A Substation Defect Location Method Based on Robot Matrix

Abstract

This invention discloses a substation defect localization method based on a robot matrix, comprising: creating a three-dimensional digital twin model of the substation, deploying a robot matrix, and generating a three-dimensional point cloud map in real time; acquiring field signals, recording the arrival time and waveform data of partial discharge signals at the robot matrix based on a full-matrix synchronous acquisition mode; acquiring the acquired data and calculating the three-dimensional coordinates of the partial discharge source, the acquired data including: the real-time spatial coordinates of the detection robots and the pulse arrival time; locating the field position based on the three-dimensional coordinates of the partial discharge source, controlling the robot matrix to perform optical scanning, and determining whether it is a deterministic defect; inputting the three-dimensional coordinates of the partial discharge source into the three-dimensional point cloud map, and highlighting the faulty equipment location and defect type in the three-dimensional digital twin model of the substation. This invention, by constructing a long-baseline robot matrix using multiple detection robots, can achieve blind-spot-free detection throughout the substation and achieve sub-meter-level accurate localization of partial discharge sources.

Description

Technical Field

[0001] This invention relates to the field of power system automation and intelligent robot technology, and in particular to a substation defect location method based on a robot matrix. Background Technology

[0002] As a critical infrastructure in the power system, substations operate under high voltage, high current, and complex electromagnetic environments for extended periods. Issues such as insulation aging, structural defects, and manufacturing or installation deviations can easily trigger partial discharge during operation. If partial discharge is not detected and accurately located in a timely manner, it often gradually evolves into insulation breakdown, equipment failure, or even power outages.

[0003] With the development of intelligent inspection technology, wheeled or tracked inspection robots are gradually being applied to substation inspection operations. However, existing inspection robots typically operate as standalone units, mainly used for image acquisition, infrared temperature measurement, or simple environmental perception. They have limited mobility in complex terrain and lack the ability to work collaboratively with multiple robots. Even when equipped with partial discharge detection devices, they are mostly deployed at single points or in small areas, failing to form large-scale, multi-angle detection arrays in space, thus limiting the improvement of positioning accuracy.

[0004] On the other hand, with the maturation of humanoid robot and multi-legged robot technologies, they have shown significant advantages in adaptability to complex terrain, flexible mobility, and multi-degree-of-freedom operation capabilities, providing a new technological foundation for multi-point collaborative detection in complex scenarios such as substations. However, existing technologies lack a systematic solution for organizing multiple humanoid robots into a collaborative detection matrix and combining high-precision time synchronization, multimodal sensing, and 3D spatial modeling technologies to achieve precise defect localization in substations. Summary of the Invention

[0005] The purpose of this invention is to provide a substation defect location method and system based on a robot matrix, which solves the problems of limited mobility and lack of collaborative work ability of inspection robots in complex terrain.

[0006] Technical solution

[0007] A substation defect localization method based on robot matrix includes the following steps:

[0008] Step 1: Create a 3D digital twin model of the substation, deploy a robot matrix, and generate a 3D point cloud map in real time. The robot matrix includes at least three detection robots, which are humanoid or multi-legged structures.

[0009] Step 2: Acquire field signals. Based on the full matrix synchronous acquisition mode, record the time and waveform data of the partial discharge signal arriving at the robot matrix.

[0010] Step 3: Acquire the collected data. Based on the long baseline TDOA positioning method, calculate the three-dimensional coordinates of the local discharge source. The collected data includes: the real-time spatial coordinates of the detection robot and the pulse arrival time.

[0011] Step 4: Based on the three-dimensional coordinates of the local discharge source, locate the on-site position, control the robot matrix to perform optical scanning, and determine whether it is a deterministic defect;

[0012] Step 5: Input the three-dimensional coordinates of the local discharge source into the three-dimensional point cloud map, and highlight the location of the faulty equipment and the type of defect in the three-dimensional digital twin model of the substation.

[0013] Preferably, the long baseline is set such that the straight-line distance between any two of the detection robots is greater than 10 meters.

[0014] Preferably, the detection robot includes: a wireless communication unit, a radio frequency acquisition unit, an infrared thermal imager, an ultraviolet imager, and a lidar, and the detection robots are connected and synchronized with each other through the wireless communication unit.

