Discrimination method for position and type of centralizer based on multi-source data fusion

Abstract

This invention provides a method for determining the placement location and type of a centralizer based on multi-source data fusion, comprising: Step 1, acquiring casing record table data before the casing operation begins; Step 2, acquiring real-time text data and real-time video data at the start of the casing operation; Step 3, determining the target casing information currently being lowered and the type of centralizer to be placed therein based on the real-time text data; Step 4, determining the type of centralizer actually placed on the target casing based on the real-time video data; Step 5, determining the specific time of centralizer placement on the target casing based on the real-time video data; Step 6, determining whether the corresponding type of centralizer was correctly placed on the target casing during the actual placement process. This method for determining the placement location and type of a centralizer based on multi-source data fusion standardizes the casing lowering operation process, effectively replaces traditional manual monitoring methods, and improves on-site safety and supervision efficiency.

Description

Technical Field

[0001] This invention relates to the field of computer vision technology, and in particular to a method for determining the placement and type of a centering device through multi-source data fusion. Background Technology

[0002] In the field of oilfield safety production, traditional centralizer identification methods that mainly rely on human vision or simple image processing algorithms have gradually revealed significant limitations: traditional methods usually require manual intervention or operation, which is time-consuming and labor-intensive, and is easily affected by human factors; identification based on human vision is easily affected by factors such as lighting and angle, resulting in limited accuracy; in complex environments, such as low light or obstruction, the identification effect of traditional methods will drop significantly; traditional methods often cannot monitor the position and status of the centralizer in real time, making it difficult to detect problems and take timely measures.

[0003] During the on-site monitoring of the centralizer installation in casing installation operations, there are problems such as the inability to achieve effective full coverage by relying solely on manpower, the inability of monitoring and identification methods to meet on-site needs, and the low efficiency and difficulty in real-time operation of existing management methods. This can no longer meet the stringent requirements of today's safety production management.

[0004] Chinese patent application CN118462147A discloses a method for visually detecting the orientation of optical fibers outside the casing, comprising the following steps: S1, setting inner circumferential angle graduation lines on the inner side of the male coupling, and setting one or more specific patterns for easy identification of the circumferential orientation angle; S2, simultaneously lowering the casing and binding the optical fiber as required by the process outside the casing, recording the clockwise or counterclockwise offset angle of the optical fiber from the orientation mark inside the casing on each outside the casing; S3, recording the deviation angle of each orientation mark inside the casing from the gyroscope orientation or gravity orientation; S4, superimposing the deviation angle of each orientation mark inside the casing with the recorded offset angle of the optical fiber outside the same casing, thereby calculating the actual orientation of the optical fiber outside the casing, and using the detected position of the cable outside the casing to guide the delineation of the safe perforation zone. This invention uses a visual method to identify downhole markings, intuitively determine the depth and orientation of the optical fiber outside the casing, guide the delineation of the safe perforation zone, and reduce the false perforation rate. This application differs from the present invention in the following ways: 1) The objects of identification are different. The object of identification in the present invention is a casing external centralizer, which is a discontinuous ring structure used to keep the casing centered. The object of identification in this application is a continuous wire tool for identifying the position of the external optical fiber. 2) The types of usage methods are different. The method of use in the present invention is mainly based on computer vision. The method of this application is a purely physical method. 3) The problems solved are different. The present invention solves the problem of poor casing centering during drilling and completion. The application solves the problem of misfiring rate during operation.

[0005] Chinese patent application CN117454591A discloses a method for optimizing the placement of a hydraulic oscillator. The method involves first collecting data, then using an axial vibration propagation model to simulate and calculate the drilling pressure. Next, it determines the effective traction force coefficient and obtains a curve showing the effective traction force coefficient changing with the placement position. Finally, it determines the placement position of the hydraulic oscillator and simulates the drilling pressure curve under drill bit forward conditions. This method considers the influence of mechanical drilling speed on the effective traction force coefficient, resulting in a more accurate structure, less data processing, and no need for complex field equipment. It effectively and accurately obtains the optimal placement position of the hydraulic oscillator, improving its friction-reducing effect in horizontal well drilling. This application differs from the focus of this invention. The invention identifies whether the casing centralizer is placed according to a pre-established plan, while this application identifies the optimal placement position under the existing wellbore environment.

