Inspection robot control method and system based on deep learning

Abstract

The application relates to the technical field of intelligent control, in particular to a deep learning-based inspection robot control method and system, which improves the reliability and safety of a power transmission system and reduces the risk of power system interruption and failure; the method comprises the following steps: performing unit division processing on cables needing to be inspected to obtain a plurality of cable units; according to an inspection plan time, controlling an inspection robot to inspect the cables, and obtaining state information of each cable unit by using the inspection robot; the state information comprises cable breakage times in the cable unit, average cable temperature, average gray value of a cable image and maximum sag distance; arranging the state information of a plurality of cable units collected by the inspection robot according to a connection order, and respectively aligning the cable breakage times, the average cable temperature, the average gray value of the cable image and the maximum sag distance to obtain a cable state feature matrix.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology, and in particular to a deep learning-based control method and system for inspection robots. Background Technology

[0002] Inspecting high-voltage overhead cables is a crucial step in ensuring the safe operation and reliable power supply of power transmission systems. Regular inspections and maintenance can reduce the risk of power system outages and faults. Using drone inspection robots for high-voltage overhead cable inspection is an advanced and efficient method that improves safety, reduces costs, and minimizes human error. This significantly enhances the reliability and safety of the power system. The high-definition cameras equipped on the drone inspection robots can clearly observe the operating status and fault types of high-voltage transmission lines.

[0003] However, existing inspection robots mostly carry out inspections of high-voltage overhead cables according to pre-defined plans. The inspection robots cannot autonomously determine the next inspection time based on the actual inspection results of the cables. When the cable condition deteriorates in the middle of the inspection plan's gap, the best opportunity for cable repair is easily missed. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a deep learning-based inspection robot control method that improves the reliability and safety of power transmission systems and reduces the risk of power system outages and failures.

[0005] In a first aspect, the present invention provides a deep learning-based control method for an inspection robot, the method comprising:

[0006] The cables that need to be inspected are divided into units to obtain multiple cable units;

[0007] According to the inspection plan schedule, the inspection robot is controlled to inspect the cable and obtain the status information of each cable unit. The status information includes the number of cable strand breaks in the cable unit, the average temperature of the cable, the average gray value of the cable image, and the maximum sag distance.

[0008] The status information of several cable units collected by the inspection robot is arranged in the connection order, and the number of broken strands, the average temperature of the cable, the average gray value of the cable image and the maximum sag distance are aligned to obtain the cable status feature matrix.

[0009] Obtain the voltage level of the cable, and retrieve the cable condition analysis model corresponding to the voltage level.

[0010] The cable condition feature matrix is ​​input into the cable condition analysis model corresponding to the voltage level, and the cable condition analysis model outputs the time node for the next cable inspection.

[0011] Compare the inspection time nodes output by the cable condition analysis model with the next inspection time node for the cable in the inspection plan:

[0012] If the inspection time node output by the cable condition analysis model is earlier than the next inspection time node in the inspection plan, then the next inspection time is the inspection time node output by the cable condition analysis model, and the inspection plan is updated accordingly.

[0013] If the inspection time node output by the cable condition analysis model is later than the next inspection time node in the inspection plan, the next inspection time will still be carried out according to the time node in the inspection plan.

[0014] Furthermore, the cable state feature matrix is ​​as follows:

[0015] Among them, B n The T represents the number of times a conductor strand breaks in the nth cable unit; n The C represents the average temperature of the nth cable unit; n The S represents the average grayscale value of the image of the nth cable unit; n The maximum sag distance of the nth cable unit is represented; the elements of each row in the cable state feature matrix represent the different state quantification feature values ​​of the cable segments between two adjacent cable towers.

[0016] Furthermore, the method for dividing cables requiring inspection into units includes:

[0017] The cables between two adjacent cable towers are divided into one cable unit;

[0018] Each cable unit is labeled with a unique identifier for the inspection robot to identify and track it;

[0019] The geographical location information of each cable unit is associated with and saved to the corresponding cable unit for navigation and positioning by the inspection robot.

[0020] Furthermore, the method for controlling the inspection robot to inspect cables includes:

[0021] Based on the pre-established inspection plan, determine the time required to inspect each cable unit;

[0022] Based on the geographical location information of the cable unit, the inspection robot is positioned to the cable unit that needs to be inspected;

[0023] Control the inspection robot to collect status information of the cable unit.

[0024] Furthermore, the method for controlling the inspection robot to collect status information of cable units includes:

[0025] Image sensors are used to acquire surface images of cable units. Image processing techniques are used to detect broken strands in the images by comparing the cable lines in the images with those in a normal state. Algorithms are used to mark and count the number of broken strands to obtain the number of times the cable unit has broken strands.

[0026] The surface temperature of the cable unit is measured at multiple locations using a temperature sensor or thermal imager, and the average temperature of the cable unit is obtained by statistically averaging the measured values.

