Method and system for positioning single-phase earth fault section of power distribution network

Abstract

The invention provides a power distribution network single-phase earth fault section positioning method and system, and belongs to the field of power system fault detection. Fault transient waveforms are collected through a transient zero sequence current transformer, and frequency band energy spectrum characteristics are extracted through wavelet packet decomposition; judging a section by adopting an adjacent monitoring point transient energy ratio method; the Bayesian framework is fused with transient-state and steady-state double criteria to output fault probability sorting; the linkage section switch automatically isolates a fault section and restores power supply, the positioning accuracy rate reaches 95%, and the processing time is shortened to 3 min.

Description

Technical Field

[0001] This invention relates to the field of power system fault detection technology, and in particular to a method and system for locating single-phase grounding fault sections in a distribution network. Background Technology

[0002] Neutral-point non-effectively grounded distribution networks are widely used in my country's 10kV distribution systems, with single-phase grounding faults accounting for over 80% of all faults. When a single-phase grounding fault occurs in a distribution network, the fault current is relatively small. If the fault section cannot be located and isolated in a timely and accurate manner, it can lead to the expansion of the fault, equipment damage, and even safety accidents such as fires and electric shocks. Therefore, rapid and accurate location of single-phase grounding faults in distribution networks is of great significance for ensuring the safe operation of the power grid and improving power supply reliability.

[0003] Chinese patent CN117233542A discloses a method and system for locating single-phase grounding fault sections in a distribution network. This method uses fault waveform recordings from the main station and substations to determine the fault section. Specifically, it calculates the first peak value of the main station current, the first peak value of the substation current, the fundamental amplitude of the zero-sequence current of the main station, the fundamental amplitude of the zero-sequence current of the substation, and the waveform correlation between the zero-sequence currents of the main station and the substation. When the above parameters meet predetermined conditions, the fault is determined to be located upstream of the current substation. This method improves the accuracy of fault location to some extent, but it still has the following technical shortcomings:

[0004] First, the aforementioned scheme primarily relies on comparing the initial peak value and fundamental amplitude of the steady-state zero-sequence current. When a high-resistance grounding fault occurs, the weak fault current amplitude and insignificant changes in steady-state characteristic quantities lead to insufficient sensitivity of the criterion, resulting in a significant decrease in location accuracy in high-resistance grounding scenarios with transition resistance exceeding 1000Ω. Second, the scheme employs a step-by-step comparison method based on a single criterion, lacking a multi-criterion fusion mechanism. When a single criterion is affected by interference or measurement errors, it is prone to misjudgment or missed judgment. Third, the scheme fails to fully utilize the transient signal characteristics at the moment of fault occurrence. The transient zero-sequence current contains rich fault information, and its spectral characteristics are closely related to the fault location, but existing methods are insufficient in mining the time-frequency characteristics of transient signals. Furthermore, the criterion system of the aforementioned scheme lacks probabilistic output and automatic execution mechanisms, requiring manual intervention for final decision-making, which affects fault handling efficiency.

[0005] Therefore, there is an urgent need for a method and system for locating single-phase grounding fault sections in distribution networks that can fully extract transient signal features, integrate multiple criteria, adapt to high-resistance grounding scenarios, and achieve automated fault isolation. Summary of the Invention

[0006] To address the technical problems existing in the prior art, such as insufficient sensitivity of high-resistance grounding, poor anti-interference capability of single criteria, insufficient utilization of transient features, and lack of automated execution mechanism, this invention provides a method and system for locating single-phase grounding fault sections in distribution networks. By extracting transient signal features and fusing multiple criteria, it achieves accurate location and rapid isolation of high-resistance grounding faults.

[0007] The first aspect of this invention provides a method for locating single-phase grounding fault sections in a distribution network, comprising: when a single-phase grounding fault is detected in a distribution network with a non-effectively grounded neutral point, acquiring transient zero-sequence current waveforms within a preset time window after the fault occurs through transient zero-sequence current transformers installed at each sectionalizing switch of the distribution line; performing wavelet packet decomposition on the transient zero-sequence current waveforms to extract the energy spectrum within a preset frequency band as a transient feature vector; determining the section based on the transient energy ratio method of adjacent monitoring points, wherein the transient zero-sequence current directions of adjacent monitoring points within the same fault section are consistent and the energy ratio is within a preset range, while the transient current directions of adjacent monitoring points in non-fault sections are opposite or the energy ratio exceeds the preset range; extracting the amplitude and phase information of the steady-state zero-sequence current as steady-state feature quantities; using a Bayesian fusion framework to integrate the transient feature vectors and the steady-state feature quantities, calculating and sorting the fault probabilities of each section; issuing an isolation command for the section with the highest fault probability, and linking the corresponding sectionalizing switch to automatically disconnect the fault section and restore power supply to the non-fault section.

[0008] A second aspect of the present invention provides a single-phase grounding fault location system for a distribution network, comprising: a transient signal acquisition module, used to acquire transient zero-sequence current waveforms within a preset time window after the fault occurs by means of transient zero-sequence current transformers installed at each section switch of the distribution line when a single-phase grounding fault is detected in a distribution network with a non-effectively grounded neutral point; a feature extraction module, used to perform wavelet packet decomposition on the transient zero-sequence current waveforms and extract the energy spectrum within a preset frequency band as a transient feature vector; and a section determination module, used to determine the section based on the transient energy ratio method of adjacent monitoring points, wherein phases within the same fault section are... The transient zero-sequence currents at adjacent monitoring points are in the same direction and their energy ratios are within a preset range. In non-fault sections, the transient currents at adjacent monitoring points are in opposite directions or their energy ratios exceed the preset range. A steady-state feature extraction module is used to extract the amplitude and phase information of the steady-state zero-sequence current as steady-state feature quantities. A Bayesian fusion module is used to combine the transient feature vectors and steady-state feature quantities using a Bayesian fusion framework to calculate and sort the fault probabilities of each section. An isolation execution module is used to issue an isolation command to the section with the highest fault probability, and to link the corresponding sectionalizing switch to automatically disconnect the faulty section and restore power supply to the non-faulty section.