[0006] Chinese patent application CN118855400A discloses a drilling tool and its control method, device, and medium, comprising: a controller, a hydraulic diameter-changing actuator, a communication unit, and at least one diameter-changing stabilizer. The controller is connected to the communication unit to acquire an inclination change command, thereby adjusting the working state of the diameter-changing stabilizer according to the inclination change command. The controller generates a diameter-changing stabilizer control signal based on the inclination change command and a build-up capability model to adjust the inclination of the drilling tool. Both the hydraulic diameter-changing actuator and the controller are connected to the diameter-changing stabilizer to control its working state according to the control signal. This application generates a diameter-changing stabilizer control signal by acquiring an inclination change command and a pre-set build-up capability model, thereby adjusting the working state of the diameter-changing stabilizer to regulate the build-up capability of the drilling tool. This allows for changing the build-up capability of the drilling tool without requiring tripping or running-in operations, improving the drilling tool's working efficiency. This application differs from the item identified in this invention. The item identified in this invention is whether the casing centralizer is installed according to a pre-established plan, while the item identified in this application is the placement and usage method of the variable diameter drill pipe centralizer under existing drill string assemblies and drilling purposes.

[0007] The existing technologies described above are significantly different from this invention. A search reveals no literature of the XY category, indicating the innovativeness of this invention. Since existing technologies lack solutions to the technical problem we aim to address, we have invented a novel method for determining the placement and type of a centralizer based on multi-source data fusion. Summary of the Invention

[0008] The purpose of this invention is to provide a multi-source data fusion method for determining the placement and type of centralizers, which automatically detects the placement of centralizers, provides real-time early warnings, reduces manpower and material costs, and efficiently monitors the correct placement of centralizers throughout the entire process of casing installation.

[0009] The objective of this invention can be achieved through the following technical measures: a method for determining the placement location and type of a centralizer based on multi-source data fusion, wherein the method includes:

[0010] Step 1: Before starting the casing operation, obtain the casing record table data;

[0011] Step 2: When the casing operation begins, acquire real-time text data and real-time video data;

[0012] Step 3: Based on real-time text data, determine the target casing information that is currently being lowered and the type of centralizer that should be installed;

[0013] Step 4: Determine the type of centralizer actually installed on the target casing based on real-time video data;

[0014] Step 5: Determine the specific time for placing the centralizer on the target casing based on real-time video data;

[0015] Step 6: Determine whether the corresponding type of centralizer was correctly installed during the actual installation of the target sleeve.

[0016] The objective of this invention can also be achieved through the following technical measures:

[0017] In step 3, based on the real-time text data, parameters such as data acquisition time, hook position, and hook load are extracted, and the casings in the casing record table are matched to determine the target casing information currently being lowered and the type of centralizer that should be installed.

[0018] Step 3 includes:

[0019] Step 301: Determine the sleeve information and lowering sequence of the centralizer to be tested using the sleeve record sheet;

[0020] Step 302: Determine the actual placement time of the target casing by analyzing the data acquisition time, hook height, and hook load variation trends in the real-time text data.

[0021] In step 4, real-time video data is input into the target detection model for centralizer placement to obtain the relative positions of the centralizer, casing wrench, and wellhead, and to determine the type of centralizer actually placed on the target casing.

[0022] Step 4 includes:

[0023] Step 401: Use the real-time bushing installation video stream as the video data input source;

[0024] Step 402: Use the stabilizer placement target detection model obtained by improving and training the Gold-YOLO-L model. The function of the stabilizer placement target detection model is to identify the type and location of stabilizers, wellheads, hydraulic tongs, and casing wrenches in the image.

