Method and device for detecting working state of three-phase cable

Abstract

The invention relates to the technical field of power detection, and discloses a three-phase cable working state detection method and device, and the method carries out the remote positioning and control based on a real-time position image fed back by an image collection device, prevents an operator from being in close contact with high-voltage equipment or carrying out the climbing operation, and improves the working efficiency. Double safety risks of electric shock and falling are effectively eliminated; a pre-trained positioning recognition model is used for assisting in judging the closing time, position information is presented in combination with a display device, dependence on manual judgment and recording is reduced, the automation degree of the detection process is improved, and personal errors are reduced; besides, three-phase current, zero-sequence current and total current can be synchronously acquired through an integrated pincerlike mutual inductor assembly, comprehensive identification of various faults is realized and corresponding early warning information is generated in combination with multi-dimensional anomaly judgment logic, so that the limitation that traditional equipment only has a basic data acquisition function is broken through, and the fault diagnosis accuracy is improved. And the deep monitoring of the working state of the cable and the early warning of potential faults are realized.

Description

Technical Field

[0001] This invention relates to the field of power testing technology, and in particular to a method and device for detecting the working status of three-phase cables. Background Technology

[0002] In power systems, the load status and leakage of three-phase cables directly affect the safety and stability of power grid operation, and are one of the core inspection contents in the field of power operation and maintenance.

[0003] Currently, the mainstream testing methods in the industry include single-phase clamp ammeters operated by workers holding insulated rods, or traditional three-phase testing methods that involve manually clamping the ammeters using climbing equipment. The operation modes are mostly based on direct contact or simple extension of the insulated rod. Data reading requires manual visual inspection of the instrument panel, and there is a lack of automated analysis and recording mechanisms.

[0004] Traditional detection techniques have many limitations that are difficult to overcome: First, workers need to be in close contact with high-voltage equipment or work at heights, inevitably facing the dual safety risks of electric shock and falls, resulting in weak safety assurance capabilities. Secondly, the detection equipment only has basic data acquisition functions and lacks the ability to deeply analyze key parameters such as current load trends and abnormal waveforms, making it unable to provide early warnings of potential faults such as insulation degradation. Furthermore, the measurement results rely entirely on manual recording and organization, making it difficult to automatically generate standardized electronic reports. Moreover, it is impossible to achieve encrypted data transmission and linkage with the back-end system, which easily leads to data silos and seriously hinders the digital transformation process of power operation and maintenance.

[0005] It is evident that existing technologies still need improvement and enhancement. Summary of the Invention

[0006] In view of the shortcomings of the prior art, the purpose of this invention is to provide a method for detecting the working status of three-phase cables. It constructs a full-process detection system with visual positioning, intelligent judgment, precise clamping, multi-dimensional analysis, and hierarchical early warning, which changes the traditional crude mode of manual visual positioning, manual clamping, and simple reading, and realizes the intelligence and automation of the detection process.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: A method for detecting the operating status of a three-phase cable, comprising a telescopic rod, a clamp-on current transformer assembly mounted on the top of the telescopic rod, and an image acquisition device, wherein the telescopic rod is equipped with a control device and a display device, and the image acquisition device, the display device, and the clamp-on current transformer assembly are electrically connected to the control device; the method includes: acquiring a real-time position image of the clamp-on current transformer assembly fed back by the image acquisition device; generating position display information based on the real-time position image and sending it to the display device; inputting the real-time position image into a pre-trained positioning recognition model to determine whether to generate a closure prompt message; controlling the clamp head of the clamp-on current transformer assembly to close when a closure command is received; acquiring real-time current information fed back by the clamp-on current transformer assembly, the real-time current information including real-time three-phase current, real-time zero-sequence current, and real-time total current; determining whether there is an abnormal operating condition based on the real-time current information, and if so, determining the corresponding abnormal operating information; acquiring a preset warning rule, and generating a warning message based on the warning rule and the abnormal operating information.

[0008] In the three-phase cable operating status detection method, the step of inputting the real-time position image into a pre-trained positioning and recognition model to determine whether to generate a closure prompt message includes: scaling and normalizing the real-time position image to obtain a pre-processed position image; inputting the pre-processed position image into the pre-trained positioning and recognition model to obtain the bounding box coordinates of the clamp head, the bounding box coordinates of the cable, and the category confidence score, wherein the positioning and recognition model is an optimized MobileNetSSD model; calculating the relative position deviation between the clamp head and the cable based on the bounding box coordinates of the clamp head and the cable, wherein the relative position deviation includes horizontal deviation, vertical deviation, and distance deviation; obtaining a preset deviation threshold; and determining whether to generate a closure prompt message based on the relative position deviation and the preset deviation threshold.

[0009] In the three-phase cable working status detection method, before inputting the real-time location image into the pre-trained positioning and recognition model, the method further includes: constructing a positioning and recognition model to be trained, wherein the positioning and recognition model to be trained includes a feature extraction module, a detection head, and a post-processing module connected in sequence, wherein the feature extraction module adopts a MobileNet V2 network, and the detection head adopts an SSD architecture; acquiring multiple sets of images of clamps and cables under different lighting conditions, different shooting distances, and different cable specifications to construct an initial sample set; performing annotation processing, data augmentation processing, unified scaling processing, and partitioning processing on the initial sample set in sequence to obtain a training set and a test set; using the TensorFlow framework and the SGD optimizer, with the goal of minimizing the loss function, iteratively training the positioning and recognition model to be trained using the training set and the test set; outputting the trained positioning and recognition model when the preset iteration stopping condition is met; and removing redundant parameters in the trained positioning and recognition model using pruning techniques to obtain the positioning and recognition model.

[0010] In the three-phase cable working status detection method, after controlling the clamp head of the clamp-type current transformer assembly to close when a closing command is received, the method further includes: using an image acquisition device to acquire a real-time closing image of the clamp-type sensor assembly after the closing command is executed; inputting the real-time closing image into the positioning recognition model to determine whether the clamp head is clamped in place; if the clamp head is clamped in place, generating clamping in place display information and controlling the clamp-type sensor assembly to lock the current state.

[0011] In the aforementioned three-phase cable operating status detection method, the step of determining whether an abnormal operating condition exists based on the real-time current information, and if so, determining the corresponding abnormal operating information, includes: calculating the three-phase unbalance based on the real-time current information, comparing the three-phase unbalance with a preset unbalance threshold to determine whether an abnormal operating condition exists; extracting high-frequency harmonic features from the real-time current information using wavelet transform, and analyzing the harmonic frequencies of the high-frequency harmonic features using Fourier transform to calculate the effective values ​​of detail coefficients, comparing the effective values ​​with a preset standard deviation threshold to determine whether an abnormal operating condition exists; inputting the real-time current information into a pre-trained time-series prediction model to obtain a future load change curve, and determining whether an abnormal operating condition exists based on the future load change curve; if an abnormal condition exists, determining the corresponding abnormal operating information, which includes one or more of the following: duration of three-phase unbalance, abnormal waveform, and load trend.

