UHV GIS breakdown point positioning method and system based on voltage wave head identification

Abstract

The disclosure provides an ultra-high voltage GIS breakdown point positioning method and system based on voltage wave head recognition, relates to the technical field of GIS breakdown point positioning, and comprises the following steps: acquiring a transient voltage traveling wave generated when GIS breaks down, and constructing double-end original voltage traveling wave time sequence data; inputting the traveling wave time sequence data into a wave head recognition model, adaptively denoising, multi-scale distortion profile recognizing and feature fusing the traveling wave time sequence through a front-end multi-scale feature extraction and cross-layer feature splicing channel of the wave head recognition model, and end-to-end outputting an accurate fault wave head position index; according to the fault wave head position index, converting the time when the traveling wave actually arrives at the first and last ends, and solving the double-end absolute fault distance by using a traveling wave distance measurement mathematical model; and obtaining the position of the fault breakdown point according to the double-end absolute fault distance, so that a high-precision GIS breakdown point positioning process is realized. The disclosure can realize low-cost and high-precision GIS breakdown point positioning.

Description

Technical Field

[0001] This disclosure relates to the field of GIS breakdown point location technology, specifically to a method and system for locating ultra-high voltage GIS breakdown points based on voltage wavefront identification. Background Technology

[0002] The statements in this section are merely background information relating to this disclosure and do not necessarily constitute prior art.

[0003] Gas-insulated metal-enclosed switchgear (GIS) is widely used in ultra-high voltage (UHV) power transmission systems due to its small footprint, minimal susceptibility to external environmental influences, and high reliability. However, because UHV systems have extremely high voltage levels and complex internal electric field distributions, insulation breakdown and flashover faults within GIS caused by insulation aging, metal particles, or foreign objects pose a serious threat to the safe and stable operation of the power grid. Therefore, rapidly and accurately locating the breakdown point after a fault occurs is crucial for quick fault diagnosis and power restoration.

[0004] Currently, ultrasonic positioning technology is commonly used in engineering practice to detect discharges and faults inside GIS (Gas Insulated Switchgear). This method mainly relies on capturing the mechanical acoustic vibration signal generated at the moment of breakdown. However, in practical applications, ultrasonic positioning has significant technical limitations: on the one hand, when ultrasonic waves propagate through the complex internal physical structure of GIS (such as interfaces between different media, basin insulators, etc.), they are prone to severe signal attenuation and distortion, leading to large errors in the final physical spatial positioning; on the other hand, the effective propagation distance of ultrasonic signals on the metal shell of GIS is extremely limited. This necessitates the dense deployment of ultrasonic probes at very short intervals along GIS pipelines in practical engineering applications. This sensor configuration significantly increases hardware procurement costs, on-site installation workload, and subsequent operation and maintenance costs.

[0005] To overcome the problems of low spatial resolution and high deployment cost of ultrasonic positioning, the two-end traveling wave ranging method based on electromagnetic transient signals is gradually becoming a more promising technical approach. Theoretically, two-end traveling wave positioning only requires installing a voltage sensor at each end of a GIS busbar segment. By utilizing the time difference between the arrival of the high-frequency voltage traveling wave generated at the moment of breakdown and the arrival at the two sensors, combined with the traveling wave propagation velocity, the precise fault location can be calculated. This method significantly reduces the number of sensors required and substantially lowers engineering costs.

[0006] However, in the complex operating environment of UHV GIS, the traditional two-end traveling wave ranging method faces the challenge of accurately extracting the wavefront. When the initial voltage traveling wave generated at the moment of breakdown propagates long distances within the GIS, it is affected by factors such as the skin effect, high-frequency losses in the gas medium, and local impedance abrupt changes, causing drastic attenuation and dispersion of the high-frequency energy of the traveling wave. This directly leads to severe distortion of the originally steep transient traveling wavefront, specifically manifested as abnormally flat rising and falling edges.