[0027] An image sensor is used to acquire an image of the cable unit surface. The pixel values ​​of the cable unit surface image are converted into grayscale values, and the average grayscale value of all pixels is calculated to obtain the average grayscale value of the cable unit image.

[0028] Using laser ranging technology, the lowest point of the cable unit is determined, and the continuous vertical distance between the lowest point of the cable unit and the two ends of the cable unit is calculated to obtain the maximum sag distance of the cable unit.

[0029] Furthermore, the method for constructing the cable condition analysis model includes:

[0030] Collect sample data of high-voltage overhead cable inspection, including cable status information under different conditions. The cable status information includes the number of cable strand breaks, the average temperature of the cable, the average gray value of the cable image, and the maximum sag distance.

[0031] The collected sample data is preprocessed, including data cleaning, standardization, and normalization; and the sample data is converted into a cable condition feature matrix format.

[0032] Based on the assessment of domain experts, cable status information is correlated with inspection time to determine when each cable status information item needs to be inspected when it reaches a set threshold.

[0033] Based on deep learning technology, a model architecture is constructed to handle cable condition analysis problems.

[0034] The selected deep learning model is trained and optimized using preprocessed sample data;

[0035] The cable condition analysis model was deployed into a real inspection system for testing to verify its predictive and generalization capabilities.

[0036] Furthermore, the model architecture of the cable condition analysis model is selected from convolutional neural networks, recurrent neural networks, or long short-term memory networks.

[0037] On the other hand, this application also provides a deep learning-based inspection robot control system, the system comprising:

[0038] The segmentation module is used to segment the cables that need to be inspected into multiple cable units.

[0039] The inspection control module is used to control the inspection robot to inspect the cable according to the inspection plan time, and to obtain the status information of each cable unit using the inspection robot. The status information includes the number of cable strand breaks in the cable unit, the average temperature of the cable, the average gray value of the cable image, and the maximum sag distance.

[0040] The data processing module is used to arrange the status information of several cable units collected by the inspection robot according to the connection order, and align the number of broken strands, the average temperature of the cable, the average gray value of the cable image, and the maximum sag distance to obtain the cable status feature matrix.

[0041] The model call module is used to obtain the voltage level of the cable and, based on the voltage level, call up the cable condition analysis model corresponding to the voltage level.

[0042] The condition analysis module is used to input the cable condition feature matrix into the cable condition analysis model corresponding to the voltage level. The cable condition analysis model outputs the time node of the next cable inspection.

[0043] The inspection plan update module is used to compare the inspection time nodes output by the cable condition analysis model with the time nodes for the next inspection of the cable in the inspection plan:

[0044] If the inspection time node output by the cable condition analysis model is earlier than the next inspection time node in the inspection plan, then the next inspection time is the inspection time node output by the cable condition analysis model, and the inspection plan is updated accordingly.

[0045] If the inspection time node output by the cable condition analysis model is later than the next inspection time node in the inspection plan, the next inspection time will still be carried out according to the time node in the inspection plan.

[0046] Thirdly, this application provides an electronic device including a bus, a transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor. The transceiver, the memory, and the processor are connected via the bus, and the computer program, when executed by the processor, implements the steps of any of the methods described above.

[0047] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the methods described above.

[0048] Compared with the prior art, the beneficial effects of the present invention are as follows: by dividing the process into units, the inspection operation of overhead bare cables is refined to each cable unit, which improves the precision of the inspection and helps to more accurately reflect the actual condition of the cable; the status information of each cable unit is collected, including the number of broken strands, average temperature, average gray value and maximum sag distance, etc., which provides comprehensive data support for subsequent status analysis and prediction.

[0049] By analyzing the cable state feature matrix using a deep learning model, the system can automatically determine the cable's health status and output the next inspection time node, reducing the tedious work and errors of manually formulating inspection plans. When comparing inspection time nodes, adjustments can be made flexibly based on actual conditions, ensuring both timeliness and stability of the inspection plan. The method is highly versatile and scalable, and can be widely applied to the inspection of overhead bare cables at various voltage levels. Through automated inspection and intelligent analysis, the inspection robot can autonomously adjust the inspection plan based on the actual condition of the cable, rather than relying entirely on a pre-determined plan. This enables more timely detection of cable problems, reduces the possibility of missed maintenance opportunities, improves the reliability and safety of power transmission systems, and reduces the risk of power system outages and failures. Attached Figure Description

[0050] Figure 1 This is a flowchart of the present invention;

[0051] Figure 2 This is a flowchart of the process of controlling the inspection robot to inspect cables;

[0052] Figure 3 This is a flowchart for constructing a cable condition analysis model;

[0053] Figure 4 This is a structural diagram of a deep learning-based inspection robot control system. Detailed Implementation

[0054] As will be apparent to those skilled in the art from the description of this application, this application can be implemented as a method, apparatus, electronic device, and computer-readable storage medium. Therefore, this application can be specifically implemented in the following forms: entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software. Furthermore, in some embodiments, this application can also be implemented as a computer program product contained in one or more computer-readable storage media, which includes computer program code.

