Switch cabinet live detection method and system based on humanoid robot

Abstract

The invention provides a switch cabinet live detection method and system based on a humanoid robot, and relates to the technical field of power grids. The target position and the detection item are automatically obtained by analyzing the detection task instruction, and real-time path planning and obstacle avoidance are performed by fusing multi-sensor data, so that the robot can autonomously move to an operation point in an optimal path, and time consumption of reciprocating walking and positioning of personnel is avoided; by controlling the humanoid mechanical arm to automatically switch probes according to detection types and execute standardized detection actions, tedious operations of manually holding, moving and replacing instruments are thoroughly replaced, and repeated measurement caused by misoperation is avoided; real-time fusion analysis and cross validation are performed on multi-source data through a built-in algorithm, and a detection result containing a fault mode is automatically generated, so that real-time intelligent live detection of the switch cabinet is realized, the technical problem of low efficiency of live detection of the switch cabinet is solved, and the detection efficiency of live detection of the switch cabinet is improved.

Description

Technical Field

[0001] This invention relates to the field of power grid technology, and in particular to a method and system for live detection of switchgear based on a humanoid robot. Background Technology

[0002] As a critical power distribution device in a power system, the reliability of switchgear's operation directly affects the safety and stability of the power grid. Partial discharge and overheating at connection points are the two most common precursors to insulation faults in switchgear. Therefore, regularly conducting live-line testing (i.e., uninterrupted testing) on ​​operating switchgear has become a core method of condition-based maintenance.

[0003] Currently, this inspection work mainly relies on manual labor. Inspectors must carry various specialized instruments into the substation to inspect the switchgear at close range. Before each inspection, personnel must obtain cumbersome work permits, undergo safety briefings, don heavy protective gear, and move multiple instruments. During the inspection, personnel must move back and forth between rows of switchgear, locate positions, and manually operate and record each inspection point, consuming a significant amount of effective working time. Furthermore, the inspection results are highly dependent on the experience and responsibility of the personnel. Inconsistencies in the pressure applied to the probe, the speed of movement, and the scanning path taken by different personnel lead to poor data consistency and make effective longitudinal comparisons difficult. To ensure reliability, critical parts often require repeated measurements, further reducing overall operational efficiency.

[0004] Therefore, the current live-line detection of switchgear suffers from low efficiency. Summary of the Invention

[0005] This invention provides a method and system for live detection of switchgear based on a humanoid robot, which solves the technical problem of low efficiency in current live detection of switchgear.

[0006] In a first aspect, the present invention provides a method for live-line detection of switchgear based on a humanoid robot. The method includes: receiving a detection task instruction for the switchgear to be tested; parsing the detection task instruction to obtain the position information and detection item type of the switchgear to be tested, wherein the detection item type includes at least one of the following: transient ground voltage detection, ultrasonic detection, infrared thermal imaging detection, and ultra-high frequency detection; performing real-time path planning and obstacle avoidance based on the position information of the switchgear to be tested and the position information of the humanoid robot, and moving to a preset working position of the switchgear to be tested; after arriving at the preset working position, controlling the humanoid robot's humanoid robotic arm to switch detection heads based on the detection item type, and performing detection on the switchgear to be tested to obtain detection data for each detection item; and performing fusion analysis and cross-validation based on the detection data for each detection item to determine the detection result of the switchgear to be tested, wherein the detection result includes the working status, fault mode, and fault location.

[0007] Secondly, embodiments of the present invention provide a live-line detection device for switchgear based on a humanoid robot. The device includes a communication module and a processing module. The communication module receives a detection task instruction from the switchgear to be tested. The processing module parses the detection task instruction to obtain the location information and detection item type of the switchgear to be tested. The detection item type includes at least one of the following: transient ground voltage detection, ultrasonic detection, infrared thermal imaging detection, and ultra-high frequency detection. Based on the location information of the switchgear to be tested and the location information of the humanoid robot, real-time path planning and obstacle avoidance are performed, moving the robot to a preset working position. After arriving at the preset working position, based on the detection item type, the humanoid robot's robotic arm is controlled to switch detection heads to detect the switchgear to be tested, obtaining detection data for each detection item. Based on the detection data for each detection item, fusion analysis and cross-validation are performed to determine the detection result of the switchgear to be tested. The detection result includes the working status, fault mode, and fault location.

[0008] Thirdly, embodiments of the present invention provide a switchgear live detection system based on a humanoid robot. The system includes an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor is used to call and run the computer program stored in the memory to perform the steps of the method as described in the first aspect and any possible implementation thereof.

[0009] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing a computer program, characterized in that, when the computer program is executed by a processor, it implements the steps of the method as described in the first aspect and any possible implementation thereof.

[0010] This invention provides a method and system for live-line detection of switchgear based on a humanoid robot. The invention automatically obtains the target location and detection items by parsing detection task instructions, replacing the inefficient process of manually processing work permits, providing safety briefings, and relying on experience to locate cabinets. By fusing multi-sensor data for real-time path planning and obstacle avoidance, the robot can autonomously and optimally move to the work point, avoiding the time wasted by personnel walking back and forth between cabinets for positioning. By controlling the humanoid robotic arm to automatically switch probes according to the detection type and execute standardized detection actions, the invention completely replaces the tedious manual handling, moving, and changing of instruments, with operational accuracy and consistency far exceeding that of manual methods, avoiding repeated measurements due to improper operation. Through built-in algorithms, the invention performs real-time fusion analysis and cross-validation of multi-source data and automatically generates detection results including fault modes, achieving real-time intelligent live-line detection of switchgear. This solves the technical problem of low efficiency in live-line detection of switchgear and improves the detection efficiency. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a flowchart illustrating a live-line detection method for switchgear based on a humanoid robot, provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of the structure of a switchgear live detection device based on a humanoid robot provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0013] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of the invention. However, those skilled in the art will understand that the invention can be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of the invention with unnecessary detail.

[0014] In the description of this invention, unless otherwise stated, " / " means "or". For example, A / B can mean A or B. The term "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone. Furthermore, "at least one" and "more than one" refer to two or more. The terms "first," "second," etc., do not limit the quantity or order of execution, and "first," "second," etc., do not necessarily imply differences.

[0015] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a specific manner to facilitate understanding.

[0016] Furthermore, the terms "comprising" and "having," and any variations thereof, used in the description of this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or modules is not limited to the steps or modules listed, but may optionally include other steps or modules not listed, or may optionally include other steps or modules inherent to such process, method, product, or device.