[0007] Traditional traveling wave positioning devices typically employ purely mathematical analytical methods such as fixed amplitude threshold, slope threshold detection, or wavelet transform singularity calibration to find the arrival time of the traveling wave. However, these methods still have the following limitations:

[0008] (1) Faced with wavefront signals that are severely distorted, slowed down and accompanied by strong electromagnetic interference at the scene, these traditional methods cannot accurately capture the true initial wavefront and are prone to misjudgment or omission of the wavefront starting point.

[0009] (2) Because electromagnetic traveling waves propagate very fast in GIS (close to the speed of light), even a nanosecond-level wavefront time calibration error will be amplified into a physical spatial positioning error of several meters or even tens of meters after being substituted into the double-end ranging formula, directly causing the ranging to fail.

[0010] (3) Existing UHV GIS fault location technology struggles to balance technical economy and ranging accuracy under complex operating conditions. How to achieve low-cost monitoring with a very small number of sensor nodes, while effectively overcoming the bottleneck of the difficulty in accurately extracting the wavefront arrival time caused by traveling wave propagation distortion, and thus achieving high-precision location of UHV GIS breakdown points, is a technical problem that urgently needs to be solved. Summary of the Invention

[0011] To address the aforementioned issues, this disclosure proposes a method and system for locating breakdown points in ultra-high voltage GIS based on voltage wavefront identification. This method utilizes a non-contact electro-optic crystal sensor to acquire localized dual-end synchronization signals and combines this with artificial intelligence algorithms to accurately identify voltage wavefronts that have undergone severe dispersion and distortion due to long-distance propagation. The arrival time of the traveling wave is precisely extracted from the disturbed transient waveform, thereby achieving low-cost, high-precision GIS breakdown point location.

[0012] According to some embodiments, the present disclosure adopts the following technical solutions:

[0013] A method for locating breakdown points in UHV GIS based on voltage wavefront identification includes:

[0014] Acquire the transient voltage traveling wave generated when GIS breaks down, and construct the time series data of the original voltage traveling wave at both ends;

[0015] The traveling wave time series data is input into the wavehead recognition model. Through the multi-scale feature extraction and cross-layer feature splicing channel of the wavehead recognition model, adaptive noise reduction, multi-scale distortion contour recognition and feature fusion are performed on the traveling wave time series, and the accurate fault wavehead location index is output end-to-end.

[0016] Based on the fault wavefront position index, the actual arrival times of the traveling wave at both ends are converted, and the absolute fault distance at both ends is calculated using the traveling wave ranging mathematical model.

[0017] The location of the breakdown point is obtained based on the absolute fault distance at both ends, thus achieving a high-precision GIS breakdown point location process.

[0018] According to some embodiments, the present disclosure adopts the following technical solutions:

[0019] The UHV GIS breakdown point location system based on voltage wavefront identification includes:

[0020] The signal acquisition module is used to acquire the transient voltage traveling wave generated when GIS breaks down and to construct the time series data of the original voltage traveling wave at both ends.

[0021] The wavefront recognition module is used to input traveling wave time series data into the wavefront recognition model. Through the multi-scale feature extraction and cross-layer feature splicing channel at the front end of the wavefront recognition model, adaptive noise reduction, multi-scale distortion contour recognition and feature fusion are performed on the traveling wave time series, and the accurate fault wavefront location index is output end-to-end.

[0022] The distance conversion module is used to convert the actual arrival time of the traveling wave at both ends based on the fault wavehead position index, and to calculate the absolute fault distance at both ends using the traveling wave ranging mathematical model.

[0023] The positioning feedback module is used to obtain the location of the fault breakdown point based on the absolute fault distance at both ends, thereby realizing a high-precision GIS breakdown point positioning process.

[0024] According to some embodiments, the present disclosure adopts the following technical solutions:

[0025] A computer program product includes a computer program that, when executed by a processor, implements the method for locating breakdown points in ultra-high voltage GIS based on voltage wavefront identification.

[0026] According to some embodiments, the present disclosure adopts the following technical solutions:

[0027] A non-transitory computer-readable storage medium is provided for storing computer instructions, which, when executed by a processor, implement the method for locating breakdown points in ultra-high voltage GIS based on voltage wavefront identification.