[0055] The aforementioned computer-readable storage medium may be any combination of one or more computer-readable storage media. Computer-readable storage media include: electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of computer-readable storage media include: portable computer disks, hard disks, random access memory, read-only memory, erasable programmable read-only memory, flash memory, optical fiber, optical disc read-only memory, optical storage devices, magnetic storage devices, or any combination thereof. In this application, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0056] This application describes the provided methods, apparatus, and electronic devices using flowcharts and / or block diagrams.

[0057] It should be understood that each block of a flowchart and / or block diagram, as well as combinations of blocks in a flowchart and / or block diagram, can be implemented by computer-readable program instructions. These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine that, when executed by a computer or other programmable data processing apparatus, creates means for implementing the functions / operations specified in the blocks of the flowchart and / or block diagram.

[0058] These computer-readable program instructions may also be stored in a computer-readable storage medium that enables a computer or other programmable data processing device to function in a particular manner. In this way, the instructions stored in the computer-readable storage medium produce an instruction apparatus product that includes the functions / operations specified in the blocks of a flowchart and / or block diagram.

[0059] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus or other device to produce a computer-implemented process, such that the instructions that execute on the computer or other programmable data processing apparatus provide a process for implementing the functions / operations specified in the blocks of the flowchart and / or block diagram.

[0060] This application will now be described with reference to the accompanying drawings. Example

[0061] like Figures 1 to 3 As shown, the deep learning-based inspection robot control method of the present invention specifically includes the following steps:

[0062] S1. Divide the cables that need to be inspected into units to obtain multiple cable units;

[0063] When inspecting high-voltage overhead cables, the cable lines typically span multiple cable towers. To effectively monitor and manage the cable's condition, it is first necessary to divide the cable segments into units. This unit division is crucial because it helps the robot monitor different parts of the cable more precisely, thereby more accurately determining the inspection frequency and timing. Step S1 specifically includes the following:

[0064] S11. Cable Line Segmentation: Cable lines need to be divided into a series of cable units; this can be achieved by the distance between cable towers, with each cable unit corresponding to a pair of adjacent cable towers; this division is based on geographical location, dividing the cable into specific segments.

[0065] S12. Cable Unit Identification: Each cable unit needs to be uniquely identified, usually using a naming convention or numbering system, so that the robot can easily identify and track each unit.

[0066] S13. Location Information Recording: In order to better navigate the inspection robot, the location information (such as GPS coordinates or other geographic reference points) of each cable unit needs to be recorded so that the robot can accurately locate and move to each unit;

[0067] S14. Cable Unit Information Collection: For each cable unit, some basic information needs to be collected, such as the start and end points of the cable, as well as other environmental factors related to the unit, such as vegetation conditions and weather conditions; this helps the robot better understand the inspection environment.

[0068] Furthermore, considering the inherent dangers of high-voltage cables, the safety of the inspection robot must be ensured. Therefore, appropriate safety measures, such as safety zones, should be implemented on each cable unit to protect the robot operator and surrounding personnel. In summary, the goal of step S1 is to divide the cable line into easily manageable units, each with a clearly defined location and relevant information. This division method will facilitate subsequent steps, such as data collection and analysis, to better manage and maintain the high-voltage overhead cable system.

[0069] S2. According to the inspection plan time, control the inspection robot to inspect the cable and use the inspection robot to obtain the status information of each cable unit; the status information includes the number of cable strand breaks in the cable unit, the average temperature of the cable, the average gray value of the cable image, and the maximum sag distance.

[0070] Step S2 is the core step of the deep learning-based inspection robot control method, involving the acquisition and processing of the actual state information of the cable. Based on the planned inspection time, the inspection robot is controlled to inspect the cable, obtaining the state information of each cable unit. Specifically, this includes the following steps:

[0071] S21. Determine the inspection schedule: Based on the pre-established inspection plan, determine the time required to inspect each cable unit; this time can be within a fixed interval or adjusted according to the actual condition of the cable.

[0072] S22. Positioning and Navigation: After the inspection time is determined, the inspection robot needs to be accurately positioned to the cable unit that needs to be inspected; this can be achieved by using GPS, LiDAR or other navigation technologies.

[0073] S23. Collect status information: After arriving at the designated location, the inspection robot uses its onboard high-definition camera to capture images of the cable unit and collect other relevant information, including but not limited to the number of cable strand breaks, the average temperature of the cable, the average grayscale value of the cable image, and the maximum sag distance.

[0074] S24. Storage and Transmission: The collected status information is stored in the inspection robot's memory and can be transmitted in real time to the back-end server or data center through the communication module for further processing and analysis.

