Insulator zero-value detection robot control method

Abstract

The application relates to the field of power equipment detection and intelligent robot control, and discloses a kind of insulator zero value detection robot control method.The method comprises the following steps: deploying a track robot on a power transmission line, synchronously collecting four-dimensional data of infrared thermal imaging, ultraviolet corona, capacitance coupling voltage and laser profile;spatial alignment and multi-modal feature extraction are performed on multi-source information;insulator unit-level health scores are generated through cross-modal attention fusion;zero value faults are determined based on the score threshold and the re-inspection or confirmation process is triggered.The system includes a double-wheel clamping drive mechanism, a multi-source sensor array and a central controller.The application realizes fully automatic, high-precision and high-safety detection under live-line conditions, significantly improving identification accuracy and work efficiency.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of power equipment detection and intelligent robot control, and particularly relates to a control method for a zero-value insulator detection robot. BACKGROUND

[0002] With the continuous improvement of the intelligent level of high-voltage transmission line operation and maintenance, as a key component to ensure the safe operation of the power grid, the real-time monitoring of the performance state of insulators has become one of the core tasks of power system maintenance. Insulators are easily affected by environmental erosion, mechanical stress, uneven electric field distribution and other factors during long-term operation, resulting in the loss of insulation capability of some insulators, forming "zero-value" or "low-value" defects. If such defects are not identified in time, they may cause flashover, breakdown or even wire breakage accidents, seriously threatening the stability of the power grid and the safety of personnel. Therefore, accurate and efficient detection of the electrical performance of insulators has become an important research direction in the field of intelligent inspection.

[0003] Based on multi-modal perception fusion and digital twin collaboration, the zero-value insulator online diagnosis technology aims to realize dynamic evaluation and autonomous decision control of the state of insulators in a non-contact and non-power-off manner. This technology relies on the joint perception of acoustic signals and ultraviolet pulses, and combines virtual models to realize real-time mapping and simulation of physical entities, in order to improve the robustness and response speed of fault identification in complex electromagnetic and meteorological environments. However, existing detection methods still have significant limitations: traditional methods rely on manual tower climbing or contact measurement after power-off, which not only has low operation efficiency, but also has high safety risks in high-voltage environments; and current mainstream single-sensor online detection schemes (such as relying only on infrared thermal imaging or ultraviolet imaging) are easily disturbed by environmental noise, background radiation or equipment aging, resulting in high misjudgment rates and difficulty in meeting high-reliability operation and maintenance requirements.

[0004] Existing technologies have not yet formed effective integration in multi-source information fusion mechanism, dynamic environment adaptability and detection-decision closed-loop control. Specifically, acoustic signals can reflect abnormal vibrations caused by internal cracks or looseness of insulators, but are easily disturbed by wind noise and wire corona; ultraviolet pulses can capture local discharge phenomena, but have insufficient sensitivity to weak discharge signals and are easily suppressed by sunlight background. If used independently, neither can accurately distinguish between zero-value insulators and normal aging states. At the same time, the lack of deep coupling with the digital twin platform makes the detection results unable to drive the robot body to perform path re-planning, sensor parameter self-adjustment or re-inspection strategy generation, resulting in a "perception-judgment" fragmented state of the system. In key scenarios such as ultra-high voltage and cross-region power transmission, the above defects directly restrict the timeliness, accuracy and operation safety of insulator defect identification, and there is an urgent need for a zero-value insulator detection robot control method that integrates acoustic-ultraviolet multi-modal perception and has real-time reasoning and autonomous control capabilities. SUMMARY

[0005] The application provides an insulator zero-value detection robot control method, aiming to solve the technical problems of low efficiency and high operation risk caused by traditional detection needing to stop or contact measurement, and single sensor being prone to misjudgment. The method realizes full-automatic, high-precision and high-safety zero-value fault identification of a power transmission line insulator string in a live operation state by constructing a multi-source heterogeneous sensor fusion architecture and a non-contact dynamic detection mechanism.

[0006] As an embodiment of the application, the insulator zero-value detection robot control method comprises the following steps: deploying a track robot with autonomous walking capability on a ground wire or a conductor of a power transmission line; synchronously collecting multi-dimensional physical field information of an insulator string by an infrared thermal imaging sensor, an ultraviolet corona detector, a distributed capacitance coupled voltage sensing array and a laser profile scanner; based on a preset time-space synchronization reference, performing timestamp alignment and space coordinate normalization processing on the multi-source sensor data;

[0007] The space coordinate normalization is realized by robot IMU data and laser geometric reference, a unified three-dimensional coordinate system with the axis of the insulator string as the reference is first constructed, and then the observation results of each sensor are mapped to the coordinate system through homogeneous coordinate transformation, ensuring the spatial alignment accuracy of the multi-source data.