[0028] According to some embodiments, the present disclosure adopts the following technical solutions:

[0029] An electronic device includes a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to implement the method for locating the breakdown point of ultra-high voltage GIS based on voltage wavefront identification.

[0030] Compared with the prior art, the beneficial effects of this disclosure are as follows:

[0031] This disclosure presents a method for locating breakdown points in UHV GIS based on voltage wavefront identification. It utilizes a non-contact electro-optic crystal sensor to synchronously acquire transient raw voltage traveling wave time series data during short-circuit faults. Using a trained U-Net deep learning model, it performs multi-scale feature extraction and cross-layer feature stitching at the front end, outputting an accurate fault wavefront location index end-to-end. The system sampling frequency is combined to calculate the absolute arrival time of the traveling wave, and the absolute fault distance at both ends is calculated using physical formulas. Finally, the location result is output, and the ranging model is optimized in a closed loop using physical location error as the core criterion. This disclosure effectively overcomes the wavefront extraction difficulties caused by the propagation distortion of the traveling wave wavelength in traditional methods, achieving high-precision location of breakdown points in UHV GIS while significantly reducing hardware configuration costs.

[0032] The UHV GIS breakdown point location method disclosed herein addresses the shortcomings of ultrasonic positioning technology, which suffers from irregular reflection and refraction during the propagation of mechanical waves, leading to positioning errors. This method utilizes electromagnetic transient traveling wave signals, whose physical propagation path is a GIS coaxial waveguide structure. The medium is uniform and the velocity is stable, thus ensuring a high accuracy limit for long-distance ranging from a mechanistic perspective.

[0033] This disclosed method for locating breakdown points in UHV GIS based on voltage wavefront identification primarily addresses the core challenge of locating the arrival time of traveling waves, which is caused by wavefront distortion due to high-frequency loss (buffering of rising and falling edges) during propagation within GIS. By inputting data into a database and driving a neural network for iterative learning, the AI ​​algorithm acquires the ability to accurately identify the true wavefront time under strong interference, providing highly reliable time delay data for dual-end positioning.

[0034] The present invention discloses a method for locating breakdown points in ultra-high voltage GIS based on voltage wavefront identification. This method addresses the problem that the currently widely used ultrasonic positioning method requires densely deploying probes on a section of GIS, resulting in excessive costs. Based on the principle of dual-end ranging combined with an electro-optic crystal sensor, this invention allows monitoring of the entire pipeline to be completed simply by placing the sensor at both ends, significantly reducing hardware investment. Attached Figure Description

[0035] The accompanying drawings, which form part of this disclosure, are used to provide a further understanding of this disclosure. The illustrative embodiments of this disclosure and their descriptions are used to explain this disclosure and do not constitute an undue limitation of this disclosure.

[0036] Figure 1 This is a schematic diagram of the breakdown point dual-end positioning system according to an embodiment of the present disclosure;

[0037] Figure 2 This is a flowchart of the method for locating breakdown points in UHV GIS based on voltage wavefront identification, according to an embodiment of this disclosure.

[0038] Figure 3 This is a flowchart illustrating the internal module structure and data processing of the traveling wave head recognition model according to an embodiment of the present disclosure.

[0039] Figure 4 The image shows the original voltage traveling wave waveforms actually collected by the sensors at both ends in this embodiment of the present disclosure. Detailed Implementation

[0040] The present disclosure will be further described below with reference to the accompanying drawings and embodiments.

[0041] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of this disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure pertains.

[0042] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this disclosure. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms “comprising” and / or “including” are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0043] Example 1

[0044] One embodiment of this disclosure provides a method for locating breakdown points in ultra-high voltage GIS based on voltage wavefront identification. The method includes the following steps:

[0045] Step 1: Obtain the transient voltage traveling wave generated when the GIS breaks down, and construct the time series data of the original voltage traveling wave at both ends;

[0046] Step 2: Input the traveling wave time series data into the wavehead recognition model. Through the multi-scale feature extraction and cross-layer feature splicing channel of the front end of the wavehead recognition model, adaptive noise reduction, multi-scale distortion contour recognition and feature fusion are performed on the traveling wave time series, and the accurate fault wavehead location index is output end-to-end.