[0025] Step 403: Input the video data into the target detection model for the centralizer placement. After detecting the wellhead, casing wrench and centralizer, determine whether the centralizer is located above the wellhead and whether the overlapping area between the centralizer and the casing wrench is greater than the threshold. If all conditions are met, save the frame image and the information of the centralizer.

[0026] In step 4, step 402 specifically includes:

[0027] Step 4a: Frame-by-frame processing of the video stream data from the oilfield operation site to obtain image data;

[0028] Step 4b: Considering that the casing operation occurs near the wellhead, the wellhead, casing wrench, centralizer, and hydraulic tongs in the image data are calibrated to obtain the image and the corresponding tag file;

[0029] Step 4c: Construct a target detection model for centering device placement based on Gold-YOLO-L within the PyTorch framework;

[0030] Step 4d: Set the training parameters for the network;

[0031] Step 4e: Train a target detection model for centering device placement using the casing operation dataset;

[0032] Step 4f: Adjust the network parameters based on the training results, and select the model with the best training results as the final model.

[0033] In step 4, the target detection model for centralizer placement based on Gold-YOLO-L includes:

[0034] The image processing module performs frame-by-frame processing on the real-time video data from the casing installation site to obtain image data;

[0035] The network module uses a centralizer to place the target detection model, analyzes and processes the obtained image, extracts the position coordinates of the wellhead, casing, hydraulic tongs, and casing tongs areas, and outputs prediction information.

[0036] The determination and storage module stores the location information and determination information of the recognition results based on the prediction information.

[0037] In step 4, the network module includes:

[0038] The backbone network uses a pre-built straightener to place the backbone network in the target detection model to extract multi-scale features of the input image;

[0039] The neck network, which integrates feature maps from different scales, improves the accuracy and robustness of the model.

[0040] The classification network uses a loss function to classify and regress image features and aggregated information of the wellhead area to obtain prediction information.

[0041] The training optimization unit adjusts the network parameters based on the training results and selects the model with the best training results as the final model.

[0042] In step 403, the target detection judgment logic of the target detection model is placed in the straightener, including the following steps:

[0043] (51) Input the real-time data image into the target detection model for the centering device;

[0044] (52) If there is no wellhead coordinate information, detect the wellhead position in the image and save it;

[0045] (53) Detect the type of stabilizer in the image, the coordinate position of the stabilizer and the casing wrench, and determine whether the stabilizer is located above the wellhead and whether the overlapping area with the casing wrench is greater than a predetermined value.

[0046] (54) If condition (53) is met, save the image and the straightening device information;

[0047] (55) Determine whether all sleeves have been lowered. If not, return to step (51). If they have been lowered, end the target detection.

[0048] In step 5, real-time video data is input into the text recognition model to determine the specific time for placing the centralizer on the target casing.

[0049] Step 5 includes:

[0050] Step 501: Apply a text recognition model to identify the time information of the real-time casing installation video and calibrate the video time;

[0051] Step 502: Identify the image that meets the conditions in the image and the information of the centralizer saved in step 4, obtain the time information of the image, and save the record.

[0052] Step 6 includes:

[0053] Step 601: Based on the data of the actual casing placement time and the centralizer placement time, determine whether the corresponding type of centralizer has been correctly placed when the casing that requires the placement of the centralizer is about to be lowered into the wellhead.

[0054] Step 602: If it is found that the centering device is not installed correctly, an abnormal alarm will be triggered to remind the operator to take corrective measures in a timely manner.

[0055] The objective of this invention can also be achieved through the following technical measures: a multi-source data fusion system for determining the placement location and type of centralizers, which uses a multi-source data fusion method to detect the quality of casing installation in oilfield production.

[0056] The multi-source data fusion method for determining the placement and type of centralizers in this invention monitors various parameters at the work site in real time. By combining preparatory work data with real-time data from the placement equipment, it accurately determines whether the centralizer has been correctly placed on the casing to be inspected. In this process, we introduce target detection computer vision technology to analyze the on-site video in real time, automatically identifying instances of incorrectly placed centralizers during on-site operations. This standardizes the casing installation process, effectively replacing traditional manual monitoring methods and improving on-site safety and supervision efficiency.