[0008] A multi-modal feature extraction network is used to analyze the original signals of each sensing channel respectively, to generate a thermal distribution feature vector, a corona discharge intensity sequence, a voltage gradient distribution atlas and a geometric structure deviation parameter; the four types of features are input into a cross-modal attention fusion module, the confidence weight of each feature channel at the insulator unit level is calculated, and the comprehensive health state score is weighted and synthesized; whether there is a zero-value insulator is determined according to the comparison between the comprehensive health state score and the preset threshold interval, and the fault position number and the confidence level are output; according to the determination result, the central controller generates a path correction instruction and a re-inspection scheduling strategy to drive the robot to perform local backtracking scanning or adjacent insulator cross-validation operation.

[0009] Further, the track robot is provided with a double-wheel clamping driving mechanism, which comprises an upper pressing wheel assembly, a lower supporting wheel assembly and a servo motor driving unit; the upper pressing wheel assembly applies a constant pre-tightening force through a spring loading mechanism, ensuring that the robot remains stable adhesion on different inclination lines; the servo motor driving unit adopts a closed-loop vector control mode, and adjusts the speed difference between the left and right wheels in real time to realize accurate displacement control along the axial direction of the conductor, with a positioning error of not more than ±5mm.

[0010] Further, the infrared thermal imaging sensor is configured as a long-wave infrared focal plane array, with a working wavelength range of 8-14 microns, a spatial resolution of 640x480 pixels, and a frame rate of no less than 25 Hz; the sensor is installed on a front-end gimbal of the robot, and automatic focusing and field coverage of the target insulator string are achieved through a two-dimensional servo mechanism, so that each insulator shed area is completely imaged.

[0011] Further, the ultraviolet corona detector adopts a sun-blind ultraviolet photomultiplier tube, with a response wavelength range of 240-280 nm and a minimum detectable photon flux of 50 photons per second; the detector is installed on the same optical axis as the visible light auxiliary camera, and an image registration algorithm is used to superimpose the ultraviolet discharge signal into the visible light image to form a corona spatial distribution heat map, which is specifically realized through camera joint calibration, ultraviolet signal quantization, affine transformation registration, and semi-transparent color superposition, i.e., first, the internal and external parameters of the two cameras are calibrated to establish coordinate mapping, then the ultraviolet photon signal is quantized into a grayscale image and mapped to the visible light coordinate system, and finally, the heat color mode is converted and superimposed.

[0012] Further, the distributed capacitance coupling voltage sensing array is composed of 9 micro-capacitance probes arranged along the transverse direction of the robot, with a probe spacing of 10 cm and a probe sensing surface facing downward toward the insulator string; each probe is connected to a 12-bit analog-to-digital converter through a high-input-impedance buffer amplifier, with a sampling frequency of 1000 Hz; the voltage distribution curve of the insulator string along the axial direction is inverted by measuring the rate of change of the potential difference between adjacent probes.

[0013] Further, the laser profile scanner adopts a triangulation ranging principle, emits a semiconductor laser beam with a wavelength of 650 nm, has a scanning frequency of 2000 Hz, and has a ranging accuracy of ±0.1 mm; the scanner performs continuous line scanning along the axial direction of the insulator string to obtain three-dimensional point cloud data of the outer diameter of the insulator shed, the exposed length of the steel foot, and the surface crack depth of each insulator.

[0014] Further, the multi-modal feature extraction network includes 4 parallel sub-networks: the first sub-network is a one-dimensional convolutional neural network for processing voltage gradient distribution maps to extract voltage drop inflection point positions and amplitude decay rates; the second sub-network is a two-dimensional residual network for analyzing infrared thermal images to identify abnormal hot spot regions and their temperature gradient features; the third sub-network is a time-series pulse counting model for counting the density of ultraviolet photon events per unit time and locating the discharge source combined with a spatial clustering algorithm; and the fourth sub-network is a point cloud semantic segmentation network based on the PointNet architecture for component-level segmentation of laser scanning point clouds to output structural integrity indicators.