[0075] Regarding the specific steps in step S23 for collecting data such as the number of cable strand breaks, the average temperature of the cable, the average grayscale value of the cable image, and the maximum sag distance, the following content is included:

[0076] a. Detection of the number of cable strand breaks: Cable strand breakage refers to the partial fracture of the insulation or conductor of a cable. To detect the number of cable strand breaks, a high-resolution camera or image sensor is usually used to acquire images of the cable surface. Image processing techniques, such as computer vision and image analysis, are used to detect the broken strands in the images. This can be achieved by comparing the cable lines in the images with those in a normal state for difference analysis. Algorithms are then used to mark and count the number of broken strands, thereby obtaining the number of cable strand breaks.

[0077] b. Detection of the average temperature of the cable: The temperature of the cable is a key parameter that can indicate whether there is a potential overload or other problem. Temperature sensors or thermal imagers are usually used to measure the surface temperature of the cable. Temperature sensors directly measure the temperature of the cable, while thermal imagers generate temperature images through infrared radiation. The average temperature of the cable can be obtained by statistically averaging the measured values, which will provide a more comprehensive understanding of the thermal condition of the cable.

[0078] c. Detection of average grayscale value of cable image: The average grayscale value of cable image is used to detect surface non-uniformity, reflecting damage or corrosion; the average grayscale value refers to the average brightness level of pixels in the image; the average grayscale value of cable image can be calculated through image processing algorithms; this usually involves converting the pixel values ​​of the image to grayscale values ​​and then calculating the average value of all pixels;

[0079] d. Detection of maximum sag distance: Sag is the maximum droop distance of a cable between two support points; its measurement requires the use of a sag measuring instrument; the sag measuring instrument can determine the sag distance of the cable through laser or other ranging technologies; the measurement of sag distance is crucial for understanding the tension of the cable and the condition of the support structure, because excessive sag may indicate a problem with the cable that requires repair.

[0080] In this step, by controlling the inspection robot to conduct inspections, automatic detection and intelligent analysis of cables can be achieved, greatly improving the efficiency and accuracy of inspections. This step can simultaneously collect multiple status information of the cable, including the number of broken strands, average temperature, average gray value, and maximum sag distance, thereby gaining a more comprehensive understanding of the cable's condition. The inspection robot can collect and transmit data in real time, enabling the cable's status information to be acquired and analyzed promptly, which is beneficial for timely detection and resolution of potential problems. The collected data can be stored in the inspection robot's memory or transmitted in real time to the backend server or data center for convenient subsequent data analysis and processing. In summary, the advantages of step S2 are mainly that it can meet the needs of automation, intelligence, real-time monitoring, high security, and data storage and transmission for cables.

[0081] S3. Arrange the status information of several cable units collected by the inspection robot according to the connection order, and align the number of broken strands, the average temperature of the cable, the average gray value of the cable image, and the maximum sag distance to obtain the cable status feature matrix.

[0082] The goal of step S3 is to organize the cable condition information collected by the inspection robot into a standardized feature matrix so that the subsequent cable condition analysis model can process and analyze this data. The following is a detailed description of step S3:

[0083] S31. Data Alignment and Arrangement: The inspection robot collects the status information of the cable units. This data needs to be arranged according to the connection order and aligned. The purpose of alignment is to ensure that the data of different cable units can be directly compared. Data alignment includes the following steps:

[0084] S311. Ensure that each cable unit has the same data items, such as the number of strand breaks, temperature, grayscale value, and sag distance;

[0085] S312. Ensure that the units of the data are consistent. For example, temperature can be expressed in degrees Celsius, while sag distance can be expressed in meters.

[0086] S313. Arrange the data according to the connection order of the cable units for subsequent processing;

[0087] S32. Generation of Cable Status Feature Matrix: After data alignment and arrangement are completed, these data can be organized into a cable status feature matrix. This matrix is ​​a table, where each row represents a cable unit and each column represents a status information item, such as the number of strand breaks, temperature, gray value, and sag distance. The elements in the matrix are the status information of the corresponding cable unit.

[0088] S33. Using the cable state feature matrix: The generated cable state feature matrix will be used for cable state analysis in subsequent steps; the purpose of this matrix is ​​to provide a simplified and standardized way to represent the state of different cable units so that they can be analyzed and predicted by deep learning models.

[0089] More specifically, the cable state feature matrix is ​​as follows:

[0090] Among them, B n The T represents the number of times a conductor strand breaks in the nth cable unit; n The C represents the average temperature of the nth cable unit; n The S represents the average grayscale value of the image of the nth cable unit; n The maximum sag distance of the nth cable unit is represented; the elements of each row in the cable state feature matrix represent the different state quantification feature values ​​of the cable segments between two adjacent cable towers.

[0091] In this step, data alignment and arrangement ensure that the status data of different cable units have the same data items, such as the number of strand breaks, temperature, grayscale value, and sag distance. This ensures data consistency, making it directly comparable and analytic. Since the status information of different cable units may use different units, such as Celsius and Fahrenheit, or meters and feet, standardizing the data units makes the data in the cable status feature matrix easier to understand and compare.

