based on a machine learning algorithm using historical material characteristic data; A predicted corrosion determination module, configured to determine a first corrosion rate prediction value according to the plurality of initial corrosion rate prediction values ​​and weight parameters of corresponding submodules; wherein the weight parameters are determined in advance by processing the historical material characteristic data using the preset regression model; A target corrosion determination module is used to process the first corrosion rate prediction value using the second module in the preset regression model to obtain a corresponding target corrosion rate prediction value; wherein the target corrosion rate prediction value is used to determine the corrosion condition of the target ultra-deep well tubular.

9. An electronic device, characterized in that: The method comprises a processor and a memory for storing instructions executable by the processor, wherein when the processor executes the instructions, the steps of the method for determining the corrosion rate of ultra-deep well tubulars according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: Computer instructions are stored thereon, and when the instructions are executed by a processor, the steps of the method for determining the corrosion rate of ultra-deep well tubulars as described in any one of claims 1 to 7 are implemented.

Citation Information

Patent Citations

  • Method for predicting end-point quality of metallurgical process based on stacking integration strategy

    CN118197463A

  • Carbon capture absorption tower CO2 concentration prediction method based on stacked ensemble learning and application

    CN118468020A

  • Wire gripping assembly for cutting surgery to treat carpal tunnel syndrome and trigger finger syndrome

    KR102672985B1

Cited By

  • System and method for evaluating corrosion state of nuclear power pipeline

    CN119804284A

  • System and method for evaluating the corrosion state of nuclear power plant piping

    CN119804284B

  • Method for predicting corrosion in pipeline based on Gaussian residual regression

    CN120030901A

  • Gaussian residual regression method for predicting internal corrosion in pipelines

    CN120030901B