[0009] Compared with the prior art, the present invention has the following beneficial effects:

[0010] First, this invention uses wavelet packet decomposition to extract the multi-band energy spectrum features of transient zero-sequence current, fully mining the time-frequency information in the fault transient signal. Compared with the method that only uses the first peak value and fundamental amplitude, it has higher sensitivity to high-resistance grounding faults, and the location accuracy can reach 95% in fault scenarios with a transition resistance of less than 3000Ω.

[0011] Second, this invention proposes a transient energy ratio method for adjacent monitoring points, which uses both energy ratio and current direction as criteria to determine the segment, overcoming the defect of a single criterion being susceptible to interference, and improving the reliability and anti-interference capability of the determination.

[0012] Third, this invention adopts a Bayesian fusion framework to integrate transient and steady-state features as dual criteria, realizes the probabilistic fusion of multi-source information, outputs the fault probability ranking of each segment, provides a quantitative basis for fault decision-making, and is more flexible and robust than deterministic judgment.

[0013] Fourth, this invention establishes a complete closed-loop system from fault detection, feature extraction, section determination to automatic isolation, reducing the average fault handling time from 2 hours to 3 minutes, significantly improving the efficiency of power distribution network fault handling and power restoration speed. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating a method for locating single-phase grounding fault sections in a power distribution network, provided in an embodiment of the present invention.

[0015] Figure 2 This is a schematic diagram of the structure of a single-phase grounding fault location system for a power distribution network provided in an embodiment of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0017] See Figure 1 As shown, the method for locating a single-phase ground fault section in a power distribution network provided in this embodiment of the invention includes the following steps:

[0018] Step S1: Acquisition of transient zero-sequence current signal. When a single-phase ground fault is detected in a distribution network with a non-effectively grounded neutral point, the transient zero-sequence current waveform within a preset time window after the fault occurs is acquired by the transient zero-sequence current transformer installed at each section switch of the distribution line.

[0019] In one embodiment of the present invention, the distribution network adopts a 10kV neutral point grounding method via an arc suppression coil, and the distribution lines are connected to each load point via multiple sectionalizing switches from the substation. A transient zero-sequence current transformer is installed at each sectionalizing switch of the distribution line. The transformer uses a high-frequency response type zero-sequence current sensor with a frequency response range covering DC to 50kHz to ensure accurate capture of high-frequency components in the fault transient signal.

[0020] Preferably, the sampling frequency of the transient zero-sequence current transformer is set to 100kHz, i.e., the sampling period is 10μs. This sampling frequency is selected based on the Nyquist sampling theorem, which states that to accurately capture the frequency components in the 0-50kHz range of the transient fault signal in the distribution network, the sampling frequency should be at least twice the highest frequency of the signal. Furthermore, considering the requirements for sampling accuracy and data storage in practical engineering, a sampling frequency of 100kHz achieves a good economic balance while ensuring signal integrity.

[0021] In one embodiment of the present invention, the length of the transient time window is set to 20 ms. The selection of this time window length takes into account the following factors: After a single-phase ground fault occurs in the distribution network, the transient process mainly concentrates within the first 1 to 2 power frequency cycles of the fault's initial stage. The 20 ms time window precisely covers one complete power frequency cycle, which can both fully capture the transient characteristics of the fault and avoid interference from steady-state components in the extraction of transient characteristics. At a sampling frequency of 100 kHz, a total of 2000 sampling points are collected within the 20 ms time window, providing sufficient data for subsequent wavelet packet decomposition.

[0022] Furthermore, the fault initiation criterion employs a zero-sequence voltage mutation detection method. When the detected zero-sequence voltage change exceeds 15% of the rated phase voltage, a single-phase ground fault is determined to have occurred, triggering transient data acquisition. This threshold setting can reliably detect both metallic grounding and low-resistance grounding faults, while also avoiding false triggering caused by zero-sequence voltage fluctuations during normal system operation. In one embodiment of the invention, after detecting the initiation signal, the fault recording device saves complete waveform data for three cycles: two cycles before initiation and one cycle after initiation. The transient analysis primarily uses the 20ms data segment after initiation.

[0023] Step S2: Wavelet packet decomposition and frequency band energy spectrum feature extraction. Perform wavelet packet decomposition on the transient zero-sequence current waveform and extract the energy spectrum within a preset frequency band as the transient feature vector.

[0024] In one embodiment of the present invention, a four-layer wavelet packet decomposition is used to perform time-frequency analysis on a transient zero-sequence current signal. Wavelet packet decomposition is a multi-resolution analysis method that simultaneously decomposes a signal into low-frequency approximation components and high-frequency detail components. Compared with traditional wavelet transform, which only iteratively decomposes the low-frequency part, wavelet packet decomposition also subdivides the high-frequency part, resulting in a more refined frequency band division. After four layers of decomposition, the original signal is decomposed into 16 frequency bands. At a sampling frequency of 100kHz, the frequency ranges of each band are 0 to 3.125kHz, 3.125 to 6.25kHz, 6.25 to 9.375kHz, 9.375 to 12.5kHz, 12.5 to 15.625kHz, 15.625 to 18.75kHz, 18.75 to 21.875kHz, and 21.875 to 25kHz, respectively, covering the entire frequency range from 0 to 50kHz.