[0017] To make the objectives, technical solutions, and advantages of the present invention clearer, the following description will be provided in conjunction with the accompanying drawings and specific embodiments.

[0018] like Figure 1 As shown, this embodiment of the invention provides a method for live detection of switchgear based on a humanoid robot. The method includes steps S101-S104.

[0019] S101. Receive the testing task instruction for the switchgear to be tested.

[0020] In some embodiments, the detection task instruction is a structured digital command, which usually comes from the background scheduling system or the handheld terminal of the operation and maintenance personnel, and contains structured information such as target device identification, detection content, and execution priority.

[0021] For example, the robot receives instructions through its communication module. The parsing process is not a simple keyword extraction, but rather a syntactic and semantic analysis of the instructions through an embedded task parser. This parser, based on a predefined ontology library for the power inspection domain, can accurately understand the deeper meaning of instructions such as performing TEV and infrared scanning on switchgear 101, and decompose them into executable operation sequences and equipment coordinates.

[0022] S102. Parse the detection task instruction to obtain the location information of the switch cabinet to be detected and the type of detection item.

[0023] In some embodiments, the detection item type includes at least one of the following: transient ground voltage detection, ultrasonic detection, infrared thermography detection, and ultra-high frequency detection.

[0024] As one possible implementation, step S102 can be specifically implemented as steps S1021-S1024.

[0025] S1021. Perform semantic analysis on the detection task instruction to identify the switchgear equipment number and keywords of the detection items to be performed in the detection task instruction.

[0026] For example, embodiments of the present invention can utilize a pre-trained dictionary in the power industry field to perform word segmentation and entity recognition on instructions. For instance, it can accurately identify that "KYN28A-12" is a switch cabinet model and that transient ground voltage is a standard test item.

[0027] S1022. Based on the switchgear equipment number and the preset substation map model, determine the global coordinates of each switchgear to be tested.

[0028] For example, this embodiment of the invention maintains a high-precision two-dimensional / three-dimensional electronic map of the substation interior, which pre-inputs the precise geographical coordinates of each switchgear. After parsing the switchgear number, this invention queries this map to convert the text number 101 switchgear into a specific coordinate point (X, Y, Z) available in the navigation system.

[0029] S1023. Based on the keywords of the test items to be performed, query the equipment testing knowledge base, verify the matching between the test items and the switchgear equipment model, and determine the test item type for each switchgear to be tested.

[0030] For example, this embodiment of the invention accesses a device testing knowledge base. This knowledge base defines a list of testing items for different models of switchgear. For instance, for fully insulated switchgear, UHF testing may not be required. This invention verifies whether the testing items requested by the instruction are applicable to this model of equipment. If an inapplicable item is requested, it will proactively report an error, thus demonstrating its intelligence and error prevention capabilities.

[0031] S1024. Based on the global coordinates and detection item type of each switchgear to be tested, as well as the shape information of the humanoid robot, a collision test is conducted in conjunction with the substation map model to determine the preset working position of each switchgear to be tested.

[0032] In some embodiments, the preset work position is not a fixed point, but a dynamic three-dimensional spatial pose based on task requirements. This ensures that the robot can clearly "see" the target while its robotic arm can operate within a dexterous, collision-free working range.

[0033] In some embodiments, real-time path planning and obstacle avoidance is a dynamic decision-making process that enables the robot to perceive and avoid suddenly appearing obstacles (such as temporarily placed ladders or other people) in real time while walking, just like a human.

[0034] For example, embodiments of the present invention can place 3D models of the robot body and robotic arm at different candidate locations to simulate detection actions. A collision detection algorithm is used to calculate whether interference with surrounding switch cabinets, walls, or other equipment will occur. Finally, an optimal location is selected that can complete all detection tasks while ensuring absolute safety as the final preset working position. This process fully considers the robot's kinematic constraints and workspace.

[0035] S103. Based on the location information of the switch cabinet to be tested and the location information of the humanoid robot, perform real-time path planning and obstacle avoidance, and move to the preset working position of the switch cabinet to be tested.

[0036] In some embodiments, the robot navigates using SLAM technology, which integrates multiple sensors. LiDAR provides precise distance information to construct the geometry of the map, while a vision camera identifies semantic information such as door numbers and switch cabinet numbers, and assists in recognizing transparent obstacles like glass. An inertial measurement unit (IMU) provides stable attitude estimation during rapid robot movement. A path planner considers multiple factors, including the shortest path, walking energy consumption, and ground flatness, to generate an optimal path, which is then fine-tuned in real-time by a local planner during movement.

[0037] As one possible implementation, step S103 can be specifically implemented as steps S1031-S1035.

[0038] S1031. Multidimensional perception data is acquired through the lidar, vision camera and inertial measurement unit carried by the humanoid robot.

[0039] In some embodiments, LiDAR, a sensor that acquires high-precision 3D point cloud data of the surrounding environment by emitting laser beams and measuring their return time, is the core of constructing the map geometry. Visual cameras, typically color or RGB-D cameras, are used to capture the texture, color, and semantic information of the environment (such as recognizing "No Entry" signs and switch cabinet numbers). Inertial Measurement Units (IMUs) are miniature sensors that measure the robot's three-axis angular velocity and acceleration, providing high-frequency, short-term pose change data during robot movement to compensate for data lag or loss from other sensors under dynamic conditions. Multidimensional perception data refers to a collection of heterogeneous data from the aforementioned different sensors, synchronized in time and calibrated in space, collectively constituting the robot's "multidimensional senses" of perceiving the environment.

[0040] For example, the data from the three sensors are synchronized via hardware timestamps and unified into the robot coordinate system through a pre-calibrated transformation matrix (extrinsic parameter). The LiDAR provides the skeletal framework of the environment, the vision camera "fills in the flesh and blood" of the framework and identifies semantic labels, and the IMU provides instantaneous and stable self-motion perception when the robot starts, turns, or bumps. The three complement each other to form a robust perception system.

[0041] S1032. Based on multi-dimensional sensing data, construct or update the environmental map of the substation in real time.

[0042] In some embodiments, an environmental map refers to a dense 3D map containing geometric and semantic information, generated in real time using Simultaneous Localization and Mapping (SLAM) technology. It is not only a static collection of obstacles but also a dynamic knowledge base that records both explored and unexplored areas.

[0043] For example, the robot inputs the fused perception data into the SLAM algorithm. This algorithm estimates the robot's current position on the map (localization) while simultaneously matching the new perception data with the existing map, adding new obstacles (such as a temporarily placed repair cart), and updating changed areas (such as a moved chair), thus ensuring the map's real-time performance and accuracy. Visual information helps distinguish between glass doors (passable) and walls (impassable), making the map more intelligent.