[0047] Step 3: Based on the fault wavefront position index, convert the actual arrival times of the traveling wave at both ends, and use the traveling wave ranging mathematical model to calculate the absolute fault distance at both ends.

[0048] Step 4: The location of the breakdown point is obtained based on the absolute fault distance at both ends, realizing a high-precision GIS breakdown point positioning process.

[0049] As one embodiment, this disclosure discloses a method for locating breakdown points in ultra-high voltage GIS based on voltage wavefront identification. This method first utilizes non-contact electro-optic crystal sensors configured at both ends of the GIS pipeline to be monitored to synchronously acquire transient raw voltage traveling wave time series data during a short-circuit fault. Then, this data is directly input into a pre-trained U-Net deep learning wavefront identification model. Through multi-scale feature extraction and cross-layer feature splicing channels at the model's front end, an accurate fault wavefront location index is output end-to-end. Next, the system sampling frequency is combined to calculate the absolute arrival time of the traveling wave, and the absolute fault distance at both ends is calculated using physical formulas. Finally, the location result is output, and the ranging model is optimized in a closed loop using physical location error as the core criterion. This disclosure effectively overcomes the wavefront extraction problem caused by the propagation distortion of the traveling wave wavelength distance in traditional methods, significantly reducing hardware configuration costs while achieving high-precision location of breakdown points in ultra-high voltage GIS. The specific implementation steps of the method are as follows:

[0050] Step 1: Obtain the transient voltage traveling wave generated when the GIS breaks down, and construct the time series data of the original voltage traveling wave at both ends;

[0051] Specifically, such as Figure 1 As shown, the main hardware components of the measurement system disclosed herein include an electro-optical crystal optical sensor 1 and an electro-optical crystal optical sensor 2 installed at the beginning M and end N of the pipeline to be monitored, and a centralized data processing host equipped with a high-precision synchronous timing module, a multi-channel high-speed data acquisition card, and a U-Net algorithm ranging unit.

[0052] When an insulation breakdown or short-circuit fault occurs within the GIS (Gas Insulation System), a high-frequency transient voltage traveling wave is generated at the fault point. Electro-optic crystal optical sensors installed at both ends of the pipeline couple this transient traveling wave signal in real time. Subsequently, after receiving a unified time reference signal triggered by a high-precision synchronization timing module, a multi-channel high-speed data acquisition card performs synchronous analog-to-digital conversion on the dual-end transient voltage traveling wave signal at a preset high-frequency sampling rate fs. The acquisition device extracts discrete data points within a set time window before and after the trigger moment, forming a one-dimensional original voltage traveling wave time series data X. m and X n .

[0053] Step 2: Input the traveling wave time series data into the wavehead recognition model. Through the multi-scale feature extraction and cross-layer feature splicing channel of the wavehead recognition model, adaptive noise reduction, multi-scale distortion contour recognition and feature fusion are performed on the traveling wave time series, and the accurate fault wavehead location index is output end-to-end.

[0054] Specifically, such as Figure 4 As shown, due to the skin effect and dielectric loss during long-distance propagation, the acquired raw waveform at this time has strong background noise and is accompanied by severe buffer distortion. This step directly feeds the unfiltered one-dimensional raw voltage traveling wave time series data into the preset wavefront identification model.

[0055] First, the wave head recognition model is based on a U-Net architecture, including a front-end feature extraction module, a low-level global feature extraction module, a back-end feature restoration module, and a wave head localization and mapping module. Specifically, it includes:

[0056] (1) Front-end feature extraction module: Receives the input voltage traveling wave time series data. To avoid excessive smoothing of the high-frequency transient change features of the traveling wave front caused by long sequence convolution, this module uses a one-dimensional residual network structure to downsample the input data, filter out high-frequency noise, and extract waveform change features. Furthermore, this module extracts shallow local spatial features that retain accurate temporal resolution from the shallow residual network and outputs these shallow features directly through the cross-layer connection channel, preparing for the feature splicing and fusion at the back end.