[0092] The cable state feature matrix integrates the state information of different cable units into a single table, with each row representing a cable unit and each column representing a state information item. This integration makes the data easier to process and analyze, while reducing the input complexity of deep learning models. The cable state feature matrix provides a simplified and standardized way to represent the state information of cable units, which helps reduce data complexity and makes it easier for deep learning models to understand and process this information. Most importantly, the goal of step S3 is to provide data for deep learning models, enabling them to perform cable state analysis and prediction. By standardizing and organizing the data, S3 provides clearer and more consistent data input for subsequent analysis models, which helps improve the performance and accuracy of the models.

[0093] In summary, step S3 helps improve the efficiency and reliability of cable condition analysis, making the data collected from the inspection robot easier to process and apply to deep learning models; these advantages help improve the maintainability and safety of cable systems and reduce the risk of cable failures.

[0094] S4. Obtain the voltage level of the cable, and retrieve the cable condition analysis model corresponding to the voltage level.

[0095] Step S4 involves obtaining the cable's voltage level and, based on that level, calling the corresponding cable condition analysis model; the specific execution process includes the following steps:

[0096] S41. Obtain the voltage rating of the cable: The voltage rating of a cable refers to the voltage level at which the cable is used for power transmission, usually expressed in kilovolts (kV). Cables of different voltage ratings have different technical specifications, standards, and operating requirements. This information can usually be found in power system documents or obtained through system databases or identifiers.

[0097] S42. Determine the cable condition analysis model: In this step, based on the cable's voltage level, a pre-established cable condition analysis model needs to be called. These models are deep learning or machine learning models that have been trained to analyze the cable's condition. Cables of different voltage levels usually require different models because they may be affected by different environments, current loads, and thermal characteristics.

[0098] S43. Retrieve the corresponding cable condition analysis model: In this step, the system will automatically select or call the corresponding cable condition analysis model based on the cable's voltage level. This can be achieved through a model selection algorithm or a model library. The model selection should be real-time because the cable condition may change over time, and it is necessary to ensure that the selected model matches the current condition and characteristics of the cable.

[0099] The cable condition analysis model is built based on deep learning technology, and the specific construction process is as follows:

[0100] S421. Collect sample data: First, it is necessary to collect a large amount of sample data from the inspection of high-voltage overhead cables, including the status information of each cable unit, such as the number of cable strand breaks, the average temperature of the cable, the average gray value of the cable image, and the maximum sag distance. These data are recorded in the form of time series to understand the evolution of the cable status over time.

[0101] S422, Data Preprocessing: Preprocess the collected sample data, including data cleaning, standardization, normalization and other operations, to eliminate noise and outliers in the data, and transform the data into a format suitable for deep learning model processing, such as the cable state feature matrix format in step S3.

[0102] S423. Characteristic Engineering: Characteristic engineering is required for cable status information. Based on the assessment of domain experts, the cable status information is correlated with the inspection time. This can be achieved by analyzing historical data and expert experience, such as analyzing the correlation between the number of cable strand breaks and the inspection time, so as to determine when the number of cable strand breaks reaches a certain threshold and an inspection is required.

[0103] S424. Constructing the Model Architecture: Based on deep learning technology, construct a model architecture suitable for handling cable condition analysis problems; common deep learning models such as convolutional neural networks, recurrent neural networks, or long short-term memory networks can be selected.

[0104] S425. Model Training and Optimization: Using the feature-engineered dataset, train the selected deep learning model. During training, the model will learn the correlation between cable status and actual inspection time. The training loss function is usually the prediction error of inspection time. During training, it is necessary to continuously evaluate the model's performance and effectiveness, and optimize the model by adjusting hyperparameters, increasing or decreasing the number of layers, changing activation functions, etc.

[0105] S426. Deployment and Testing: After training, the model is deployed to the actual inspection system for testing to verify the model's predictive and generalization capabilities.

[0106] S427. Iterative Updates: The cable condition analysis model needs to be updated regularly to adapt to the evolution of cable conditions and changes in model performance; this can be achieved by periodically retraining the model and using new data.

[0107] It is important to note that when constructing a cable condition analysis model, the characteristics and inspection requirements of high-voltage overhead cables must be fully considered, and appropriate deep learning algorithms and model structures should be selected to improve the model's prediction accuracy and generalization ability. Furthermore, for different voltage levels and cable types, the model needs to be adjusted and optimized according to the actual situation. In summary, the construction of a cable condition analysis model involves multiple steps, including data acquisition, preprocessing, feature engineering, model selection, training, evaluation, hyperparameter tuning, deployment, and regular updates. The goal of this model is to predict the state evolution of cables using deep learning technology to improve the efficiency of inspection robots and the reliability of power systems.

[0108] S5. Input the cable condition feature matrix into the cable condition analysis model corresponding to the voltage level. The cable condition analysis model outputs the time node for the next cable inspection.