[0015] Further, the cross-modal attention fusion module adopts a gated cross-attention mechanism, and the input of the module is the feature tensors output by the four sub-networks; the module first linearly projects each feature tensor to a unified embedding space, then takes the voltage distribution feature as a query vector, and the remaining three types of features correspond to key vectors and value vectors respectively, calculates the correlation weight through the query vector and each key vector, and then realizes fusion by weighted sum of each value vector to calculate the attention weight; finally, the output is the health state embedding vector of the insulator unit after weighted fusion, and the dimension is 256.

[0016] Further, the comprehensive health state score is obtained by mapping the health state embedding vector through a fully connected regression layer, and the value range is 0 to 100; the preset threshold interval includes a first threshold 70 and a second threshold 90; when the score is lower than 70, the insulator is determined to be zero value; when the score is between 70 and 90, the insulator is marked as suspected failure, triggering the re-inspection process; when the score is higher than 90, the insulator is determined to be normal.

[0017] Further, the re-inspection scheduling strategy includes: controlling the robot to retreat to the position of the previous insulator and re-executing multi-source data acquisition; adjusting the integration time of the infrared and ultraviolet sensors to improve the signal-to-noise ratio of weak signals; starting the high-density scanning mode of the laser scanner, and the sampling point spacing is reduced to 0.5 millimeters; if the two detection results are consistent, the fault is confirmed; otherwise, continue to detect the next insulator and record the contradictory data for background analysis.

[0018] Further, the central controller is built-in with a task state machine, and the states include initialization, cruise detection, local re-inspection, fault confirmation, and return standby; the state transition is driven by the detection result and the environment feedback; in the cruise detection state, the robot advances at a speed of 3 meters per minute; in the local re-inspection state, the speed is reduced to 0.5 meters per minute, and the high-precision positioning mode is enabled.

[0019] Further, the method further includes a communication guarantee mechanism: the robot establishes a bidirectional data link with the ground base station through a power line carrier communication module; the key detection data is transmitted in the form of groups after compression encoding, and each group contains the insulator number, the four types of original sensor fragments, the feature vector and the preliminary determination result; the ground base station receives and performs secondary verification, and can remotely issue intervention instructions.

[0020] Compared with the prior art, the beneficial effects of the present application are:

[0021] By constructing an infrared, ultraviolet, capacitively coupled voltage and laser profile four-source heterogeneous sensing system, the thermal, optical, electrical and shape four-dimensional physical field information of the insulator is synchronously acquired under non-power-off and non-contact conditions, which fundamentally avoids the safety risks and efficiency bottlenecks of manual tower climbing or power-off detection; the feature fusion strategy based on the cross-modal attention mechanism is adopted, which effectively suppresses the misjudgment caused by environmental interference or device drift of a single sensor, and significantly improves the accuracy and robustness of zero-value insulator identification; the closed-loop recheck scheduling and task state machine control logic are introduced, so that the robot has the ability of autonomous decision-making and abnormal processing, and realizes the full-process automation from data acquisition to fault confirmation; the whole method supports stable operation on an ultra-high voltage alternating current or direct current transmission line, the detection speed is more than 5 times higher than that of the traditional method, the false positive rate is less than 2%, and the false negative rate is less than 0.5%, which provides reliable technical support for intelligent operation and maintenance of power grids. BRIEF DESCRIPTION OF DRAWINGS

[0022] Fig. 1 is the overall technical scheme architecture schematic diagram of the present application;

[0023] Fig. 2 is the core principle framework schematic diagram of the cross-modal attention fusion module in the present application;

[0024] Fig. 3 is the logic flow framework diagram of the multi-source heterogeneous sensing data acquisition and preprocessing stage in the present application;

[0025] Fig. 4 is the parallel processing logic framework diagram of the multi-modal feature extraction network in the present application;

[0026] Fig. 5 is the detection and recheck scheduling logic flow framework diagram driven by the robot task state machine in the present application;

[0027] Fig. 6 is the multi-level interaction relationship and data flow schematic diagram of the track-type robot and the ground base station in the present application. DETAILED DESCRIPTION

[0028] Please refer to the accompanying Figs. 1 to 6 The present application provides a kind of insulator zero detection robot control method, its core is to build a closed-loop control system for realizing full-automatic, non-contact, high-precision identification of zero-value insulator under the condition of live transmission line operation. The method is to deploy track-type robot with autonomous walking ability on ground wire or conductor of transmission line, synchronously acquire multi-dimensional physical field information of insulator string, and complete dynamic evaluation and fault determination of insulator health status based on multi-source heterogeneous sensing fusion and cross-modal feature interaction mechanism.