[0057] The method for determining the position and type of centralizer based on multi-source data fusion in this invention provides a new solution for oilfield safety production management with its high efficiency, accuracy and real-time characteristics. It improves the efficiency of oilfield on-site management and greatly reduces manpower and material resources, providing a strong guarantee for safe production and efficient management at oilfield operation sites. Attached Figure Description

[0058] Figure 1 A flowchart of a specific embodiment of the method for determining the placement location and type of a centralizer based on multi-source data fusion according to the present invention;

[0059] Figure 2 This is a framework diagram of a target detection model for centering device placement based on Gold-YOLO-L in a specific embodiment of the present invention;

[0060] Figure 3 This is a flowchart of a method for determining the position and type of a centralizer based on multi-source data fusion in a specific embodiment of the present invention. Detailed Implementation

[0061] 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.

[0062] 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 scope of exemplary embodiments according to the 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.

[0063] like Figure 1 As shown, Figure 1 This is a flowchart of the method for determining the placement location and type of a centralizer based on multi-source data fusion according to the present invention. The method includes:

[0064] Step 1: Before starting the casing operation, obtain the casing record table data.

[0065] Step 2: When the casing operation begins, acquire real-time text data and real-time video data.

[0066] Step 3: Based on real-time text data, extract parameters such as data acquisition time, hook position, and hook load, and match them with the casing in the casing record table. Determine the target casing information currently being lowered and the type of centralizer that should be installed;

[0067] Step 4: Input the real-time video data into the target detection model for centralizer placement to obtain the relative positions of the centralizer, casing wrench, wellhead and other operating equipment, and determine the type of centralizer actually placed on the target casing.

[0068] Step 5: Input the real-time video data into the text recognition model to determine the specific time for placing the centralizer on the target casing;

[0069] Step 6: Determine whether the corresponding type of centralizer was correctly installed during the actual installation of the target sleeve, and save the test results.

[0070] The specific implementation steps of step (3) are as follows:

[0071] (301) Determine the sleeve information and lowering sequence of the centralizer to be tested through the sleeve record sheet;

[0072] (302) The actual placement time of the target casing is determined by the data acquisition time, hook height and hook load change trends in the real-time text data.

[0073] The specific implementation steps of step (4) are as follows:

[0074] (401) Use the real-time bushing installation site video stream as the video data input source;

[0075] (402) A target detection model for centralizer placement is obtained by improving and training the Gold-YOLO-L model. The function of the target detection model for centralizer placement is to identify the type and location of centralizer, wellhead, hydraulic tongs and casing tongs in the image.

[0076] The specific implementation steps are as follows:

[0077] ① The video stream data from the oilfield operation site is processed into frames to obtain image data;

[0078] ② Considering that casing operations generally occur near the wellhead, Label Img software was used to calibrate the "wellhead", "casing wrench", "centralizer", and "hydraulic tongs" in the image data to obtain the images and corresponding label files;

[0079] ③ Construct a target detection model for centering device placement based on Gold-YOLO-L within the PyTorch framework: a backbone network for extracting multi-scale features from the input image; a neck network for processing the extracted features and fusing features from different scales; and a classification network for outputting the detected category and location information.

[0080] ④ Set the network training parameters: set the maximum number of iterations to 200; initialize the learning rate to 0.001, reduce it to 0.0001 in the 10th round, and reduce it to 0.00001 in the 50th round; use Gold_l_pre_dist.pt for the pre-trained model; set the momentum to 0.9 and the weight decay to 0.0001.

[0081] ⑤ Train a Gold-YOLO-L target detection model using the casing operation dataset;

[0082] ⑥ Adjust the network parameters based on the training results, and select the model with the best training results as the final target detection model for the straightener placement.

[0083] (403) Input the video data into the model. After detecting the wellhead, casing wrench and centralizer, determine whether the centralizer is located above the wellhead and whether the overlapping area between the centralizer and the casing wrench is greater than the threshold. If all conditions are met, save the frame image and the information of the centralizer.