[0109] Step S5 is a crucial step in analyzing the cable condition and predicting the next inspection time using a deep learning model; the specific process is as follows:

[0110] S51. Obtain the cable state feature matrix: In step S3, the state information of several cable units collected by the inspection robot is arranged according to the connection order, and the number of cable strand breaks, the average temperature of the cable, the average gray value of the cable image, and the maximum sag distance are aligned to form the cable state feature matrix; this matrix contains important information about the cable state and can be used as input to a deep learning model.

[0111] S52. Retrieve the corresponding cable condition analysis model: In step S4, the cable condition analysis model corresponding to the voltage level of the cable was retrieved; this model is a pre-trained deep learning model that has the ability to analyze and predict the cable condition.

[0112] S53. Input the cable state feature matrix into the model: Input the cable state feature matrix obtained in step S3 into the cable state analysis model retrieved in step S4;

[0113] S54. Model predicts the next inspection time: The cable condition analysis model infers the next inspection time based on the input cable condition feature matrix. This time is predicted based on the current cable condition and historical data and can serve as an important reference for the inspection plan.

[0114] In this step, by using a deep learning model, the cable condition can be monitored in real time, and the next inspection time node can be predicted based on the model, providing an important basis for formulating a reasonable inspection plan. The deep learning-based inspection robot control method can automatically determine the next inspection time node based on the actual inspection results of the cable, avoiding the tediousness and inaccuracy of manual planning. At the same time, through model training and optimization, the accuracy and precision of prediction can be improved, making the inspection plan more reasonable and efficient. The deep learning-based inspection robot control method has scalability and flexibility. With the increase of datasets and the development of technology, the model can be continuously optimized and improved to improve the accuracy and precision of prediction. In addition, this method can also be applied to different types of cables and inspection robots, and has broad application prospects. In summary, step S5, which uses a deep learning model to analyze the cable condition and predict the next inspection time node, is a key step that realizes real-time monitoring and prediction of cable condition, improves inspection efficiency and accuracy, reduces costs and risks, and has scalability and flexibility.

[0115] S6. Compare the inspection time node output by the cable condition analysis model with the next inspection time node in the inspection plan: If the inspection time node output by the cable condition analysis model is earlier than the next inspection time node in the inspection plan, the next inspection time is the inspection time node output by the cable condition analysis model, and the inspection plan is updated accordingly; if the inspection time node output by the cable condition analysis model is later than the next inspection time node in the inspection plan, the next inspection time is still according to the time node in the inspection plan.

[0116] The purpose of step S6 is to compare the inspection time nodes output by the cable condition analysis model with the corresponding cable's next inspection time nodes in the inspection plan to determine the actual time for the next inspection. Specifically, step S6 includes the following:

[0117] S61. Obtain the inspection time node output by the cable condition analysis model: In step S5, the cable condition feature matrix is ​​input into the cable condition analysis model corresponding to the voltage level, and the model outputs the predicted inspection time node.

[0118] S62. Obtain the inspection time nodes in the inspection plan: In the inspection plan, there are predetermined inspection time nodes for each cable and voltage level.

[0119] S63. Compare the two time points: Compare the inspection time points output by the cable condition analysis model with the time points in the inspection plan; specifically, the comparison results are handled in the following two ways:

[0120] A. The inspection time output by the cable condition analysis model is earlier than the time in the inspection plan: If the output of the cable condition analysis model indicates that the next inspection needs to be carried out in a short period of time, such as because the cable condition is deteriorating rapidly or there is an emergency, then this new time will be adopted to ensure the safety and reliability of the cable. In this case, the inspection plan should be updated to set the next inspection time of the cable to the inspection time output by the cable condition analysis model.

[0121] B. The inspection time output by the cable condition analysis model is later than the time in the inspection plan: If the cable condition analysis model believes that the next inspection is not needed for a long time, then the original time in the inspection plan is still valid; in this case, the inspection plan should be maintained and the original time in the inspection plan should be kept unchanged.

[0122] In this step, by comparing the output of the cable condition analysis model with the inspection plan, the actual time for the next inspection can be updated in a timely manner. This allows the inspection plan to be dynamically adjusted according to the latest cable condition, improving the timeliness and flexibility of the inspection. Incorporating the prediction results of the cable condition analysis model into the decision-making process can more accurately reflect the actual condition of the cable. This can avoid the situation where the inspection plan based on a fixed time interval may be too early or too late.

[0123] When the inspection time output by the cable condition analysis model is earlier than the time in the inspection plan, potential problems with the cable can be detected in a timely manner, thereby reducing potential safety risks and improving the reliability of the cable system. When the cable condition deteriorates rapidly or an emergency occurs, necessary measures can be taken quickly to ensure the normal operation of the cable system. When the inspection time output by the cable condition analysis model is later than the time in the inspection plan, maintaining the original inspection plan can avoid excessive intervention and maintain the stability of the system.