[0057] (2) Bottom-level global feature extraction module: Located at the deepest layer of the network architecture, it receives the output from the front-end feature extraction module. This module obtains the deepest, global distortion semantic features of the traveling wave sequence through deep convolution operations. These features contain the attenuation and distortion information formed after the traveling wave has traveled a long distance, which can help the model to lock the approximate wavefront occurrence range under interference.

[0058] (3) Backend Feature Restoration Module: Receives deep features from the underlying global feature extraction module and performs layer-by-layer upsampling on these features to achieve restoration. During the restoration process, it receives shallow local spatial features from the bypass output of the frontend feature extraction module and performs feature splicing and fusion. Furthermore, to suppress environmental electromagnetic noise carried in the shallow features, an attention gating mechanism is introduced, using deep macroscopic features as guiding signals to purify the shallow local spatial features and eliminate the positioning ambiguity caused by waveform distortion and the superposition of ambient noise.

[0059] (4) Wavehead localization and mapping module: This module receives the fused features output by the backend feature restoration module. Through a fully connected layer and a classification activation function, this module maps the fused features to a time-series mutation probability distribution sequence. The final model directly searches on the time axis and outputs the precise wavehead position with the highest mutation probability as the localization index.

[0060] As one embodiment, this disclosure utilizes a one-dimensional residual U-Net architecture with an integrated attention mechanism. It introduces a one-dimensional residual module to address the problem in standard convolutional networks where the "high-frequency abrupt change characteristics" of the traveling wavefront are excessively smoothed as the number of layers increases, leading to blurred wavefront origins. An attention gating mechanism is introduced to address the issue in standard U-Net where strong electromagnetic background noise carried in the shallow, original traveling wavefront is directly copied to the back end during skip connections, thus achieving "denoising and purification" of the time scale.

[0061] The one-dimensional raw voltage traveling wave time series data obtained in step 1 is input into the U-Net front-end feature extraction module of the wavefront recognition model. The ordinary one-dimensional convolutional layers in the front-end multi-scale feature extractor are replaced with one-dimensional residual modules. Each residual module contains two cascaded one-dimensional convolutional layers, and a cross-layer identity mapping path is added between the module input and output. After the traveling wave time series data is input into the encoder, the feature mapping formula for the data in the l-th residual module is improved as follows:

[0062]

[0063] in, These are the input features for the previous layer. , and , These are the weight matrix and bias vector of the two convolutional layers within the residual module, respectively. It is a non-linear activation function, and the subsequent... This is the identity mapping operation. While extracting the macroscopic distortion profile of the traveling wave, this data processing procedure forcibly preserves the extremely weak "step change" high-frequency components in the original signal through identity mapping, preventing the attenuation of wavefront features caused by network deepening.

[0064] Secondly, an attention gating unit is connected in series in the originally directly connected cross-layer feature splicing channels. This unit simultaneously receives micro-features from the shallow front-end layers and macro-features from the deep layers of the network. The former carries high temporal resolution but contains noise, while the latter has a global perspective but low resolution.

[0065] In the l-th layer of the decoder, let the shallow front-end features be... The gated signal transmitted from the deep upsampling at the back end is The attention gating unit first calculates the spatial attention weight coefficient matrix. for:

[0066]

[0067] in, These represent the weight matrices for linear transformation of shallow features and gated signals, respectively. This represents the convolution kernel matrix and bias term used inside the attention gating module to calculate the final weights; These represent different nonlinear activation functions (usually...) Use the ReLU function. (Using the Sigmoid function)

[0068] Subsequently, the shallow features are weighted and activated using this weight matrix to obtain the purified shallow features.

