Oil well indicator diagram image data enhancement method

Abstract

The invention provides an oil well indicator diagram image data enhancement method. The method comprises the following steps: step 1, sorting and classifying working condition types lacking working condition samples; 2, performing enhancement processing on displacement load data in the scarce working condition type displacement load data sample library; 3, drawing new indicator diagrams, and classifying the new indicator diagrams; step 4, performing integration to form a training sample set; 5, constructing a VAE (variational auto-encoder) model, and inputting the training sample set into the model for learning and training; step 6, identifying the indicator diagram sample set of the scarce working condition type to generate similar indicator diagram images; and 7, inputting the newly generated indicator diagram sample into a working condition diagnosis system for verification. According to the oil well indicator diagram image data enhancement method, the comprehensiveness and accuracy of the working condition intelligent diagnosis system on the diagnosis of the working condition of the category are improved, the capability of the working condition intelligent diagnosis system on the working condition of the target oil field block is greatly improved, and huge benefits can be brought to the oil field.

Description

Technical Field

[0001] This invention relates to the field of oil well condition diagnosis technology, and in particular to a method for enhancing oil well dynamometer image data. Background Technology

[0002] Rod-operated pumps are currently the most widely used mechanical oil extraction method both domestically and internationally. These pumps operate year-round in high-temperature, high-pressure environments thousands of meters underground, facing harsh working conditions and frequent malfunctions. The dynamometer card (DVC) is a crucial indicator of the well's operating status, comprehensively reflecting the pumping unit's operational condition. A DVC is a planar graph plotted from two data points: displacement and load. Due to the complex downhole environment and the diverse malfunctions experienced by rod-operated pumps, different dynamometer cards correspond to different malfunctions, and the image characteristics of the DVC image reflect specific operating conditions.

[0003] Currently, analyzing the features of oil well dynamometer cards is a common method for diagnosing downhole faults in rod-pumped oil systems. Among these methods, diagnostic systems based on neural network algorithms are particularly efficient and accurate. However, such diagnostic methods require training the neural network algorithm with a large sample set of dynamometer cards from various oil well conditions. Due to factors such as the geographical location and environment of the oilfield, some condition categories have limited sample sets. This significantly impacts the learning and diagnostic accuracy of intelligent condition diagnostic systems, resulting in incomplete diagnostic results and hindering the realization of good benefits for oilfield production.

[0004] Chinese patent application CN109767440A discloses a method for augmenting image data for deep learning model training and learning, comprising the following steps: a) first, determining the data type, identifying CT or MRI image data; b) then, for the image data, determining whether a Region of Interest (ROI) has been defined, and combining this with the size of the tumor region, selecting an appropriate method to construct the image dataset; c) then using basic image transformation methods to train the image dataset to obtain a preliminary training dataset; d) finally, augmenting the preliminary training dataset, then using a network model for deep training, and finally performing probability prediction. This invention, based on deep learning in artificial intelligence, applies a series of data augmentation methods to the training of deep models in the field of medical image processing, addressing the impact of abnormal data caused by the heterogeneity of medical image data, thus aiding computer-aided diagnosis and improving diagnostic efficiency and accuracy. However, this method, which uses image transformation, is not suitable for augmenting oil well dynamometer images. Traditional image transformation can cause the newly obtained dynamometer image to be inconsistent with the specific operating conditions of the corresponding oil well, leading to poor accuracy in neural network training.

[0005] The paper "Research and Implementation of UAV Image Enhancement Method" describes a method for dividing bright and dark regions in an image obtained from a UAV by calculating the mean and variance of grayscale pixels and a brightness coefficient. Correction coefficients are then calculated based on the pixel mean of a sliding window and a set threshold to correct image brightness. Verification shows that the algorithm effectively improves local image brightness, enhances image contrast, and preserves contour edge texture information. However, this paper, while also processing grayscale images, differs fundamentally in its effect on oil well dynamometer maps and UAV-acquired grayscale images. Improving image contrast and local brightness through image enhancement methods cannot fundamentally change the oil well dynamometer map, thus offering little benefit for expanding the rare type sample library.

[0006] Chinese patent application CN201810648470.7 discloses a method and apparatus for checking the operating conditions of pumping unit wells. The method includes: extracting feature descriptors from target pumping unit well dynamometer card data and sample dynamometer card data according to a specified operator; wherein the sample dynamometer card data represents the dynamometer card data of a pumping unit well with known corresponding operating conditions; the feature descriptors are used to distinguish the feature information of the dynamometer card data; generating a sample dataset using the feature descriptors of the sample dynamometer card data as elements; classifying the feature descriptors in the sample dataset according to the operating conditions corresponding to the sample dynamometer card data; selecting at least one feature descriptor from the sample dataset that satisfies a specified condition with the feature descriptors of the target pumping unit well dynamometer card data; and checking the operating conditions of the target pumping unit well based on the number of selected feature descriptors in each category. This method can achieve rapid checking of operating conditions.