[0084] The specific implementation steps of step (5) are as follows:

[0085] (501) Use a text recognition model to identify the time information of the real-time casing installation video and calibrate the video time;

[0086] (502) Identify the information and image of the stabilizer saved in step 4 (specifically step 403), use the text recognition model to identify the time information in the upper right corner of the image, and save the record.

[0087] The specific implementation steps of step (6) are as follows:

[0088] (601) Based on the data of the actual casing placement time and the centralizer placement time, determine whether the corresponding type of centralizer has been correctly placed when the casing that needs to be placed is about to be lowered into the wellhead;

[0089] (602) If the system finds that the centering device is not properly installed, it will immediately trigger an alarm to remind the operator to take corrective measures in time.

[0090] This invention aims to improve the efficiency of current casing operation site centralizer placement and detection by providing a centralizer placement and detection method that automatically extracts features, thereby increasing detection efficiency and reducing manpower and material costs. It employs multi-source data fusion technology, combining casing information from the casing record sheet, parameters such as hook position and hook load from real-time data, and target detection and text recognition results from video, achieving comprehensive integration of multi-source data. Using real-time data processing and intelligent discrimination methods, it achieves real-time monitoring and intelligent discrimination of the centralizer placement process by acquiring and processing data in real time, combined with preset discrimination logic or algorithms. This method significantly improves the timeliness and accuracy of discrimination, effectively solving the problems of untimely manual monitoring and insufficient perspective. It integrates target detection and text recognition models to accurately identify the relative positions of operating equipment such as centralizers, casing clamps, and wellheads, as well as the correct placement time of the centralizer from on-site video. The application of this integrated technology greatly improves the automation level of data processing and the accuracy of identification. Real-time alarms and cloud data push: When an anomaly is detected during the installation of the centering device, this method can immediately issue a real-time alarm and push relevant data to the cloud. This real-time alarm and cloud data push mechanism helps to promptly detect and handle problems, improving the efficiency and accuracy of supervision.

[0091] The following are several specific embodiments of the application of the present invention.

[0092] Example 1

[0093] like Figure 2 As shown, Embodiment 1 of the present invention provides a framework diagram for the placement and detection of a centralizer based on an improved Gold-YOLO-L model, which is used for... Figure 1 The object detection model framework in the document includes:

[0094] The image processing module performs frame-by-frame processing on the real-time video data from the casing installation site to obtain image data;

[0095] The network module is used to analyze and process the obtained images by placing the target detection model with the centralizer, extract the position coordinates of the wellhead, casing, hydraulic tongs, and casing tongs areas, and output prediction information.

[0096] The determination and storage module stores the location information and determination information of the recognition results based on the prediction information.

[0097] In this embodiment 1, the network model includes a backbone network, a neck network, a classification network, and a training optimization unit.

[0098] The backbone network is used to extract multi-scale features of the input image using the backbone network in the pre-constructed straightener placement target detection model;

[0099] The neck network is used to fuse feature maps at different scales to improve the accuracy and robustness of the model;

[0100] The classification network is used to classify and regress the image features and the aggregated information of the wellhead area using a loss function to obtain prediction information.

[0101] The training optimization unit adjusts the network parameters based on the training results and selects the model with the best training results as the final model.

[0102] Example 2

[0103] like Figure 3 As shown, Embodiment 2 of the present invention provides a flowchart of the process for placing the centralizer during target detection and confirmation of casing placement. Figure 1 The target detection and judgment logic in the code includes the following steps:

[0104] (51) Input the real-time data image into the target detection model for the centering device;

[0105] (52) If there is no wellhead coordinate information, detect the wellhead position in the image and save it;

[0106] (53) Detect the type of stabilizer in the image, the coordinate position of the stabilizer and the casing wrench, and determine whether the stabilizer is located above the wellhead and the overlap area with the casing wrench is greater than 0.5.