[0069]

[0070] at last, These are shallow features; the purified shallow features... The deep gating signal (deep features) from the backend is spliced ​​and fused along the channel dimension to finally output a fused feature map. Represented as:

[0071]

[0072] Deep Gated Signals It indicates the approximate range of wavefront occurrence, and the weights are calculated using a formula. It automatically suppresses high-frequency electromagnetic noise in non-wavehead regions of shallow features, allowing only true wavehead abrupt change regions to pass through the channel, thus improving anti-interference capabilities.

[0073] Furthermore, in the wavefront localization and mapping module, the fused feature map is processed by the convolutional layer at the end and the Softmax classifier to calculate the mutation probability. The calculation process is as follows:

[0074] First, a convolutional layer is used to compress the multi-channel feature map into a single-channel time-series logical value. The calculation formula is as follows:

[0075]

[0076] in, The weight vector of the terminal mapping layer. For the fused feature map For end bias, Given a one-dimensional sequence of data points of length T, where T is the total length of the time series data points, each element in the sequence... Z i This represents the unnormalized score at which a sudden change in the wavefront occurs at that moment.

[0077] Subsequently, the logical value sequence Z is input into a Softmax classifier, which maps it to a mutation probability distribution sequence P over time. For the i-th sampling point in the sequence, its probability as the arrival time of the true fault wavefront is... P i The calculation formula is:

[0078]

[0079] Ultimately, each element in the sequence Z i The unnormalized score representing the wavefront abrupt change at that moment. T The total length of the data points. Represents the first in the sequence j The unnormalized score of each sampling point where a sudden change in wavefront occurs is used to directly retrieve the index coordinates of the maximum probability value on the generated one-dimensional probability distribution sequence P. The precise wavefront position index, Index, is output through the Argmax function. The localization formula is as follows:

[0080]

[0081] The output index value is the location of the data point with the highest probability of mutation in the entire time series, which is the precise index value of the arrival time of the actual fault wavefront.

[0082] Step 3: Based on the fault wavefront position index, convert the actual arrival times of the traveling wave at both ends, and use the traveling wave ranging mathematical model to calculate the absolute fault distance at both ends.

[0083] Specifically, after extracting the precise wavefront index, the centralized data processing host, in conjunction with the sampling frequency fs of the data acquisition card and the trigger zero-point time T0, converts the discrete wavefront position index values ​​into absolute physical time. The specific conversion relationship is as follows:

[0084] Absolute arrival time at the beginning: tm = T0 + Indexm / fs;

[0085] Terminal absolute arrival time: tn = T0 + Indexn / fs.

[0086] Where tm and tn are the converted absolute arrival times of the first and last ends, T0 is the trigger zero point time, fs is the sampling frequency, and Indexm and Indexn are the wavefront position index values ​​of the first and last ends, respectively.

[0087] Subsequently, by substituting the absolute arrival time difference of the traveling waves at the beginning and end into the classic two-end ranging physical formula, the absolute physical distance D from the measuring end to the actual fault breakdown point is calculated:

[0088]

[0089] in, D The distance from the measuring end to the fault point. L This represents the actual total length of the pipeline between the sensors at both ends. v The propagation speed of electromagnetic traveling waves in the gas inside a GIS. tm and tn These are the converted absolute arrival times of the first and last ends, respectively.

[0090] Step 4: Obtain the location of the fault breakdown point based on the absolute fault distance at both ends, thus achieving a high-precision GIS breakdown point location process.

[0091] Specifically, after completing the physical conversion of distance, the final positioning result is output, and the ranging model is evaluated and optimized in a closed loop.

[0092] In the actual online monitoring application of the system, the centralized data processing host directly outputs the physical distance D calculated in the above steps as the final fault breakdown point location result through a visual interface to guide on-site troubleshooting.

[0093] Meanwhile, during the offline training and finalization phase of the wave head recognition model disclosed in this paper, the system needs to calculate the predicted distance. D pred Compared with a large number of known real fault breakdown locations D true By comparing and calculating, the absolute physical positioning error is obtained. .

[0094] This disclosure uses the positioning error as the core feedback criterion for the loss function, continuously calculates the network gradient through the backpropagation algorithm, and iteratively adjusts all weights and bias parameters within the wavehead recognition model. When the positioning error on the overall training set is lower than the set accuracy requirements for power grid engineering measurements, the model is considered to have converged and reached its final form, thereby achieving online, fixed deployment within a centralized processing host.