[0012] In the aforementioned three-phase cable operating status detection method, the step of acquiring preset early warning rules and generating early warning information based on the early warning rules and the abnormal operating information includes: if the duration of the three-phase imbalance is ≥ a preset duration threshold, there are no abnormal waveforms, and the load trend is normal, then a first-level early warning information is generated; if an abnormal waveform is detected, the duration of the three-phase imbalance is < a preset duration threshold, and the load trend is normal, then a second-level early warning information is generated; if both an abnormal waveform and the duration of the three-phase imbalance are ≥ a preset duration threshold, or the load trend is abnormal, then a third-level early warning information is generated.

[0013] In the three-phase cable operating status detection method, after generating early warning information based on the early warning rules and the abnormal operating information, the method further includes: summarizing the data to be transmitted, which includes measurement data, status data, and control commands; classifying the data to be transmitted to distinguish between key control commands and batch data; encrypting the key control commands using the SM2 elliptic curve asymmetric encryption algorithm, and then digitally signing the encrypted key control commands using a preset private key to obtain encrypted commands with signature values; encrypting the batch data in blocks using the SM4 block cipher algorithm to generate encrypted data blocks and corresponding data verification codes; and generating wireless communication information based on the encrypted commands with signature values, encrypted data blocks, and corresponding data verification codes.

[0014] The present invention also provides a three-phase cable operating status detection device, wherein the three-phase cable operating status detection device uses any of the three-phase cable operating status detection methods described above to achieve operating control; the three-phase cable operating status detection device includes a telescopic rod, a clamp-on current transformer assembly disposed on the top of the telescopic rod, and an image acquisition device; the telescopic rod is provided with a control device and a display device; the image acquisition device, the display device, and the clamp-on current transformer assembly are electrically connected to the control device; the clamp-on current transformer assembly includes five synchronously opening and closing clamp-on current transformers, and the five clamp-on current transformers are respectively used to detect the three-phase current, zero-sequence current, and total current.

[0015] In the aforementioned three-phase cable operating status detection device, an operation panel is provided at the bottom of the telescopic rod, and the display device is embedded in the operation panel; the operation panel is also provided with a current transformer position adjustment rocker and a telescopic rod control rocker, both of which are electrically connected to the control device; a telescopic rod drive motor is provided in the middle of the telescopic rod, and the drive motor is electrically connected to the control device, the control device adjusting the operating state of the telescopic rod drive motor based on the control information of the telescopic rod control rocker; a current transformer position control servo is provided at the top of the telescopic rod, the servo is electrically connected to the control device and drively connected to the clamp-type current transformer assembly, the control device adjusting the operating state of the current transformer position control servo based on the control information of the current transformer position adjustment rocker.

[0016] In the aforementioned three-phase cable working status detection device, the control device includes a controller and a power management module electrically connected to the controller. The controller integrates an artificial intelligence processing chip, a national cryptographic security chip, a wireless communication module, and a WiFi module.

[0017] Beneficial effects: This invention provides a method for detecting the working status of three-phase cables. By extending the working distance with a telescopic pole and combining real-time position images from an image acquisition device, remote positioning and control are achieved, avoiding close contact with high-voltage equipment or working at heights for operators, effectively eliminating the dual safety risks of electric shock and falls. A pre-trained positioning recognition model assists in determining the closing timing, and a display device presents the position information, reducing reliance on manual judgment and recording, improving the automation level of the detection process, and reducing human error. Furthermore, through an integrated clamp-on current transformer assembly, three-phase current, zero-sequence current, and total current can be simultaneously collected. Combined with multi-dimensional anomaly judgment logic, comprehensive identification of various faults such as cable load imbalance, partial insulation discharge, and long-term heavy load is achieved, generating corresponding early warning information. This overcomes the limitations of traditional equipment that only has basic data acquisition functions, enabling in-depth monitoring of cable working status and early warning of potential faults. Attached Figure Description

[0018] Figure 1 A logic flowchart of the three-phase cable operating status detection method provided by the present invention; Figure 2 This is a schematic diagram of the structure of the three-phase cable working status detection device provided by the present invention.

[0019] Explanation of main component symbols: 1-Telescopic rod, 2-Clamp-type current transformer assembly, 3-Image acquisition device, 4-Operation panel, 5-Telescopic rod drive motor, 6-Control device. Detailed Implementation

[0020] This invention provides a method and device for detecting the working status of three-phase cables. To make the objectives, technical solutions and effects of this invention clearer and more explicit, the invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0021] In the description of this invention, it should be understood that the terms "top," "bottom," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing this invention and simplifying the description, and should not be construed as limiting this invention; in addition, the terms "installation," "connection," etc. should be interpreted broadly, and those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0022] Please see Figure 1 This invention provides a method for detecting the operating status of a three-phase cable. The method comprises a telescopic rod 1, a clamp-on current transformer assembly 2 disposed on the top of the telescopic rod 1, and an image acquisition device 3. The telescopic rod 1 is equipped with a control device 6 and a display device. The image acquisition device 3, the display device, and the clamp-on current transformer assembly 2 are electrically connected to the control device 6. The method for detecting the operating status of a three-phase cable includes: 101. Obtain the real-time position image of the clamp-on current transformer component 2 fed back by the image acquisition device 3, generate position display information based on the real-time position image and send it to the display device; In this embodiment, the image acquisition device 3 uses a high-definition industrial miniature camera, installed next to the clamp-shaped current transformer assembly 2 at the top of the telescopic rod 1, to acquire real-time images of the spatial position of the clamp head and the target cable; the control device 6 has a built-in ARM Cortex-A72 architecture processor and WiFi module, and after receiving the video stream transmitted by the camera, it performs real-time decoding processing on the image to generate position display information that conforms to the display device's adaptation format, and sends it to the OLED display device via a wired connection to realize visual observation at the operator's end; the position display information includes deviation auxiliary indicators superimposed on the image, which makes it easy for the operator to intuitively judge the adjustment direction.

[0023] 102. Input the real-time location image into the pre-trained localization and recognition model to determine whether to generate a closure prompt message; In this embodiment, the positioning recognition model is an optimized MobileNet SSD model. The control device 6 preprocesses the real-time position image according to preset specifications and inputs it into the positioning recognition model. The positioning recognition model extracts features at multiple scales and detects targets, and outputs the relative positional relationship between the clamp head and the cable and the clamping compatibility judgment result. If the judgment result satisfies that the cable falls completely into the clamping range of the clamp head and the position deviation is within the allowable threshold, a closure prompt message is generated. The closure prompt message includes a green pop-up window displayed on the OLED display device and a buzzer prompt.