[0124] In summary, step S6, by combining the output of the cable condition analysis model with the predetermined inspection plan, makes the inspection strategy more targeted and real-time, thereby improving the operating efficiency and safety of the cable system. Example

[0125] like Figure 4 As shown, the deep learning-based inspection robot control system of the present invention specifically includes the following modules;

[0126] The segmentation module is used to segment the cables that need to be inspected into multiple cable units.

[0127] The inspection control module is used to control the inspection robot to inspect the cable according to the inspection plan time, and to obtain the status information of each cable unit using the inspection robot. The status information includes the number of cable strand breaks in the cable unit, the average temperature of the cable, the average gray value of the cable image, and the maximum sag distance.

[0128] The data processing module is used to arrange the status information of several cable units collected by the inspection robot according to the connection order, and align the number of broken strands, the average temperature of the cable, the average gray value of the cable image, and the maximum sag distance to obtain the cable status feature matrix.

[0129] The model call module is used to obtain the voltage level of the cable and, based on the voltage level, call up the cable condition analysis model corresponding to the voltage level.

[0130] The condition analysis module is used to input the cable condition feature matrix into the cable condition analysis model corresponding to the voltage level. The cable condition analysis model outputs the time node of the next cable inspection.

[0131] The inspection plan update module is used to compare the inspection time nodes output by the cable condition analysis model with the time nodes for the next inspection of the cable in the inspection plan:

[0132] If the inspection time node output by the cable condition analysis model is earlier than the next inspection time node in the inspection plan, then the next inspection time is the inspection time node output by the cable condition analysis model, and the inspection plan is updated accordingly.

[0133] If the inspection time node output by the cable condition analysis model is later than the next inspection time node in the inspection plan, the next inspection time will still be carried out according to the time node in the inspection plan.

[0134] In this embodiment, the system can autonomously control the actions of the inspection robot and adjust the inspection plan according to the actual condition of the cable, rather than relying solely on a pre-defined plan. This improves the system's automation level, reduces human intervention, and thus increases efficiency. Through the status information collected by the inspection robot, the system can understand the status of each cable unit in real time, including the number of broken strands, temperature, image features, etc., thereby enabling faster detection of cable problems and reducing the possibility of missed fault detection.

[0135] The data processing module can integrate and align the status information of multiple cable units collected to construct a cable status feature matrix. This helps to comprehensively analyze the overall status of the cable system, rather than individual cable units, thereby gaining a better understanding of the system's health status. The system can select the appropriate analysis model based on the voltage level of different cables through the model calling module and the status analysis module, making it suitable for high-voltage overhead cables of various voltage levels.

[0136] The system can output the next inspection time node based on the actual status of the cable, making the inspection plan more flexible and optimized. This helps to detect problems in a timely manner, reduce maintenance costs, and improve the reliability and safety of the power system. By reducing human risk and providing real-time status information, the system can improve the safety of the inspection process and reduce the risk of personnel needing to come into contact with high-voltage cables.

[0137] The various variations and specific embodiments of the deep learning-based inspection robot control method in the aforementioned Embodiment 1 are also applicable to the deep learning-based inspection robot control system of this embodiment. Through the foregoing detailed description of the deep learning-based inspection robot control method, those skilled in the art can clearly understand the implementation method of the deep learning-based inspection robot control system of this embodiment. Therefore, for the sake of brevity, it will not be described in detail here.

[0138] In addition, this application also provides an electronic device, including a bus, a transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor. The transceiver, the memory, and the processor are respectively connected via the bus. When the computer program is executed by the processor, it implements the various processes of the above-described method embodiment for controlling output data and achieves the same technical effect. To avoid repetition, it will not be described again here.

[0139] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A deep learning-based control method for an inspection robot, characterized in that, The method includes: The cables that need to be inspected are divided into units to obtain multiple cable units, including: dividing the cable between two adjacent cable towers into one cable unit; marking each cable unit with a unique identifier for the inspection robot to identify and track; associating and saving the geographical location information of each cable unit with the corresponding cable unit for the inspection robot to navigate and locate. According to the inspection plan schedule, the inspection robot is controlled to inspect the cable and obtain the status information of each cable unit. The status information includes the number of cable strand breaks in the cable unit, the average temperature of the cable, the average gray value of the cable image, and the maximum sag distance. The status information of several cable units collected by the inspection robot is arranged in the connection order, and the number of broken strands, the average temperature of the cable, the average gray value of the cable image and the maximum sag distance are aligned to obtain the cable status feature matrix. Obtain the voltage level of the cable, and retrieve the cable condition analysis model corresponding to the voltage level. The cable condition feature matrix is ​​input into the cable condition analysis model corresponding to the voltage level, and the cable condition analysis model outputs the time node for the next cable inspection. Compare the inspection time nodes output by the cable condition analysis model with the next inspection time node for the cable in the inspection plan: If the inspection time node output by the cable condition analysis model is earlier than the next inspection time node in the inspection plan, then the next inspection time is the inspection time node output by the cable condition analysis model, and the inspection plan is updated accordingly. If the inspection time node output by the cable condition analysis model is later than the next inspection time node in the inspection plan, the next inspection time will still be carried out according to the time node in the inspection plan.