[0007] Chinese patent application CN201911339973.7 discloses a method and apparatus for identifying the operating conditions of a pumping unit well. The method includes: acquiring a dynamometer card (DTC) of the pumping unit well to be identified; inputting the DTC card into a pre-trained machine learning model, and outputting a probability distribution of the operating condition type corresponding to the DTC card; wherein the machine learning model is trained using each DTC card sample in a DTC card sample set as training samples, and using the operating condition type corresponding to each DTC card sample as a sample label. This invention has the advantage of not relying on standard DTC cards, and can accurately identify the operating conditions of pumping unit wells using actually measured DTC card samples.

[0008] The above-mentioned existing technologies are all quite different from the present invention and have failed to solve the technical problem we want to solve. Therefore, we have invented a new method for enhancing oil well dynamometer image data. Summary of the Invention

[0009] The purpose of this invention is to provide an oil well dynamometer image data enhancement method that can be applied to the problem of limited training samples for operating conditions, thereby improving the ability of operating condition diagnostic systems to handle single rare operating condition problems.

[0010] The objective of this invention can be achieved through the following technical measures: a method for enhancing oil well dynamometer image data, comprising:

[0011] Step 1: Organize and classify the working condition types that lack working condition samples to obtain a normal working condition type sample set and a scarce working condition type sample set;

[0012] Step 2: Enhance the original displacement load data of the indicator diagram samples in the scarce working condition type sample library by using image enhancement methods;

[0013] Step 3: Draw new dynamometer diagrams from the enhanced data points and classify the new dynamometer diagram types.

[0014] Step 4: Integrate the newly expanded dynamometer diagram sample set (scarce type) with the normal type dynamometer diagram sample set to form a training sample set;

[0015] Step 5: Construct a variational autoencoder (VAE) model and feed the training sample set into the model for learning and training;

[0016] Step 6: Use the trained VAE model to identify the dynamometer diagram sample set of scarce working conditions to generate similar dynamometer diagram images.

[0017] Step 7: Input the newly generated dynamometer diagram samples into the operating condition diagnostic system for verification, and re-input the newly generated samples into the training sample set to further improve the model performance.

[0018] The objective of this invention can also be achieved through the following technical measures:

[0019] In step 1, based on the existing oil well dynamometer diagram sample library in the oilfield information platform's big data database, the working condition types for which working condition samples are lacking are sorted and classified, and displacement and load data of the original oil well dynamometer diagrams are collected.

[0020] In step 1, using the oil well dynamometer diagram sample library already collected and organized in the big data database of the oilfield information platform, the number of samples for each type of working condition is counted through manual calibration. Working condition types with few samples are classified separately, and the original displacement and load data of the dynamometer diagram samples for each scarce working condition type are collected and organized to form a new sample library of displacement and load data for scarce working condition types.

[0021] In step 2, the displacement load data in the scarce working condition type displacement load data sample library is enhanced by image enhancement method, and the new data points generated after enhancement are organized and drawn into a new oil well dynamometer diagram sample.

[0022] In step 2, the methods for enhancing the dynamometer image of a scarce working condition oil well specifically include: translation, scaling, and mirroring.

[0023] Step 2, the translational image enhancement specifically includes: for operating conditions with scarce samples, selecting a small number of existing samples, and acquiring the original displacement load data points of the oil well dynamometer diagram sample, including the following steps:

[0024] A1: By keeping the original position data unchanged and uniformly adding or subtracting a fixed parameter from the original load data, the dynamometer card of this sample type can be shifted vertically upward or downward. This method can obtain dynamometer card samples of oil wells under different load conditions under the same displacement conditions, while keeping the scarce working condition type unchanged, thus achieving the effect of expanding the sample library of this working condition type.

[0025] A2: By keeping the load data unchanged and uniformly adding or subtracting a fixed parameter from the original displacement data, the dynamometer card of this sample type can be shifted laterally to the left or right. This method can obtain dynamometer card samples of oil wells under different displacement conditions under the same load conditions, while keeping the scarce working condition type unchanged, thus achieving the effect of expanding the sample library of this working condition type.