[0029] The method first performs a robot deployment step. The track robot is installed on the ground wire or conductor of the power transmission line, and its structure includes a double-wheel clamping driving mechanism. The mechanism is composed of an upper pressing wheel assembly, a lower supporting wheel assembly, and a servo motor driving unit. The upper pressing wheel assembly applies a constant pre-tightening force through a spring-loaded mechanism to ensure that the robot remains stable on different inclined lines, preventing derailment due to wind load or line vibration. The servo motor driving unit uses a closed-loop vector control mode to adjust the speed difference between the left and right wheels in real time to achieve precise displacement control along the conductor axis, with a positioning error of not more than ±5 mm. After the robot starts, it enters an initialization state, completes self-checking, sensor calibration, and communication link establishment.

[0030] Subsequently, a multi-source sensor data synchronous acquisition step is performed. The robot carries four types of core sensors: an infrared thermal imaging sensor, an ultraviolet corona detector, a distributed capacitively coupled voltage sensing array, and a laser profile scanner. The infrared thermal imaging sensor is configured as a long-wave infrared focal plane array with a working wavelength range of 8 to 14 microns, a spatial resolution of 640x480 pixels, and a frame rate of not less than 25 Hz. The sensor is installed on the robot's front-end gimbal and achieves automatic focusing and field coverage of the target insulator string through a two-dimensional servo mechanism, ensuring that each insulator shed area is completely imaged.

[0031] The ultraviolet corona detector uses a sun-blind ultraviolet photomultiplier tube with a response wavelength range of 240 to 280 nanometers and a minimum detectable photon flux of 50 photons per second. The detector is installed on the same optical axis as the visible light auxiliary camera, and the ultraviolet discharge signal is superimposed into the visible light image through an image registration algorithm to form a corona spatial distribution heat map.

[0032] The specific implementation process is as follows: First, jointly calibrate the ultraviolet camera and the visible light camera to obtain the intrinsic parameters (focal length, principal point coordinates) and extrinsic parameters (relative translation, rotation angle) of the two cameras, and establish the pixel coordinate mapping relationship; then, quantize the discharge photon number signal collected by the ultraviolet detector into a gray value (the higher the discharge intensity, the larger the gray value), forming a single-channel ultraviolet gray map; subsequently, use an affine transformation algorithm to map the ultraviolet gray map to the visible light image coordinate system according to the calibration parameters, and correct the coordinate deviation through bilinear interpolation; finally, convert the ultraviolet gray map to a heat map color mode (discharge intensity from low to high corresponds to blue, green, yellow, and red gradual change), and superimpose it on the corresponding area of the visible light image in a semi-transparent mode with 50% transparency, ultimately forming a heat map that directly reflects the spatial distribution of corona.

[0033] The distributed capacitively coupled voltage sensing array consists of nine miniature capacitive probes arranged laterally along the robot, with a spacing of 10 cm between each probe and the sensing surface of the probe facing the insulator string below. Each probe is connected to a 12-bit analog-to-digital converter via a high input impedance buffer amplifier, with a sampling frequency of 1000 Hz. By measuring the rate of change of potential difference between adjacent probes, the voltage distribution curve of the insulator string along the axial direction is inverted.

[0034] The laser profile scanner employs the triangulation principle, emitting a semiconductor laser beam with a wavelength of 650 nanometers and a scanning frequency of 2000 Hz, achieving a ranging accuracy of ±0.1 mm. This scanner performs continuous line scanning along the insulator string axis, acquiring three-dimensional point cloud data of the outer diameter of the sheds, the exposed length of the steel feet, and the depth of surface cracks for each insulator. All four types of sensors operate synchronously under a unified clock source, ensuring strict time alignment of all raw data.

[0035] After data acquisition, the system performs spatiotemporal synchronization and coordinate normalization. Based on a preset spatiotemporal synchronization benchmark, the system timestamps infrared image frames, ultraviolet event streams, voltage sampling sequences, and laser point cloud trajectories. Time alignment uses a hardware trigger signal as a common reference point; all sensor data streams are interpolated and corrected with this reference point as their zero point, unifying data from different sampling frequencies onto the same time grid. The time resolution is set to 10 milliseconds. Spatial coordinate normalization, based on the robot's current position and attitude information, combined with the geometric benchmark provided by the laser scanner, maps the sensor observations to a unified three-dimensional coordinate system with the insulator string axis as the reference.

