Distribution network hidden danger grounding fault detection method, device, system, equipment and medium

Abstract

The invention discloses a distribution network hidden danger grounding fault detection method, device, system and equipment and a medium, and relates to the technical field of distribution network fault detection, and the method comprises the steps: collecting a grounding fault current magnetic field signal through a preset high-frequency current transformer, carrying out the signal processing of the grounding fault current magnetic field signal based on a signal conditioning circuit, and obtaining a processed signal; transmitting the processed signal to a main control module through a digital interface circuit, and determining a filtered signal according to the processed signal through the main control module based on a Sage-Husa adaptive Kalman filtering algorithm; extracting a target feature of the filtered signal through a main control module to obtain a target feature vector, and performing similarity matching on the target feature vector and a feature vector in a preset feature library to obtain a corresponding similarity score; and the main control module determines a distribution network grounding fault type corresponding to the distribution network line based on the similarity score, and alarms the distribution network line according to the fault type. According to the invention, accurate identification of faults is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of distribution network fault detection, in particular to a distribution network hidden ground fault detection method, device, system, equipment and medium. BACKGROUND

[0002] As the terminal link of power transmission, the distribution network system has a wide coverage and a complex environment, and hidden ground faults occur frequently. The traditional ground fault detection device has the following defects: first, the detection precision is low, and it is difficult to capture weak characteristic current signals; second, the signal conditioning adaptability is poor, and it cannot be compatible with different amplitude and frequency fault currents; third, the fault type recognition is single, and it is difficult to distinguish complex hidden dangers such as intermittent arc grounding and high resistance grounding; fourth, the data transmission efficiency is low, and the main station computing power is large due to the lack of edge computing; fifth, the installation adaptability is weak, and the protection level is insufficient, which cannot meet the use requirements in outdoor harsh environments.

[0003] In some prior art, the aircraft area is divided into three categories and connected with the to-be-tested cable after testing using three different connection methods, solving the problem that the use of only one connection method of distributed test box + process adapter cable makes some process adapter cables longer and more concentrated, resulting in a longer test preparation process, and improving the test efficiency. However, the above prior art only mentions the use of Kalman filtering, but does not solve the problem of fixed parameters and poor adaptability in complex noise environments. Therefore, how to accurately detect the distribution network ground fault in a complex noise environment is a problem to be solved at present. SUMMARY

[0004] Therefore, the purpose of the present application is to provide a distribution network hidden ground fault detection method, device, system, equipment and medium, which can realize dynamic matching of time-varying noise environment, has better de-bouncing and anti-burst interference ability compared with fixed parameter Kalman filtering, improves the signal-to-noise ratio from the signal source, and significantly improves the recognition accuracy and classification reliability of complex faults and weak faults. The specific scheme is as follows:

[0005] In a first aspect, the present application discloses a distribution network hidden ground fault detection method, comprising:

[0006] The ground fault current magnetic field signal of the distribution network line is collected by a pre-set high-frequency current transformer, and the ground fault current magnetic field signal is processed by a signal conditioning circuit to obtain a processed signal.

[0007] The processed signal is transmitted to a master module through a digital interface circuit, process noise covariance matrix and observation noise covariance matrix are updated based on a Sage-Husa adaptive Kalman filtering algorithm through the master module, and a filtered signal is determined according to the updated process noise covariance matrix, the updated observation noise covariance matrix and the processed signal; the master module is a micro control unit or a system level chip;

[0008] Target features of the filtered signal are extracted through the master module to obtain a corresponding target feature vector, and the target feature vector is weighted according to characteristics of the filtered signal to obtain a corresponding weighted feature vector, and the weighted feature vector is matched with feature vectors in a preset feature library to obtain a corresponding similarity score;

[0009] The master module determines a distribution network line corresponding distribution network grounding fault type based on the similarity score, and alarms the distribution network line according to the distribution network grounding fault type.

[0010] Optionally, the signal conditioning circuit includes an instrument amplifier, a low-pass filter and a voltage follower;

[0011] Correspondingly, the signal conditioning circuit processes the grounding fault current magnetic field signal to obtain a processed signal, including:

[0012] The instrument amplifier amplifies the grounding fault current magnetic field signal by a target multiple to obtain an amplified current signal;

[0013] The low-pass filter performs high-frequency filtering on the amplified current signal based on a preset cutoff frequency to obtain a filtered current signal;

[0014] The voltage follower converts the filtered current signal into an analog signal;

[0015] An analog-to-digital converter converts the analog signal into a digital signal to obtain the processed signal.

[0016] Optionally, the master module updates process noise covariance matrix and observation noise covariance matrix based on a Sage-Husa adaptive Kalman filtering algorithm, and determines a filtered signal according to the updated process noise covariance matrix, the updated observation noise covariance matrix and the processed signal, including:

[0017] The current state estimation value of the current signal is determined according to the product of the last time state estimation value of the current signal and a state transition matrix;

[0018] The state covariance matrix determination formula at the current moment determines the state covariance matrix of the current moment of the current signal based on the state transition matrix and the state covariance matrix of the last moment; the state covariance matrix determination formula at the current moment is:

[0019] ;

[0020] Wherein, The state covariance matrix at the current moment of the current signal; F is the state transition matrix; The state covariance matrix of the last moment of the current signal; The transpose matrix of the state transition matrix; The process noise covariance matrix of the last moment;

[0021] The Kalman gain determination formula determines the Kalman gain according to the state covariance matrix of the current moment of the current signal and the observation matrix corresponding to the processed signal; the Kalman gain determination formula is:

[0022] ;

[0023] Wherein, The Kalman gain at the current moment; The state covariance matrix at the current moment; H is the observation matrix corresponding to the processed signal; The transpose matrix of the observation matrix corresponding to the processed signal; inv is the inverse operation of the matrix; The observation noise covariance matrix of the last moment;

[0024] The updated state estimation value determination formula updates the state estimation value at the current moment based on the Kalman gain to obtain the updated state estimation value; the updated state estimation value determination formula is:

[0025] ;

[0026] Wherein, The updated state estimation value; The state estimation value at the current moment; The Kalman gain at the current moment; H is the observation matrix corresponding to the processed signal; The processed signal;

[0027] The updated state covariance matrix determination formula updates the state covariance matrix at the current moment based on the Kalman gain to obtain the updated state covariance matrix; the updated state covariance matrix determination formula is:

[0028] ;

[0029] wherein, is the updated state covariance matrix; I is an identity matrix; is the Kalman gain at the current time; is the state covariance matrix at the current time; H is an observation matrix corresponding to the processed signal;

[0030] updating, by a Sage-Husa algorithm, the process noise covariance matrix and the observation noise covariance matrix based on the updated state estimate value and the updated state covariance matrix to obtain a corresponding updated process noise covariance matrix and an updated observation noise covariance matrix;

[0031] determining the Kalman gain at the current time and the state estimate value at the current time according to the updated process noise covariance matrix and the updated observation noise covariance matrix;

[0032] obtaining the updated state estimate value according to the Kalman gain at the current time and the state estimate value at the current time;

[0033] judging whether there is a new processed signal;

[0034] if there is, then jumping back to the step of determining the state estimate value at the current time of the current signal according to the product of the state estimate value at the last time of the current signal and the state transition matrix;

[0035] if there is not, then ending the filtering to determine the updated state estimate value as the filtered signal.