[0026] A3: By combining the above two image translation methods and simultaneously changing the displacement and load data of the original oil well dynamometer diagram sample, and uniformly adding or subtracting a fixed parameter to the original displacement data and uniformly adding or subtracting a fixed parameter to the original load data, it is possible to move the oil well dynamometer diagram of this sample type in different directions while maintaining the rare working condition type unchanged, thus achieving the effect of expanding the sample library of this working condition type.

[0027] Step 2, image enhancement using mirroring specifically includes: for operating conditions with scarce samples, selecting a small number of existing samples, and acquiring the original displacement load data points of the oil well dynamometer diagram sample, including the following steps:

[0028] B1: Keep the displacement data points unchanged, only mirror the load data points to obtain a new load data list. This method may change the working condition type of the original oil well dynamometer. The newly generated oil well dynamometer can be classified by manual sorting.

[0029] B2: Based on the new oil well dynamometer sample obtained by the translation method, a secondary image enhancement method using mirroring is used to obtain a new oil well dynamometer sample.

[0030] Step 2, image enhancement using scaling specifically includes: for operating conditions with scarce samples, selecting a small number of existing samples to acquire the original oil well dynamometer image for that operating condition, including the following steps:

[0031] C1: Based on the acquired original oil well dynamometer image, first obtain the original size of the image. If the width of the original oil well dynamometer image remains unchanged, multiplying the image length by a fixed parameter can compress or enlarge the original oil well dynamometer sample in the horizontal direction.

[0032] C2: Based on the acquired original oil well dynamometer image, first obtain the original size of the image. If the length of the original oil well dynamometer image remains unchanged, multiplying the image width by a fixed parameter can compress or enlarge the original oil well dynamometer sample in the vertical direction.

[0033] C3: Based on the new oil well dynamometer sample obtained by the above translation and mirroring methods, a secondary image enhancement is performed by scaling image enhancement method to obtain a new oil well dynamometer sample.

[0034] In step 2, when using image enhancement by scaling or panning, if the fixed parameter value is appropriate, the scarce working condition type can be kept unchanged, thus expanding the sample library of the working condition type. If the fixed parameter value is too small or too large, the working condition type of the oil well dynamometer will change. The newly generated oil well dynamometer can be classified by manual sorting.

[0035] In step 3, the working condition type of the new oil well dynamometer diagram sample is determined by manual sorting and then classified and organized.

[0036] In step 4, the newly classified dynamometer sample set and the original sample set are integrated to form a new training sample set.

[0037] In step 5, a variational autoencoder (VAE) model is constructed, and the prepared training sample set is fed into the model for training. The model generates similar graphics by acquiring image features from the captured sample set.

[0038] In step 6, for the sample set of scarce working conditions, the trained VAE model is used to identify and generate samples of that scarce working condition type.

[0039] In step 7, the new dynamometer diagram sample generated by the VAE model is put into the working condition diagnosis system for verification to check whether it meets the rare working condition type. If it meets the requirements, the dynamometer diagram is expanded into the training sample set for further training to improve model performance and also expand the sample set of the rare working condition type.

[0040] The objective of this invention can also be achieved through the following technical measures: an oil well dynamometer image data enhancement system, which uses an oil well dynamometer image data enhancement method to perform data enhancement processing on dynamometer images of rare working conditions, thereby expanding the sample set of scarce working condition categories.

[0041] To address the challenge of limited samples for complex operating conditions, which hinders the training of intelligent operating condition diagnostic systems, this invention provides an image data augmentation-based method. This method solves the problem of lacking advanced dynamometer image enhancement techniques to expand the sample library of scarce oil well dynamometer image types. It performs data augmentation on dynamometer images for rare operating conditions, initially expanding the sample set for scarce operating condition categories. A variational autoencoder (VAE) model is constructed to train the dynamometer image sample set, capturing dynamometer image features. By identifying dynamometer image samples of scarce operating condition types, new dynamometer images for that operating condition category are generated. Unlike traditional image enhancement techniques, due to the unique characteristics of oil well dynamometer images, this invention aims to expand the sample library while preserving the original image features, minimizing significant changes to the shape of the target sample image. If dynamometer images for other operating condition types are generated simultaneously, the sample library for those other types can also be expanded.