2. The deep learning-based inspection robot control method as described in claim 1, characterized in that, The cable state feature matrix is ​​as follows: Wherein, Bn represents the number of times the conductor strand breaks in the nth cable unit; Tn represents the average temperature of the nth cable unit; Cn represents the average gray value of the image of the nth cable unit; Sn represents the maximum sag distance of the nth cable unit; and the elements of each row in the cable state feature matrix represent the different state quantification feature values ​​of the cable segments between two adjacent cable towers.

3. The deep learning-based inspection robot control method as described in claim 2, characterized in that, Methods for controlling inspection robots to inspect cables include: Based on the pre-established inspection plan, determine the time required to inspect each cable unit; Based on the geographical location information of the cable unit, the inspection robot is positioned to the cable unit that needs to be inspected; Control the inspection robot to collect status information of the cable unit.

4. The deep learning-based inspection robot control method as described in claim 3, characterized in that, Methods for controlling the inspection robot to collect status information of cable units include: Image sensors are used to acquire surface images of cable units. Image processing techniques are used to detect broken strands in the images by comparing the cable lines in the images with those in a normal state. Algorithms are used to mark and count the number of broken strands to obtain the number of times the cable unit has broken strands. The surface temperature of the cable unit is measured at multiple locations using a temperature sensor or thermal imager, and the average temperature of the cable unit is obtained by statistically averaging the measured values. An image sensor is used to acquire an image of the cable unit surface. The pixel values ​​of the cable unit surface image are converted into grayscale values, and the average grayscale value of all pixels is calculated to obtain the average grayscale value of the cable unit image. Using laser ranging technology, the lowest point of the cable unit is determined, and the continuous vertical distance between the lowest point of the cable unit and the two ends of the cable unit is calculated to obtain the maximum sag distance of the cable unit.

5. The deep learning-based inspection robot control method as described in claim 2, characterized in that, The method for constructing the cable condition analysis model includes: Collect sample data of high-voltage overhead cable inspection, including cable status information under different conditions. The cable status information includes the number of cable strand breaks, the average temperature of the cable, the average gray value of the cable image, and the maximum sag distance. The collected sample data is preprocessed, including data cleaning, standardization, and normalization; and the sample data is converted into a cable condition feature matrix format. Based on the assessment of domain experts, cable status information is correlated with inspection time to determine when each cable status information item needs to be inspected when it reaches a set threshold. Based on deep learning technology, a model architecture is constructed to handle cable condition analysis problems. The selected deep learning model is trained and optimized using preprocessed sample data; The cable condition analysis model was deployed into a real inspection system for testing to verify its predictive and generalization capabilities.

6. The deep learning-based inspection robot control method as described in claim 5, characterized in that, The model architecture of the cable condition analysis model is selected from convolutional neural networks, recurrent neural networks, or long short-term memory networks.

7. A deep learning-based inspection robot control system, characterized in that, The system, applied to the deep learning-based inspection robot control method as described in any one of claims 1-6, comprises: The segmentation module is used to segment the cables that need to be inspected into multiple cable units. The inspection control module is used to control the inspection robot to inspect the cable according to the inspection plan time, and to obtain the status information of each cable unit using the inspection robot. The status information includes the number of cable strand breaks in the cable unit, the average temperature of the cable, the average gray value of the cable image, and the maximum sag distance. The data processing module is used to arrange the status information of several cable units collected by the inspection robot according to the connection order, and align the number of broken strands, the average temperature of the cable, the average gray value of the cable image, and the maximum sag distance to obtain the cable status feature matrix. The model call module is used to obtain the voltage level of the cable and, based on the voltage level, call up the cable condition analysis model corresponding to the voltage level. The condition analysis module is used to input the cable condition feature matrix into the cable condition analysis model corresponding to the voltage level. The cable condition analysis model outputs the time node of the next cable inspection. The inspection plan update module is used to compare the inspection time nodes output by the cable condition analysis model with the time nodes for the next inspection of the cable in the inspection plan: If the inspection time node output by the cable condition analysis model is earlier than the next inspection time node in the inspection plan, then the next inspection time is the inspection time node output by the cable condition analysis model, and the inspection plan is updated accordingly. If the inspection time node output by the cable condition analysis model is later than the next inspection time node in the inspection plan, the next inspection time will still be carried out according to the time node in the inspection plan.

8. A deep learning-based inspection robot control electronic device, comprising a bus, a transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the transceiver, the memory, and the processor are connected via the bus, characterized in that, When the computer program is executed by the processor, it implements the steps of the method as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1-6.

Images