[0107] (54) If condition (53) is met, save the image and the straightening device information;

[0108] (55) Determine whether all sleeves have been lowered. If not, return to step (51). If they have been lowered, end the target detection.

[0109] In summary, the present invention has the following technical advantages:

[0110] In terms of intelligent detection and early warning, through multi-source data fusion and target detection technology, this invention can achieve intelligent detection of the placement of the centralizer and provide real-time early warnings for incorrect placement. This greatly improves the safety and management efficiency of the work site and reduces interference and errors caused by human factors. During the centralizer identification process, the detection accuracy rates for the wellhead, hydraulic tongs, centralizer, and casing tongs were 97.73%, 98.40%, 96.36%, and 95.27%, respectively.

[0111] Regarding real-time performance and accuracy, by utilizing real-time data analysis, this invention can accurately determine the sequence and timing of casing placement and compare it with the placement time of the centralizer, thereby ensuring that the centralizer is placed at the correct time. This real-time performance and accuracy are crucial for oilfield operations, preventing potential safety hazards and production delays.

[0112] In terms of application costs, compared to traditional manual monitoring methods, this invention significantly reduces the investment of manpower and resources through automated detection and early warning. Operators only need to pay attention to the system's alarm information to quickly locate problems and take corresponding measures, reducing workload and costs.

[0113] Regarding the workflow, by using a text recognition model to identify time information in the on-site video, this invention can record the precise time of the centering device's placement, providing data support for optimizing the workflow. This helps analyze bottlenecks and problems during the operation, further improving operational efficiency and safety.

[0114] In terms of scalability and flexibility, the technical solution of this invention is based on a modular design, which can be flexibly configured and adjusted according to different operational needs and environmental conditions. Furthermore, as technology continues to develop and update, this invention can also be upgraded and optimized accordingly to adapt to more complex and diverse operational scenarios.

[0115] 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.

[0116] 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 judging the placement position and type of a centralizer through multi-source data fusion, characterized in that, The method for determining the placement location and type of the centralizer through multi-source data fusion includes: Step 1: Before starting the casing operation, obtain the casing record table data; Step 2: When the casing operation begins, acquire real-time text data and real-time video data; Step 3: Based on real-time text data, determine the target casing information that is currently being lowered and the type of centralizer that should be installed; Step 4: Determine the type of centralizer actually installed on the target casing based on real-time video data; Step 5: Determine the specific time for placing the centralizer on the target casing based on real-time video data; Step 6: Determine whether the corresponding type of centralizer was correctly installed during the actual installation of the target sleeve.

2. The method according to claim 1, wherein In step 3, based on the real-time text data, parameters such as data acquisition time, hook position, and hook load are extracted, and the casings in the casing record table are matched to determine the target casing information that is currently being lowered and the type of centralizer that should be installed.

3. The method for determining the placement location and type of a centralizer based on multi-source data fusion according to claim 2, characterized in that, Step 3 includes: Step 301: Determine the sleeve information and lowering sequence of the centralizer to be tested using the sleeve record sheet; Step 302: Determine the actual placement time of the target casing by analyzing the data acquisition time, hook height, and hook load variation trends in the real-time text data.

4. The method for determining the placement location and type of a centralizer based on multi-source data fusion according to claim 1, characterized in that, In step 4, real-time video data is input into the target detection model for centralizer placement to obtain the relative positions of the centralizer, casing wrench, and wellhead, and to determine the type of centralizer actually placed on the target casing.

5. The method for determining the placement location and type of a centralizer based on multi-source data fusion according to claim 4, characterized in that, Step 4 includes: Step 401: Use the real-time bushing installation video stream as the video data input source; Step 402: Use the stabilizer placement target detection model obtained by improving and training the Gold-YOLO-L model. The function of the stabilizer placement target detection model is to identify the type and location of stabilizers, wellheads, hydraulic tongs, and casing wrenches in the image. Step 403: Input the video data into the target detection model for the centralizer placement. After detecting the wellhead, casing wrench and centralizer, determine whether the centralizer is located above the wellhead and whether the overlapping area between the centralizer and the casing wrench is greater than the threshold. If all conditions are met, save the frame image and the information of the centralizer.