[0042] The beneficial effects of this invention are as follows: while retaining the value of oil well dynamometer diagrams, it greatly expands the training library of oil well dynamometer diagrams with a small sample size, improves the comprehensiveness and accuracy of the intelligent working condition diagnosis system for this type of working condition, and greatly enhances the ability of the intelligent working condition diagnosis system to diagnose the working conditions of the target oilfield block, which can bring huge benefits to the oilfield. Attached Figure Description

[0043] Figure 1 This is a schematic diagram of the basic process of a dynamometer image data enhancement method according to an embodiment of the present invention;

[0044] Figure 2 An example diagram illustrating a dynamometer image data enhancement method according to an embodiment of the present invention, using the "gas influence" working condition as an example;

[0045] Figure 3 An example diagram illustrating a method for enhancing dynamometer image data according to an embodiment of the present invention, which involves translating the original dynamometer sample longitudinally;

[0046] Figure 4 An example diagram illustrating a dynamometer image data enhancement method provided in an embodiment of the present invention, which involves laterally shifting the original dynamometer sample;

[0047] Figure 5An example diagram illustrating a method for enhancing dynamometer image data according to an embodiment of the present invention, which involves translating the original dynamometer sample in different directions;

[0048] Figure 6 This is an example diagram illustrating a method for enhancing dynamometer image data according to an embodiment of the present invention, which involves mirroring the displacement data points of the original dynamometer sample.

[0049] Figure 7 An example diagram illustrating a method for enhancing dynamometer image data according to an embodiment of the present invention, which involves scaling the original dynamometer sample horizontally;

[0050] Figure 8 An example diagram illustrating a method for enhancing dynamometer image data according to an embodiment of the present invention, which involves scaling the original dynamometer sample in the longitudinal direction;

[0051] Figure 9 This is a comparison of examples of a dynamometer image data enhancement method provided in an embodiment of the present invention, which involves vertically magnifying a sample of a "normal" operating condition. Detailed Implementation

[0052] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0053] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments of the present invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, and / or combinations thereof.

[0054] like Figure 1 As shown, Figure 1 This is a flowchart of the oil well dynamometer image data enhancement method of the present invention. The oil well dynamometer image data enhancement method includes the following steps:

[0055] S1: Based on the existing oil well dynamometer diagram sample library in the big data database of the oilfield information platform, sort and classify the working condition types that lack working condition samples, and collect displacement and load data of the original oil well dynamometer diagrams;

[0056] Furthermore, for operating condition types lacking sample data, the process involves organizing and classifying them, and collecting displacement and load data from original oil well dynamometer diagrams. Specifically, this includes: utilizing the oil well dynamometer diagram sample library already collected and organized in the oilfield information platform's big data database, manually calibrating the sample quantity for each operating condition type, classifying the operating condition types with scarce sample quantities separately, and collecting and organizing the original displacement and load data of the dynamometer diagram samples for each scarce operating condition type to form a new sample library of displacement and load data for scarce operating condition types.

[0057] S2: By using image enhancement methods, the displacement load data in the scarce working condition type displacement load data sample library is enhanced, and the new data points generated after enhancement are organized and drawn into a new oil well dynamometer diagram sample.

[0058] Furthermore, methods for enhancing dynamometer images of oil wells operating under scarce conditions specifically include: translation, scaling, and mirroring.

[0059] Specifically, the translational image enhancement includes: for operating conditions with scarce samples, selecting a small number of existing samples and collecting the original displacement load data points of the oil well dynamometer diagram sample, including the following steps:

[0060] A1: By keeping the original position data unchanged and uniformly adding or subtracting a fixed parameter, such as 10 or 15, from the original load data, the dynamometer card of this sample type can be shifted vertically upward or downward. This method can obtain dynamometer card samples of oil wells under different load conditions under the same displacement conditions, while keeping the scarce working condition type unchanged, thus expanding the sample library of this working condition type.

[0061] A2: By keeping the load data constant and uniformly adding or subtracting a fixed parameter, such as 1 or 2, from the original displacement data, the dynamometer card of this sample type can be shifted laterally to the left or right. This method can obtain dynamometer card samples of oil wells under the same load conditions but with different displacement conditions, while maintaining the scarce operating condition type unchanged, thus expanding the sample library for this operating condition type.

[0062] A3: By combining the image translation methods a and b, and simultaneously changing the displacement and load data of the original oil well dynamometer diagram sample, a fixed parameter, such as 1 or 2, is uniformly added to or subtracted from the original displacement data, and a fixed parameter, such as 10 or 15, is uniformly added to or subtracted from the original load data. This allows the oil well dynamometer diagram of this sample type to be moved in different directions while maintaining the rare working condition type, thus expanding the sample library for this working condition type.