[0044] S1033. Starting from the position information of the humanoid robot and taking the preset working position of each switch cabinet to be tested as the target point, perform global path planning in the environment map to determine multiple candidate movement paths.

[0045] S1034. Construct a multi-objective optimization function based on total path length, smoothness, energy consumption, and safety margin. Combine multi-dimensional perception data and environmental map to conduct dynamic obstacle avoidance evaluation and determine the optimal path among multiple candidate movement paths.

[0046] In some embodiments, a multi-objective optimization function is used to comprehensively evaluate the merits of a path. It considers multiple metrics simultaneously, including total path length (efficiency), smoothness (reducing robot energy consumption and sway), energy consumption (terrain-dependent), and safety margin (distance maintained from obstacles), seeking an optimal balance. Dynamic obstacle avoidance evaluation, based on a coarsely planned global path and combined with real-time perception data, assesses and avoids previously unknown or suddenly appearing dynamic obstacles (such as walking workers), ensuring local path safety.

[0047] For example, in this embodiment of the invention, a comprehensive score is calculated for each candidate path. For instance, a short path that is too close to a wall will have a lower safety margin score; another slightly longer path that is wide and flat may have a higher comprehensive score. Simultaneously, the invention continuously compares real-time perception data with the map. Once a new obstacle is detected on a candidate path, the score of that path is immediately reduced or it is eliminated. Ultimately, the invention selects the path with the highest comprehensive score and currently no dynamic obstacles as the optimal path.

[0048] S1035. Based on the optimal path, control the humanoid robot to move to the preset working position of each switch cabinet to be tested.

[0049] In some embodiments, an advanced control strategy for humanoid robots does not control the legs or torso individually, but treats the robot as a whole, coordinating its bipedal gait, arm swing, and torso posture to complete a predetermined movement task while maintaining dynamic balance. The optimal path is a series of waypoints. The robot's control system decomposes this path into specific gait sequences and body posture commands. Through a whole-body control algorithm, the robot can walk stably like a human, automatically adjusting its ankle and knee angles to maintain balance when encountering slight unevenness in the ground, ultimately stopping precisely at a preset work position and adjusting its overall body orientation to prepare for subsequent robotic arm detection operations.

[0050] S104. After arriving at the preset work position, based on the type of inspection item, control the humanoid robot's humanoid robotic arm to switch inspection heads, inspect the switch cabinet to be inspected, and obtain inspection data for each inspection item.

[0051] In some embodiments, the humanoid robotic arm is a robotic arm that mimics the structure and degrees of freedom of a human arm, typically having six or more joints, and capable of complex posture adjustments and fine manipulations in three-dimensional space. The detection head consists of various sensor modules integrated at the end of the robotic arm, which are grasped, locked, and connected to circuitry by an automatic quick-change device.

[0052] As one possible implementation, step S104 can be specifically implemented as steps S1041-S1044.

[0053] S1041. For any switch cabinet to be tested, based on the equipment model and historical data of the switch cabinet to be tested, determine the testing parameters of the humanoid robot for each testing item when testing the switch cabinet to be tested.

[0054] In some embodiments, equipment model and historical data refer to the specific specifications of the switchgear (such as KYN28A-12) and its time-series data such as inspection records and maintenance files.

[0055] For example, this embodiment of the invention does not use a fixed set of parameters to test all switchgear. Once the target switchgear model is determined, the invention retrieves the standard testing procedures for that model from the background database, such as the standard threshold for TEV testing and the standard scanning speed for ultrasonic testing. Simultaneously, it analyzes the unique historical data of the device. If a historical overheating record is found in a certain part, a lower alarm threshold will be automatically set for the infrared detection of that part, achieving dynamic parameter adjustment based on the device's health status.

[0056] S1042. Determine the sequence of detection items based on the type of detection item.

[0057] In some embodiments, the test sequence refers to the order in which multiple tests are performed on a single switchgear. The planning principle is to maximize operational efficiency and ensure data quality.

[0058] For example, sequence planning follows an intelligent principle of moving from static to dynamic and from commonalities to individual characteristics. Typically, infrared thermal imaging is performed first because it is a non-contact, wide-area, rapid scan that can initially locate abnormal heat points. Next, transient ground voltage (TEV) detection is performed, as it is also non-contact and allows for rapid electromagnetic screening of the cabinet. Finally, ultrasonic (AE) and ultra-high frequency (UHF) detection, requiring precise manipulation, are performed to accurately locate and verify suspected areas identified in the previous steps. This sequence avoids frequent probe changes by the robotic arm, improving overall efficiency.

[0059] S1043. Based on the sequence of detection items, control the humanoid robot's robotic arm to switch to the detection head corresponding to the current detection item.

[0060] In some embodiments, the quick-change device is integrated into a universal interface at the end of the humanoid robotic arm, which has functions such as automatic locking, electrical connection and pneumatic connection, allowing the robotic arm to automatically pick up and replace different detection probes like a human changing tools.

[0061] For example, the robot carries an integrated inspection head library. When it decides to perform the next inspection, the robotic arm moves to the head library and precisely mates with the target inspection head (such as an AE probe) via a quick-change device at the end effector, locking it in place. The entire process is fully automated, accompanied by a connection success feedback signal to ensure the reliability of the electrical and physical connections.

[0062] S1044. Based on the detection parameters of the humanoid robot for the current detection item, set up the humanoid robot to perform detection on the switch cabinet to be detected, and obtain the detection data of the current detection item.

[0063] In some embodiments, the present invention can send detection parameters (such as force control threshold and scanning speed) to the underlying controller of the robotic arm. Based on these parameters, the robotic arm calls its built-in dedicated control algorithms (such as force control and visual servoing) to perform specific detection actions, ensuring that each operation meets the preset accuracy and quality requirements.

[0064] For example, step S1044 can be specifically implemented as steps A1-A4.

[0065] A1. If the current detection item is transient ground voltage detection, the robotic arm is controlled to hold the TEV probe and, based on the feedback from the force sensor, a constant pressure perpendicular to the cabinet surface is applied, so that the TEV probe is in close contact with the cabinet surface and moves along a predetermined trajectory to obtain the detection data of transient ground voltage detection.