[0015] Preferably, the full-matrix synchronous acquisition mode is triggered when any of the detection robots detects a partial discharge pulse signal exceeding the acquisition threshold.

[0016] Preferably, in step 4, the condition for determining a deterministic defect is that an anomaly is detected in the optical image within the same coordinate region.

[0017] Preferably, the detection robot is equipped with a multi-degree-of-freedom robotic arm and an end effector, and the radio frequency acquisition unit is located at the end of the robotic arm.

[0018] Preferably, the robot matrix adopts a combined GPS positioning and UWB timing positioning method, using GPS positioning in outdoor areas and automatically switching to UWB positioning in areas where satellite signals are blocked.

[0019] The positioning error is less than 10cm, and the time synchronization error is less than 1ns.

[0020] Preferably, when the signal-to-noise ratio of the detected partial discharge signal is lower than a threshold, the robot matrix automatically calculates the optimal observation position and schedules the detection robot to move to the new coordinates.

[0021] Preferably, in step S1, the robot matrix collaborative SLAM generates the 3D point cloud map, and the specific steps are as follows:

[0022] S1.1, The detection robot independently performs local real-time SLAM, generates a local point cloud map and estimates its own pose;

[0023] S1.2, the detection robot uploads the local map and pose information to the central holographic analysis platform via wireless network;

[0024] S1.3, the central holographic analysis platform adopts a graph-optimized multi-robot map fusion algorithm to align and merge the local maps into a globally consistent 3D point cloud map;

[0025] S1.4, Combining visual recognition with a pre-set equipment model library, perform semantic segmentation on the point cloud, identify and label key equipment in the substation, and form a 3D map with semantic labels;

[0026] S1.5 continuously monitors environmental changes, compares real-time scan data with existing maps, detects newly added or removed objects, dynamically updates the 3D point cloud map, and marks temporary dynamic obstacles as temporary layers.

[0027] Preferably, the 3D point cloud map is further converted into a grid model or voxel map, and associated with the extracted equipment topology relationship, and output as a substation holographic model.

[0028] Beneficial effects

[0029] This invention constructs a long-baseline robot matrix using multiple inspection robots, possessing strong obstacle-crossing and terrain adaptability capabilities. It enables blind-spot-free detection across the entire substation area, achieving sub-meter-level precise positioning of local discharge sources, effectively distinguishing defects between adjacent devices, and significantly improving TDOA positioning accuracy. By integrating data from multiple sources such as radio frequency, infrared, ultraviolet, and acoustic sensors, it achieves comprehensive detection of internal and external defects, improving the accuracy and reliability of defect identification. The robot's position is optimized in real time based on signal quality and environmental changes, automatically adjusting the detection baseline length and array geometry to adapt to complex electromagnetic environments and variable inspection scenarios. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of the method flow of the present invention; Detailed Implementation

[0031] The technical solution of the present invention will be further described below with reference to the embodiments and accompanying drawings.

[0032] Collaborative SLAM, Simultaneous Localization and Mapping.

[0033] A substation defect localization method based on robot matrix includes the following steps:

[0034] Step 1: Create a 3D digital twin model of the substation, deploy a robot matrix, and generate a 3D point cloud map in real time. The robot matrix includes at least three inspection robots. These robots are humanoid or multi-legged and possess the ability to autonomously navigate and overcome obstacles in the complex terrain of the substation. Each inspection robot is equipped with a LiDAR, a vision camera, an inertial measurement unit, and a wireless communication unit.

[0035] Step 2: Acquire field signals. Based on the full matrix synchronous acquisition mode, record the time and waveform data of the partial discharge signal arriving at the robot matrix.

[0036] Step 3: Acquire the collected data. Based on the long baseline TDOA positioning method, calculate the three-dimensional coordinates of the local discharge source. The collected data includes: the real-time spatial coordinates of the detection robot and the pulse arrival time.

[0037] Step 4: Based on the three-dimensional coordinates of the local discharge source, locate the on-site position, control the robot matrix to perform optical scanning, and determine whether it is a deterministic defect;

[0038] Step 5: Input the three-dimensional coordinates of the local discharge source into the three-dimensional point cloud map of the central holographic analysis platform, and highlight the location of the faulty equipment and the type of defect in the three-dimensional digital twin model of the substation.