[0063] Specifically, image enhancement using mirroring includes: for operating conditions with scarce samples, selecting a small number of existing samples and collecting the original displacement load data points of the oil well dynamometer diagram sample, including the following steps:

[0064] B1: Keep the displacement data points unchanged, only mirror the load data points to obtain a new load data list. This method may change the working condition type of the original oil well dynamometer. The newly generated oil well dynamometer can be classified by manual sorting.

[0065] B2: Based on the new oil well dynamometer sample obtained by the translation method, a secondary image enhancement method using mirroring is used to obtain a new oil well dynamometer sample.

[0066] Specifically, image enhancement using scaling includes: for operating conditions with scarce samples, selecting a small number of existing samples to acquire the original oil well dynamometer image for that operating condition, including the following steps:

[0067] C1: Based on the acquired original oil well dynamometer image, first obtain the original size of the image. If the width of the original oil well dynamometer image remains unchanged, multiply the image length by a fixed parameter such as 0.9, 2, 15, etc., to compress or enlarge the original oil well dynamometer sample in the horizontal direction.

[0068] C2: Based on the acquired original oil well dynamometer image, first obtain the original size of the image. If the length of the original oil well dynamometer image remains unchanged, multiply the image width by a fixed parameter such as 0.9, 2, 15, etc., to compress or enlarge the original oil well dynamometer sample in the vertical direction.

[0069] C3: Based on the new oil well dynamometer sample obtained by the above translation and mirroring methods, a secondary image enhancement is performed by scaling image enhancement method to obtain a new oil well dynamometer sample.

[0070] If the fixed parameter value is appropriate, the scarce working condition type can be kept unchanged, thus expanding the sample library of this working condition type. If the fixed parameter value is too small or too large, the working condition type of the oil well dynamometer will change. The newly generated oil well dynamometer can be classified by manual sorting.

[0071] S3: Use manual sorting to determine the operating condition type of the new oil well dynamometer diagram samples and classify and organize them.

[0072] S4: Integrate the newly classified dynamometer sample set with the original sample set to form a new training sample set.

[0073] S5: Construct a variational autoencoder (VAE) model, and feed the sorted training sample set into the model for training. The model generates similar graphics by acquiring image features from the captured sample set.

[0074] S6: For a sample set of scarce working conditions, use the trained VAE model to identify and generate samples of that scarce working condition type.

[0075] S7: The new dynamometer diagram sample generated by the VAE model is put into the working condition diagnosis system for verification to check whether it meets the rare working condition type. If it meets the requirements, the dynamometer diagram is expanded into the training sample set to further train and improve the model performance, and also expand the sample set of the rare working condition type.

[0076] Furthermore, the newly generated oil well dynamometer diagram samples through image enhancement technology are manually classified, and the classified samples are put into the intelligent diagnostic system for verification and training. Samples that meet the requirements can be repeatedly image enhanced, and the sample library of scarce working conditions can be continuously expanded.

[0077] The following are several specific embodiments of the application of the present invention.

[0078] Example 1:

[0079] Reference Figure 2-5 This is an embodiment of the present invention. Taking the original oil well dynamometer diagram sample with gas-affected working conditions as an example, this embodiment provides a translational image enhancement method, which is scientifically demonstrated through specific implementation methods and implementation effects.

[0080] The specific details of this embodiment are as follows:

[0081] Based on dynamometer diagram samples of oil wells actually collected from a certain oilfield, taking the dynamometer diagram sample of the "gas-affected" operating condition as an example, the specific dynamometer diagram image is as follows: Figure 1 As shown.

[0082] Reference Figure 1 Based on the original oil well indicator diagram samples collected, the original displacement load data points were further obtained. Some displacement load data are shown in Table 1.

[0083] Table 1. Displacement load data points from the original oil well indicator diagram (partial)

[0084]

[0085]

[0086] Based on the displacement load data of the original oil well dynamometer diagrams obtained through further acquisition, the displacement load data of the original oil well dynamometer diagrams were processed using a translational image enhancement method. The resulting sample of the new oil well dynamometer diagram after translational processing is shown below. Figure 3-5 As shown.

[0087] Reference Figure 3 (1), where A is the original dynamometer image and B1 is the new oil well dynamometer sample obtained after image enhancement processing. Specifically, by keeping the displacement data points unchanged, the original load data is uniformly increased by 10, and the original oil well dynamometer sample is shifted upward, keeping the "gas influence" working condition type unchanged and expanding the sample library of this working condition type.

[0088] Reference Figure 3 (2), where A is the original dynamometer image and B2 is the new oil well dynamometer sample obtained after image enhancement processing. Specifically, by keeping the displacement data points unchanged, the original load data is uniformly reduced by 10, and the original oil well dynamometer sample is shifted downward, keeping the "gas influence" working condition type unchanged and expanding the sample library of this working condition type.