[0036] The specific mapping process is as follows:

[0037] Basic data acquisition: The robot acquires its own position coordinates in real time through an inertial measurement unit (IMU). And attitude angles (roll angle, pitch angle, yaw angle); the laser scanner scans the hardware, steel cap and other feature points of the insulator string to obtain the three-dimensional coordinates of the feature points, and establishes the geometric reference by taking the line connecting the center points of the hardware at both ends of the insulator string as the initial axis;

[0038] Unified coordinate system construction: With the insulator string axis as the Z-axis, the center point of one end fitting as the origin O, the horizontal direction perpendicular to the axis as the X-axis, and the vertical direction perpendicular to the axis as the Y-axis, a unified three-dimensional coordinate system O-XYZ is constructed.

[0039] Coordinate transformation mapping: Sensor data mapping is achieved through a homogeneous coordinate transformation matrix. First, the original sensor observation coordinates are converted to world coordinates based on the robot's position and attitude information. Then, translation and rotation operations are performed using the laser geometric reference to map the world coordinates to the O-XYZ coordinate system. The transformation formula is as follows: ,in a target point coordinate in the unified coordinate system, a raw observation coordinate, a rotation matrix corresponding to an attitude angle, a translation vector of the robot relative to the origin, a coordinate of an origin of the unified coordinate system in the world coordinate system.

[0040] Each insulator is assigned a unique number, and its physical position is determined by the axial distance measured by the laser scanner, with an error of within ±2 mm. The normalized data is sliced by insulator unit to form a multi-channel input block with single insulator as the processing granularity.

[0041] Next, a multi-modal feature extraction step is performed. The system calls a multi-modal feature extraction network, which is composed of 4 parallel sub-networks. The first sub-network is a one-dimensional convolutional neural network for processing voltage gradient distribution maps. This sub-network receives the normalized 9-point voltage sequence, and gradually extracts local potential change features through three one-dimensional convolutional layers, finally outputs the voltage drop inflection point position and amplitude decay rate, forming a one-dimensional voltage feature vector with a dimension of 64. The second sub-network is a two-dimensional residual network for analyzing infrared thermal images. This sub-network takes the infrared image block corresponding to a single insulator as input, extracts spatial thermal distribution patterns through residual block stacking, identifies abnormal hot spot areas and their temperature gradient features, and outputs a two-dimensional thermal feature tensor, which is compressed into a 64-dimensional vector after global average pooling. The third sub-network is a time series pulse counting model for counting the density of ultraviolet photon events per unit time. This model performs sliding window counting on the photon arrival time series output by the ultraviolet detector, and combines spatial clustering algorithm to aggregate discrete photon events into discharge source clusters, calculates the center coordinates and intensity weight of each cluster, and finally generates a 32-dimensional corona feature vector. The fourth sub-network is a point cloud semantic segmentation network based on the PointNet architecture, which performs part-level segmentation on the laser scanning point cloud.

[0042] This network receives the point cloud subset corresponding to a single insulator, extracts point-level features through shared multi-layer perceptron, and obtains global shape descriptors through maximum pooling to output structure integrity indicators, including umbrella skirt deformation degree, steel foot exposure ratio and crack coverage rate, forming a 64-dimensional geometric feature vector. The four types of feature vectors together constitute a multi-modal representation set of the insulator unit.

[0043] Subsequently, a cross-modal attention fusion step is performed. The system calls a cross-modal attention fusion module, which adopts a gated cross-attention mechanism. The input is the feature vectors output by the above-mentioned 4 sub-networks, with a total dimension of 224. The module first projects each feature vector to a unified embedding space, with a projected dimension of 64, forming query vectors, key vectors and value vectors. The voltage distribution features are taken as the query vector, and the other three types of features are taken as the key vector and the value vector, respectively, to calculate the attention weight.

[0044] The key and value vectors of the other three types of features are assigned as follows: the thermal distribution feature vector is used as the first key vector and the first value vector, the corona discharge intensity sequence is used as the second key vector and the second value vector, and the geometric structure deviation parameter is used as the third key vector and the third value vector. The query vector (voltage distribution feature projection vector) is used to calculate the correlation weight with each key vector. Then, the weighted value vector is obtained by calculating the sum of the products of each correlation weight and the corresponding value vector, so as to achieve the targeted fusion of the three types of features with the voltage feature.

[0045] Attention weight Calculated using the scaled dot product formula:

[0046]

[0047] in The voltage feature projection vector, For the first Key vectors of non-voltage characteristics, For the first Key vectors of non-voltage characteristics, The embedding dimension is 64. The calculated attention weights reflect the contribution of thermal, optical, and shape features to fault determination under the current voltage anomaly background. Subsequently, each value vector is weighted and summed according to its corresponding weight, and then concatenated with the query vector and nonlinearly fused through a gated linear unit. The final output is a 256-dimensional insulator unit-level health state embedding vector.