[0025] Preferably, this invention selects the db4 wavelet basis function as the mother wavelet for wavelet packet decomposition. The db4 wavelet belongs to the Daubechies wavelet family, possessing compact support and good time-frequency localization characteristics. Its vanishing moment is 4, effectively suppressing polynomial trend interference. Compared to other wavelet basis functions, the db4 wavelet exhibits better resolution in both the time and frequency domains, making it suitable for analyzing non-stationary signals such as transient signals from distribution network faults. Comparative experiments have verified that the db4 wavelet outperforms commonly used wavelet bases such as db2, db6, sym3, and coif2 in decomposing transient signals from single-phase ground faults in distribution networks.

[0026] In one embodiment of the present invention, the energy of the third to eighth frequency bands is selected as the feature vector from the 16 frequency bands obtained by four-layer wavelet packet decomposition. The selection of this frequency band range is based on the following theoretical analysis and experimental verification: When a single-phase ground fault occurs in a distribution network, the transient processes of the upstream and downstream sections of the fault point are independent. Due to differences in line parameters and network structure, the energy of the fault transient signal is mainly concentrated within a specific frequency band. Experimental data shows that in a 10kV distribution network, the main energy components of the single-phase ground fault transient signal are distributed in the frequency range of 6.25 to 18.75 kHz, corresponding to the third to eighth frequency bands of the four-layer wavelet packet decomposition. Selecting the energy of these six frequency bands to form the feature vector can not only fully characterize the transient characteristics of the fault but also reduce the feature dimension and improve computational efficiency.

[0027] The formula for calculating the energy of the k-th frequency band is: ,in: is the energy value of the k-th frequency band, in J, representing the total energy of the signal within that frequency band; is the coefficient of the nth wavelet packet in the kth frequency band, which is dimensionless and directly obtained from wavelet packet decomposition; The summation term represents the total number of wavelet packet coefficients in the given frequency band. Under the conditions of four-level decomposition and 2000 sampling points, each frequency band contains 125 coefficients. The summation symbol indicates the accumulation of the squares of all coefficients in the given frequency band. This formula is based on Parseval's theorem, ensuring the equivalence between the energy of the time-domain signal and the energy of the frequency-domain coefficients.

[0028] The feature vectors are constructed using normalization. For the j-th monitoring point, its transient feature vector is: ,in: Let be the six-dimensional transient feature vector of the j-th monitoring point, which is dimensionless; The energy of the k-th frequency band at the j-th monitoring point is expressed in J. The total energy of the third to eighth frequency bands is expressed in J. Normalization eliminates the influence of differences in transformer characteristics and fault current amplitude at different monitoring points, making the feature vectors comparable.

[0029] Step S3: Segment determination based on the transient energy ratio method of adjacent monitoring points. Segment determination is performed based on the transient energy ratio method of adjacent monitoring points. In the same fault segment, the transient zero-sequence current direction of adjacent monitoring points at both ends is consistent and the energy ratio is within the preset range. In non-fault segments, the transient current direction of adjacent monitoring points is opposite or the energy ratio exceeds the preset range.

[0030] In one embodiment of the present invention, the physical basis of the transient energy ratio method of adjacent monitoring points lies in the following: when a single-phase ground fault occurs in a certain section, the transient zero-sequence currents detected by the monitoring points at both ends of the section are fault currents flowing from the fault point in their respective directions, thus the current directions are consistent and the energy is similar; while for non-faulty sections, the transient zero-sequence currents detected by the monitoring points at both ends originate from capacitive currents or through currents from other sections, and the current directions are opposite or the energy differences are significant. This characteristic provides a reliable basis for determining the faulty section.

[0031] For the i-th and j-th adjacent monitoring points on the distribution line, the transient energy ratio is defined as: ,in: Let be the transient energy ratio between the i-th and j-th monitoring points, which is dimensionless; and These represent the total energy of the third to eighth frequency bands at the i-th and j-th monitoring points, respectively, in J. When "Time" indicates that the energy of the i-th monitoring point is greater than that of the j-th monitoring point, and vice versa.

[0032] Simultaneously, the directional relationship of transient zero-sequence currents at adjacent monitoring points is calculated. The current direction is determined using the first half-wave polarity method: the first peak point of the zero-sequence current waveform within the transient time window is taken; if the peak is positive, the current direction is determined to be positive; if it is negative, the current direction is determined to be negative. The consistency of current directions between two adjacent monitoring points is characterized by a direction factor. ,in: This is the direction factor, which takes the value of +1 or -1. and These are the first peak values ​​of the transient zero-sequence current at the i-th and j-th monitoring points, respectively, in A; This is a sign function that returns +1 or -1. When... When the current direction at the two monitoring points is the same, When the current is reversed, it indicates that the direction of the current is opposite.

[0033] The criteria for determining fault sections take into account both energy ratio and direction factor. ,in: The lower limit threshold for the energy ratio is set to 0.3 in one embodiment of the present invention; The upper limit threshold for the energy ratio is set to 3.0 in one embodiment of the present invention. The selection of this threshold range is based on statistical analysis of a large amount of simulation and measured data: within the fault section, due to factors such as the difference in electrical distance between the monitoring points at both ends and the fault point, the influence of the line distributed capacitance, and the measurement error of the current transformer, the energy ratio usually fluctuates within the range of 0.3 to 3.0; while in the non-fault section, due to different current sources, the energy ratio often exceeds this range or the current direction is opposite.