[0024] 103. When a closing command is received, the clamp head of the clamp-on current transformer assembly 2 is controlled to close; In this embodiment, the closing command is triggered by the operator through the physical button or joystick on the operation panel 4 at the bottom of the telescopic rod 1. After receiving the command, the control device 6 drives the current transformer position control servo motor through the PWM signal. The current transformer position control servo motor drives the five clamp heads of the clamp-type current transformer assembly 2 to open and close synchronously through the linkage rod. The clamp head closing process adopts a graded driving strategy. The speed is faster in the initial stage and decelerates when approaching the closing end point to avoid impact damage to the cable or current transformer.

[0025] 104. Obtain the real-time current information fed back by the clamp-on transformer assembly 2, wherein the real-time current information includes real-time three-phase current, real-time zero-sequence current and real-time total current; In this embodiment, the clamp-on current transformer assembly 2 integrates five high-precision clamp-on current transformers, three of which collect the three-phase currents A, B, and C respectively, one collects the zero-sequence leakage current, and one collects the total current. The clamp-on current transformers convert the current signals into 0-5V analog signals, which are then converted into digital signals by a 16-bit AD converter and transmitted to the control device 6 via wired connection. In addition, the real-time current information also includes a timestamp of the acquisition, providing a time reference for subsequent multi-parameter timing analysis. The control device 6 uses a mean filtering algorithm to filter the received current data to remove high-frequency noise interference.

[0026] 105. Based on the real-time current information, determine whether there is an abnormal working condition. If so, determine the corresponding abnormal working condition information. In this embodiment, the control device 6 has a built-in three-level anomaly judgment logic: First, the three-phase unbalance is calculated based on the industry standard formula and compared with the preset unbalance threshold; second, the high-frequency harmonic features of the current signal are extracted by db4 wavelet transform and the abnormal waveform is analyzed by Fourier transform; finally, the load trend for the next 24 hours is output by the LSTM time series prediction model to determine whether there is a long-term heavy load; if any judgment result exceeds the normal range, the corresponding anomaly type, occurrence time, numerical deviation and other work anomaly information are recorded.

[0027] 106. Obtain preset early warning rules, and generate early warning information based on the early warning rules and the work anomaly information; In this embodiment, the preset early warning rules are stored in the storage module of the control device 6. A hierarchical early warning mechanism is adopted, and the rule content is strongly correlated with the anomaly type, duration, and load background. The early warning information includes the early warning level, anomaly description, and suggested handling measures. It is visualized through an OLED display device and an electrical signal is generated simultaneously to realize audible and visual alarms. The generated early warning information is accompanied by a unique identification code, which facilitates data traceability and association with the background system.

[0028] The three-phase cable working status detection method disclosed in this invention extends the working distance by using a telescopic pole 1 and combines it with real-time position images fed back by an image acquisition device 3 for remote positioning and control. This avoids workers from having close contact with high-voltage equipment or performing work at heights, effectively eliminating the dual safety risks of electric shock and falls. A pre-trained positioning recognition model assists in judging the closing timing, and the display device presents the position information, reducing reliance on manual judgment and recording, improving the automation level of the detection process, and reducing human error. In addition, through the integrated clamp-on current transformer component 2, three-phase current, zero-sequence current, and total current can be collected simultaneously. Combined with multi-dimensional anomaly judgment logic, it can achieve comprehensive identification of various faults such as cable load imbalance, insulation partial discharge, and long-term heavy load, and generate corresponding early warning information. This breaks through the limitations of traditional equipment that only has basic data acquisition functions, and realizes in-depth monitoring of cable working status and early warning of potential faults.

[0029] Furthermore, in this embodiment of the invention, the step of inputting the real-time location image into a pre-trained localization and recognition model to determine whether to generate a closure prompt message includes: 201. The real-time location image is scaled and normalized to obtain a preprocessed location image; In this embodiment, the scaling process uses a bilinear interpolation algorithm to uniformly scale the real-time position image to 300×300 pixels, avoiding model inference errors caused by inconsistent image sizes; the normalization process maps the image pixel values ​​from [0, 255] to the [0, 1] interval, reducing the numerical computation complexity in the model training and inference process and improving the convergence speed; in addition, the preprocessing process also includes image channel conversion, specifically, converting the RGB image to BGR format to adapt to the input requirements of the TensorFlow framework, and making slight contrast adjustments to enhance the contour features of the clamp head and cable in the image.

[0030] 202. Input the preprocessed location image into the pre-trained localization recognition model to obtain the bounding box coordinates of the clamp head, the bounding box coordinates of the cable, and the category confidence. The localization recognition model is an optimized MobileNet SSD model. In this embodiment, the optimized MobileNet SSD model uses MobileNet V2 as the feature extraction module and SSD as the detection head. The feature extraction module employs depthwise separable convolution and inverse residual structure, and the detection head performs parallel detection on six multi-scale feature maps. The bounding box coordinates output by the localization and recognition model are formatted to accurately describe the position range of the pliers and cable in the image. The category confidence represents the probability that the localization and recognition model determines the target to be a pliers or a cable, with a value range of [0, 1], and is used to filter valid detection results.

[0031] 203. Based on the bounding box coordinates of the clamp head and the bounding box coordinates of the cable, calculate the relative positional deviation between the clamp head and the cable, wherein the relative positional deviation includes horizontal deviation, vertical deviation and distance deviation; In this embodiment, the horizontal deviation is defined as the difference between the x-coordinate of the center of the clamp head bounding box and the x-coordinate of the center of the cable bounding box, and the vertical deviation is defined as the difference between the y-coordinate of the center of the clamp head bounding box and the y-coordinate of the center of the cable bounding box. Both are in pixels. The distance deviation is calculated based on the ratio between the pixel size of the cable in the image and its actual size. It is determined by a preset pixel-actual size mapping table corresponding to different cable specifications and is used to reflect the actual spatial distance between the clamp head and the cable.

[0032] 204. Obtain a preset deviation threshold, and based on the relative position deviation and the preset deviation threshold, determine whether to generate a closure prompt message; In this embodiment, the preset deviation thresholds were determined through extensive experimental verification. The specific values ​​are: horizontal deviation ≤ 5mm, vertical deviation ≤ 5mm, and distance deviation ≤ 10cm. The control device 6 compares the calculated relative position deviations with the preset deviation thresholds one by one. If all three deviations meet the preset deviation threshold requirements, and the positioning recognition model outputs an open jaw state with a category confidence level ≥ 0.7, a closure prompt message is generated. If any deviation exceeds the preset deviation threshold or the jaw state is closed, no prompt is generated, or a position adjustment prompt is generated.