[0066] In some embodiments, force sensor feedback typically refers to a six-dimensional force / torque sensor mounted on the wrist of the robotic arm, which accurately measures the force and torque in three directions experienced by the probe when it contacts the cabinet surface. Constant pressure: To ensure the comparability and accuracy of the detection data, signal errors caused by fluctuations in contact force need to be eliminated.

[0067] For example, the robotic arm doesn't rigidly press the probe against the cabinet surface; instead, it uses adaptive impedance control. When the force sensor detects that the contact force is less than a set value, the control algorithm instructs the robotic arm to move forward slightly; when the force is too great, it retracts slightly. At the same time, the robotic arm mimics a human hand, making smooth posture adjustments based on the slight undulations of the cabinet surface to ensure that the entire bottom surface of the probe is evenly and firmly attached, avoiding signal attenuation due to localized suspension.

[0068] A2. If the current testing item is ultrasonic testing, control the robotic arm to hold the AE probe and guide the AE probe to scan along the cabinet door gap and sleeve interface of the switch cabinet at a constant distance and speed to obtain the ultrasonic testing data.

[0069] In some embodiments, constant distance and speed are crucial for ensuring the consistency and repeatability of ultrasonic signal acquisition. Uneven speed can lead to signal spectrum distortion, while changes in distance can alter signal strength.

[0070] For example, the robotic arm plans a three-dimensional spatial trajectory along the gap in the cabinet door based on a pre-input three-dimensional model of the switch cabinet. During the movement, the distance between the probe and the cabinet surface is monitored in real time through visual servo or distance sensors, and dynamic compensation is performed to maintain it at a fixed optimal detection distance. The robot's servo control system ensures that the end effector moves at a constant speed along the trajectory, a level of precision that is difficult to achieve manually.

[0071] A3. If the current inspection item is infrared thermal imaging inspection, control the robotic arm to adjust the angle and distance of the infrared thermal imager integrated at the end to avoid reflection and obtain infrared thermal images of key parts; key parts include electrical connection points and circuit breakers.

[0072] In some embodiments, the smooth surface of the anti-reflective switch cabinet may reflect heat sources in the environment (such as lights or people), forming halos or cold reflections. These false thermal images can seriously interfere with the judgment of the real hot spots.

[0073] For example, the robotic arm does not simply take pictures from the front. It controls the infrared thermal imager to take pictures from multiple angles. The invention analyzes the images in real time, and if it detects obvious reflective areas, it automatically adjusts the posture of the robotic arm and changes the shooting axis to find an optimal observation angle without reflections. In addition, it automatically adjusts the lens focal length and thermal imaging range to ensure clear thermal images and accurate temperature measurements of critical connection points.

[0074] A4. If the current testing item is UHF testing, control the robotic arm to hold the UHF sensor and align the UHF sensor with the basin insulator or testing interface of the switchgear to be tested to collect signals and obtain the UHF testing data.

[0075] In some embodiments, a basin-type insulator is a specific component on a gas-insulated switchgear, serving as a window through which ultra-high frequency signals generated by internal partial discharge can escape. Detection interfaces are standardized interfaces specifically reserved for UHF detection in some switchgear.

[0076] For example, the robotic arm holds the UHF sensor, aligning its antenna end face with and bringing it close to the basin insulator or detection interface. The control system maintains this posture for a period of time to continuously monitor the signal and acquire patterns, ensuring that any intermittent partial discharge signals can be captured.

[0077] S105. Based on the test data of each test item, perform fusion analysis and cross-validation to determine the test results of the switchgear to be tested. The test results include the working status, fault mode and fault location.

[0078] As one possible implementation, step S105 can be specifically implemented as steps S1051-S1054.

[0079] S1051. Based on the detection data of each detection item, perform spatiotemporal registration for the same part and the same time period during the detection process of the switch to be tested, and determine the registration data of each detection item.

[0080] In some embodiments, spatiotemporal registration refers to the process of unifying data collected at different times, locations, and from different sensors to the same time reference and spatial coordinate system. This is a prerequisite for effective comparison and fusion of multi-source data.

[0081] In some embodiments, spatial registration: This invention utilizes the kinematic model of the robotic arm and pre-calibrated sensor extrinsic parameters to accurately calculate the specific position of each detection data point in the unified three-dimensional coordinate system of the switch cabinet. For example, it can determine that a certain TEV abnormal signal and a certain ultrasonic signal both originate from the same bolt at the upper left corner of the cabinet door.

[0082] In some embodiments, time registration: all detection data are tagged with a high-precision timestamp at the time of acquisition. This embodiment of the invention aligns data from different detection items on the timeline according to the execution order of the tasks and the timestamps, ensuring that the analysis focuses on the same state of the device within the same time period and avoiding misjudgments.

[0083] S1052. Based on the registration data of each detection item, use threshold criteria, pattern recognition or deep learning model analysis to perform preliminary abnormality diagnosis on each type of detection data and obtain the initial diagnosis result.

[0084] In some embodiments, the initial diagnostic results are preliminary conclusions drawn from independent analysis of various test data, providing multiple possible clues to the fault, but which have not yet been cross-verified.

[0085] For example, in this embodiment of the invention, multiple dedicated analysis modules run in parallel: a threshold criterion module compares TEV amplitude, infrared temperature, etc., with safety thresholds specified by national or factory standards to quickly filter out data exceeding the limits; a pattern recognition module analyzes the waveform of ultrasonic signals or the PRPD spectrum of ultra-high frequency signals and matches it with "typical partial discharge patterns" in its feature library to identify specific discharge types; and a deep learning model uses a trained convolutional neural network to analyze infrared thermal images and automatically identify tiny hot spots or abnormal temperature distribution patterns that are difficult to detect with the naked eye. Each module outputs its independent suspected location and suspected fault type, forming a preliminary diagnostic results list.

[0086] S1053. Based on the initial diagnosis results, a data fusion algorithm is used to perform cross-validation and comprehensive judgment to obtain the fusion result.

[0087] In some embodiments, cross-validation involves cross-validating the conclusions of different detection methods (based on different physical principles such as electricity, sound, light, and heat). This is crucial for improving the reliability and confidence of the diagnosis. Fusion results: The comprehensive diagnostic conclusion obtained after fusing multi-source information has a much higher reliability than the diagnostic results of any single method.

[0088] For example, this embodiment of the invention performs correlation analysis on the initial diagnostic results. It follows a core principle: single evidence is insufficient, while multiple chains of evidence are conclusive. For instance, if the initial diagnostic results for a certain area show that TEV detection indicates the presence of a high-frequency electromagnetic signal (possible discharge), and ultrasound detection also captures a hissing spectral characteristic (possible discharge) at the same area, then these two pieces of evidence from different principles form a strong chain of evidence, and this invention will confirm the presence of a "partial discharge" fault at that area. Conversely, if only one method triggers an alarm, it may be considered interference, requiring further observation.