[0095] Example 2

[0096] One embodiment of this disclosure provides a UHV GIS breakdown point location system based on voltage wavefront identification, comprising:

[0097] The signal acquisition module is used to acquire the transient voltage traveling wave generated when GIS breaks down and to construct the time series data of the original voltage traveling wave at both ends.

[0098] The wavefront recognition module is used to input traveling wave time series data into the wavefront recognition model. Through the multi-scale feature extraction and cross-layer feature splicing channel at the front end of the wavefront recognition model, adaptive noise reduction, multi-scale distortion contour recognition and feature fusion are performed on the traveling wave time series, and the accurate fault wavefront location index is output end-to-end.

[0099] The distance conversion module is used to convert the actual arrival time of the traveling wave at both ends based on the fault wavehead position index, and to calculate the absolute fault distance at both ends using the traveling wave ranging mathematical model.

[0100] The positioning feedback module is used to obtain the location of the fault breakdown point based on the absolute fault distance at both ends, thereby realizing a high-precision GIS breakdown point positioning process.

[0101] Example 3

[0102] One embodiment of this disclosure provides a computer program product, including a computer program that, when executed by a processor, implements the UHV GIS breakdown point location method based on voltage wavefront identification.

[0103] Example 4

[0104] One embodiment of this disclosure provides a non-transitory computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, they implement the method for locating breakdown points in ultra-high voltage GIS based on voltage wavefront identification.

[0105] Example 5

[0106] One embodiment of this disclosure provides an electronic device, including a processor, a memory, and a computer program; wherein the processor is connected to the memory, and the computer program is stored in the memory. When the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to implement the UHV GIS breakdown point location method based on voltage wavefront identification.

[0107] This disclosure is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0108] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0109] While the specific embodiments of this disclosure have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of this disclosure. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of this disclosure are still within the scope of protection of this disclosure.

Claims

1. A method for locating breakdown points in ultra-high voltage GIS based on voltage wavefront identification, characterized in that, include: Acquire the transient voltage traveling wave generated when GIS breaks down, and construct the time series data of the original voltage traveling wave at both ends; Traveling wave time series data is input into a wavefront recognition model. Through the front-end multi-scale feature extraction and cross-layer feature concatenation channel of the wavefront recognition model, adaptive noise reduction, multi-scale distortion contour recognition, and feature fusion are performed on the traveling wave time series, outputting an accurate fault wavefront location index end-to-end. The wavefront recognition model is a U-Net architecture. The original voltage traveling wave time series data first enters the front-end multi-scale feature extractor for front-end downsampling feature extraction. The ordinary one-dimensional convolutional layer in the front-end multi-scale feature extractor is replaced with a one-dimensional residual module. Each residual module contains two cascaded one-dimensional convolutional layers, and a cross-layer identity mapping path is added between the module input and output. After the traveling wave time series data is input into the encoder, it is processed in the l-th layer residual module. In the process, feature mapping of the data is performed. In the originally directly connected cross-layer feature splicing channels, an attention gating unit is connected in series. This unit simultaneously receives micro-features from the shallow front layer and macro-features from the deep network layer, and splices and fuses them in the channel dimension. The spliced ​​and fused feature map is then passed to the convolutional layer at the end and the Softmax classifier to calculate the mutation probability mapping. The convolutional layer compresses the multi-channel feature map into a single-channel time series logical value. The logical value sequence Z is input into the Softmax classifier, which maps it to a mutation probability distribution sequence on the time series. Finally, the index coordinates of the maximum probability value are directly retrieved on the generated one-dimensional probability distribution sequence, and the precise wavefront position index is output through the Argmax function. Based on the fault wavefront position index, the actual arrival times of the traveling wave at both ends are converted, and the absolute fault distance at both ends is calculated using a traveling wave ranging mathematical model; the process of converting the actual arrival times of the traveling wave at both ends based on the fault wavefront position index and calculating the absolute fault distance at both ends using a traveling wave ranging mathematical model includes: Combining the sampling frequency of the data acquisition card and the trigger zero point time, the discrete wavefront position index values ​​are converted into absolute physical time. The specific conversion relationship is as follows: the absolute arrival time of the first end tm = T0 + Indexm / fs, and the absolute arrival time of the last end tn = T0 + Indexn / fs, where tm and tn are the converted absolute arrival times of the first and last ends, respectively, T0 is the trigger zero point time, fs is the sampling frequency, and Indexm and Indexn are the wavefront position index values ​​of the first and last ends, respectively. Subsequently, by substituting the absolute arrival time difference of the traveling waves at the beginning and end into the classic double-end ranging physical formula, the absolute fault distance from the measuring end to the actual fault breakdown point is calculated. The location of the breakdown point is obtained based on the absolute fault distance at both ends, thus achieving a high-precision GIS breakdown point location process.