[0048] Based on this embedding vector, a comprehensive health status score generation step is performed. The system calls a fully connected regression layer, which consists of two perceptron layers with a hidden layer dimension of 128 and an output layer activation function of sigmoid. This layer maps the health status embedding vector to a single scalar score, with a value ranging from 0 to 100. This score comprehensively reflects the electrical performance, thermal stability, discharge activity, and structural integrity of the insulator. The preset threshold range includes a first threshold of 70 and a second threshold of 90. When the score is below 70, it is judged as a zero-value insulator; when the score is between 70 and 90, it is marked as a suspected fault; when the score is above 90, it is judged as normal.

[0049] After the judgment result is generated, the fault judgment and output steps are executed. The system iterates through the scores of all detected insulator units and classifies them according to the threshold range. For insulators judged to have a zero value, their location number, comprehensive score, and confidence level of the four types of original features are recorded; for suspected faults, a re-inspection flag is set. The judgment result, along with the original sensor segments, feature vectors, and location information, is packaged and prepared for subsequent control decisions.

[0050] According to the determination result, a path correction and re-inspection scheduling step is performed. The central controller has a built-in task state machine, and the states thereof include initialization, cruise detection, local re-inspection, fault confirmation, and return standby. In the cruise detection state, the robot uniformly advances at a speed of 3 meters per minute, and continuously performs the above-mentioned collection and determination process. Once a suspected fault or a zero-value insulator is detected, the state machine immediately enters the local re-inspection state. In this state, the speed of the robot is reduced to 0.5 meters per minute, and a high-precision positioning mode is enabled.

[0051] The re-inspection scheduling strategy specifically includes: controlling the robot to retreat to the position of the previous piece of insulator, and re-performing multi-source data collection; adjusting the integration time of the infrared and ultraviolet sensors, the infrared integration time being extended to 1.5 times the original value, and the ultraviolet integration time being extended to 2 times the original value, so as to improve the signal-to-noise ratio of weak signals; starting the high-density scanning mode of the laser scanner, and reducing the sampling point spacing to 0.5 millimeters, so as to improve the geometric detail resolution. After the re-inspection is completed, feature extraction and score generation are performed again. If the two detection results are consistent, the fault is confirmed, the state machine enters the fault confirmation state, and the final determination result is recorded; otherwise, the re-inspection flag is cleared, the next insulator is continuously detected, and the contradictory data is marked as a sample to be analyzed in the background.

[0052] During the entire detection process, a communication guarantee mechanism is performed. The robot establishes a bidirectional data link with the ground base station through a power line carrier communication module. Key detection data is transmitted in the form of packets after being compressed and encoded, and each packet contains the insulator number, four types of original sensor fragments, feature vectors, and preliminary determination results. Lossless differential encoding and Huffman encoding are combined for compression to ensure that key features are not distorted. After receiving the data, the ground base station performs secondary verification, including data integrity verification and logical consistency check. If abnormalities are found, the base station can remotely issue intervention instructions, such as forced re-inspection, task suspension, or emergency return. The communication link uses frequency hopping spread spectrum technology, has strong anti-electromagnetic interference ability, and is suitable for extra-high voltage alternating current or direct current transmission line environments.

[0053] In actual operation of the method on an extra-high voltage transmission line, the robot cruises along the ground wire at a speed of 3 meters per minute, and it takes about 8 seconds to detect each piece of insulator. Four types of sensors synchronously collect data, and feature extraction and fusion are completed by a local edge computing unit, with a total delay of less than 500 milliseconds. The cross-modal attention mechanism effectively suppresses single sensor misjudgment: for example, in strong sunlight, false hot spots may appear in infrared images, but if the voltage distribution is normal and there is no ultraviolet discharge, the attention weight automatically reduces the contribution of the thermal feature, avoiding false positives; conversely, when the ultraviolet signal is attenuated in rainy and foggy weather, if the voltage gradient drops significantly and the geometric structure is intact, the system can still accurately identify zero-value faults. Experimental data show that in a 5000 kilometer line test, the false positive rate of the method is less than 2%, the false negative rate is less than 0.5%, and the detection speed is more than 5 times higher than that of the traditional manual tower climbing method.