[0036] Optionally, the extracting, by the main control module, the target feature of the filtered signal to obtain a corresponding target feature vector comprises:

[0037] extracting, by the main control module, a peak value, a frequency and a duty cycle of the filtered signal, and constructing the target feature vector according to the peak value, the frequency and the duty cycle.

[0038] Optionally, before the similarity matching of the weighted feature vector with the feature vectors in the preset feature library, the method further comprises:

[0039] constructing a preset feature library according to the feature vectors of waveforms of various types of distribution network line faults.

[0040] Optionally, the determining, by the main control module, the distribution network line corresponding distribution network grounding fault type based on the similarity score comprises:

[0041] If the similarity score is greater than or equal to a preset threshold, a distribution network line corresponding distribution network grounding fault type is determined according to a feature vector corresponding to the preset feature library.

[0042] If the similarity score is less than the preset threshold, it is determined that the distribution network line has no fault or the distribution network line corresponding distribution network grounding fault type is an unknown type.

[0043] Optionally, the corresponding alarm of the distribution network line according to the distribution network grounding fault type comprises:

[0044] The distribution network grounding fault type, the similarity score and the fault occurrence time are uploaded to the master station through the communication module, so that the master station performs corresponding alarm.

[0045] In a second aspect, the application discloses a distribution network hidden danger grounding fault detection device, comprising:

[0046] A preset high-frequency current transformer is configured to collect a grounding fault current magnetic field signal of a distribution network line.

[0047] A signal conditioning circuit connected with the preset high-frequency current transformer and the analog-to-digital converter module is configured to perform signal processing on the grounding fault current magnetic field signal to obtain a processed signal.

[0048] The analog-to-digital converter module comprises a digital interface circuit, an analog front-end circuit, a conversion circuit and a clock circuit.

[0049] A master control module connected with the analog-to-digital converter module is configured to update a process noise covariance matrix and an observation noise covariance matrix based on a Sage-Husa adaptive Kalman filtering algorithm, determine a filtered signal according to the updated process noise covariance matrix, the updated observation noise covariance matrix and the processed signal, extract a target feature of the filtered signal to obtain a corresponding target feature vector, perform attention feature weighting on the target feature vector according to the characteristics of the filtered signal to obtain a corresponding weighted feature vector, perform similarity matching on the weighted feature vector and a feature vector in a preset feature library to obtain a corresponding similarity score, determine a distribution network line corresponding distribution network grounding fault type based on the similarity score, and perform corresponding alarm of the distribution network line according to the distribution network grounding fault type.

[0050] A communication module connected with the master control module;

[0051] A master station connected with the communication module;

[0052] A power module connected with the preset high-frequency current transformer, the signal conditioning circuit, the analog-to-digital converter module, the main control module and the communication module.

[0053] In a third aspect, the application discloses a distribution network hidden ground fault detection system, comprising:

[0054] A signal processing module is configured to collect a ground fault current magnetic field signal of a distribution network line through a preset high-frequency current transformer, and perform signal processing on the ground fault current magnetic field signal based on a signal conditioning circuit to obtain a processed signal.

[0055] A filtered signal determination module is configured to transmit the processed signal to a main control module through a digital interface circuit, update a process noise covariance matrix and an observation noise covariance matrix based on a Sage-Husa adaptive Kalman filtering algorithm through the main control module, and determine a filtered signal according to the processed signal, the updated process noise covariance matrix and the updated observation noise covariance matrix. The main control module is a micro control unit or a system on chip.

[0056] A similarity matching module is configured to extract target features of the filtered signal through the main control module to obtain a corresponding target feature vector, perform attention feature weighting on the target feature vector according to the characteristics of the filtered signal to obtain a corresponding weighted feature vector, and perform similarity matching on the weighted feature vector and a feature vector in a preset feature library to obtain a corresponding similarity score.

[0057] An alarm module is configured to determine a distribution network ground fault type corresponding to the distribution network line based on the similarity score through the main control module, and perform corresponding alarm on the distribution network line according to the distribution network ground fault type.

[0058] In a fourth aspect, the application discloses an electronic device, comprising:

[0059] A memory is configured to store a computer program.

[0060] A processor is configured to execute the computer program to implement the distribution network hidden ground fault detection method as described above.

[0061] In a fifth aspect, the application discloses a computer readable storage medium, and the computer readable storage medium stores a computer program. When the computer program is executed by a processor, the distribution network hidden ground fault detection method as described above is implemented.

[0062] The application collects the ground fault current magnetic field signal of the distribution network line through a preset high-frequency current transformer, processes the ground fault current magnetic field signal based on a signal conditioning circuit to obtain a processed signal, transmits the processed signal to a master control module through a digital interface circuit, updates a process noise covariance matrix and an observation noise covariance matrix based on a Sage-Husa adaptive Kalman filtering algorithm through the master control module, determines a filtered signal according to the updated process noise covariance matrix, the updated observation noise covariance matrix and the processed signal, the master control module is a micro control unit or a system level chip, extracts a target feature of the filtered signal through the master control module to obtain a corresponding target feature vector, performs attention feature weighting on the target feature vector according to the characteristics of the filtered signal to obtain a corresponding weighted feature vector, and performs similarity matching on the weighted feature vector and the feature vectors in a preset feature library to obtain a corresponding similarity score, determines the distribution network line corresponding distribution network ground fault type based on the similarity score through the master control module, and alarms the distribution network line according to the distribution network ground fault type. It can be seen that the application filters and processes the processed signal based on the Sage-Husa adaptive Kalman filtering algorithm, adjusts the noise covariance matrix in real time, dynamically matches the time-varying noise environment, has better de-bouncing and anti-burst interference ability compared with the fixed parameter Kalman filter, and solves the dynamic noise adaptability problem. And through the multi-dimensional feature extraction and matching of the attention mechanism weighting, the intelligent focusing of the key features of the fault waveform is realized, the signal-to-noise ratio is improved from the signal source, the problem that some features are covered by noise or have insufficient weight under the fixed rule is avoided, and the fault recognition accuracy and classification reliability are significantly improved. BRIEF DESCRIPTION OF DRAWINGS

[0063] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only embodiments of the present application, and those skilled in the art can obtain other drawings according to the provided drawings without creative labor.

[0064] Figure 1 A flow chart of a distribution network hidden ground fault detection method disclosed by the present application;

[0065] Figure 2 A flow chart of a specific distribution network hidden ground fault detection method disclosed by the present application;

[0066] Figure 3 A structural schematic diagram of a distribution network hidden ground fault detection device disclosed by the present application;

[0067] Figure 4 Another structure schematic diagram of the power distribution network hidden ground fault detection device disclosed in the application is disclosed.