2. The method for locating breakdown points in UHV GIS based on voltage wavefront identification as described in claim 1, characterized in that, The acquisition of the transient voltage traveling wave generated during GIS breakdown and the construction of the original voltage traveling wave time series data at both ends include: The transient voltage traveling wave signal is coupled in real time using electro-optic crystal sensors installed at both ends of the pipeline; After receiving the unified time reference signal triggered by the high-precision synchronous timing module, the multi-channel high-speed data acquisition card performs synchronous analog-to-digital conversion on the dual-end transient voltage traveling wave signal at a preset high-frequency sampling rate. Discrete data points within a set time window before and after the trigger moment are extracted to form a one-dimensional original voltage traveling wave time series data.

3. The method for locating breakdown points in UHV GIS based on voltage wavefront identification as described in claim 1, characterized in that, The process of obtaining the location of the fault breakdown point based on the absolute fault distance at both ends, thereby achieving high-precision GIS breakdown point positioning, includes: The absolute fault distance is used as the final fault breakdown point location result and output to the visualization interface. At the same time, during the offline training and modeling stage, the predicted distance is compared with a large number of known real fault breakdown locations to calculate the absolute physical location error. This location error is used as the core feedback judgment index of the loss function. The network gradient is continuously calculated through the backpropagation algorithm, and all weights and bias parameters inside the wave head recognition model are iteratively adjusted.

4. A UHV GIS breakdown point location system based on voltage wavefront identification, characterized in that, Specifically, the method for locating the breakdown point of ultra-high voltage GIS based on voltage wavefront identification as described in any one of claims 1-3 includes: The signal acquisition module is used to acquire the transient voltage traveling wave generated when GIS breaks down and to construct the time series data of the original voltage traveling wave at both ends. The wavefront recognition module is used to input traveling wave time series data into the wavefront recognition model. Through the multi-scale feature extraction and cross-layer feature splicing channel at the front end of the wavefront recognition model, adaptive noise reduction, multi-scale distortion contour recognition and feature fusion are performed on the traveling wave time series, and the accurate fault wavefront location index is output end-to-end. The distance conversion module is used to convert the actual arrival time of the traveling wave at both ends based on the fault wavehead position index, and to calculate the absolute fault distance at both ends using the traveling wave ranging mathematical model. The positioning feedback module is used to obtain the location of the fault breakdown point based on the absolute fault distance at both ends, thereby realizing a high-precision GIS breakdown point positioning process.

5. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the method for locating the breakdown point of ultra-high voltage GIS based on voltage wavefront identification as described in any one of claims 1-3.

6. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium is used to store computer instructions, which, when executed by a processor, implement the method for locating ultra-high voltage GIS breakdown points based on voltage wavefront identification as described in any one of claims 1-3.

7. An electronic device, characterized in that, include: The device includes a processor, a memory, and a computer program; wherein the processor is connected to the memory, and the computer program is stored in the memory. When the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to perform the method for locating the breakdown point of ultra-high voltage GIS based on voltage wavefront identification as described in any one of claims 1-3.

Images