[0054] The mechanical structure of the track-mounted robot ensures stable operation on a 30-degree incline, while the dual-wheel gripping mechanism provides an adhesion force greater than 50 Newtons, capable of withstanding winds up to level 10. The servo motor drive unit employs incremental encoder feedback, combined with a Kalman filter to estimate position, achieving millimeter-level positioning accuracy. The sensor gimbal has two degrees of freedom: pitch and yaw, and can automatically plan the optimal observation angle before detection, ensuring unobstructed imaging of each insulator.

[0055] The distributed capacitively coupled voltage sensing array inverts the voltage distribution by measuring the rate of change of potential difference between probes; its mathematical model is based on the principle of capacitive voltage division. Let the... With the The potential difference between the probes is Then the voltage gradient of the corresponding section on the insulator string is ,in The probe spacing is 10 cm. The voltage gradient distribution spectrum is the sequence. Zero-value insulators exhibit this characteristic in a certain section. The gradient is close to 0, while the gradient of adjacent segments increases abruptly. .

[0056] The point cloud data acquired by the laser contour scanner is processed by denoising, registration, and segmentation before being input into the PointNet network. This network extracts the local geometric features of each point using a shared MLP and then aggregates them into a global descriptor using a symmetric function. The structural integrity index is a weighted composite of three sub-indices: the skirt deformation degree is defined as the absolute value of the relative deviation between the measured outer diameter and the standard value; the exposed length of the steel foot is calculated by the axial extension of the metal portion in the point cloud; and the crack depth is obtained by fitting the abrupt change point of the laser reflection intensity. The three indices are normalized and then weighted and summed, with weights of 0.4, 0.3, and 0.3, respectively.

[0057] The state transitions of the task state machine are driven by both detection results and environmental feedback. For example, in a partial re-inspection state, if the re-inspection result is still suspected, a maximum of two re-inspection attempts are allowed; if the third result is still inconsistent, it is marked as a difficult sample and the process continues. After the fault confirmation state, the robot can choose to continue detection or return immediately, depending on the remaining battery power and task priority. The return path adopts the original route return strategy, with the speed increased to 4 meters per minute to save energy.

[0058] In summary, the insulator zero-value detection robot control method of the present invention achieves high-precision and high-robustness identification of zero-value insulators under uninterrupted power supply and non-contact conditions through multi-source heterogeneous sensor fusion, cross-modal attention mechanism and closed-loop re-inspection scheduling, which significantly improves the safety and efficiency of power grid operation and maintenance.

Claims

1. An insulator zero value detection robot control method, characterized by, The application relates to a method for detecting zero-value insulators on a power transmission line. The method comprises the following steps: deploying a track robot with autonomous walking capability on a ground wire or a conductor of a power transmission line; synchronously collecting multi-source sensing data: synchronously collecting multi-dimensional physical field information of an insulator string through an infrared thermal imaging sensor, an ultraviolet corona detector, a distributed capacitive coupled voltage sensing array and a laser profile scanner; based on a preset time-space synchronization reference, performing timestamp alignment and space coordinate normalization processing on the multi-source sensing data; the space coordinate normalization is realized through cooperation of robot IMU data and laser geometric reference, a unified three-dimensional coordinate system taking an insulator string axis as a reference is constructed, and then the observation results of various sensors are mapped to the coordinate system through homogeneous coordinate transformation; a multi-modal feature extraction network is used to analyze original signals of various sensing channels respectively, to generate a thermal distribution feature vector, a corona discharge intensity sequence, a voltage gradient distribution atlas and a geometric structure deviation parameter; the thermal distribution feature vector, the corona discharge intensity sequence, the voltage gradient distribution atlas and the geometric structure deviation parameter are input into a cross-modal attention fusion module, confidence weights of various feature channels at an insulator unit level are calculated, and a comprehensive health state score is synthesized by weighting; according to comparison between the comprehensive health state score and a preset threshold interval, it is determined whether there is a zero-value insulator, and a fault position number and a confidence level are output; according to the determination result, a path correction instruction and a rechecking scheduling strategy are generated by a central controller, to drive the robot to perform local backtracking scanning or adjacent insulator cross verification operation; the multi-modal feature extraction network comprises four parallel sub-networks: a first sub-network is a one-dimensional convolutional neural network, used for processing the voltage gradient distribution atlas, extracting a voltage drop inflection point position and an amplitude decay rate; a second sub-network is a two-dimensional residual network, used for analyzing an infrared thermal image, identifying an abnormal hot spot area and a temperature gradient feature thereof; a third sub-network is a time series pulse counting model, used for counting an ultraviolet photon event density per unit time, and positioning a discharge source in combination with a space clustering algorithm; and a fourth sub-network is a point cloud semantic segmentation network, based on a PointNet architecture, used for performing component-level segmentation on laser scanning point clouds, and outputting a structure integrity index; the cross-modal attention fusion module adopts a gated cross-attention mechanism, and the input is feature tensors output by the four sub-networks; the cross-modal attention fusion module firstly projects the feature tensors to a unified embedding space linearly, then takes the voltage distribution feature as a query vector, and takes the other three types of features as key vectors and value vectors respectively, calculates the correlation weight of the query vector and each key vector, and realizes fusion by weighted summation of each value vector to calculate the attention weight; finally, the output is an insulator unit level health state embedding vector after weighted fusion, and the dimension is 256.