[0039] In a further implementation of this embodiment, the long baseline is set such that the straight-line distance between any two detection robots is greater than 10 meters. The robot matrix has been dispersed in step 1. This large-aperture array can effectively reduce the error amplification effect caused by the geometric accuracy factor. The specific positioning method is as follows:

[0040] Construct a system of TDOA equations with the speed of light c as a constant, c⋅(t i -t j )= Solving this overdetermined system of equations using Chan's algorithm, Taylor series expansion, etc., yields the three-dimensional coordinates of the local discharge source in the global coordinate system. , , ).

[0041] In a further embodiment of this invention, the detection robot includes: a wireless communication unit, a radio frequency acquisition unit, an infrared thermal imager, an ultraviolet imager, and a lidar. The detection robots are connected and synchronized with each other through the wireless communication unit.

[0042] In a further implementation of this embodiment, within a preset time window, the full-matrix synchronous acquisition mode ensures that at least two spatially separated detection robots simultaneously detect signal pulses exceeding their respective dynamic adaptive thresholds, and the time-domain waveforms or frequency-domain characteristics of the pulses possess a preset correlation. The acquisition threshold is a dynamic adaptive threshold value. In a substation environment with strong electromagnetic interference, this ensures reliable capture of real partial discharge pulses while maximally suppressing noise-induced false triggering. The definition method is as follows:

[0043] During the baseline noise learning phase, an electromagnetic background noise baseline is established in the current environment. Periodically, during periods without known discharge activity, all detection robots are controlled to perform a 60-second synchronous background monitoring. The central holographic analysis platform collects the monitoring data from each detection robot, calculates the statistical characteristics of the electromagnetic signal amplitude received by the entire matrix, and determines the mean μ and standard deviation σ of the noise amplitude.

[0044] Threshold V th Defined as: V th (t)=μ noise +k⋅σ noise +ΔV adaptive , where μ noise and σ noise This is a real-time estimate of the current ambient noise; k is a configurable sensitivity coefficient, ranging from 3 to 5, representing the margin of the trigger threshold relative to the noise fluctuation range, which is remotely adjusted according to the detection task; ΔV adaptive For adaptive adjustment items, specifically;

[0045] A true partial discharge pulse has specific spectral characteristics. The system performs a real-time Fast Fourier Transform on the signal. If the detected signal peak is high but the main energy distribution is in the power frequency harmonics or communication frequency band, then ΔV is increased. adaptive This makes the acquisition threshold less likely to be triggered, thus suppressing switching operations or communication interference.

[0046] If only one robot detects an "exceeding the limit" signal, while neighboring robots do not detect the relevant signal, it may be due to local interference. In this case, the trigger threshold for that area should be temporarily increased.

[0047] For equipment with a high incidence of partial discharge, a lower baseline threshold should be set, i.e., the k value or ΔV should be reduced. adaptive For equipment areas that have never had problems before, the threshold can be appropriately increased;

[0048] In previous detection cycles, if localization was triggered but multimodal verification ultimately failed, the system would record the "features" of such signals as suspected interference. Subsequent encounters of signals with similar characteristics would trigger a fine-tuning of ΔV. adaptive This makes it less likely to be triggered.

[0049] In this embodiment, the threshold is no longer a preset fixed value, but can sense changes in environmental noise and adjust automatically, ensuring the stability of the system sensitivity in different substations, at different times, and under different electromagnetic environments; the detection robot can make comprehensive decisions by combining multi-dimensional information such as amplitude, spectrum, spatial distribution, and equipment context, which greatly reduces the false alarm rate and makes the system triggering smarter; by using historical verification results to optimize the threshold strategy, the system has the ability to learn and continuously improve itself.

[0050] In a further implementation of this embodiment, in step 4, the condition for determining a deterministic defect is that an anomaly is detected in the optical image within the same coordinate region.

[0051] In a further embodiment of this invention, the detection robot is equipped with a multi-degree-of-freedom robotic arm and an end effector. The radio frequency acquisition unit is located at the end of the robotic arm. When performing precise positioning, the detection robot adjusts the posture of the robotic arm to change the height and polarization direction of the antenna, thereby constructing a three-dimensional detection baseline in the vertical dimension.