[0033] In this embodiment, multi-dimensional relative positional deviations are calculated using bounding box coordinates to achieve quantitative judgment of the positional relationship between the clamp head and the cable, avoiding the subjective errors of traditional manual positioning and meeting the requirements of high-precision detection. By establishing a multi-condition closed-loop prompt logic based on deviation threshold, confidence level, and clamp head status, the effectiveness of the closing action is ensured from multiple dimensions, improving the clamping success rate and solving the industry pain points of poor visibility and difficulty in alignment when hanging cables at high altitudes.

[0034] Furthermore, in this embodiment of the invention, before inputting the real-time location image into the pre-trained localization and recognition model, the method further includes: 301. Construct a localization and recognition model to be trained. The localization and recognition model to be trained includes a feature extraction module, a detection head, and a post-processing module connected in sequence. The feature extraction module adopts the MobileNet V2 network, and the detection head adopts the SSD architecture. In this embodiment, the feature extraction module adopts the MobileNet V2 architecture, which includes an initial convolutional layer and 19 inverse residual blocks, ultimately outputting six feature maps at different scales. Specifically, the initial convolutional layer is a 3×3 convolution with a stride of 2 and 32 output channels. The 19 inverse residual blocks are divided into 7 groups, with dilation factors of 1, 6, 6, 6, 6, 6, and 6, respectively. By processing each group of inverse residual blocks sequentially, the scale of the feature maps ranges from 38×38×256 to 1×1×1024. The detection head adopts an SSD architecture, and each feature map... Corresponding to independent classification and regression convolutional layers, the classification convolutional layer is a 3×3 convolution with the number of output channels equal to the number of prior boxes multiplied by 3; the regression convolutional layer is also a 3×3 convolution with the number of output channels equal to the number of prior boxes multiplied by 4. The post-processing module includes non-maximum suppression, confidence filtering, and bounding box coordinate calibration, which are connected in series. Non-maximum suppression, or NMS, has a threshold set to 0.5; the confidence filtering threshold is set to 0.7. The localization and recognition model is built using the TensorFlow 2.8 framework, with an input of a 300×300×3 RGB image and an output of target location, category, and confidence information. In addition, a batch normalization layer is set between the feature extraction module and the detection head to accelerate model training convergence.

[0035] 302. Collect multiple sets of images of clamps and cables under different lighting conditions, shooting distances, and cable specifications to construct an initial sample set; In this embodiment, the lighting conditions cover three scenarios: strong light, weak light, and nighttime; the shooting distance covers 1-5m with a step size of 1m, covering the mainstream operating distance for on-site testing; the cable specifications include two commonly used high-voltage cables, 10kV and 35kV, and also include cable samples with different degrees of wear to ensure sample diversity; the initial sample set contains 5000 images, of which 2500 are in the open / closed state of the clamps, and the relative position of the cable and the clamps covers various scenarios such as alignment, offset, and misalignment.

[0036] 303. Perform annotation processing, data augmentation processing, uniform scaling processing, and partitioning processing on the initial sample set in sequence to obtain the training set and the test set; In this embodiment, the labeling process is performed using the LabelImg tool, and the following content is labeled according to the Pascal VOC format: clamp head and corresponding open / closed status label, cable bounding box coordinates, and cable category label; The data augmentation process includes the following operations: random cropping with a cropping ratio of 0.7 to 1.0; horizontal or vertical flipping with a flipping probability of 0.5; brightness and contrast adjustment with a brightness adjustment range of ±20% and a contrast adjustment range of ±15%; Gaussian blurring with a standard deviation of 0 to 0.5; through the aforementioned data augmentation operations, the sample size is expanded to 20,000 images. The unified scaling process uses a bilinear interpolation algorithm to scale all images to 300×300 pixels; The sample splitting process uses stratified sampling, randomly dividing the training set and the test set in an 8:2 ratio. The training set contains 16,000 samples, and the test set contains 4,000 samples to ensure the independence of training and testing.

[0037] 304. Based on the TensorFlow framework, using the SGD optimizer, with the goal of minimizing the loss function, the localization and recognition model to be trained is iteratively trained using the training set and the test set; In this embodiment, the training parameters are set as follows: initial learning rate 0.001, momentum parameter 0.9, weight decay coefficient 0.0005, batch size 32, and 50 training epochs; the loss function is a weighted sum of position loss and classification loss with a weight ratio of 1:1; after each epoch during training, the model performance is verified using a test set, and the change in loss value is recorded; the learning rate decays by 10% every 10 epochs to avoid oscillations in later training; overfitting is monitored in real time during training, and an early stopping strategy is adopted when the accuracy of the test set does not improve for 5 consecutive epochs to avoid model overfitting.

[0038] 305. When the preset iteration stopping condition is met, output the trained localization and recognition model; In this embodiment, the iteration stopping condition for training the localization and recognition model is set to any of the following: first, the mAP of the test set is ≥95% and there is no performance improvement after 5 consecutive training rounds; second, the number of training rounds is ≥50. The trained localization and recognition model must meet the following performance indicators: clamp head recognition accuracy ≥96%; cable recognition accuracy ≥97%; bounding box positioning error ≤2 pixels; single frame inference time ≤30 milliseconds. The trained localization and recognition model is output in TensorFlowSavedModel format, which facilitates the subsequent deployment of the model to embedded hardware devices.

[0039] 306. Use pruning techniques to remove redundant parameters from the trained localization and recognition model to obtain the localization and recognition model; In this embodiment, structured pruning technology is employed. Based on the L1 regularization criterion, convolutional kernels in the trained localization recognition model are screened to remove redundant convolutional kernels with absolute weight values ​​less than a threshold of 0.01. The pruning process is performed in stages, with each stage pruning by 10%. After each stage of pruning, three rounds of fine-tuning training are conducted to ensure that the model performance does not significantly degrade. The final model size can be compressed to less than 5MB while maintaining a test set mAP ≥ 94%, thus meeting the storage and computing resource constraints of the embedded control device 6. Through pruning and quantization optimization, the localization recognition model is made lightweight and efficient while ensuring its recognition accuracy, adapting to the hardware resource requirements of the embedded control device 6. This solves the technical bottleneck problem of traditional deep learning models being difficult to deploy on portable detection devices.

[0040] Furthermore, in this embodiment of the invention, after controlling the clamp head of the clamp-on transformer assembly 2 to close when a closing command is received, the method further includes: 401. The image acquisition device 3 is used to acquire the real-time closing image of the clamp sensor assembly after the closing command is executed; In this embodiment, after the closing command is executed, the control device 6 delays for 100 milliseconds to trigger the image acquisition device 3 to take pictures, ensuring that the clamp head is in a fully closed and stable state, and continuously acquires 10 frames of images to avoid misjudgment caused by jitter or interference in a single frame image; the acquisition parameters of the real-time closing image are consistent with the acquisition parameters of the real-time position image to ensure the consistency of image quality; the acquired real-time closing images are all accompanied by a timestamp and a closing command execution identifier to facilitate correlation analysis with previously acquired real-time position images; the control device 6 performs deduplication processing on the 10 frames of images, and the processed images are used for subsequent recognition and verification steps.