[0034] Furthermore, this invention implements an adaptive adjustment mechanism for the energy ratio threshold. Considering the differences in line parameters and operating conditions across different distribution networks, the baseline range of the energy ratio needs to be corrected based on historical data. Specifically, during normal system operation, the transient energy ratio distribution characteristics of adjacent monitoring points in each section are statistically analyzed, and their mean is calculated. and standard deviation Then the threshold is set as follows: ,in: This is the statistical mean of the energy ratio under normal operating conditions, and is dimensionless. The statistical standard deviation of the energy ratio is dimensionless; coefficient 2 corresponds to an approximately 95% confidence interval. This adaptive mechanism allows the decision threshold to adapt to the characteristics of different distribution networks, improving the accuracy and applicability of the decision.

[0035] Step S4: Steady-state zero-sequence current amplitude and phase extraction. The amplitude and phase information of the steady-state zero-sequence current are extracted as steady-state characteristic quantities. In one embodiment of the present invention, the steady-state zero-sequence current characteristic extraction is performed in the time period from 60ms to 100ms after the fault occurs, during which time the transient process of the fault has basically decayed and the system enters the steady-state stage. The zero-sequence current during this period is subjected to spectrum analysis using Fast Fourier Transform (FFT) to extract the amplitude and phase of the fundamental component.

[0036] The formula for calculating the amplitude of the steady-state zero-sequence current fundamental wave is:

[0037] ,

[0038] in: This represents the fundamental amplitude of the steady-state zero-sequence current, in amperes (A). The zero-sequence current instantaneous value at the nth sampling point, in A; To determine the number of sampling points used in the calculation, we take 4000 points for two power frequency cycles. This is the power frequency, with a value of 50Hz. The sampling period is 0.00001s, which is 10μs. This formula is based on the principle of fundamental component extraction of discrete Fourier transform. It obtains the real and imaginary parts of the fundamental component through sine and cosine correlation operations, and then calculates the amplitude.

[0039] The formula for calculating the phase of the fundamental wave of the steady-state zero-sequence current is: ,in: The phase of the steady-state zero-sequence current is expressed in degrees (°) and ranges from -180° to +180°. The phase value is referenced to the zero-sequence voltage. The phase information reflects the phase relationship between the zero-sequence current and the zero-sequence voltage, and plays an important role in determining the direction of fault current flow.

[0040] Preferably, the present invention performs change processing on steady-state characteristic quantities to eliminate the influence of unbalanced current before the fault. Specifically, the method involves: saving the zero-sequence current data for one cycle before the fault and calculating its fundamental amplitude. and phase Then, the change is obtained by subtracting the pre-fault value from the post-fault steady-state characteristic value:

[0041] ,in: This represents the change in the amplitude of the zero-sequence current fundamental wave, expressed in A. This represents the phase change of the zero-sequence current fundamental wave, measured in degrees. Processing this change effectively filters out inherent system-wide three-phase imbalances and transformer zero-drift interference, improving the reliability of steady-state characteristics.

[0042] Step S5: Bayesian fusion framework for multi-criteria synthesis and fault probability calculation. The transient feature vector and the steady-state feature quantity are synthesized using the Bayesian fusion framework to calculate and sort the fault probability of each segment.

[0043] In one embodiment of this invention, the core idea of ​​the Bayesian fusion framework is to treat transient and steady-state criteria as two independent sources of evidence, and to fuse the information of the two types of evidence based on Bayesian probabilistic inference to output the posterior probability of failure in each segment. Compared with traditional logical AND-OR judgment methods, Bayesian fusion can quantify the credibility of each criterion, achieve soft decision output, and improve the robustness of the judgment result.

[0044] First, the fault confidence level for each segment is calculated based on transient criteria. For the m-th segment (defined by the m-th and (m+1)-th monitoring points), its transient fault confidence level is defined as:

[0045] ,in: Let f be the fault confidence level of the m-th segment based on transient criteria, with a value ranging from 0 to 1; This represents the ratio of the transient energy at monitoring points m and (m+1)th. Direction factor; The Gaussian kernel bandwidth parameter is set to 0.5 in one embodiment of this invention. The physical meaning of this formula is: when the current directions are consistent, the closer the energy ratio is to 1 (indicating that the energy at the two monitoring points is equal), the higher the fault confidence. When the current directions are opposite, the confidence is set to a lower baseline value of 0.1 instead of 0, in order to retain a certain fault tolerance to cope with extreme situations.

[0046] Secondly, the fault confidence level of each section is calculated based on the steady-state criterion. The steady-state criterion is determined using the ratio of zero-sequence current amplitude and phase difference between adjacent monitoring points:

[0047] ,in: Let f be the fault confidence level of the m-th segment based on the steady-state criterion, with a value ranging from 0 to 1; and These are the absolute values ​​of the zero-sequence current amplitude changes at the m-th and m+1-th monitoring points, respectively, in A; and These are the phase changes of the zero-sequence current at the m-th and m+1-th monitoring points, respectively, in radians; and The weighting coefficients are 2.0 and 1.5 respectively in one embodiment of the present invention. The formula uses the Sigmoid function to map the amplitude ratio and phase difference to the interval between 0 and 1. When the fault is in the m-th segment, the current amplitude at the upstream monitoring point is usually greater than that at the downstream, and the phase difference is close to 0, making the confidence level approach 1.

[0048] Then, the comprehensive failure probability of each section is calculated using the Bayesian fusion formula:

[0049] ,in: Given the confidence levels of transient and steady-state criteria, this represents the posterior probability of a fault occurring in the m-th segment. Let f(m) be the likelihood probability of observing the current confidence level under the condition that a fault occurs in the m-th segment. Let be the prior fault probability of the m-th segment, which is uniformly distributed when there is no other information. This is the normalization constant.