6. The method for determining the placement location and type of a centralizer based on multi-source data fusion according to claim 5, characterized in that, Step 402 specifically includes: Step 4a: Frame-by-frame processing of the video stream data from the oilfield operation site to obtain image data; Step 4b: Considering that the casing operation occurs near the wellhead, the wellhead, casing wrench, centralizer, and hydraulic tongs in the image data are calibrated to obtain the image and the corresponding tag file; Step 4c: Construct a target detection model for centering device placement based on Gold-YOLO-L within the PyTorch framework; Step 4d: Set the training parameters for the network; Step 4e: Train a target detection model for centering device placement using the casing operation dataset; Step 4f: Adjust the network parameters based on the training results, and select the model with the best training results as the final model.

7. The method for determining the placement location and type of a centralizer based on multi-source data fusion according to claim 6, characterized in that, In step 4, the target detection model for centralizer placement based on Gold-YOLO-L includes: The image processing module performs frame-by-frame processing on the real-time video data from the casing installation site to obtain image data; The network module uses a centralizer to place the target detection model, analyzes and processes the obtained image, extracts the position coordinates of the wellhead, casing, hydraulic tongs, and casing tongs areas, and outputs prediction information. The determination and storage module stores the location information and determination information of the recognition results based on the prediction information.

8. The method for determining the placement location and type of a centralizer based on multi-source data fusion according to claim 7, characterized in that, In step 4, the network module includes: The backbone network uses a pre-built straightener to place the backbone network in the target detection model to extract multi-scale features of the input image; The neck network, which integrates feature maps from different scales, improves the accuracy and robustness of the model. The classification network uses a loss function to classify and regress image features and aggregated information of the wellhead area to obtain prediction information. The training optimization unit adjusts the network parameters based on the training results and selects the model with the best training results as the final model.

9. The method for determining the placement location and type of a centralizer based on multi-source data fusion according to claim 5, characterized in that, In step 403, the target detection judgment logic of the target detection model is placed in the straightener, including the following steps: (51) Input the real-time data image into the target detection model for the centering device; (52) If there is no wellhead coordinate information, detect the wellhead position in the image and save it; (53) Detect the type of stabilizer in the image, the coordinate position of the stabilizer and the casing wrench, and determine whether the stabilizer is located above the wellhead and whether the overlapping area with the casing wrench is greater than a predetermined value. (54) If condition (53) is met, save the image and the straightening device information; (55) Determine whether all sleeves have been lowered. If not, return to step (51). If they have been lowered, end the target detection.

10. The method for determining the placement location and type of a centralizer based on multi-source data fusion according to claim 1, characterized in that, In step 5, real-time video data is input into the text recognition model to determine the specific time for placing the centralizer on the target casing.

11. The method for determining the placement location and type of a centralizer based on multi-source data fusion according to claim 10, characterized in that, Step 5 includes: Step 501: Apply a text recognition model to identify the time information of the real-time casing installation video and calibrate the video time; Step 502: Identify the image that meets the conditions in the image and the information of the centralizer saved in step 4, obtain the time information of the image, and save the record.

12. The method for determining the placement location and type of a centralizer based on multi-source data fusion according to claim 1, characterized in that, Step 6 includes: Step 601: Based on the data of the actual casing placement time and the centralizer placement time, determine whether the corresponding type of centralizer has been correctly placed when the casing that requires the placement of the centralizer is about to be lowered into the wellhead. Step 602: If it is found that the centering device is not installed correctly, an abnormal alarm will be triggered to remind the operator to take corrective measures in a timely manner.

13. A system for determining the placement location and type of a centralizer based on multi-source data fusion, characterized in that, The multi-source data fusion-based system for determining the placement and type of centralizers uses the multi-source data fusion-based method for determining the placement and type of centralizers as described in any one of claims 1-12 to detect the quality of casing installation operations in oilfield production.

Images