[0089] Reference Figure 4 (1), where A is the original dynamometer image and C1 is the new oil well dynamometer sample obtained after image enhancement processing. Specifically, by keeping the load data points unchanged, adding 2 to the original displacement data, and shifting the original oil well dynamometer sample to the right, the "gas influence" working condition type is kept unchanged, and the sample library of this working condition type is expanded.

[0090] Reference Figure 4 (2), where A is the original dynamometer image and C2 is the new oil well dynamometer sample obtained after image enhancement processing. Specifically, by keeping the load data points unchanged, adding 2 to the original displacement data, and shifting the original oil well dynamometer sample to the right, the "gas influence" working condition type is kept unchanged, and the sample library of this working condition type is expanded.

[0091] Reference Figure 5 Where A is the original dynamometer image, and D1-D4 are new oil well dynamometer samples obtained after image enhancement processing. By uniformly adding or subtracting 2 from the original displacement data and uniformly adding or subtracting 10 from the original load data, translation in four directions was achieved, keeping the "gas influence" working condition type unchanged and expanding the sample library for this working condition type.

[0092] Example 2:

[0093] Reference Figure 6This is an embodiment of the present invention. Taking the original oil well dynamometer diagram sample with gas-affected working conditions as an example, this embodiment provides a mirror image enhancement method, which is scientifically demonstrated through specific implementation methods and implementation effects.

[0094] The specific details of this embodiment are as follows:

[0095] Based on the displacement load data of the original oil well dynamometer diagrams collected in Table 1, the displacement load data of the original oil well dynamometer diagrams were processed using a mirroring image enhancement method. The resulting sample of the new oil well dynamometer diagram after translation processing is shown below. Figure 6 As shown.

[0096] Reference Figure 6 Where A is the original dynamometer image, and E is the new oil well dynamometer image sample obtained after image enhancement processing. Keeping the displacement data points unchanged, only the load data points are mirrored to obtain a new dynamometer image for the new operating condition type. The newly generated oil well dynamometer image is inconsistent with the original operating condition type. Through manual sorting, it can be determined that this operating condition type is a mirror flip due to acquisition errors, and the sample library for this type of operating condition can be further expanded.

[0097] Example 3:

[0098] Reference Figure 7-9 As an embodiment of the present invention, this embodiment takes the original oil well dynamometer diagram sample with gas-affected operating conditions as an example, and provides a scaling image enhancement method, which is scientifically demonstrated through specific implementation methods and implementation effects.

[0099] The specific details of this embodiment are as follows:

[0100] according to Figure 2 The original oil well dynamometer image was acquired, and its original dimensions were obtained. If the width of the original oil well dynamometer image remained unchanged, a scaling image enhancement method was used to process the original oil well dynamometer image. The resulting new oil well dynamometer sample after scaling is shown below. Figure 7-9 As shown.

[0101] Reference Figure 7 (1) Keeping the width of the original oil well dynamometer image unchanged, the new oil well dynamometer sample is obtained by multiplying the image length by 0.9. This achieves a horizontal compression of the original oil well dynamometer sample by 0.9 times. The sample keeps the "gas influence" working condition type unchanged and expands the sample library of this working condition type.

[0102] Reference Figure 7(2) Keeping the width of the original oil well dynamometer image unchanged, the new oil well dynamometer sample is obtained by multiplying the image length by 2, which realizes that the original oil well dynamometer sample is horizontally enlarged by 2 times. The sample keeps the "gas influence" working condition type unchanged and expands the sample library of this working condition type.

[0103] Reference Figure 8 (1) Keeping the length of the original oil well dynamometer image unchanged, the image width is multiplied by 0.9 to obtain a new oil well dynamometer sample, which achieves a longitudinal compression of the original oil well dynamometer sample by 0.9 times. This sample keeps the "gas influence" working condition type unchanged and expands the sample library of this working condition type.

[0104] Reference Figure 8 (2) Keeping the length of the original oil well dynamometer image unchanged, the image width is multiplied by 2 to obtain a new oil well dynamometer sample, which realizes that the original oil well dynamometer sample is vertically enlarged by 2 times. The sample keeps the "gas influence" working condition type unchanged and expands the sample library of this working condition type.