[0041] 402. Input the real-time closed image into the positioning and recognition model to determine whether the clamp head is in place; In this embodiment, the preprocessed real-time closed image is input into the positioning and recognition model, and the model outputs the bounding box coordinates and overlap rate of the clamp head and the cable. The judgment criteria for clamping in place are: the overlap rate between the cable bounding box and the clamp head bounding box is ≥90%, and the clamp head state output by the positioning and recognition model is closed, with a category confidence level ≥0.8; if the overlap rate is <70%, it is judged as clamping misalignment; if 70%≤overlap rate<90%, it is judged as clamping deviation.

[0042] 403. If the clamping head is in place, generate a clamping position display message and control the clamp sensor assembly to lock the current state; In this embodiment, the clamping position display information includes a blue pop-up window on the OLED display device, which displays the message "Clamping in place, start measurement"; it also includes a buzzer prompt, which is a single long beep with a duration of 0.5 seconds; When the current state is locked, the control device 6 sends a locking signal to the current transformer position control servo motor. The servo motor drives the mechanical lock to fix the clamp head and prevent the clamp head from loosening during the detection process. If the judgment result is that the clamping is misaligned or the clamping deviation is detected, a red pop-up window is generated, which displays the message "Clamping abnormal, please operate again". At the same time, the current transformer position control servo motor is driven to open the clamp head and return to the positioning stage.

[0043] In this embodiment, the copper drum adds a clamping position verification step after closure, forming a closed-loop control of positioning, closure and verification. This avoids improper clamping due to mechanical errors and operational deviations, ensuring the accuracy of current acquisition. Compared with detection methods without verification procedures, the measurement error is reduced. The locking mechanism after clamping ensures stable contact between the clamp head and the cable during the detection process, avoiding poor contact caused by external forces, and improving the stability and consistency of the detection data. It is especially suitable for detection operations in complex outdoor environments.

[0044] Furthermore, in this embodiment of the invention, the step of determining whether there is an abnormal working condition based on the real-time current information, and if so, determining the corresponding abnormal working condition information, includes: 501. Calculate the three-phase unbalance based on the real-time current information, and compare the three-phase unbalance with a preset unbalance threshold to determine whether there is an abnormal working condition. In this embodiment, the three-phase current data in the real-time current information is used to calculate the three-phase imbalance according to the standard formula of the power industry; the preset imbalance threshold is set to 10% by default, and can be adjusted by the background system according to the power grid operation requirements, with an adjustment range of 5%-15%; the control device 6 calculates the three-phase imbalance once per second, records the calculation results for 10 consecutive times, and takes the average value as the current imbalance value. If the average value is greater than the threshold and the duration is greater than or equal to 10 minutes, it is determined to be an abnormal three-phase current imbalance.

[0045] 502. High-frequency harmonic features are extracted from the real-time current information using wavelet transform, and the harmonic frequencies of the high-frequency harmonic features are analyzed using Fourier transform to calculate the effective values ​​of detail coefficients. The effective values ​​are compared with a preset standard deviation threshold to determine whether there is an abnormal working condition. In this embodiment, the db4 wavelet is used to decompose the real-time current time series data into three layers to extract the d3 layer detail coefficients; Fourier transform is performed on the d3 layer detail coefficients to obtain the harmonic spectrum and identify the characteristic peaks in the spectrum; the effective value of the detail coefficients is calculated according to the root mean square formula; the preset standard deviation threshold is three times the effective value of the detail coefficients of the normal current waveform; if the effective value exceeds the threshold for five consecutive sampling points, it is determined to be an abnormal current waveform, and the corresponding harmonic frequency, amplitude and phase of occurrence are recorded.

[0046] 503. Input the real-time current information into the pre-trained time-series prediction model to obtain the future load change curve, and determine whether there is any abnormal operation based on the future load change curve; In this embodiment, the pre-trained time series prediction model is an LSTM model, whose input is the three-phase current data of the same period in the last 7 days and the output is the load change curve of the next 24 hours; the time series prediction model training process uses historical load data of the power grid, and the training index is average absolute error ≤ 5%; when the load is ≥ 80% of the rated load and the duration is ≥ 6 hours in the next 24 hours, it is judged as a long-term heavy load anomaly.

[0047] 504. If an abnormal situation exists, the corresponding abnormal work information shall be determined. The abnormal work information includes one or more of the following: duration of three-phase imbalance, abnormal waveform, and load trend. In this embodiment, the operational anomaly information is stored in a structured data format, including anomaly type, occurrence time, key parameter values, phase, and fault probability. If multiple anomalies exist simultaneously, they are sorted by priority, with high-priority anomaly information being recorded first. The priority order is: abnormal waveform > three-phase imbalance > long-term heavy load. The operational anomaly information is accompanied by a data traceability identifier, which is associated with the corresponding original current data and image data, facilitating subsequent fault review and analysis, avoiding blind repairs, reducing maintenance costs, and improving fault handling efficiency.

[0048] In this embodiment, a three-level anomaly judgment system is constructed to comprehensively cover cable operation anomaly scenarios from three dimensions: static parameters, dynamic characteristics, and future trends. Compared with traditional single-parameter judgment, the anomaly identification accuracy is improved. The anomaly judgment process is automated and real-time, requiring no manual intervention. This solves the problems of strong subjectivity and delayed response in traditional detection that rely on human experience to judge faults, shortens the anomaly response time, and buys time for early intervention in faults.

[0049] Furthermore, in this embodiment of the invention, the step of obtaining a preset early warning rule and generating early warning information based on the early warning rule and the work anomaly information includes: 601. If the duration of the three-phase imbalance is greater than or equal to the preset duration threshold, there are no abnormal waveforms, and the load trend is normal, then a first-level early warning message will be generated. In this embodiment, the preset duration threshold is set to 10 minutes by default, and can be adjusted according to the power grid operation and maintenance requirements, with an adjustment range of 5-30 minutes; the specific content of the first-level warning information is "First-level warning: Three-phase current imbalance, current imbalance degree is XX%, duration is XX minutes, load trend is normal, no abnormal waveform; it is recommended to optimize load allocation and review regularly"; the warning information is presented as a yellow pop-up window on an OLED display device, without a buzzer alarm.