[0068] Figure 5 A specific structure schematic diagram of the power distribution network hidden ground fault detection device disclosed in the application is disclosed.

[0069] Figure 6 A detection principle schematic diagram disclosed in the application is disclosed.

[0070] Figure 7 A structure schematic diagram of the power distribution network hidden ground fault detection system disclosed in the application is disclosed.

[0071] Figure 8 An electronic device structure diagram disclosed in the application is disclosed.

[0072] The figure mark: 1, device body; 2, high-frequency current transformer; 3, signal conditioning circuit; 4, analog-to-digital converter module; 5, main control module; 6, communication module; 7, main station; 8, power module. DETAILED DESCRIPTION

[0073] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application.

[0074] The prior art only mentions using Kalman filtering, but does not solve the problems of fixed parameters and poor adaptability in complex noise environment. In order to solve the above technical problems, the application discloses a power distribution network hidden ground fault detection method, device, system, equipment and medium, which can realize dynamic matching of time-varying noise environment, has better de-bouncing and anti-burst interference ability compared with fixed parameter Kalman filtering, improves signal-to-noise ratio from signal source, and significantly improves recognition accuracy and classification reliability of complex faults and weak faults.

[0075] Referring to Figure 1 The embodiment of the application discloses a power distribution network hidden ground fault detection method, which comprises:

[0076] Step S11, the ground fault current magnetic field signal of the power distribution network line is collected through a pre-set high-frequency current transformer, and the ground fault current magnetic field signal is processed based on a signal conditioning circuit to obtain a processed signal.

[0077] In this embodiment, the ground fault current magnetic field signal of the distribution network line is collected by the preset high-frequency current transformer, and then the signal processing circuit is used for signal processing of the ground fault current magnetic field signal. The signal conditioning circuit includes an instrument amplifier, a low-pass filter and a voltage follower. Therefore, when processing the signal, the ground fault current magnetic field signal is amplified by the instrument amplifier by a target multiple to obtain an amplified current signal; the amplified current signal is high-frequency filtered by the low-pass filter based on a preset cutoff frequency to obtain a filtered current signal; the filtered current signal is converted into an analog signal by the voltage follower; and the analog signal is converted into a digital signal by an analog-to-digital converter to obtain the processed signal. In a specific embodiment, when the distribution network line has an intermittent arc grounding fault, the high-frequency current transformer senses the intermittent ground fault current magnetic field signal, and outputs a voltage signal with an amplitude of 10 μV-10 mV; the instrument amplifier of the signal conditioning circuit amplifies the signal by 100 times to 1 mV-1 V, the low-pass filter sets the cutoff frequency to 10 kHz to filter out the high-frequency noise generated by the arc discharge, and the voltage follower outputs a stable analog signal; and the high-precision ADC (Analog To Digital Converter) module converts the analog signal into a 24-bit digital signal at a sampling rate of 50 kSPS (kiloSamples per Second).

[0078] In step S12, the processed signal is transmitted to the host module through the digital interface circuit. The process noise covariance matrix and the observation noise covariance matrix are updated based on the Sage-Husa adaptive Kalman filtering algorithm through the host module. The filtered signal is determined according to the updated process noise covariance matrix, the updated observation noise covariance matrix and the processed signal. The host module is a micro control unit or a system on chip.

[0079] In this embodiment, after the signal is preliminarily processed, the processed signal is transmitted to the host module through the digital interface circuit. The adaptive Kalman filter applies Kalman filtering to the sampled discrete current sequence z(k). The Sage-Husa adaptive algorithm is used to estimate and update the process noise covariance matrix Q(k) and the measurement noise covariance matrix R(k) in real time. Specifically, the current state estimation value at the current time is determined according to the product of the last time state estimation value of the current signal and the state transition matrix. The current state covariance matrix is determined by the current state covariance matrix determination formula based on the state transition matrix and the last time state covariance matrix. The current state covariance matrix determination formula is:

[0080] ;

[0081] in, F is the state covariance matrix of the current signal at the current moment; F is the state transition matrix. Let be the state covariance matrix of the current signal at the previous moment; This is the transpose of the state transition matrix; The process noise covariance matrix is ​​given by the previous time step. The Kalman gain is determined using the Kalman gain determination formula based on the current signal's current state covariance matrix and the observation matrix corresponding to the processed signal. The Kalman gain determination formula is as follows:

[0082] ;

[0083] in, The Kalman gain at the current moment; H is the state covariance matrix at the current moment; H is the observation matrix corresponding to the processed signal; is the transpose of the observation matrix corresponding to the processed signal; inv is the matrix inversion operation; Let be the observation noise covariance matrix of the previous time step; the current time step state estimate is updated based on the Kalman gain using the updated state estimate determination formula to obtain the updated state estimate; the updated state estimate determination formula is:

[0084] ;

[0085] in, This is the updated state estimate; This is the estimated state value at the current moment; is the Kalman gain at the current moment; H is the observation matrix corresponding to the processed signal; The processed signal is used; the state covariance matrix at the current time is updated based on the Kalman gain using the updated state covariance matrix determination formula to obtain the updated state covariance matrix; the updated state covariance matrix determination formula is:

[0086] ;

[0087] in, Let I be the updated state covariance matrix; I is the identity matrix; The Kalman gain at the current moment; the updated state estimation value and the updated state covariance matrix to obtain a corresponding updated process noise covariance matrix and an updated observation noise covariance matrix; determining a Kalman gain at the current time and a state estimation value at the current time according to the updated process noise covariance matrix and the updated observation noise covariance matrix; obtaining the updated state estimation value according to the Kalman gain at the current time and the state estimation value at the current time; determining whether there is a new processed signal; if there is, then jumping back to the step of determining the state estimation value at the current time of the current signal according to the product of the state estimation value at the last time of the current signal and the state transition matrix; if there is not, then ending the filtering to determine the updated state estimation value as the filtered signal. It should be noted that the Sage-Husa adaptive algorithm will first initialize the state estimation, the covariance matrix, the process noise covariance matrix and the observation noise covariance matrix to obtain the initial state estimation value, the initial covariance matrix, the process noise covariance initial value and the observation noise covariance initial value. Thus, when the pulse interference occurs, the Sage-Husa algorithm can quickly perceive the mutation of the measurement innovation and dynamically increase the estimation value of the observation noise covariance matrix R(k). This makes the gain of the Kalman filter decrease, and more rely on the predicted value of the system, thereby significantly suppressing the manifestation of the pulse interference in the output signal, and obtaining a smooth current curve that retains the fault characteristics.

[0088] In step S13, the target feature of the filtered signal is extracted by the main control module to obtain a corresponding target feature vector. According to the characteristic degree of the filtered signal, the target feature vector is weighted to obtain a corresponding weighted feature vector. The weighted feature vector is matched with the feature vector in the preset feature library to obtain a corresponding similarity score.