[0089] For example, step S1053 can be specifically implemented as steps B1-B4.

[0090] B1. The amplitude of TEV detection, the signal spectrum characteristics of ultrasound detection, the temperature distribution characteristics of infrared thermography, and the PRPD spectrum characteristics of UHF detection in the initial diagnosis results are normalized and feature extracted respectively, and transformed into independent evidence.

[0091] In some embodiments, an independent evidence body refers to a standardized data unit, after normalization and feature extraction, that represents the diagnostic conclusion of a single sensor and can be input into the fusion model. Each evidence body contains information on the sensor's "degree of support" for various failure modes.

[0092] For example, in embodiments of the present invention, different types of feature data (such as the amplitude of TEV and the spectral features of ultrasound) are mapped to a unified 0, 10, 1 interval using mathematical methods, and each piece of evidence is assigned a basic level of confidence for various fault modes (such as discharge and overheating) according to preset rules.

[0093] B2. Input the independent evidence into the multi-source information fusion model based on DS evidence theory, perform inference calculations, and determine the joint support probability and confidence interval of each candidate fault mode in the initial diagnosis results under the current multi-source evidence.

[0094] In some embodiments, DS evidence theory—a mathematical theory for dealing with uncertainty—is able to synthesize “evidence” from different sources on the same problem and calculate the overall support for various hypotheses. Joint support probability and confidence interval: The joint support probability reflects the combined support of all evidence for a certain failure mode; the confidence interval characterizes the reliable range of this support.

[0095] For example, after receiving all independent pieces of evidence, the fusion model performs calculations according to the combination rules of the DS theory. These rules cleverly handle consistency and conflict between pieces of evidence. For instance, when both TEV and ultrasound evidence strongly support "partial discharge," the model calculates a high joint support probability for this failure mode; if the evidence contradicts each other, the model provides a wider confidence interval, indicating a higher degree of uncertainty in the conclusion. This mimics the thought process of human experts making judgments after synthesizing information from multiple sources.

[0096] B3. Based on the joint support probability and confidence interval of each candidate fault mode, sort them according to the part of the switchgear to be tested, and generate fusion results.

[0097] In some embodiments, the fusion result includes the failure modes of each part, supporting independent evidence bodies for the failure modes of each part.

[0098] In some embodiments, after the model calculation is completed, the present invention will obtain a quantitative diagnostic list sorted by location. For example: Location A: Partial discharge, probability of support 95%, confidence interval [92%, 98%]; Location B: Overheating, probability of support 85%, confidence interval [80%, 90%]. This fusion result is no longer a vague suspicion, but a precise diagnosis with a quantified confidence level, providing maintenance personnel with an extremely clear and reliable basis for decision-making. The present invention will associate and store this result with the corresponding original evidence to ensure the traceability of the diagnostic conclusion.

[0099] S1054. If the fusion results show that at least two different detection methods indicate an abnormality in the same part, then it is confirmed that there is a fault in the same part of the switchgear to be tested.

[0100] S1055. Based on the data characteristics of the same location in the initial diagnosis results and fusion results, determine the fault mode and severity level of the same location.

[0101] In some embodiments, the failure modes include partial discharge, surface discharge, or contact overheating.

[0102] S1056. Based on the fault modes and severity levels of each part in each switchgear to be tested, determine the working status and fault location of each switchgear to be tested.

[0103] S1057. Based on the working status and fault location of each switchgear to be tested, as well as the fault mode and severity level of each part, generate the test results.

[0104] In some embodiments, after confirming a fault, the present invention integrates all relevant data features to accurately determine the fault mode (e.g., surface discharge, internal discharge, contact overheating) and severity level (e.g., minor, warning, severe). Finally, embodiments of the present invention automatically generate a structured inspection report. This report not only includes the original inspection data and graphs, but also the conclusions of intelligent diagnosis—clearly indicating "which part of which switchgear has what type of fault and its severity," and may include handling suggestions, completely replacing the process of manual report analysis.

[0105] This invention provides a method for live-line testing of switchgear based on a humanoid robot. It automatically obtains the target location and testing items by parsing the testing task instructions, replacing the inefficient process of manually processing work permits, providing safety briefings, and relying on experience to locate the cabinet. By fusing multi-sensor data for real-time path planning and obstacle avoidance, the robot can autonomously and optimally move to the work point, avoiding the time wasted by personnel walking back and forth between cabinets for positioning. By controlling the humanoid robotic arm to automatically switch probes according to the testing type and execute standardized testing actions, it completely replaces the tedious manual handling, moving, and changing of instruments, with operational accuracy and consistency far exceeding that of manual methods, avoiding repeated measurements due to improper operation. Through a built-in algorithm, it performs real-time fusion analysis and cross-validation of multi-source data and automatically generates testing results including fault modes, achieving real-time intelligent live-line testing of switchgear. This solves the technical problem of low efficiency in live-line testing of switchgear and improves the testing efficiency.

[0106] Optionally, the live detection method for switchgear based on humanoid robots provided in this embodiment of the invention further includes steps S201-S205.

[0107] S201. During the process of collecting test data, calculate the quality assessment index of the current test data in real time.

[0108] In some embodiments, quality evaluation metrics include: signal-to-noise ratio of transient ground voltage detection data, effective pulse count of ultrasonic detection signal, edge sharpness of infrared thermogram, and temperature measurement stability.

[0109] In some embodiments, quality assessment metrics are key parameters used to quantify the validity and usability of single-sample detection data. They directly determine the reliability and diagnostic value of the data. Signal-to-noise ratio (SNR): The ratio of useful signal strength to background noise intensity. A low SNR means the effective signal is overwhelmed by noise, rendering the data invalid. Number of effective pulses: In ultrasonic testing, the number of pulses exceeding a set threshold within a specific time period that can be considered genuine discharge pulses. Too many or too few pulses may indicate interference or poor sensor contact. Edge sharpness: In infrared thermography, the sharpness of the edges of device components. Blurred images usually indicate inaccurate focusing or interference from steam or dust, leading to inaccurate temperature measurements.

[0110] For example, in embodiments of the present invention, a dedicated quality analysis algorithm is launched in the background while data is being collected. For instance, a Fast Fourier Transform is performed on the TEV signal to analyze the ratio of signal power in a specific frequency band to noise power across the entire frequency band, and the signal-to-noise ratio is output in real time. For infrared thermal images, computer vision algorithms are used to calculate the gradient values ​​of the edges of key components such as switch contacts in the image to quantify their sharpness.