[0105] Reference Figure 9 For special cases, the dynamometer diagram of an oil well with the selected operating condition type as normal is shown below. Figure 9 (1); refer to Figure 9 (2) Keeping the width of the original oil well dynamometer image unchanged, the image length is multiplied by 15 to obtain a new oil well dynamometer sample, which realizes that the original oil well dynamometer sample is horizontally enlarged by 15 times. The working condition type of this sample is inconsistent with that of the original oil well dynamometer. It can be known by manual sorting that the newly generated working condition type is rod breakage, which can further expand the sample library of this type of working condition.

[0106] Example 4:

[0107] In one embodiment 4 of the present invention, taking the original oil well dynamometer diagram sample with gas-affected working conditions as an example, a method for generating similar images using a VAE model is provided, and the specific implementation method and implementation effect are scientifically demonstrated.

[0108] The specific details of this embodiment are as follows:

[0109] The new dynamometer samples obtained through data augmentation in Examples 1-3 above are organized and integrated with the original sample set to form a training sample set. A variational autoencoder (VAE) model is constructed, and the specific architecture of the model is shown in Table 2.

[0110] Table 2. Variational Autoencoder (VAE) Model Architecture Parameters

[0111]

[0112] The collected training sample set is fed into the model for training. The trained VAE model is then used to identify scarce operating condition types and generate similar graphics. The newly generated dynamometer images are then fed into an intelligent operating condition diagnosis system for verification. If they match the corresponding operating condition type, the sample set for that scarce operating condition type is further expanded. Simultaneously, the new samples are added to the training sample set to retrain the VAE model, further improving its performance.

[0113] Compared with existing technologies, this invention has significant advantages. While retaining the value of oil well dynamometer cards, it solves the problem of insufficient training sample sets, thus greatly improving the comprehensiveness and accuracy of the intelligent diagnostic system for this type of operating condition. This significantly enhances the system's ability to diagnose the operating conditions of target oilfield blocks, bringing substantial benefits to the oilfield.

[0114] Finally, it should be noted that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0115] Except for the technical features described in the specification, all other technologies are known to those skilled in the art.

Claims

1. A method for enhancing oil well dynamometer image data, characterized in that, The data augmentation method for the oil well indicator image includes: Step 1: Organize and classify the working condition types that lack working condition samples to obtain a normal working condition type sample set and a scarce working condition type sample set; Step 2: Enhance the original displacement load data of the indicator diagram samples in the scarce working condition type sample library by using image enhancement methods; Step 3: Draw new dynamometer diagrams from the enhanced data points and classify the new dynamometer diagram types. Step 4: Integrate the scarce working condition type sample set and the normal type dynamometer sample set of the newly expanded dynamometer diagram to form a training sample set; Step 5: Construct a variational autoencoder (VAE) model and input the training sample set into the model for learning and training; Step 6: Use the trained VAE model to identify the dynamometer diagram sample set of scarce working conditions to generate similar dynamometer diagram images. Step 7: Input the newly generated dynamometer diagram samples into the operating condition diagnostic system for verification, and re-input the newly generated samples into the training sample set to further improve the model performance.

2. The oil well dynamometer image data enhancement method according to claim 1, characterized in that, In step 1, based on the existing oil well dynamometer diagram sample library in the big data database of the oilfield information platform, the working condition types lacking working condition samples are sorted and classified to obtain the normal working condition type sample set and the scarce working condition type sample set, and the displacement and load data of the original oil well dynamometer diagrams in the scarce working condition type sample set are collected.

3. The oil well dynamometer image data enhancement method according to claim 1, characterized in that, In step 1, using the oil well dynamometer diagram sample library already collected and organized in the big data database of the oilfield information platform, the number of samples for each type of working condition is counted through manual calibration. Working condition types with few samples are classified separately, and the original displacement and load data of the dynamometer diagram samples for each scarce working condition type are collected and organized to form a new sample library of displacement and load data for scarce working condition types.

4. The oil well dynamometer image data enhancement method according to claim 1, characterized in that, In step 2, the displacement load data in the scarce working condition type displacement load data sample library is enhanced by image enhancement method, and the new data points generated after enhancement are organized and drawn into a new oil well dynamometer diagram sample.

5. The oil well dynamometer image data enhancement method according to claim 4, characterized in that, In step 2, the methods for enhancing the dynamometer image of a scarce working condition oil well specifically include: translation, scaling, and mirroring.