[0089] In this embodiment, the peak value, frequency and duty cycle of the filtered signal are extracted by the main control module, and the target feature vector is constructed according to the peak value, frequency and duty cycle. And the target feature vector is weighted according to the characteristic degree of the filtered signal to obtain a corresponding weighted feature vector, so as to perform similarity matching between the weighted feature vector and the feature vector in the preset feature library to obtain a corresponding similarity score. The attention mechanism feature weighting is automatically calculated according to the overall characteristics of the current signal, and different weights are given to different feature dimensions. For example, for a fault with rich transient state, the algorithm automatically gives higher weights to "transient energy" and "main resonance frequency"; for a fault with obvious steady state characteristics, higher weights are given to "power frequency component" and "harmonic ratio". The multi-dimensional features are calculated from the filtered signal. At this time, since the pulse interference has been filtered out, the calculated "waveform complexity" and "kurtosis" feature values can more truly reflect the fault itself rather than interference. In a specific embodiment, the attention mechanism calculates that the "main resonance frequency" and "transient energy" features are most critical for diagnosis under the current state, and gives them high weights. The weighted feature vector is matched with the feature library. In this application, the DTW (Dynamic Time Warping) matching is used to calculate the similarity between the attention-weighted feature vector to be diagnosed and the template in the feature library. The DTW algorithm can effectively overcome the slight stretching and deformation of the waveform on the time axis, and improve the matching accuracy. The target feature vector is matched with the feature vector in the preset feature library to obtain a corresponding similarity score. That is, the main control module extracts the multi-dimensional feature vector of the digital signal, such as peak value, frequency, duty cycle, etc. The multi-dimensional feature extraction extracts the multi-dimensional features corresponding to the feature library definition from the filtered signal to form a feature vector to be diagnosed; and matches with the "intermittent arc grounding" feature vector in the built-in feature library. Therefore, before similarity matching, a preset feature library is first constructed according to the feature vectors of the waveforms of various types of distribution network line faults. In this way, not only can the corresponding fault point be warned in advance through the very small leakage current or partial discharge signal collected before the fault occurs, but also the fault can be warned after the distribution network hidden ground fault occurs. The "attention mechanism weighted multi-dimensional feature extraction and matching" method can realize intelligent focusing on the key features of the fault waveform, avoid the problem that some features are covered by noise or have insufficient weight under fixed rules, and significantly improve the recognition accuracy and classification reliability of complex faults and weak faults.

[0090] In step S14, the main control module determines the distribution network grounding fault type corresponding to the distribution network line based on the similarity score, and performs corresponding warning on the distribution network line according to the distribution network grounding fault type.

[0091] In this embodiment, as shown in Figure 2 If the similarity score is greater than or equal to the preset threshold, the corresponding distribution network line fault type of the corresponding distribution network line is determined according to the feature vector in the preset feature library. If the similarity score is less than the preset threshold, it is determined that the distribution network line has no fault or the corresponding distribution network line fault type is unknown. In a specific embodiment, the master control module extracts the multi-dimensional feature vector of the peak value, frequency, duty cycle, etc. of the digital signal, matches it with the "intermittent arc grounding" feature vector in the built-in feature library, the matching degree is 95%, which is greater than the preset threshold, so it is determined as intermittent arc grounding fault. After determining the fault type, the distribution network grounding fault type, the similarity score and the fault occurrence time are uploaded to the master station through the communication module, so that the master station can perform corresponding alarm. Specifically, the master control module compresses the feature data through edge computing, the communication module uploads the fault type, similarity score, occurrence time and other information to the master station through the 4G network, and the master station triggers an alarm prompt after receiving. The master station triggers an audible and visual alarm and pushes an operation and maintenance work order. In this way, adaptive filtering and intelligent feature matching are combined, so that integrated optimization from signal preprocessing to final diagnosis can be realized, and the distribution network hidden danger grounding fault detection device has expert-level diagnostic ability to work stably and reliably in complex field environment.

[0092] In summary, the application collects the grounding fault current magnetic field signal of the distribution network line through the preset high-frequency current transformer, processes the grounding fault current magnetic field signal based on the signal conditioning circuit to obtain a processed signal, transmits the processed signal to a master control module through a digital interface circuit, updates the process noise covariance matrix and the observation noise covariance matrix based on the Sage-Husa adaptive Kalman filtering algorithm through the master control module, determines a filtered signal according to the corresponding updated process noise covariance matrix, the updated observation noise covariance matrix and the processed signal, the master control module is a micro control unit or a system level chip, extracts the target feature of the filtered signal through the master control module to obtain a corresponding target feature vector, performs attention feature weighting on the target feature vector according to the characteristic of the filtered signal to obtain a corresponding weighted feature vector, and performs similarity matching on the weighted feature vector and the feature vector in the preset feature library to obtain a corresponding similarity score, determines the distribution network grounding fault type corresponding to the distribution network line based on the similarity score through the master control module, and alarms the distribution network line according to the distribution network grounding fault type. It can be seen that the application filters and processes the processed signal based on the Sage-Husa adaptive Kalman filtering algorithm, adjusts the noise covariance matrix in real time, dynamically matches the time-varying noise environment, has better de-bouncing and anti-burst interference ability compared with the fixed parameter Kalman filter, and solves the dynamic noise adaptability problem. And through the multi-dimensional feature extraction and matching of the attention mechanism weighting, the intelligent focusing of the key features of the fault waveform is realized, the signal-to-noise ratio is improved from the signal source, the problem that some features are covered by noise or have insufficient weight under the fixed rule is avoided, and the fault recognition accuracy and classification reliability are significantly improved.

[0093] Referring to Figure 3 , Figure 4 and Figure 5 , the embodiment of the application discloses a distribution network hidden danger grounding fault detection device, which comprises a device body 1, and the device body 1 comprises:

[0094] a preset high-frequency current transformer 2 for collecting a grounding fault current magnetic field signal of a distribution network line;

[0095] a signal conditioning circuit 3 connected with the preset high-frequency current transformer 2 and an analog-to-digital converter module 4, for processing the grounding fault current magnetic field signal to obtain a processed signal;

[0096] the analog-to-digital converter module 4, comprising a digital interface circuit, an analog front-end circuit, a conversion circuit and a clock circuit; wherein the digital interface circuit is used for transmitting the processed signal to a master control module;

[0097] A master module 5 connected with the analog-to-digital converter module 4, used for updating the process noise covariance matrix and the observation noise covariance matrix based on the Sage-Husa adaptive Kalman filtering algorithm, determining a filtered signal according to the updated process noise covariance matrix, the updated observation noise covariance matrix and the processed signal, extracting a target feature of the filtered signal to obtain a corresponding target feature vector, performing attention feature weighting on the target feature vector according to the characteristics of the filtered signal to obtain a corresponding weighted feature vector, performing similarity matching between the weighted feature vector and a feature vector in a preset feature library to obtain a corresponding similarity score, determining a distribution network line corresponding distribution network grounding fault type based on the similarity score, and performing corresponding warning on the distribution network line according to the distribution network grounding fault type; the master module is a micro control unit or a system level chip;

[0098] A communication module 6 connected with the master module 5;

[0099] A master station 7 connected with the communication module 6;

[0100] A power supply module 8 connected with the preset high-frequency current transformer, the signal conditioning circuit, the analog-to-digital converter module, the master module and the communication module.