[0050] 602. If an abnormal waveform is detected, the duration of the three-phase imbalance is less than the preset duration threshold, and the load trend is normal, a level 2 early warning message will be generated. In this embodiment, the specific content of the secondary warning information is "Secondary warning: An abnormal current waveform has been detected, corresponding to phase XX, the three-phase imbalance is normal, and the load trend is normal; suspected partial discharge or poor contact, it is recommended to re-check and test within 24 hours"; the warning information of the secondary warning information is displayed in an orange pop-up window on an OLED display device, accompanied by a 1Hz buzzer alarm for 3 seconds.

[0051] 603. If abnormal waveforms exist simultaneously and the duration of three-phase imbalance is greater than or equal to the preset duration threshold, or if the load trend is abnormal, a level three early warning message will be generated. In this embodiment, the abnormal load trend refers to the LSTM model predicting a long-term heavy load in the next 24 hours, i.e., a load ≥ 80% of the rated load for ≥ 6 hours. The specific content of the Level 3 warning information is "Level 3 Warning: High-risk insulation degradation! Abnormal current waveform detected, three-phase imbalance XX% / long-term heavy load in the next 24 hours; it is recommended to immediately stop the machine for inspection to avoid the fault from escalating." The warning information of the Level 3 warning information is presented by a red pop-up window flashing on an OLED display device, accompanied by a 3Hz buzzer alarm that continues until the operator confirms, while simultaneously sending an emergency alarm signal including location information to the background system.

[0052] In this embodiment, a tiered early warning mechanism is established to generate early warning information based on the severity of anomalies and risk levels. This avoids the use of homogenized early warning modes, allowing maintenance personnel to quickly distinguish fault priorities, rationally allocate maintenance resources, and improve the targeted nature of maintenance responses. Furthermore, it ensures that high-risk faults receive focused attention and emergency handling, preventing the escalation of faults due to untimely early warnings and significantly improving the safety of power grid operation. Each level of early warning information includes specific abnormal parameters, fault location, risk assessment results, and handling suggestions. Compared to traditional vague fault alarm methods, this provides more comprehensive decision support, lowers the professional threshold for maintenance personnel, and is particularly suitable for grassroots maintenance scenarios.

[0053] Furthermore, in this embodiment of the invention, after generating the warning information based on the warning rules and the work anomaly information, the method further includes: 701. Summarize the data to be transmitted, including measurement data, status data, and control commands; In this embodiment, the measurement data includes real-time current information, three-phase imbalance calculation results, abnormal waveform identification records, load trend prediction curves, and early warning information; the status data includes equipment operating status, battery power, signal strength, and clamping status; the control commands include closing / releasing commands, emergency stop commands, and detection mode switching commands issued by the operator, as well as parameter adjustment commands and early warning rule update commands issued by the background system; the data to be transmitted is summarized in timestamp order, and each frame of data is accompanied by a unique frame number to form a structured data frame, which facilitates packet loss retransmission and data sorting during transmission.

[0054] 702. Classify the data to be transmitted to distinguish between critical control instructions and batch data; In this embodiment, the key control commands are defined as commands that directly affect the safety of equipment operation and the effectiveness of detection, specifically including: clamp opening and closing control commands, equipment start and stop commands, emergency stop commands, detection mode switching commands, and data upload trigger commands; batch data is defined as the massive amount of data continuously generated during the detection process, specifically including: real-time video streams, current timing data, anomaly records, equipment status parameters, and early warning information; the classification of data to be transmitted is achieved through a data type identifier field, and the control device 6 automatically parses the identifier field to complete the classification after receiving the data to be transmitted; the classification rules are stored in the national cryptographic security chip of the control device 6 and support remote updates through the background system to ensure the flexibility of the classification logic.

[0055] 703. The key control command is encrypted using the SM2 elliptic curve asymmetric encryption algorithm, and then the encrypted key control command is digitally signed using a preset private key to obtain an encrypted command with a signature value. In this embodiment, the SM2 encryption algorithm follows the GM / T0003-2012 standard and uses 256-bit elliptic curve parameters. The control device 6 has a built-in national cryptographic security chip that generates a device-specific public / private key pair. The private key is stored in the chip's secure storage area and cannot be exported. The public key is pre-distributed to the backend system. When the data to be transmitted is a critical control command, the critical control command is first processed in plaintext to remove redundant fields. Then, the command is asymmetrically encrypted using the backend system's public key to generate an encrypted command. Finally, the encrypted command is digitally signed using the device's private key to obtain an encrypted command with a signature value.

[0056] 704. The batch data is encrypted in blocks using the SM4 block cipher algorithm to generate encrypted data blocks and corresponding data check codes; In this embodiment, the SM4 encryption algorithm follows the GM / T0004-2012 standard, uses a 128-bit key, and sets the encryption mode to ECB mode. When encrypting batch data, the batch data is first divided into blocks of 16 bytes each. Then, the national cryptographic security chip calls the preset SM4 key to perform encryption on each data block and outputs the encrypted data block. Finally, CRC32 checksum calculation is performed on all encrypted data blocks to generate the corresponding data checksum, and the data checksum is combined with the encrypted data block. The SM4 key is generated by a hardware random number generator to ensure the randomness and security of the key.

[0057] 705. Generate wireless communication information based on the encrypted instructions with signature values, the encrypted data blocks, and the corresponding data verification codes; In this embodiment, the wireless communication information adopts a unified transmission protocol frame format, which is as follows: Frame header: 0xAA55; Frame type: optional type is critical control instruction or batch data; Data length; Data content; Frame tail: 0x55AA; The control device 6 fills the encrypted instruction or encrypted data block with the signature value and the corresponding data verification code into the data content field, and sends it to the background monitoring center via the 4G / 5G industrial-grade wireless communication module.

[0058] In this embodiment, based on the differences in characteristics between key control instructions and batch data, a classification encryption strategy is adopted to adapt the corresponding encryption algorithm. Specifically, SM2 asymmetric encryption is used to ensure the immutability and non-repudiation of key instructions; SM4 symmetric encryption is used to balance the encryption efficiency of batch data, so as to solve the technical problem that traditional single encryption algorithms cannot balance security and efficiency; by adopting national cryptographic standard algorithms and dedicated national cryptographic security chips, the compliance and security of encrypted data transmission are guaranteed, meeting the strict requirements of the power industry for data confidentiality, and effectively preventing data from being eavesdropped, tampered with or forged during transmission.

[0059] Please see Figure 2The present invention also provides a three-phase cable operating status detection device, wherein the three-phase cable operating status detection device uses any of the three-phase cable operating status detection methods described above to achieve operating control; the three-phase cable operating status detection device includes a telescopic rod 1 and a clamp-type current transformer assembly 2 and an image acquisition device 3 disposed on the top of the telescopic rod 1; a control device 6 and a display device are disposed on the telescopic rod 1; the image acquisition device 3, the display device and the clamp-type current transformer assembly 2 are electrically connected to the control device 6 respectively; the clamp-type current transformer assembly 2 includes five synchronously opening and closing clamp-type current transformers, and the five clamp-type current transformers are respectively used to detect three-phase current, zero-sequence current and total current.