[0050] Assuming the transient and steady-state criteria are conditionally independent, the likelihood probability can be decomposed as follows:

[0051] Substituting into Bayes' formula and performing normalization, we obtain the failure probability of each segment: ,in: Let f be the failure probability of the m-th segment, with a value ranging from 0 to 1, and the sum of the probabilities of all segments is 1; This represents the total number of distribution line sections. This formula achieves probabilistic fusion of transient and steady-state criteria, while ensuring the mathematical rationality of the probabilities through normalization.

[0052] In one embodiment of the present invention, a reliability weighting mechanism is also introduced to adapt to different fault scenarios. When a high-resistance ground fault is detected (judged by the zero-sequence voltage amplitude), the weight of the transient criterion is increased; when the transient signal-to-noise ratio is lower than a preset signal-to-noise ratio threshold (set to 20dB in one embodiment of the present invention), the weight of the steady-state criterion is increased. The weighted fusion formula is as follows: ,in: This is the weighting index for the transient criterion, with a value ranging from 0.5 to 2.0; The steady-state criterion weighting index ranges from 0.5 to 2.0; when The process degenerates into equal-weighted fusion. The weighting coefficients are adaptively adjusted based on the fault scenario: in high-resistance grounding scenarios, the following settings are applied. , To enhance the role of the transient criterion; when the signal-to-noise ratio of the transient signal is below 20dB, set... , To enhance the role of the steady-state criterion.

[0053] Finally, the fault probabilities of each section are sorted in descending order to form a fault probability ranking list. The first item in the ranking list is the section with the highest fault probability, which is the first choice for fault location; at the same time, the probability information of other sections is retained to provide a reference for subsequent verification and manual decision-making.

[0054] Step S6: Fault isolation command execution and power restoration. An isolation command is issued for the section with the highest fault probability, and the corresponding sectionalizing switch is linked to automatically disconnect the faulty section and restore power to the non-faulty section.

[0055] In one embodiment of the present invention, after the fault probability ranking list is generated, the system automatically checks whether the fault probability of the first-ranked segment exceeds the isolation action threshold. Isolation action threshold The threshold is set to 0.7. This threshold selection is based on a comprehensive trade-off between the risks of erroneous activation and failure to activate: a threshold that is too low increases the risk of erroneous activation, leading to the incorrect removal of healthy sections; a threshold that is too high increases the risk of failure to activate, leading to the failure to isolate faulty sections in a timely manner. When the faulty section is automatically disconnected, the system sends a trip command to the sectionalizing switches at both ends of the m-th section.

[0056] The operating logic of the sectionalizing switch follows these principles: First, disconnect the sectionalizing switch downstream of the fault section to cut off the connection between the fault point and the downstream load. Then, disconnect the sectionalizing switch upstream of the fault section to isolate the connection between the fault point and the power supply. The operating interval between the two switches is set to 100ms to avoid potential closing problems caused by simultaneous operation. The sectionalizing switch uses a vacuum circuit breaker or a load switch, and its breaking capacity should meet the requirements of the distribution network fault current. In one embodiment of this invention, a vacuum circuit breaker with a rated short-circuit breaking current of 12.5kA is used.

[0057] After the faulty section is isolated, the system automatically performs power restoration operations on the non-faulty sections. The restoration strategy adopts a segmented, step-by-step closing method: starting from the substation outgoing line, the sectionalizing switches of each healthy section are closed sequentially according to the line topology until the upstream boundary of the faulty section is reached; simultaneously, starting from the end of the line, the sectionalizing switches of each healthy section downstream of the faulty section are closed sequentially, and power is transferred from adjacent feeders through tie switches. This restoration strategy ensures that power is restored to the largest possible area in the shortest time, while avoiding closing at the fault point.

[0058] Furthermore, this invention includes a power restoration confirmation mechanism. After each sectionalizing switch is closed, the system monitors the current and voltage parameters of that section. If an overcurrent or overvoltage abnormality is detected, the system immediately trips the switch and triggers an alarm to prevent the fault from spreading. In one embodiment of this invention, the overcurrent protection setting for power restoration is set to 1.5 times the normal load current, and the action time limit is set to 200ms.

[0059] In one embodiment of the present invention, a manual confirmation mode is also provided as an alternative to automatic execution. When the probability value of the segment with the highest fault probability is in the range of 0.5 to 0.7, the system only issues a fault alarm and displays a fault probability ranking list, which is then manually executed by maintenance personnel after confirmation. This mode is suitable for complex scenarios where fault characteristics are not obvious or where there are multiple suspected fault segments, reserving space for manual judgment while ensuring safety.

[0060] After extensive simulation and field testing, the single-phase grounding fault location method for distribution networks provided by this invention achieves a location accuracy of 95% in various fault scenarios with a transition resistance of less than 3000Ω. The fault handling time is reduced from an average of 2 hours to 3 minutes using traditional methods, significantly improving the efficiency of distribution network fault handling and power supply reliability.

[0061] In one embodiment of the present invention, a 10kV distribution network simulation model is established using PSCAD / EMTDC electromagnetic transient simulation software for method verification. The simulation model includes one 35kV / 10kV substation and three 10kV distribution lines, each 10km long, using a mixed overhead and cable laying method. The overhead lines use LGJ-120 type conductors, and the cables use YJV22-8.7 / 15-3×120 type power cables. The neutral point of the distribution network uses an arc suppression coil grounding method, with an arc suppression coil capacity of 300kVA and a residual current compensation degree set to 5%.

[0062] The simulation test included six fault scenarios: metallic grounding fault (transition resistance 0Ω), low-resistance grounding fault (transition resistance 100Ω), medium-resistance grounding fault (transition resistance 500Ω), high-resistance grounding fault (transition resistance 1000Ω), ultra-high-resistance grounding fault (transition resistance 2000Ω), and extremely high-resistance grounding fault (transition resistance 3000Ω). Each fault type was tested at different locations on the three feeders (the beginning, middle, and end of the line), totaling 54 simulated operating conditions.