[0101] In the present application, the distribution network hidden danger grounding fault detection device comprises a device body 1, a high-frequency current transformer 2, a signal conditioning circuit 3, a high-precision analog-to-digital converter module 4, a master module 5 and a communication module 6. The signal conditioning circuit 3, the high-precision analog-to-digital converter module 4 and the master module 5 are arranged in the device body 1, the high-frequency current transformer 2 is arranged outside the device body 1, the high-frequency current transformer 2 is connected with the high-precision analog-to-digital converter module 4 through the signal conditioning circuit 3, the high-precision analog-to-digital converter module 4 is connected with the master module 5, and the master module 5 is uploaded to the master station 7 through the communication module 6. It should be noted that the distribution network hidden danger grounding fault detection device in the present application can be a handheld device.

[0102] The communication module is one or a combination of 4G, 5G, or LoRa (Long Range, a wireless communication technology based on spread spectrum modulation) modules, and supports edge computing data preprocessing before transmission. The main control module is an MCU (Microcontroller Unit) or SoC (System on Chip) chip, integrating a non-volatile memory unit for storing a dynamic waveform feature library, which contains multi-dimensional feature vectors of typical faults such as stable grounding, intermittent arc grounding, and high-resistance grounding. The signal conditioning circuit includes an instrumentation amplifier, a low-pass filter, and a voltage follower connected in sequence. The gain of the instrumentation amplifier is adjustable from 10 to 1000 times, and the cutoff frequency of the low-pass filter can be configured between 1kHz and 100kHz. The device body has an embedded integrated structure, and the shell adopts an IP65 (Ingress Protection 65) protection design, which can be directly installed on the distribution network line or tower through a snap-on or bolt-on structure. The device has a built-in power module that provides power to the high-frequency current transformer, signal conditioning circuit, high-precision ADC module, main control module, and communication module. The high-precision analog-to-digital converter module includes an analog front-end circuit, a conversion circuit, a reference and power supply module, a digital interface circuit, and a clock circuit connected in sequence.

[0103] When installing the distribution network grounding fault detection device of this application, select either a snap-on or bolt-on installation method according to the actual situation of the distribution network line. Fix the device body 1 to the pole crossarm or line support, and install the high-frequency current transformer on the distribution network line. Connect the device body to the signal input interface through a shielded cable. Parameter configuration: Configure the instrument amplifier gain (e.g., set a gain of 1000 times for weak signals) and the low-pass filter cutoff frequency (e.g., set a cutoff frequency of 5kHz for power frequency fault current) through the remote or local debugging interface of the main station. Feature library update: Receive the dynamic waveform feature library update file issued by the main station through the communication module and store it in the non-volatile storage unit of the main control module to ensure the timeliness of fault identification.

[0104] Working process of the power distribution network grounding fault detection device:

[0105] When an intermittent arc-to-ground fault occurs in the distribution network line, the high-frequency current transformer senses the intermittently changing ground fault current magnetic field signal and outputs a voltage signal with an amplitude of 10μV-10mV. The instrumentation amplifier in the signal conditioning circuit amplifies the signal by 100 times to 1mV-1V, and the low-pass filter is set to a cutoff frequency of 10kHz to filter out high-frequency noise generated by arc discharge. The voltage follower outputs a stable analog signal. The high-precision ADC module converts the analog signal into a 24-bit digital signal at a sampling rate of 50kSPS and transmits it to the main control module through the digital interface circuit. The main control module extracts multi-dimensional feature vectors such as peak value, frequency, and duty cycle of the digital signal and matches them with the "intermittent arc-to-ground" feature vector in the built-in feature library. If the matching degree reaches more than 95%, it is determined to be an intermittent arc-to-ground fault. The main control module compresses the feature data through edge computing, and the communication module uses a 4G network to upload information such as fault type, feature parameters, and occurrence time to the main station. After receiving the information, the main station triggers an alarm.

[0106] This device can be widely used in 10kV-35kV medium and low voltage distribution network systems to achieve early detection and accurate alarm of potential grounding faults, reduce distribution network outage time, and improve the reliability of power supply. The main control module employs adaptive noise reduction filtering and waveform feature focusing. Its filtering steps include: acquiring zero-sequence current signals, signal preprocessing (preliminary filtering and standardization), adaptive Kalman noise reduction filtering, real-time estimation and correction of the Q and R matrices using the Sage-Husa adaptive algorithm, outputting the optimal estimated signal, constructing a dynamic waveform feature library, multi-dimensional feature extraction (transient energy, dominant frequency component, waveform complexity, etc.), attention mechanism feature weighting, dynamic time warping similarity matching with the feature library, and outputting the fault type and confidence level. Figure 6 As shown, based on the principle of magnetic fields generated by energized conductors, the magnetic field plane generated by the line load current is orthogonal at 90 degrees to the magnetic field plane generated by the grounding current. High-precision current sensors can be used for non-contact, close-range detection of energized towers, thus achieving accurate acquisition of leakage current without electrical contact. This provides a safe and efficient technical means for determining tower status, detecting abnormal current characteristics, and alarming potential faults.

[0107] In a specific embodiment of the application of the device of this application, the component selection and parameters are as follows:

[0108] High-frequency current transformer: A Rogowski coil transformer with a bandwidth of 10Hz~1MHz is selected, with a transformation ratio accuracy of [missing information]. It can sense zero-sequence current signals in the distribution network line from microamps to amperes, and the output signal is transmitted to the device body through a shielded cable.

[0109] In the signal conditioning circuit, the instrumentation amplifier is AD8421, with an adjustable gain range of 10 to 1000 times, and supports remote gain configuration via the main control module (such as setting a gain of 500 times for weak arc signals).

[0110] Low-pass filter: An active filter circuit is constructed using OPA847, and the cutoff frequency can be configured between 1kHz and 100kHz (in this embodiment, it is set to 10kHz to filter out high-frequency noise from the electric arc).

[0111] Voltage follower: The LMV324 is selected to achieve signal isolation and buffering, ensuring that the conditioned signal is stably input to the ADC module.

[0112] High-precision ADC module: Uses ADS1256 chip, 24-bit conversion accuracy, 50kSPS sampling rate, built-in analog front-end anti-aliasing filter circuit, and paired with REF5040 high-precision reference source (temperature drift). This ensures the accuracy of signal digitization.

[0113] Main control module: adopts STM32H743SoC chip, integrates 2MB Flash (stores dynamic waveform feature library, including 12 types of fault feature vectors such as intermittent arc grounding and high resistance grounding) and 1MB RAM, and supports hardware floating-point operation to accelerate feature matching.