[0052] In a further implementation of this embodiment, the robot matrix adopts a combined GPS positioning and UWB timing positioning method. GPS positioning is used in outdoor areas, and UWB positioning is automatically switched in areas where satellite signals are blocked.

[0053] The positioning error is less than 10cm, and the time synchronization error is less than 1ns.

[0054] In a further implementation of this embodiment, when the signal-to-noise ratio of the detected partial discharge signal is lower than a threshold, the robot matrix automatically calculates the optimal observation position and schedules the detection robot to move to the new coordinates, in order to increase the detection baseline length or optimize the array geometry.

[0055] In a further implementation of this embodiment, step S1 employs collaborative SLAM technology, which, through the lidar, vision sensor, and inertial measurement unit mounted on the detection robot, achieves real-time, adaptive 3D reconstruction of the substation environment. The specific steps are as follows:

[0056] S1.1, The inspection robot independently performs local real-time SLAM using LiDAR, vision camera and inertial measurement unit to generate local point cloud map and estimate its own pose;

[0057] S1.2, the detection robot uploads the local map and pose information to the central holographic analysis platform via a 5G private network or a high-speed self-organizing network;

[0058] S1.3 The central holographic analysis platform adopts a multi-robot map fusion algorithm based on graph optimization to align and merge the local maps into a globally consistent 3D point cloud map, forming a globally consistent, non-repeating, high-precision 3D point cloud map.

[0059] LiDAR point clouds provide accurate geometric structures, while visible light cameras provide texture and color information. The fusion of these two technologies generates a 3D semantic map with device category labels, which facilitates the subsequent association of defect locations with specific devices.

[0060] S1.4, combining visual recognition with a pre-set equipment model library, calls a deep learning-based visual recognition model to perform semantic segmentation on the point cloud, identify and label key equipment in the substation, such as transformers, circuit breakers, insulators, etc., to form a 3D map with semantic labels.

[0061] S1.5 continuously monitors environmental changes, compares real-time scan data with existing maps, detects newly added or removed objects, dynamically updates the 3D point cloud map, marks temporary dynamic obstacles as temporary layers and excludes them from the persistent map, ensuring the stability and reliability of the map.

[0062] In a further implementation of this embodiment, the 3D point cloud map is further converted into a mesh model or voxel map, and associated with the extracted equipment topology relationship, and output as a substation holographic model.

[0063] In a further implementation of this embodiment, the central holographic analysis platform is responsible for coordinating the robot matrix, processing massive heterogeneous data, executing core algorithms, and generating intuitive visualization results. It integrates edge-cloud collaborative computing, multimodal data fusion, intelligent analysis and decision-making, and real-time 3D rendering into a comprehensive software and hardware platform, including:

[0064] The data fusion and spatiotemporal alignment module receives multi-source asynchronous data streams from all detection robots and performs standardization, timestamp alignment, and spatial coordinate system unification; it adds a unified nanosecond-level time tag to all access data and transforms it to the global coordinate system; it performs preliminary cleaning of the raw data, such as bandpass filtering of radio frequency signals, denoising and enhancement of images, downsampling of point clouds, and outlier removal.

[0065] The collaborative perception scheduling and array optimization module dynamically allocates tasks, formulates collaborative perception strategies, and performs real-time geometric optimization of the detection array. Based on the inspection plan or manual instructions, it assigns initial positions, movement paths, and perception tasks to each robot, calculates the geometric accuracy factor of the current robot array and the signal quality of each node in real time, and automatically triggers the array reconstruction process when the signal-to-noise ratio is detected to be lower than the threshold or the GDOP deteriorates.

[0066] The intelligent defect calculation and localization module is used to extract features from fused data, calculate defect locations, and perform comprehensive diagnosis. It implements a dynamic threshold triggering mechanism to perform cross-robot correlation analysis on pulses exceeding the threshold to confirm whether they are the same real discharge event. It uses the TDOA algorithm, combined with the robot's precise coordinates, to calculate the two-dimensional / three-dimensional spatial coordinates of the discharge signal. The TDOA localization results are then fused with data from infrared, ultraviolet, visible light, and ultrasound at the pixel-level / feature-level in time and space.