[0060] In this embodiment, the telescopic rod 1 has a segmented structure. Its top is fixed with a modular quick-release structure to a clamp-type current transformer assembly 2 and an image acquisition device 3. The camera of the image acquisition device 3 is mounted next to the clamp-type current transformer assembly 2 via a bracket, maintaining a fixed distance from the clamp head. A drive motor for the telescopic rod 1 is built into the middle of the telescopic rod 1, and this drive motor is connected to the telescopic mechanism of the telescopic rod 1. An operation panel 4 is fixed to the bottom of the telescopic rod 1. The image acquisition device 3 is responsible for environmental visual perception, converting position information into image signals, and transmitting these images to the control device 6 via a WiFi module. The clamp-type current transformer assembly 2 is responsible for current signal acquisition, converting physical current into electrical signals. The display device is used for information output, providing visual feedback to the operator. The control device 6 interacts remotely with the back-end monitoring center via a wireless communication module and outputs control commands after processing the data using a built-in algorithm to drive the actuator.

[0061] The three-phase cable operating status detection device disclosed in this invention integrates image acquisition, current measurement, remote control, intelligent analysis, and secure communication functional modules into a single telescopic pole structure through an integrated design. Compared with traditional distributed detection equipment, it is smaller and lighter, making it easier to carry and operate on-site, thus solving the problems of bulky and cumbersome operation of traditional detection equipment. The modular integration and synchronous opening and closing design of the five clamp-on current transformers enable simultaneous acquisition of three-phase current, zero-sequence current, and total current. Compared with traditional step-by-step measurement methods, the detection efficiency is improved, and the data synchronization is significantly enhanced, providing an accurate data foundation for multi-parameter collaborative analysis.

[0062] Furthermore, an operation panel 4 is provided at the bottom of the telescopic rod 1, and the display device is embedded in the operation panel 4; the operation panel 4 is also provided with a current transformer position adjustment rocker and a telescopic rod 1 control rocker, both of which are electrically connected to the control device 6; a telescopic rod 1 drive motor is provided in the middle of the telescopic rod 1, and the drive motor is electrically connected to the control device 6. The control device 6 adjusts the working state of the telescopic rod 1 drive motor based on the control information of the telescopic rod 1 control rocker; a current transformer position control servo is provided at the top of the telescopic rod 1, and the servo is electrically connected to the control device 6 and drivenly connected to the clamp-type transformer assembly 2. The control device 6 adjusts the working state of the current transformer position control servo based on the control information of the current transformer position adjustment rocker.

[0063] In this embodiment, the operation panel 4 is fixed to the bottom of the telescopic rod 1 and adopts an embedded installation design, which fits seamlessly with the outer shell of the telescopic rod 1; the display device is embedded in the middle of the operation panel 4, and the current transformer position adjustment rocker and the telescopic rod 1 control rocker are symmetrically arranged on both sides of the display device; a physical control button for performing closing or releasing, emergency stop, and mode switching functions is also provided below the operation panel 4. The drive motor of the telescopic rod 1 is installed in the internal cavity of the middle part of the telescopic rod 1. The output shaft of the motor is connected to the telescopic sleeve of the telescopic rod 1 through a gear transmission mechanism. The control signal input terminal of the motor is electrically connected to the control device 6. The current transformer position control servo is installed next to the clamp-shaped current transformer assembly 2 at the top of the telescopic rod 1. The output shaft of the current transformer position control servo is connected to the opening and closing mechanism of the five clamps through a linkage rod. The control signal input terminal of the current transformer position control servo is electrically connected to the control device 6. The current transformer position adjustment rocker is a three-dimensional rocker. The operator's displacement operation is converted into a voltage signal. The control device 6 analyzes the amplitude and direction of the voltage signal and outputs the corresponding PWM control signal to drive the current transformer position control servo to rotate, thereby adjusting the spatial position of the clamps through the linkage rod. The telescopic rod 1 control rocker is a one-dimensional rocker. The displacement signal is converted into a voltage signal. The control device 6 drives the drive motor of the telescopic rod 1 to rotate forward or reverse, realizing the telescopic rod 1's extension and retraction.

[0064] Furthermore, the control device 6 includes a controller and a power management module electrically connected to the controller. The controller integrates an artificial intelligence processing chip, a national cryptographic security chip, a wireless communication module, and a WiFi module. In this embodiment, the control device 6 adopts a modular integrated design, with the controller as its core component. Each functional module is electrically connected to the controller via PCB board circuitry, forming the core for data processing and control. The power management module is electrically connected to the power input terminal of the controller and supplies power to each working component via branch power supply lines. The power management module has a built-in lithium battery and charging management circuit. The artificial intelligence processing chip is interconnected with the controller via a PCIe interface, focusing on inference calculations for the positioning and recognition model and anomaly analysis of current data to reduce the computational load on the controller. The national cryptographic security chip is electrically connected to the controller via an SPI interface, with a built-in encryption algorithm engine and secure storage area. It receives encryption or signature commands from the controller, performs encryption operations, and feeds back the results to the controller. The wireless communication module is electrically connected to the controller via a UART interface, responsible for remote data interaction with the background monitoring center. The WiFi module is electrically connected to the controller via an SDIO interface, responsible for short-range image transmission with the image acquisition device 3.

[0065] In this embodiment, a distributed computing architecture combining a controller and a dedicated chip is adopted. Computationally intensive tasks such as intelligent analysis are assigned to the artificial intelligence processing chip. The controller focuses on generating control instructions and coordinating modules. Compared with the traditional single controller design, the data processing speed is significantly improved, thereby ensuring the real-time performance of the detection process. The integrated design of the national cryptographic security chip and the wireless communication module realizes the hardware-based and automated data encryption and transmission. The encryption process does not occupy the computing resources of the controller, and the key is stored in the chip's secure area, effectively avoiding the risk of key leakage in software encryption methods and meeting the high requirements of the power industry for secure data transmission.

[0066] It is understood that those skilled in the art can make equivalent substitutions or changes to the technical solution and inventive concept of the present invention, and all such changes or substitutions should fall within the protection scope of the present invention.

Claims

1. A method for detecting the operating status of a three-phase cable, characterized in that, The three-phase cable working status detection device includes a telescopic rod, a clamp-on current transformer assembly and an image acquisition device installed on the top of the telescopic rod. The telescopic rod is equipped with a control device and a display device. The image acquisition device, the display device and the clamp-on current transformer assembly are electrically connected to the control device. The method for detecting the operating status of three-phase cables includes: The real-time position image of the clamp-on current transformer component fed back by the image acquisition device is obtained, and position display information is generated based on the real-time position image and sent to the display device. The real-time location image is input into a pre-trained localization and recognition model to determine whether to generate a closure prompt message; When a closing command is received, the clamp head of the clamp-on current transformer assembly is controlled to close. The real-time current information fed back by the clamp-on transformer assembly is obtained, and the real-time current information includes real-time three-phase current, real-time zero-sequence current and real-time total current. Based on the real-time current information, determine whether there is any abnormal operation; if so, determine the corresponding abnormal operation information. Obtain preset warning rules, and generate warning information based on the warning rules and the work anomaly information.