[0114] Communication module: It adopts the combination of "EC200S4G module + SX1278LoRa module" to support edge computing data preprocessing (such as compressing fault feature data by ≥80% before transmission).

[0115] Device body and power module: The device has an embedded integrated structure with IP65 protection, a built-in wide input (9~36V) linear regulator, and outputs 3.3V, 5V and 12V multiple power supplies with ripple less than or equal to 10mV.

[0116] During physical installation, a bolt-type installation structure is adopted to fix the device body to the crossarm of the 10kV distribution network tower. The high-frequency current transformer is sleeved on the A-phase conductor of the overhead distribution line and connected to the signal input interface of the device body through a shielded cable.

[0117] When configuring parameters, remotely log in to the main station system and send configuration commands to the device through the communication module: set the instrumentation amplifier gain to 500 times, the low-pass filter cutoff frequency to 10kHz, and the ADC sampling rate to 50kSPS.

[0118] The dynamic waveform feature library is updated synchronously to ensure that the feature vector (including transient energy, main frequency distribution, waveform complexity and other multi-dimensional parameters) of "intermittent arc grounding" is the latest version.

[0119] Then, the first step is signal acquisition: when the distribution network line experiences intermittent arc grounding due to insulation aging, the high-frequency current transformer senses the intermittently changing zero-sequence current signal and outputs a voltage signal with an amplitude of 20μV~5mV.

[0120] Next, the signal is conditioned: the signal is amplified 500 times by the AD8421 instrumentation amplifier (the amplitude becomes 10mV~2.5V), then the high-frequency interference generated by the electric arc is filtered out by the 10kHz low-pass filter, and finally the standardized analog signal is output through the LMV324 voltage follower.

[0121] Secondly, analog-to-digital conversion: The ADS1256 module converts the analog signal into a 24-bit digital signal at a rate of 50kSPS, and transmits it to the STM32H743 main control module through the SPI interface.

[0122] Then, fault identification is performed: the main control module extracts the feature vectors such as the peak value, duty cycle, and frequency domain energy distribution of the digital signal, and matches them with the feature vector of "intermittent arc grounding" in the built-in dynamic waveform feature library, with a matching degree of 98%.

[0123] Finally, the data is uploaded: The main control module performs edge computing preprocessing on the fault characteristic data (the data volume is reduced by 85% after compression), and uploads information such as "intermittent arc grounding fault, confidence level 98%, occurrence time ×××" to the main station through the 4G module. The main station then triggers an audible and visual alarm and pushes a maintenance work order.

[0124] This embodiment achieves accurate detection and real-time alarm of intermittent arc grounding faults in distribution networks through the coordinated work of various modules, and verifies the reliability and practicality of the device in complex fault scenarios. It can be widely promoted to the investigation of grounding fault hazards in 10kV~35kV medium and low voltage distribution network systems.

[0125] In this way, this application achieves accurate acquisition and conversion of weak characteristic current signals through the combination of a high-frequency current transformer, an adjustable gain instrumentation amplifier, and a high-precision analog-to-digital converter module, capable of capturing fault currents as small as microamperes. It features a built-in dynamic waveform feature library, supporting the identification of various typical faults such as stable grounding, intermittent arc grounding, and high-resistance grounding, solving the problem of single-identification in traditional devices. The gain and filtering frequency of the signal conditioning circuit can be flexibly configured to adapt to grounding fault current magnetic field signals of different amplitudes and frequencies. The device itself has an IP65 protection rating, supports multiple installation methods, and is suitable for harsh outdoor environments. The communication module supports multiple protocol combinations, combined with edge computing preprocessing functions, reducing data transmission volume, improving real-time transmission performance, and reducing the pressure on the main station. Its embedded integrated structure design is compact, and the snap-on / bolt-type installation method eliminates the need for complex construction, facilitating later maintenance.

[0126] See Figure 7As shown in the figure, an embodiment of the present invention discloses a distribution network hidden grounding fault detection system, comprising:

[0127] Signal processing module 11 is used to acquire the ground fault current magnetic field signal of the distribution network line through a preset high-frequency current transformer, and to process the ground fault current magnetic field signal based on the signal conditioning circuit to obtain the processed signal.

[0128] The filtered signal determination module 12 is used to transmit the processed signal to the main control module through a digital interface circuit. The main control module updates the process noise covariance matrix and the observation noise covariance matrix based on the Sage-Husa adaptive Kalman filtering algorithm, and determines the filtered signal based on the updated process noise covariance matrix, the updated observation noise covariance matrix, and the processed signal. The main control module is a microcontroller unit or a system-on-a-chip.

[0129] The similarity matching module 13 is used to extract the target features of the filtered signal through the main control module to obtain the corresponding target feature vector, perform attention feature weighting on the target feature vector according to the characteristics of the filtered signal to obtain the corresponding weighted feature vector, and perform similarity matching between the weighted feature vector and the feature vector in the preset feature library to obtain the corresponding similarity score.

[0130] The alarm module 14 is used to determine the type of distribution network grounding fault corresponding to the distribution network line through the main control module based on the similarity score, and to issue corresponding alarms to the distribution network line according to the type of distribution network grounding fault.

[0131] In summary, this application acquires the ground fault current magnetic field signal of the distribution network line through a preset high-frequency current transformer, processes the ground fault current magnetic field signal based on a signal conditioning circuit to obtain a processed signal, and transmits the processed signal to the main control module through a digital interface circuit. The main control module updates the process noise covariance matrix and the observation noise covariance matrix based on the Sage-Husa adaptive Kalman filtering algorithm, and determines the filtered signal based on the updated process noise covariance matrix, the updated observation noise covariance matrix, and the processed signal. The control module is a microcontroller unit or a system-on-a-chip. The main control module extracts the target features of the filtered signal to obtain a corresponding target feature vector. Based on the characteristics of the filtered signal, the target feature vector is weighted by attention features to obtain a weighted feature vector. The weighted feature vector is then matched with feature vectors in a preset feature library to obtain a similarity score. The main control module determines the distribution network grounding fault type corresponding to the distribution network line based on the similarity score and issues corresponding alarms to the distribution network line according to the distribution network grounding fault type. Therefore, this application uses the Sage-Husa adaptive Kalman filter algorithm to filter the processed signal, adjusting the noise covariance matrix in real time to achieve dynamic matching to time-varying noise environments. Compared to fixed-parameter Kalman filtering, it has superior jitter reduction and anti-sudden interference capabilities, solving the problem of dynamic noise adaptability. Furthermore, through attention-weighted multidimensional feature extraction and matching, intelligent focusing on key features of fault waveforms is achieved, improving the signal-to-noise ratio from the source of the signal and avoiding the problem that some features are masked by noise or have insufficient weight under fixed rules, thus significantly improving the accuracy of fault identification and the reliability of classification.