[0111] S202. Compare the quality assessment indicators with the preset quality thresholds to obtain the comparison results.

[0112] In some embodiments, the quality threshold is a pre-defined quality pass line based on different testing items and technical standards. It is an objective standard for determining whether data can be adopted. The preset strategy library is an expert knowledge base that stores "if item [X] has quality problem [Y], then take adjustment measure [Z]". It encapsulates the debugging experience of senior testing personnel.

[0113] For example, in this embodiment of the invention, the calculated quality indicators are compared with thresholds in real time. If a non-compliance is detected, it does not simply report an error, but instead initiates an intelligent decision-making process. The invention uses the specific detection item and the specific indicator that is non-compliant as a key to query the optimal correction scheme from the strategy library. For example, if "ultrasonic detection - low effective pulse count" is identified, the strategy library may output the instruction to "reduce the probe distance and repeat the scan".

[0114] S203. If there are substandard indicators in the comparison results, then according to the substandard indicators of the current test item, match and generate a robot arm control parameter adjustment strategy from the preset strategy library.

[0115] S204. Based on the robotic arm control parameter adjustment strategy, control the humanoid robotic arm to make dynamic adjustments.

[0116] In some embodiments, dynamic adjustment includes: for transient ground voltage detection, increasing the contact pressure and reducing the moving speed; for ultrasonic detection, reducing the distance between the probe and the cabinet surface and repeatedly scanning the abnormal section; for infrared thermal imaging detection, adjusting the focal length and angle and turning on the anti-glare filter before re-shooting.

[0117] In some embodiments, adaptive adjustment refers to the behavior of a robot actively changing its own operating parameters based on feedback (in this case, quality feedback), which reflects intelligent and closed-loop control capabilities.

[0118] For example, in this embodiment of the invention, the adjustment strategy is parsed into specific robotic arm movement commands. Increase TEV probe pressure: By increasing the target force setting value in the force control loop, the robotic arm presses the probe more forcefully against the cabinet surface, ensuring tight contact. Reduce AE probe movement speed: Directly adjust the speed parameters of the trajectory planner to make the probe move slower, so as to collect richer and more stable signals per unit time. Activate anti-glare filter: Send a command to the infrared thermal imager to activate its built-in anti-glare function (such as installing a polarizing filter) to filter out mirror reflections. Simultaneously, the robotic arm carries the thermal imager to several different preset angles for re-shooting, completely avoiding reflections through multi-view synthesis technology.

[0119] S205. After adjusting the control parameters, re-test and calculate the quality assessment index of the newly collected data; if the new index is higher than the preset threshold, replace the original substandard test data with the newly collected high-quality data.

[0120] In some embodiments of the invention, previously substandard data is overwritten or replaced with newly acquired, high-quality data. This is a crucial step in achieving data autonomy, ensuring that the final dataset used for analysis is of high overall quality.

[0121] For example, after adjustments and retesting, this embodiment of the invention will recalculate the quality indicators of the new data. Only after the new indicators are confirmed to be qualified will this invention perform a replacement operation—marking the old data file as "invalid" or deleting it directly in the data storage area, and binding and archiving the new data file with metadata such as device ID and testing time. If the quality still does not meet the standards after several adjustments, this invention will record a "data acquisition anomaly" log and may report it to remote maintenance personnel for intervention.

[0122] Thus, by introducing a closed-loop control mechanism for real-time data quality assessment and adaptive adjustment, this invention significantly improves the reliability and validity of test data. Its technical effect lies in transforming the traditional data quality control process, which relies on manual experience and rework, into a fully automated, online, and intelligent process. This ensures the accuracy and consistency of data from the source, providing a high-quality data foundation for subsequent intelligent diagnosis, thereby enhancing the overall automation level and reliability of the testing system.

[0123] Optionally, the live detection method for switchgear based on humanoid robots provided in this embodiment of the invention further includes steps S301-S306.

[0124] S301. For any switch cabinet to be tested, based on the test results of the switch cabinet to be tested and the historical test results, perform spatiotemporal alignment to construct the feature parameter sequence of each part of the switch cabinet to be tested at different time points.

[0125] In some embodiments, the characteristic parameter sequence refers to the sequence formed by arranging key status indicators (such as infrared temperature and TEV amplitude) of the same equipment part (such as the A-phase connector of a circuit breaker) in chronological order at different historical time points. This is the data basis for trend analysis.

[0126] For example, in this embodiment of the invention, based on the currently detected switchgear number and location identifier, all past detection records for that location are automatically retrieved from the historical database. Using the unique equipment location code and a timestamp accurate to the second, the currently detected characteristic parameters such as temperature and discharge level are precisely aligned with historical data in both time and space dimensions, forming a complete historical-current data sequence.

[0127] S302. Based on the characteristic parameter sequence of each part at different time points, calculate the key health index of each part at different time points. The key health index characterizes the degree of insulation degradation of the equipment.

[0128] S303. Based on the key health indices of each part at different time points, a time series prediction algorithm is used to model the evolution trend of the key health indices and predict the change trajectory of the key health indices of each part in a future set period.

[0129] In some embodiments, a key health index is a normalized numerical value (e.g., 0-100 points) synthesized by an algorithm to comprehensively quantify the health status of a part of a device. It is typically calculated by weighting multiple detection parameters (such as temperature, discharge signal, and insulation resistance), providing a more comprehensive reflection of the device's condition than a single parameter. A time series forecasting algorithm is a mathematical model capable of predicting future data trends based on historical data sequences.

[0130] For example, this embodiment of the invention does not view a single detection data point in isolation. Instead, it inputs the aligned sequence of feature parameters into a health assessment model to calculate the health index corresponding to each time point. Subsequently, using prediction algorithms such as autoregressive integral moving average or more advanced long short-term memory networks, the time series of the health index is learned to capture the pattern of its slow deterioration, thereby predicting the trajectory of the index's changes over the next few weeks or months.

[0131] S304. Compare the changes in key health indices of each part over a future set period with preset safe operating thresholds.

[0132] S305. When the predicted value of a key health index of a certain part exceeds the safe operation threshold, calculate the remaining safe operation time of that part.