[0067] The 3D holographic twin and visualization module constructs and drives a "digital twin" synchronized with the physical substation, enabling holographic and immersive display of monitoring results.

[0068] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; 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; and these 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.

Claims

1. A substation defect location method based on a robot matrix, characterized in that, Includes the following steps: Step 1: Create a 3D digital twin model of the substation, deploy a robot matrix, and generate a 3D point cloud map in real time. The robot matrix includes at least three detection robots, which are humanoid or multi-legged structures. Step 2: Acquire field signals. Based on the full matrix synchronous acquisition mode, record the time and waveform data of the partial discharge signal arriving at the robot matrix. Step 3: Acquire the collected data. Based on the long baseline TDOA positioning method, calculate the three-dimensional coordinates of the local discharge source. The collected data includes: the real-time spatial coordinates of the detection robot and the pulse arrival time. Step 4: Based on the three-dimensional coordinates of the local discharge source, locate the on-site position, control the robot matrix to perform optical scanning, and determine whether it is a deterministic defect; Step 5: Input the three-dimensional coordinates of the local discharge source into the three-dimensional point cloud map, and highlight the location of the faulty equipment and the type of defect in the three-dimensional digital twin model of the substation.

2. The substation defect location method based on robot matrix according to claim 1, characterized in that, The long baseline is set such that the straight-line distance between any two of the detection robots is greater than 10 meters.

3. The substation defect location method based on robot matrix according to claim 1, characterized in that, The detection robot includes a wireless communication unit, a radio frequency acquisition unit, an infrared thermal imager, an ultraviolet imager, and a lidar. The detection robots are connected and synchronized with each other through the wireless communication unit.

4. The substation defect location method based on robot matrix according to claim 1, characterized in that, The full-matrix synchronous acquisition mode is triggered when any of the detection robots detects a partial discharge pulse signal exceeding the acquisition threshold, whereby the acquisition threshold is a dynamically adaptive threshold value.

5. The substation defect location method based on robot matrix according to claim 1, characterized in that, In step 4, the condition for determining a deterministic defect is that an anomaly is detected in the optical image within the same coordinate region.

6. The substation defect location method based on robot matrix according to claim 3, characterized in that, The detection robot is equipped with a multi-degree-of-freedom robotic arm and an end effector, and the radio frequency acquisition unit is located at the end of the robotic arm.

7. The substation defect location method based on robot matrix according to claim 1, characterized in that, The robot matrix adopts a combined GPS positioning and UWB timing positioning method. Specifically, GPS positioning is used in outdoor areas, and UWB positioning is automatically switched in areas where satellite signals are blocked. The positioning error is less than 10cm, and the time synchronization error is less than 1ns.

8. The substation defect location method based on robot matrix according to claim 1, characterized in that, When the signal-to-noise ratio of the detected partial discharge signal is lower than the critical value, the robot matrix automatically calculates the optimal observation position and schedules the detection robot to move to the new coordinates.

9. The substation defect location method based on robot matrix according to claim 1, characterized in that, In step S1, the robot matrix collaborative SLAM generates the 3D point cloud map. The specific steps are as follows: S1.1, the detection robot independently performs local real-time SLAM, generates a local point cloud map and estimates its own pose; S1.2, the detection robot uploads the local map and pose information via a wireless network; S1.3, the central holographic analysis platform adopts a graph-optimized multi-robot map fusion algorithm to align and merge the local maps into a globally consistent 3D point cloud map; S1.4, Combining visual recognition with a pre-set equipment model library, perform semantic segmentation on the point cloud, identify and label key equipment in the substation, and form a 3D map with semantic labels; S1.5 continuously monitors environmental changes, compares real-time scan data with existing maps, detects newly added or removed objects, dynamically updates the 3D point cloud map, and marks temporary dynamic obstacles as temporary layers.

10. The substation defect location method based on robot matrix according to claim 9, characterized in that, The 3D point cloud map is further converted into a mesh model or voxel map, and associated with the extracted equipment topology, outputting a holographic model of the substation.

Images