[0132] In some specific embodiments, the signal processing module 11 can be used to amplify the ground fault current magnetic field signal by a target factor based on the instrumentation amplifier to obtain an amplified current signal; to perform high-frequency filtering on the amplified current signal based on a preset cutoff frequency using the low-pass filter to obtain a filtered current signal; to convert the filtered current signal into an analog signal using the voltage follower; and to convert the analog signal into a digital signal using an analog-to-digital converter to obtain the processed signal.

[0133] In some specific embodiments, the filtered signal determination module 12 can be used to determine the current state estimate of the current signal based on the product of the previous state estimate and the state transition matrix; and to determine the current state covariance matrix of the current signal based on the state transition matrix and the previous state covariance matrix using the current state covariance matrix determination formula; the current state covariance matrix determination formula is as follows:

[0134] ;

[0135] in, F is the state covariance matrix of the current signal at the current moment; F is the state transition matrix. Let be the state covariance matrix of the current signal at the previous moment; This is the transpose of the state transition matrix; The process noise covariance matrix of the previous time step;

[0136] The Kalman gain is determined using the Kalman gain determination formula based on the state covariance matrix of the current signal at the current moment and the observation matrix corresponding to the processed signal; the Kalman gain determination formula is:

[0137] ;

[0138] in, The Kalman gain at the current moment; H is the state covariance matrix at the current moment; H is the observation matrix corresponding to the processed signal; is the transpose of the observation matrix corresponding to the processed signal; inv is the matrix inversion operation; This is the observation noise covariance matrix of the previous time step;

[0139] The updated state estimate is determined by updating the current state estimate based on the Kalman gain using the updated state estimate determination formula; the updated state estimate determination formula is as follows:

[0140] ;

[0141] in, This is the updated state estimate; This is the estimated state value at the current moment; is the Kalman gain at the current moment; H is the observation matrix corresponding to the processed signal; The processed signal;

[0142] The updated state covariance matrix is ​​determined by updating the current state covariance matrix based on the Kalman gain using the formula for determining the updated state covariance matrix; the formula for determining the updated state covariance matrix is ​​as follows:

[0143] ;

[0144] in, Let I be the updated state covariance matrix; I is the identity matrix; The Kalman gain at the current moment; H is the state covariance matrix at the current moment; H is the observation matrix corresponding to the processed signal;

[0145] The Sage-Husa algorithm is used to update the process noise covariance matrix and observation noise covariance matrix based on the updated state estimate and the updated state covariance matrix, to obtain the corresponding updated process noise covariance matrix and updated observation noise covariance matrix. The Kalman gain and the current state estimate are determined based on the updated process noise covariance matrix and the updated observation noise covariance matrix. An updated state estimate is obtained based on the current Kalman gain and the current state estimate. It is then determined whether a new processed signal exists. If it does, the process jumps back to the step of determining the current state estimate of the current signal based on the product of the previous state estimate and the state transition matrix. If it does not exist, the filtering process ends, and the updated state estimate is determined as the filtered signal.

[0146] In some specific embodiments, the similarity matching module 13 can be used to extract the peak value, frequency and duty cycle of the filtered signal through the main control module, and construct the target feature vector based on the peak value, the frequency and the duty cycle.

[0147] In some specific embodiments, the device can also be used to construct a preset feature library based on the feature vectors of waveforms of various types of distribution network line faults.

[0148] In some specific embodiments, the alarm module 14 can be used to determine the distribution network grounding fault type corresponding to the distribution network line based on the distribution network line fault type corresponding to the feature vector in the preset feature library if the similarity score is greater than or equal to a preset threshold; if the similarity score is less than the preset threshold, it is determined that the distribution network line has no fault or the distribution network grounding fault type corresponding to the distribution network line is an unknown type.

[0149] In some specific embodiments, the device can also be used to upload the distribution network grounding fault type, the similarity score, and the fault occurrence time to the master station via a communication module, so that the master station can issue corresponding alarms.

[0150] Furthermore, embodiments of this application also disclose an electronic device, Figure 8 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content of the diagram should not be construed as limiting the scope of this application.

[0151] Figure 8This is a schematic diagram of the structure of an electronic device 20 provided in an embodiment of this application. Specifically, the electronic device 20 may include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 stores a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the distribution network hidden grounding fault detection method disclosed in any of the foregoing embodiments. Furthermore, the electronic device 20 in this embodiment may specifically be a computer.

[0152] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and is not specifically limited here; the input / output interface 25 is used to acquire external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.

[0153] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk or optical disk, etc. The resources stored thereon can include operating system 221, computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0154] The operating system 221 is used to manage and control the various hardware devices on the electronic device 20 and the computer program 222, which may be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program capable of performing the distribution network hidden grounding fault detection method executed by the electronic device 20 as disclosed in any of the foregoing embodiments, the computer program 222 may further include a computer program capable of performing other specific tasks.

[0155] Furthermore, this application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the aforementioned method for detecting potential grounding faults in distribution networks. Specific steps of this method can be found in the corresponding content disclosed in the foregoing embodiments, and will not be repeated here.

[0156] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0157] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0158] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0159] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0160] The technical solutions provided in this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for detecting potential grounding faults in a distribution network, characterized in that, include: The ground fault current magnetic field signal of the distribution network line is collected by a preset high-frequency current transformer, and the ground fault current magnetic field signal is processed by a signal conditioning circuit to obtain the processed signal. The processed signal is transmitted to the main control module via a digital interface circuit. The main control module updates the process noise covariance matrix and the observation noise covariance matrix based on the Sage-Husa adaptive Kalman filtering algorithm. The filtered signal is determined based on the updated process noise covariance matrix, the updated observation noise covariance matrix, and the processed signal. The main control module is a microcontroller unit or a system-on-a-chip. The main control module extracts the target features of the filtered signal to obtain the corresponding target feature vector. Based on the characteristics of the filtered signal, the target feature vector is weighted by attention features to obtain the corresponding weighted feature vector. The weighted feature vector is then matched with the feature vectors in the preset feature library to obtain the corresponding similarity score. The main control module determines the distribution network grounding fault type corresponding to the distribution network line based on the similarity score, and issues corresponding alarms to the distribution network line according to the distribution network grounding fault type.

2. The method for detecting hidden grounding faults in distribution networks according to claim 1, characterized in that, The signal conditioning circuit includes an instrumentation amplifier, a low-pass filter, and a voltage follower; Accordingly, the signal conditioning circuit performs signal processing on the ground fault current magnetic field signal to obtain a processed signal, including: The ground fault current magnetic field signal is amplified by a target factor based on the instrumentation amplifier to obtain the amplified current signal; The amplified current signal is subjected to high-frequency filtering based on a preset cutoff frequency by the low-pass filter to obtain the filtered current signal. The voltage follower is used to convert the filtered current signal into an analog signal; The analog signal is converted into a digital signal using an analog-to-digital converter to obtain the processed signal.