2. The method for detecting the working status of a three-phase cable according to claim 1, characterized in that, The step of inputting the real-time location image into a pre-trained localization and recognition model to determine whether to generate a closure prompt includes: The real-time location image is scaled and normalized to obtain a preprocessed location image; The preprocessed location image is input into the pre-trained localization and recognition model to obtain the bounding box coordinates of the clamp head, the bounding box coordinates of the cable, and the class confidence. The localization and recognition model is an optimized MobileNet SSD model. Based on the bounding box coordinates of the clamp head and the bounding box coordinates of the cable, the relative positional deviation between the clamp head and the cable is calculated, and the relative positional deviation includes horizontal deviation, vertical deviation and distance deviation. Obtain a preset deviation threshold, and based on the relative position deviation and the preset deviation threshold, determine whether to generate a closure prompt message.

3. The method for detecting the working status of a three-phase cable according to claim 2, characterized in that, Before inputting the real-time location image into the pre-trained localization and recognition model, the method further includes: A localization and recognition model to be trained is constructed. The localization and recognition model to be trained includes a feature extraction module, a detection head and a post-processing module connected in sequence. The feature extraction module adopts the MobileNet V2 network and the detection head adopts the SSD architecture. Multiple sets of images of clamps and cables under different lighting conditions, shooting distances, and cable specifications were collected to construct an initial sample set. The initial sample set is sequentially subjected to annotation processing, data augmentation processing, uniform scaling processing, and partitioning processing to obtain the training set and the test set; Based on the TensorFlow framework, the SGD optimizer is used to minimize the loss function, and the localization and recognition model to be trained is iteratively trained using the training set and the test set. When the preset iteration stopping condition is met, the trained localization and recognition model is output. Redundant parameters in the trained localization and recognition model are removed using pruning techniques to obtain the localization and recognition model.

4. The method for detecting the working status of a three-phase cable according to claim 1, characterized in that, After the clamp head of the clamp-on current transformer assembly is closed upon receiving a closing command, the process further includes: An image acquisition device is used to acquire real-time closed images of the clamp sensor assembly after the closure command is executed; The real-time closed image is input into the positioning and recognition model to determine whether the clamp head is in place; If the clamping head is in place, a clamping position display message is generated, and the clamping sensor assembly is controlled to lock the current state.

5. The method for detecting the working status of a three-phase cable according to claim 1, characterized in that, The step of determining whether there is an abnormal working condition based on the real-time current information, and if so, determining the corresponding abnormal working condition information, includes: The three-phase unbalance is calculated based on the real-time current information, and the three-phase unbalance is compared with a preset unbalance threshold to determine whether there is an abnormal working condition. The wavelet transform method is used to extract high-frequency harmonic features from the real-time current information, and the harmonic frequency of the high-frequency harmonic features is analyzed by Fourier transform to calculate the effective value of the detail coefficients. The effective value is compared with the preset standard deviation threshold to determine whether there is any abnormal operation. The real-time current information is input into a pre-trained time-series prediction model to obtain a future load change curve, and the existence of any abnormal operating conditions is determined based on the future load change curve. If an abnormal situation exists, the corresponding abnormal work information is determined. The abnormal work information includes one or more of the following: duration of three-phase imbalance, abnormal waveform, and load trend.

6. The method for detecting the working status of a three-phase cable according to claim 5, characterized in that, The step of obtaining preset early warning rules and generating early warning information based on the early warning rules and the work anomaly information includes: If the duration of the three-phase imbalance is greater than or equal to the preset duration threshold, there are no abnormal waveforms, and the load trend is normal, then a level one early warning message will be generated. If an abnormal waveform is detected, the duration of the three-phase imbalance is less than the preset duration threshold, and the load trend is normal, a level 2 early warning message will be generated. If abnormal waveforms exist simultaneously and the duration of three-phase imbalance is greater than or equal to the preset duration threshold, or if the load trend is abnormal, a level three early warning message will be generated.

7. The method for detecting the working status of a three-phase cable according to claim 1, characterized in that, After generating the early warning information based on the early warning rules and the work anomaly information, the method further includes: Summarize the data to be transmitted, which includes measurement data, status data, and control commands; The data to be transmitted is classified to distinguish between critical control commands and batch data; The key control instructions are encrypted using the SM2 elliptic curve asymmetric encryption algorithm, and then digitally signed using a preset private key to obtain encrypted instructions with signature values. The batch data is encrypted in blocks using the SM4 block cipher algorithm to generate encrypted data blocks and corresponding data check codes. Wireless communication information is generated based on the encrypted instructions with signature values, encrypted data blocks, and corresponding data verification codes.

8. A three-phase cable operating status detection device, characterized in that, The three-phase cable operating status detection device uses the three-phase cable operating status detection method as described in any one of claims 1-7 to achieve operating control; the three-phase cable operating status detection device includes a telescopic rod, a clamp-type current transformer assembly and an image acquisition device disposed on the top of the telescopic rod, a control device and a display device disposed on the telescopic rod, and the image acquisition device, the display device and the clamp-type current transformer assembly are electrically connected to the control device respectively; the clamp-type current transformer assembly includes five clamp-type current transformers that open and close synchronously, and the five clamp-type current transformers are respectively used to detect the three-phase current, the zero-sequence current and the total current.

9. A three-phase cable operating status detection device according to claim 8, characterized in that, An operation panel is provided at the bottom of the telescopic rod, and the display device is embedded in the operation panel. The operation panel is also provided with a current transformer position adjustment rocker and a telescopic rod control rocker, both of which are electrically connected to the control device. A telescopic rod drive motor is provided in the middle of the telescopic rod, and the drive motor is electrically connected to the control device. The control device adjusts the working state of the telescopic rod drive motor based on the control information of the telescopic rod control rocker. A current transformer position control servo is provided at the top of the telescopic rod. The servo is electrically connected to the control device and drivenly connected to the clamp-type current transformer assembly. The control device adjusts the working state of the current transformer position control servo based on the control information of the current transformer position adjustment rocker.

10. A three-phase cable operating status detection device according to claim 8, characterized in that, The control device includes a controller and a power management module electrically connected to the controller. The controller integrates an artificial intelligence processing chip, a national cryptographic security chip, a wireless communication module, and a WiFi module.

Images