2. The insulator null detection robot control method of claim 1, wherein The track robot is provided with a double-wheel clamping driving mechanism, which comprises an upper pressing wheel assembly, a lower load wheel assembly and a servo motor driving unit; the upper pressing wheel assembly applies constant pre-tightening force through a spring loading mechanism, ensuring that the robot remains stable attachment on different inclination lines; the servo motor driving unit adopts a closed-loop vector control mode, and adjusts the speed difference between the left and right wheels in real time to realize accurate displacement control along the axis of the guide wire, with a positioning error of not more than ±5mm.

3. The insulator null detection robot control method of claim 2, wherein, The infrared thermal imaging sensor is configured as a long-wave infrared focal plane array, with a working waveband of 8-14 microns, a spatial resolution of 640×480 pixels, and a frame rate of not less than 25 Hz; the sensor is installed on the front end of the robot, and automatic focusing and field coverage of the target insulator string are realized through a two-dimensional servo mechanism, ensuring that the umbrella skirt area of each insulator is completely imaged.

4. The insulator null detection robot control method of claim 3, wherein, The ultraviolet corona detector adopts a solar blind ultraviolet photomultiplier tube, with a response wavelength range of 240-280 nm and a minimum detectable photon flux of 50 photons per second; the detector is installed on the same optical axis as the visible light auxiliary camera, and the ultraviolet discharge signal is superimposed into the visible light image through an image registration algorithm to form a corona spatial distribution heat map, which is realized by camera joint calibration, ultraviolet signal quantization, affine transformation registration and semi-transparent color superposition, i.e., first calibrate the internal and external parameters of the two cameras to establish coordinate mapping, then quantize the ultraviolet photon signal into a grayscale image and map it to the visible light coordinate system, and finally convert it into a heat color mode for superposition.

5. The insulator null detection robot control method of claim 4, wherein, The distributed capacitance coupling voltage sensing array is composed of 9 micro-capacitance probes arranged along the lateral direction of the robot, with a probe spacing of 10 cm and the probe sensing surface facing downward toward the insulator string; each probe is connected to a 12-bit analog-to-digital converter through a high-input-impedance buffer amplifier, with a sampling frequency of 1000 Hz; the voltage distribution curve of the insulator string along the axial direction is obtained by measuring the rate of change of the potential difference between adjacent probes.

6. The insulator null detection robot control method of claim 5, wherein, The laser profile scanner adopts a triangulation ranging principle, emits a semiconductor laser beam with a wavelength of 650 nm, and has a scanning frequency of 2000 Hz and a ranging accuracy of ±0.1 mm; the scanner performs continuous line scanning along the axial direction of the insulator string to obtain three-dimensional point cloud data of the outer diameter of the umbrella skirt, the exposed length of the steel foot and the surface crack depth of each insulator.

7. The insulator null detection robot control method of claim 6, wherein, The comprehensive health status score is obtained by mapping the health status embedding vector through a fully connected regression layer, with a value range of 0-100; the preset threshold interval includes a first threshold of 70 and a second threshold of 90; when the score is lower than 70, it is determined as a zero-value insulator; when the score is between 70 and 90, it is marked as a suspected fault, triggering a re-inspection process; when the score is higher than 90, it is determined as normal.

8. The insulator null detection robot control method of claim 7, wherein, The re-inspection scheduling strategy includes: controlling the robot to retreat to the position of the previous insulator and re-executing multi-source data acquisition; adjusting the integration time of the infrared and ultraviolet sensors to improve the signal-to-noise ratio of weak signals; starting the high-density scanning mode of the laser scanner, with a sampling point spacing reduced to 0.5 mm; if the results of 2 detections are consistent, the fault is confirmed; otherwise, the next insulator is detected forward and the contradictory data are recorded for background analysis.

Images