[0133] In some embodiments, the safe operating threshold refers to a critical safety value for a health index or key parameter (such as temperature). When the predicted trajectory reaches this threshold, it means the device has moved from a "watchful" state to a "risk" state. Remaining safe operating time: Based on the predicted trajectory, the estimated time from the current moment until the health index reaches the safe operating threshold. This is the core output of predictive maintenance.

[0134] For example, this embodiment of the invention compares the future health index trajectory with a preset safety threshold line. Using mathematical interpolation, the intersection point of the trajectory and the threshold line is accurately calculated. The difference between the time point corresponding to this intersection and the current time is the remaining safe operating time. Based on the remaining time and the rate of deterioration, this invention automatically classifies the risk level (e.g., high, medium, low) and generates a report containing specific warning information.

[0135] S306. Based on the key health indices of each part of the switchgear under test at different time points, the change trajectory of these indices over a future set period, and the remaining safe operating time, generate risk warning information.

[0136] In some embodiments, dynamic updating refers to the process of retraining and fine-tuning the prediction model using the latest detection data, enabling the model to track new trends in device state changes and maintain prediction accuracy.

[0137] For example, after each new detection is completed, the present invention automatically adds the newly collected data points to the historical sequence, forming a longer dataset containing the latest status information. Subsequently, the present invention uses this updated dataset to perform incremental learning or periodic retraining on the internal prediction model, enabling the model to adapt to subtle changes in the aging process of the equipment, much like an increasingly experienced doctor continuously improving the accuracy of its diagnosis and prognosis.

[0138] Thus, this invention realizes a paradigm shift from passive response maintenance to proactive predictive maintenance. By quantifying equipment health trends and estimating remaining lifespan, it provides forward-looking data support for operation and maintenance decisions, thereby enabling scientific scheduling of maintenance plans, effective avoidance of sudden failures, significant reduction in operation and maintenance costs, and improvement of power grid reliability.

[0139] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0140] The following are device embodiments of the present invention. For details not described in detail, please refer to the corresponding method embodiments described above.

[0141] Figure 2 A schematic diagram of a switchgear live-line detection device based on a humanoid robot, provided by an embodiment of the present invention, is shown. The detection device 400 includes a communication module 401 and a processing module 402.

[0142] Communication module 401 is used to receive the testing task instructions from the switchgear to be tested. The processing module 402 is used to parse the detection task instruction to obtain the location information and detection item type of the switchgear to be tested. The detection item type includes at least one of the following: transient ground voltage detection, ultrasonic detection, infrared thermal imaging detection, and ultra-high frequency detection. Based on the location information of the switchgear to be tested and the location information of the humanoid robot, it performs real-time path planning and obstacle avoidance, and moves to the preset working position of the switchgear to be tested. After arriving at the preset working position, based on the detection item type, it controls the humanoid robot's humanoid robotic arm to switch detection heads to detect the switchgear to be tested and obtain the detection data of each detection item. Based on the detection data of each detection item, it performs fusion analysis and cross-validation to determine the detection result of the switchgear to be tested. The detection result includes the working status, fault mode, and fault location.

[0143] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. The electronic device 500 includes: a processor 501, a memory 502, and a computer program 503 stored in the memory 502 and executable on the processor 501. When the processor 501 executes the computer program 503, it implements the steps in the above-described method embodiments. Alternatively, when the processor 501 executes the computer program 503, it implements the functions of each module / unit in the above-described device embodiments.

[0144] For example, the computer program 503 may be divided into one or more modules / units, which are stored in the memory 502 and executed by the processor 501 to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program 503 in the electronic device 500.

[0145] The processor 501 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0146] The memory 502 can be an internal storage unit of the electronic device 500, such as a hard disk or memory of the electronic device 500. The memory 502 can also be an external storage device of the electronic device 500, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the electronic device 500. Furthermore, the memory 502 can include both internal and external storage units of the electronic device 500. The memory 502 is used to store the computer program and other programs and data required by the terminal. The memory 502 can also be used to temporarily store data that has been output or will be output.

[0147] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for detecting live electrical signals in a switchgear based on a humanoid robot, characterized in that, include: Receive the testing task instruction for the switchgear to be tested; The detection task instruction is parsed to obtain the location information of the switchgear to be tested and the type of detection item. The type of detection item includes at least one of the following: transient ground voltage detection, ultrasonic detection, infrared thermal imaging detection and ultra-high frequency detection. Based on the location information of the switch cabinet to be tested and the location information of the humanoid robot, real-time path planning and obstacle avoidance are performed, and the robot moves to the preset working position of the switch cabinet to be tested. After arriving at the preset work position, based on the type of inspection item, the humanoid robot's robotic arm is controlled to switch inspection heads to inspect the switch cabinet to be inspected, and inspection data for each inspection item is obtained. Based on the test data of each test item, fusion analysis and cross-validation are performed to determine the test results of the switchgear to be tested. The test results include the working status, fault mode and fault location.

2. The method for live detection of switchgear based on humanoid robots according to claim 1, characterized in that, The process of parsing the detection task instruction yields the location information of the switchgear to be inspected and the type of detection item, including: Semantic analysis is performed on the detection task instructions to identify the switchgear equipment number and keywords of the detection items to be performed in the detection task instructions; Based on the switchgear equipment number and the preset substation map model, the global coordinates of each switchgear to be tested are determined. Based on the keywords of the test items to be performed, the equipment testing knowledge base is queried, and the matching between the test items and the switch cabinet equipment model is verified to determine the test item type for each switch cabinet to be tested. Based on the global coordinates and detection item type of each switchgear to be tested, as well as the shape information of the humanoid robot, a collision test is conducted using the substation map model to determine the preset working position of each switchgear to be tested.

3. The method for live detection of switchgear based on humanoid robots according to claim 1, characterized in that, The step of performing real-time path planning and obstacle avoidance based on the location information of the switch cabinet to be tested and the location information of the humanoid robot, and moving to the preset working position of the switch cabinet to be tested, includes: Multidimensional perception data is acquired through lidar, vision cameras, and inertial measurement units mounted on humanoid robots; Based on the multi-dimensional sensing data, the environmental map of the substation is constructed or updated in real time. Starting from the location information of the humanoid robot and taking the preset working position of each switch cabinet to be tested as the target point, global path planning is performed in the environmental map to determine multiple candidate movement paths; A multi-objective optimization function is constructed based on the total path length, smoothness, energy consumption, and safety margin. Combined with the multi-dimensional perception data and environmental map, dynamic obstacle avoidance evaluation is performed to determine the optimal path among multiple candidate movement paths. Based on the optimal path, the humanoid robot is controlled to move to the preset working position of each switch cabinet to be tested.