[0063] Simulation results show that in fault scenarios with a transition resistance below 1000Ω, the location accuracy of the method of this invention reaches 100%; in the range of transition resistance from 1000Ω to 2000Ω, the location accuracy is 98.1%; and in the range of transition resistance from 2000Ω to 3000Ω, the location accuracy is 94.4%. Compared with the method of CN117233542A, in high-resistance grounding scenarios with a transition resistance exceeding 1000Ω, the location accuracy of the method of this invention is improved by approximately 25 percentage points.

[0064] In one embodiment of the present invention, a field application test was conducted on a 10kV distribution network in a provincial capital city. The test distribution network covered an urban area of ​​35km. 2The scope includes 15 10kV feeders, with a total line length of approximately 180km. Transient zero-sequence current transformers and data acquisition terminals are installed at the sectionalizing switch of each feeder, totaling 78 monitoring devices. The main station system is deployed in the distribution automation main station room and is connected to each monitoring terminal via a fiber optic communication network.

[0065] During the field testing period (January 2024 to June 2024), a total of 32 naturally occurring single-phase grounding fault events were recorded. For each fault event, the method of this invention was applied to locate the fault section, and the location was compared with the actual fault location confirmed by manual line inspection. Test results show that: of the 32 fault events, the method of this invention correctly located 30, with a location accuracy rate of 93.75%; the two misjudged events both occurred in extremely high resistance grounding scenarios with a transition resistance exceeding 2500Ω, and the fault points were located in areas with relatively small capacitive current at the end of the line.

[0066] The on-site test also statistically analyzed the fault handling time. Using the method of this invention, the average time from fault occurrence to completion of fault section isolation was 2.8 minutes, the average time from fault isolation to power restoration of the non-faulty section was 4.2 minutes, and the overall average fault handling time was 7 minutes. Compared to the average fault handling time of 2 hours using the traditional manual line inspection method, the efficiency is improved by approximately 94%.

[0067] See Figure 2 As shown, this embodiment of the invention also provides a single-phase grounding fault location system for a distribution network, used to implement the functions of each step in the above method embodiment. The system includes a transient signal acquisition module 1, a feature extraction module 2, a section determination module 3, a steady-state feature extraction module 4, a Bayesian fusion module 5, and an isolation execution module 6. The modules are interconnected via a data bus and a control bus.

[0068] The transient signal acquisition module 1 is used to acquire the transient zero-sequence current waveform within a preset time window after the fault occurs when a single-phase ground fault is detected in a distribution network with a non-effectively grounded neutral point. This is achieved by using transient zero-sequence current transformers installed at each sectionalizing switch of the distribution line. In one embodiment of the invention, the module includes multiple distributed data acquisition units. Each data acquisition unit is installed at a sectionalizing switch and is equipped with a high-frequency response zero-sequence current transformer and a 16-bit high-speed AD converter. The sampling frequency is 100kHz, and the data is uploaded to the master station via fiber optic Ethernet. The acquisition unit has a local data caching function, and the cache capacity supports the continuous storage of complete waveform data for 10 fault events. The fault initiation criterion adopts zero-sequence voltage mutation detection, with an initiation threshold of 15% of the rated phase voltage. After initiation, waveform data for three cycles before and after the fault are automatically saved.

[0069] Feature extraction module 2 is used to perform wavelet packet decomposition on the transient zero-sequence current waveform and extract the energy spectrum within a preset frequency band as a transient feature vector. In one embodiment of the present invention, this module adopts a heterogeneous computing architecture of FPGA and DSP. The FPGA is responsible for the parallel computation of wavelet packet decomposition, and the DSP is responsible for energy spectrum extraction and feature vector construction. The wavelet packet decomposition uses the db4 mother wavelet for four-level decomposition, extracting the energy of the third to eighth frequency bands to form a six-dimensional feature vector. The feature vector is normalized and then output to the segment determination module 3. The processing delay of this module is less than 10ms, which meets the real-time requirements.

[0070] The section determination module 3 is used to determine sections based on the transient energy ratio method of adjacent monitoring points. In one embodiment of the present invention, this module receives the transient feature vectors of each monitoring point, calculates the energy ratio and direction factor between adjacent monitoring points, and outputs the transient fault confidence of each section according to the determination criteria described in step S3 of the method embodiment. The energy ratio threshold supports adaptive adjustment, and the threshold parameter is stored in a configurable parameter table, which can be updated online according to the field operation data.

[0071] The steady-state feature extraction module 4 is used to extract the amplitude and phase information of the steady-state zero-sequence current as steady-state feature quantities. In one embodiment of the present invention, this module extracts zero-sequence current data during a time period of 60ms to 100ms after the fault occurs, calculates the amplitude and phase of the fundamental component using the FFT algorithm, and performs change processing to eliminate the influence of the unbalanced current before the fault. The FFT calculation uses a data window of 4000 points across two cycles, with a frequency resolution of 25Hz.

[0072] Bayesian fusion module 5 is used to integrate the transient feature vector and the steady-state feature quantity using a Bayesian fusion framework, calculate the fault probability of each segment, and sort them. In one embodiment of the present invention, this module receives the transient fault confidence and the steady-state fault confidence, calculates the comprehensive fault probability of each segment according to the Bayesian fusion formula described in step S5 of the method embodiment, and outputs a fault probability ranking list in descending order. This module supports adaptive adjustment of the criterion weights and automatically selects a transient-dominated or steady-state-dominated fusion strategy according to the fault scenario.