3. The method for detecting potential grounding faults in distribution networks according to claim 1, characterized in that, The process noise covariance matrix and observation noise covariance matrix are updated by the main control module based on the Sage-Husa adaptive Kalman filtering algorithm. The filtered signal is then determined based on the updated process noise covariance matrix, the updated observation noise covariance matrix, and the processed signal. This process includes: The current state estimate of the current signal is determined by multiplying the previous state estimate of the current signal with the state transition matrix. The state covariance matrix of the current signal at the current moment is determined using a formula based on the state transition matrix and the state covariance matrix of the previous moment; the formula for determining the state covariance matrix at the current moment is: ； in, F is the state covariance matrix of the current signal at the current moment; F is the state transition matrix. Let be the state covariance matrix of the current signal at the previous moment; This is the transpose of the state transition matrix; The process noise covariance matrix of the previous time step; The Kalman gain is determined using the Kalman gain determination formula based on the state covariance matrix of the current signal at the current moment and the observation matrix corresponding to the processed signal; the Kalman gain determination formula is: ； in, The Kalman gain at the current moment; H is the state covariance matrix at the current moment; H is the observation matrix corresponding to the processed signal; is the transpose of the observation matrix corresponding to the processed signal; inv is the matrix inversion operation; This is the observation noise covariance matrix of the previous time step; The updated state estimate is determined by updating the current state estimate based on the Kalman gain using the updated state estimate determination formula; the updated state estimate determination formula is as follows: ； in, This is the updated state estimate; This is the estimated state value at the current moment; is the Kalman gain at the current moment; H is the observation matrix corresponding to the processed signal; The processed signal; The updated state covariance matrix is ​​determined by updating the current state covariance matrix based on the Kalman gain using the formula for determining the updated state covariance matrix; the formula for determining the updated state covariance matrix is ​​as follows: ； in, Let I be the updated state covariance matrix; I is the identity matrix; The Kalman gain at the current moment; H is the state covariance matrix at the current moment; H is the observation matrix corresponding to the processed signal; The process noise covariance matrix and observation noise covariance matrix are updated based on the updated state estimate and the updated state covariance matrix using the Sage-Husa algorithm to obtain the corresponding updated process noise covariance matrix and updated observation noise covariance matrix. The Kalman gain and the state estimate at the current time are determined based on the updated process noise covariance matrix and the updated observation noise covariance matrix. The updated state estimate is obtained based on the Kalman gain and the current state estimate. Determine if a new post-processing signal exists; If it exists, then jump back to the step of determining the current state estimate of the current signal based on the product of the previous state estimate of the current signal and the state transition matrix; If it does not exist, then the filtering ends, so that the updated state estimate is determined as the filtered signal.

4. The method for detecting hidden grounding faults in distribution networks according to claim 1, characterized in that, The step of extracting the target features of the filtered signal through the main control module to obtain the corresponding target feature vector includes: The main control module extracts the peak value, frequency, and duty cycle of the filtered signal, and constructs the target feature vector based on the peak value, frequency, and duty cycle.

5. The method for detecting hidden grounding faults in distribution networks according to claim 1, characterized in that, Before performing similarity matching between the weighted feature vector and feature vectors in a preset feature library, the method further includes: A pre-defined feature library is constructed based on the feature vectors of waveforms from various types of distribution network line faults.

6. The method for detecting hidden grounding faults in distribution networks according to claim 1, characterized in that, The step of determining the distribution network grounding fault type corresponding to the distribution network line through the main control module based on the similarity score includes: If the similarity score is greater than or equal to a preset threshold, the distribution network grounding fault type corresponding to the distribution network line is determined according to the distribution network line fault type corresponding to the feature vector in the preset feature library. If the similarity score is less than the preset threshold, it is determined that the distribution network line is fault-free or the distribution network grounding fault type corresponding to the distribution network line is unknown.

7. The method for detecting hidden grounding faults in distribution networks according to any one of claims 1 to 6, characterized in that, The step of issuing corresponding alarms to the distribution network lines based on the distribution network grounding fault type includes: The distribution network grounding fault type, similarity score, and fault occurrence time are uploaded to the main station via the communication module so that the main station can issue corresponding alarms.

8. A device for detecting potential grounding faults in a distribution network, characterized in that, include: A high-frequency current transformer is pre-set to collect the ground fault current magnetic field signal of the distribution network line; The signal conditioning circuit connected to the preset high-frequency current transformer and analog-to-digital converter module is used to process the ground fault current magnetic field signal to obtain the processed signal. The analog-to-digital converter module includes a digital interface circuit, an analog front-end circuit, a conversion circuit, and a clock circuit; wherein, the digital interface circuit is used to transmit the processed signal to the main control module; The main control module, connected to the analog-to-digital converter module, is used to update the process noise covariance matrix and the observation noise covariance matrix based on the Sage-Husa adaptive Kalman filtering algorithm. It determines the filtered signal based on the updated process noise covariance matrix, the updated observation noise covariance matrix, and the processed signal. It extracts the target features of the filtered signal to obtain a corresponding target feature vector. It then performs attention feature weighting on the target feature vector based on the characteristics of the filtered signal to obtain a weighted feature vector. The weighted feature vector is matched with feature vectors in a preset feature library to obtain a similarity score. Based on the similarity score, it determines the distribution network grounding fault type corresponding to the distribution network line and issues corresponding alarms to the distribution network line according to the distribution network grounding fault type. The main control module is a microcontroller unit or a system-on-a-chip (SoC). The communication module connected to the main control module; The main station connected to the communication module; A power supply module connected to the preset high-frequency current transformer, the signal conditioning circuit, the analog-to-digital converter module, the main control module, and the communication module.

9. A distribution network grounding fault detection system, characterized in that, include: The signal processing module is used to acquire the ground fault current magnetic field signal of the distribution network line through a preset high-frequency current transformer, and to process the ground fault current magnetic field signal based on the signal conditioning circuit to obtain the processed signal. The filtered signal determination module is used to transmit the processed signal to the main control module through a digital interface circuit. The main control module updates the process noise covariance matrix and the observation noise covariance matrix based on the Sage-Husa adaptive Kalman filtering algorithm, and determines the filtered signal based on the updated process noise covariance matrix, the updated observation noise covariance matrix, and the processed signal. The main control module is a microcontroller unit or a system-on-a-chip. The similarity matching module is used to extract the target features of the filtered signal through the main control module to obtain the corresponding target feature vector, perform attention feature weighting on the target feature vector according to the characteristics of the filtered signal to obtain the corresponding weighted feature vector, and perform similarity matching between the weighted feature vector and the feature vector in the preset feature library to obtain the corresponding similarity score. The alarm module is used to determine the type of distribution network grounding fault corresponding to the distribution network line based on the similarity score by the main control module, and to issue corresponding alarms to the distribution network line according to the type of distribution network grounding fault.

10. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing a computer program to implement the method for detecting potential grounding faults in a distribution network as described in any one of claims 1 to 7.

11. A computer-readable storage medium, characterized in that, A computer program is stored on a computer-readable storage medium, and when executed by a processor, the computer program implements the method for detecting potential grounding faults in a distribution network as described in any one of claims 1 to 7.

Images