4. The method for live detection of switchgear based on humanoid robots according to claim 1, characterized in that, Based on the type of inspection item, the humanoid robot's robotic arm is controlled to switch inspection heads to inspect the switch cabinet to be inspected, obtaining inspection data for each inspection item, including: For any switch cabinet to be tested, based on the equipment model and historical data of the switch cabinet, the humanoid robot determines the testing parameters for each testing item when testing the switch cabinet; Based on the types of detection items, a sequence of detection items is determined; Based on the sequence of detection items, control the humanoid robot's humanoid robotic arm to switch to the detection head corresponding to the current detection item; Based on the detection parameters of the humanoid robot for the current detection project, the humanoid robot is set up to detect the switch cabinet to be tested, and the detection data of the current detection project is obtained.

5. The method for live detection of switchgear based on humanoid robots according to claim 4, characterized in that, The process involves setting up a humanoid robot to perform inspections on the switch cabinet to be inspected, based on the inspection parameters of the current inspection item, to obtain the inspection data for the current inspection item, including: If the current detection item is transient ground voltage detection, the robotic arm is controlled to hold the TEV probe and, based on the feedback from the force sensor, a constant pressure perpendicular to the cabinet surface is applied, so that the TEV probe is in close contact with the cabinet surface and moves along a predetermined trajectory to obtain the detection data of transient ground voltage detection. If the current testing item is ultrasonic testing, the robotic arm is controlled to hold the AE probe and guide it to scan along the cabinet door gap and sleeve interface of the switch cabinet at a constant distance and speed to obtain the ultrasonic testing data. If the current detection item is infrared thermal imaging detection, the robotic arm is controlled to adjust the angle and distance of the infrared thermal imager integrated at the end to avoid reflection and obtain infrared thermal images of key parts; the key parts include electrical connection points and circuit breakers. If the current testing item is UHF testing, the robotic arm is controlled to hold the UHF sensor and align it with the basin insulator or testing interface of the switchgear to be tested to collect signals and obtain the UHF testing data.

6. The method for live detection of switchgear based on humanoid robots according to claim 1, characterized in that, The process involves fusing and cross-validating the test data from each test item to determine the test results for the switchgear under test, including: Based on the detection data of each detection item, spatiotemporal registration is performed on the same part and the same time period during the detection process of the switch to be tested to determine the registration data of each detection item; Based on the registration data of each detection item, the threshold criterion, pattern recognition or deep learning model analysis method is used to perform preliminary anomaly diagnosis on each type of detection data to obtain the initial diagnosis result. Based on the initial diagnosis results, a data fusion algorithm was used to perform cross-validation and comprehensive judgment to obtain the fusion result; If the fusion results show that at least two different detection methods indicate an abnormality in the same part, then it is confirmed that there is a fault in the same part of the switchgear under test. Based on the data characteristics of the same location in the initial diagnosis and fusion results, the fault mode and severity level of the same location are determined; the fault mode includes partial discharge, surface discharge, or contact overheating. Based on the fault modes and severity levels of each part in each switchgear to be tested, the working status and fault location of each switchgear to be tested are determined. The test results are generated based on the working status and fault location of each switchgear to be tested, as well as the fault mode and severity level of each part.

7. The method for live detection of switchgear based on humanoid robots according to claim 6, characterized in that, Based on the initial diagnosis results, a data fusion algorithm is used to perform cross-validation and comprehensive judgment to obtain the fusion result, including: The amplitude of TEV detection, the signal spectrum features of ultrasound detection, the temperature distribution features of infrared thermography, and the PRPD spectrum features of UHF detection in the initial diagnosis results are normalized and feature extracted respectively, and transformed into independent evidence. The independent evidence is input into a multi-source information fusion model based on DS evidence theory for inference calculation to determine the joint support probability and confidence interval of each candidate fault mode in the initial diagnosis result under the current multi-source evidence. Based on the joint support probability and confidence interval of each candidate fault mode, the parts of the switchgear to be tested are sorted to generate a fusion result. The fusion result includes the fault modes of each part and the independent evidence supporting the fault modes of each part.

8. The method for live detection of switchgear based on a humanoid robot according to any one of claims 1 to 7, characterized in that, The method further includes: During the data acquisition process, the quality assessment indicators of the current data are calculated in real time. The quality assessment indicators include: the signal-to-noise ratio of transient ground voltage detection data, the number of effective pulses of ultrasonic detection signal, the edge clarity of infrared thermal image and temperature measurement stability. The quality assessment indicators are compared with preset quality thresholds to obtain the comparison results; If there are any non-compliant indicators in the comparison results, then based on the non-compliant indicators of the current test item, a robot arm control parameter adjustment strategy will be matched and generated from the preset strategy library. Based on the aforementioned robotic arm control parameter adjustment strategy, the humanoid robotic arm is controlled to make dynamic adjustments; the dynamic adjustments include: for transient ground voltage detection, increasing the contact pressure and reducing the moving speed; for ultrasonic detection, reducing the distance between the probe and the cabinet surface and repeatedly scanning the abnormal section; for infrared thermal imaging detection, adjusting the focal length and angle, turning on the anti-glare filter, and then re-shooting. After adjusting the control parameters, the test is repeated, and the quality assessment index of the newly collected data is calculated. If the new index is higher than the preset threshold, the original substandard test data is replaced with the newly collected high-quality data.

9. The method for live detection of switchgear based on a humanoid robot according to any one of claims 1 to 7, characterized in that, The method further includes: For any switch cabinet to be tested, based on the test results and historical test results of the switch cabinet to be tested, spatiotemporal alignment is performed to construct the feature parameter sequence of each part of the switch cabinet to be tested at different time points; Based on the characteristic parameter sequence of each part at different time points, the key health index of each part at different time points is calculated, and the key health index characterizes the degree of insulation degradation of the equipment. Based on the key health indices of each part at different time points, a time series prediction algorithm is used to model the evolution trend of the key health indices and predict the change trajectory of the key health indices of each part in a future set period. The changes in key health indices for each part over a future set period are compared with preset safe operating thresholds. When the predicted value of a key health index of a certain part exceeds the safe operation threshold, calculate the remaining safe operation time of that part. Based on the key health indices of various parts of the switchgear under test at different time points, their changing trajectory over a future set period, and the remaining safe operating time, risk warning information is generated.

10. A live-line detection system for switchgear based on a humanoid robot, characterized in that, The system includes an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor is configured to invoke and run the computer program stored in the memory to perform the method as described in any one of claims 1 to 9.

Images