[0073] The isolation execution module 6 is used to issue isolation commands to the section with the highest fault probability, and to coordinate with the corresponding sectionalizing switch to automatically disconnect the faulty section and restore power supply to the non-faulty section. In one embodiment of the invention, this module receives a fault probability ranking list, and when the probability of the first-ranked section exceeds 0.7, it automatically issues a tripping command to the corresponding sectionalizing switch. After the tripping action is completed, the power supply restoration strategy is executed. This module interacts with the intelligent terminal of the sectionalizing switch through the IEC 61850 communication protocol, supporting GOOSE fast messages to achieve millisecond-level remote control command transmission. After the isolation execution is completed, the module uploads the fault location results and action records to the distribution master station for archiving.

[0074] In one embodiment of the present invention, the above modules can be integrated into the master station server and implemented in software, or they can be distributed and deployed in the intelligent terminals at each sectional switch. The centralized architecture is suitable for distribution networks with good communication conditions and has the advantage of centralized algorithm maintenance; the distributed architecture is suitable for distribution networks with limited communication conditions and has the advantages of local autonomy and rapid response. Preferably, a hybrid architecture combining centralized and distributed approaches is adopted: transient signal acquisition and preliminary feature extraction are completed in the distributed intelligent terminals, feature fusion and fault probability calculation are completed centrally at the master station, and the generation and execution of isolation commands are completed collaboratively by the master station and the terminals.

[0075] In one embodiment of the present invention, the hardware configuration of the distributed intelligent terminal is as follows: the main processor adopts an ARM Cortex-A9 dual-core processor with a main frequency of 1GHz, and is used in conjunction with an FPGA coprocessor (Xilinx Zynq-7020) for hardware acceleration of wavelet packet decomposition; the memory is configured with 2GB DDR3 memory and 16GB eMMC flash memory, supporting complete waveform storage for 10 fault events; the communication interface includes dual-channel fiber optic Ethernet interface (supporting IEC 61850 communication protocol), dual-channel RS485 serial interface (for transformer data acquisition), and a 4G wireless communication module (as a backup channel for fiber optic communication); the power supply adopts a dual-power supply design for AC and DC, with both AC 220V and DC 110V input, and a built-in backup battery supports continuous operation for 30 minutes after a power outage.

[0076] The main server's hardware configuration is as follows: The server uses dual Intel Xeon Gold 6248R processors with a main frequency of 3.0GHz, a total of 48 cores and 96 threads; the memory configuration is 256GB ECC DDR4, supporting large-scale parallel computing; the storage uses an all-flash array with an effective capacity of 20TB and is configured with RAID6 redundancy protection; the server is deployed in the power distribution automation main station data center, and a dual-machine hot standby architecture is adopted to ensure system reliability.

[0077] In one embodiment of the present invention, the system software adopts a layered architecture design, divided from bottom to top into a device access layer, a data processing layer, a business logic layer, and a human-machine interaction layer. The device access layer is responsible for communication connections and data acquisition with distributed intelligent terminals, supporting the IEC 61850 standard protocol stack to achieve millisecond-level data transmission latency. The data processing layer is responsible for waveform data preprocessing, feature extraction, and fusion calculation. The core algorithm is implemented in C++ to ensure computational efficiency, and the parallel computing framework is based on OpenMP and CUDA acceleration libraries. The business logic layer is responsible for fault determination, isolation decisions, and the execution of power restoration strategies, employing a combination of a rule engine and a machine learning model to achieve intelligent decision-making. The human-machine interaction layer provides a web-based monitoring interface, supporting functions such as real-time display of fault events, historical data query, parameter configuration, and report generation.

[0078] In one embodiment of the present invention, the system software further includes a self-learning and parameter optimization module. This module performs offline optimization of the judgment threshold and fusion weights based on historical fault data, using a particle swarm optimization algorithm or a genetic algorithm to search for the optimal parameter combination. The optimization objective function comprehensively considers three indicators: location accuracy, false alarm rate, and refusal to operate rate, constructing a multi-objective optimization problem through a weighted summation method. The optimization results are manually reviewed and then updated to the parameter configuration table of the online system. This self-learning mechanism enables the system to continuously adapt to changes in distribution network operating conditions, maintaining optimal fault location performance.

[0079] In one embodiment of this invention, the system employs the IEC 61850 communication protocol to achieve information exchange between the master station and intelligent terminals. Fault waveform data is transmitted via MMS (Manufacturing Message Specification) service, with a single transmission time not exceeding 500ms. Fault location results and isolation commands are transmitted using GOOSE (General Object-Oriented Substation Events) fast messages, with a transmission delay of less than 4ms, meeting the real-time requirements for rapid fault isolation. Peer-to-peer communication between intelligent terminals also uses GOOSE messages, enabling data sharing and collaborative judgment between adjacent terminals.

[0080] The system communication network adopts a dual-ring redundant topology, with the main network using fiber optic Ethernet and the backup network using 4G wireless communication. Network switching time is less than 50ms, ensuring high availability of the communication link. In one embodiment of this invention, the communication network bandwidth is designed to be 1Gbps, meeting the peak bandwidth requirements of 78 monitoring devices simultaneously uploading waveform data.

[0081] In one embodiment of the present invention, the system parameter configuration adopts a hierarchical management mechanism, including three levels: global parameters, feeder-level parameters, and section-level parameters. Global parameters apply to the entire distribution network and include basic configurations such as sampling frequency, time window length, wavelet basis function type, and frequency band selection range. Feeder-level parameters are set differently for the line characteristics of different feeders, including energy ratio threshold benchmark values ​​and fault probability action thresholds. Section-level parameters are finely adjusted for the operating conditions of specific sections, including transformer correction coefficients, historical energy ratio mean and standard deviation.