6. The oil well dynamometer image data enhancement method according to claim 5, characterized in that, Step 2, the translational image enhancement specifically includes: for operating conditions with scarce samples, selecting a small number of existing samples, and acquiring the original displacement load data points of the oil well dynamometer diagram sample, including the following steps: A1: By keeping the original position data unchanged and uniformly adding or subtracting a fixed parameter from the original load data, the dynamometer card of this sample type can be shifted vertically upward or downward. This method can obtain dynamometer card samples of oil wells under different load conditions under the same displacement conditions, while keeping the scarce working condition type unchanged, thus achieving the effect of expanding the sample library of this working condition type. A2: By keeping the load data unchanged and uniformly adding or subtracting a fixed parameter from the original displacement data, the dynamometer card of this sample type can be shifted laterally to the left or right. This method can obtain dynamometer card samples of oil wells under different displacement conditions under the same load conditions, while keeping the scarce working condition type unchanged, thus achieving the effect of expanding the sample library of this working condition type. A3: By combining the above two image translation methods and simultaneously changing the displacement and load data of the original oil well dynamometer diagram sample, and uniformly adding or subtracting a fixed parameter to the original displacement data and uniformly adding or subtracting a fixed parameter to the original load data, it is possible to move the oil well dynamometer diagram of this sample type in different directions while maintaining the rare working condition type unchanged, thus achieving the effect of expanding the sample library of this working condition type.

7. The oil well dynamometer image data enhancement method according to claim 6, characterized in that, Step 2, image enhancement using mirroring specifically includes: for operating conditions with scarce samples, selecting a small number of existing samples, and acquiring the original displacement load data points of the oil well dynamometer diagram sample, including the following steps: B1: Keep the displacement data points unchanged, only mirror the load data points to obtain a new load data list. This method may change the working condition type of the original oil well dynamometer. The newly generated oil well dynamometer can be classified by manual sorting. B2: Based on the new oil well dynamometer sample obtained by the translation method, a secondary image enhancement method using mirroring is used to obtain a new oil well dynamometer sample.

8. The oil well dynamometer image data enhancement method according to claim 7, characterized in that, Step 2, image enhancement using scaling specifically includes: for operating conditions with scarce samples, selecting a small number of existing samples to acquire the original oil well dynamometer image for that operating condition, including the following steps: C1: Based on the acquired original oil well dynamometer image, first obtain the original size of the image. If the width of the original oil well dynamometer image remains unchanged, multiplying the image length by a fixed parameter can compress or enlarge the original oil well dynamometer sample in the horizontal direction. C2: Based on the acquired original oil well dynamometer image, first obtain the original size of the image. If the length of the original oil well dynamometer image remains unchanged, multiplying the image width by a fixed parameter can compress or enlarge the original oil well dynamometer sample in the vertical direction. C3: Based on the new oil well dynamometer diagram samples obtained by the above translation and mirroring methods, secondary image enhancement is performed by scaling image enhancement method to obtain new oil well dynamometer diagram samples.

9. The oil well dynamometer image data enhancement method according to claim 8, characterized in that, In step 2, when using image enhancement by scaling or panning, if the fixed parameter value is appropriate, the scarce working condition type can be kept unchanged, thus expanding the sample library of the working condition type. If the fixed parameter value is too small or too large, the working condition type of the oil well dynamometer will change. The newly generated oil well dynamometer can be classified by manual sorting.

10. The method for enhancing oil well dynamometer image data according to claim 1, characterized in that, In step 3, the working condition type of the new oil well dynamometer diagram sample is determined by manual sorting and then classified and organized.

11. The method for enhancing oil well dynamometer image data according to claim 1, characterized in that, In step 4, the newly classified dynamometer sample set and the original sample set are integrated to form a new training sample set.

12. The method for enhancing oil well dynamometer image data according to claim 1, characterized in that, In step 5, a variational autoencoder (VAE) model is constructed, and the prepared training sample set is fed into the model for training. The model generates similar graphics by acquiring image features from the captured sample set.

13. The oil well dynamometer image data enhancement method according to claim 1, characterized in that, In step 6, for the sample set of scarce working conditions, the trained VAE model is used to identify and generate samples of that scarce working condition type.

14. The oil well dynamometer image data enhancement method according to claim 1, characterized in that, In step 7, the new dynamometer diagram sample generated by the VAE model is put into the working condition diagnosis system for verification to check whether it meets the rare working condition type. If it meets the requirements, the dynamometer diagram is expanded into the training sample set for further training to improve model performance and also expand the sample set of the rare working condition type.

15. An oil well dynamometer image data enhancement system, characterized in that, The oil well dynamometer image data enhancement system uses the oil well dynamometer image data enhancement method described in any one of claims 1-14 to perform data enhancement processing on the dynamometer images of rare working conditions, thereby expanding the sample set of scarce working condition categories.

Images