[0082] The parameter configuration interface employs a wizard-driven design, guiding maintenance personnel through parameter settings. For newly connected distribution lines, the system provides a parameter self-tuning function: during normal operation of the distribution line, it automatically collects transient and steady-state characteristic data from each monitoring point, statistically analyzes their distribution characteristics, and recommends appropriate threshold parameters. Maintenance personnel can accept the recommended parameters or manually modify them and confirm their effectiveness. Parameter modification records are automatically archived, supporting parameter version rollback and historical comparison analysis.

[0083] In one embodiment of the present invention, the recommended configuration ranges for key parameters are as follows: the sampling frequency is recommended to be configured to 100kHz to 200kHz, the transient time window length is recommended to be configured to 10ms to 30ms, the wavelet packet decomposition level is recommended to be configured to 3 to 5 levels, the lower limit threshold of the energy ratio is recommended to be configured to 0.2 to 0.5, the upper limit threshold of the energy ratio is recommended to be configured to 2.0 to 5.0, and the fault probability action threshold is recommended to be configured to 0.6 to 0.8. The parameter configurations should be optimized and adjusted according to the actual characteristics and operation and maintenance requirements of the distribution network.

[0084] In one embodiment of the present invention, the system employs multi-level anti-interference measures to ensure the reliability of fault location. At the hardware level, the transient zero-sequence current transformer adopts a differential signal transmission method to effectively suppress common-mode interference; the analog front end of the data acquisition terminal adopts a multi-stage low-pass filter design with a cutoff frequency set to 50kHz to filter out aliasing signals higher than the Nyquist frequency; a transient suppression diode is added to the input of the AD converter to prevent damage to the acquisition circuit from overvoltages such as lightning strikes.

[0085] At the signal processing level, wavelet packet decomposition itself has good denoising properties; by selecting specific frequency bands where signal energy is concentrated, noise components can be automatically filtered out. In one embodiment of the present invention, median filtering is also used to smooth the energy spectrum feature vector, eliminating the influence of occasional pulse interference. For steady-state feature extraction, FFT calculation is performed using data from two power frequency cycles, and random noise is effectively suppressed through time-domain averaging.

[0086] At the criterion fusion level, the probabilistic output of the Bayesian fusion framework itself has the ability to resist the failure of a single criterion. When a criterion is disturbed and produces an outlier, the normal output of another criterion can compensate for its impact, ensuring the reliability of the overall failure probability. In one embodiment of the present invention, a criterion validity detection mechanism is also set: when the transient energy of a monitoring point is lower than the noise background threshold or the steady-state amplitude change is close to zero, the data of that monitoring point is determined to be invalid and is removed or downweighted in the fusion calculation.

[0087] At the system decision-making level, multiple verification mechanisms are implemented to prevent accidental activation. In addition to the fault probability threshold judgment, the following verification conditions are added: the zero-sequence voltage change exceeds the activation threshold, the data from the monitoring points at both ends of the fault section are valid, and the difference in fault probability with adjacent sections exceeds the significance threshold. Only when all verification conditions are met will the system issue an automatic isolation command; otherwise, it switches to manual confirmation mode, and the final decision is made by the maintenance personnel.

[0088] In one embodiment of the present invention, in addition to supporting the location of single-phase grounding fault sections, the system can also be extended to support the identification and location of other fault types. By analyzing the characteristic combinations of transient zero-sequence current and negative-sequence current, the following fault types can be distinguished: stable single-phase grounding fault, intermittent arcing grounding fault, two-phase grounding fault, and open-circuit fault. For intermittent arcing grounding faults, the system adopts a continuous monitoring and cumulative judgment strategy. When multiple short-duration grounding signals are detected cumulatively within a set time period (e.g., 10 minutes), it is determined to be an intermittent arcing grounding fault, and the possible range of the arcing channel is given.

[0089] In one embodiment of the present invention, the system also supports fault development trend analysis. Through statistical analysis of historical fault data, high-fault-rate sections and high-risk periods are identified, providing a reference for distribution network operation and maintenance. The fault trend analysis employs time series analysis, combined with meteorological data (temperature, humidity, wind speed, lightning activity, etc.) and load data, to establish a fault occurrence probability prediction model. The prediction results are displayed on the monitoring interface in the form of a heat map, helping maintenance personnel to deploy preventative measures in advance.

[0090] In one embodiment of the present invention, the system adopts a distributed data storage architecture to achieve long-term storage and efficient querying of fault data. Raw waveform data is stored in a distributed file system (e.g., HDFS or object storage), supporting fast indexing by timestamp and feeder number; feature data and judgment results are stored in a relational database (e.g., PostgreSQL), supporting combined queries with complex conditions; real-time status data is stored in a time-series database (e.g., InfluxDB), supporting high-frequency writes and rolling overwrites. The data retention period can be configured as needed. In one embodiment of the present invention, raw waveform data is stored for 90 days, feature data and judgment results are stored for 5 years, and statistical report data is permanently stored.

[0091] The system provides rich reporting functions, including daily, weekly, monthly, and annual reports on fault events, statistical reports on fault type distribution, section fault rate ranking reports, statistical reports on judgment accuracy, and equipment operation status reports. Reports support automatic generation and scheduled push notifications; maintenance personnel can receive report summaries via email or mobile devices. In one embodiment of this invention, the report format supports PDF, Excel, and HTML output formats to meet the needs of different scenarios.

[0092] In one embodiment of the present invention, the system is integrated with existing distribution automation systems through standard interfaces. Integration with SCADA systems utilizes the IEC 104 protocol or IEC 61850 MMS service to achieve bidirectional interaction of remote signaling, telemetry, and remote control information; integration with GIS geographic information systems uses a WebService interface to achieve visual location of fault points on a map; integration with production management systems (PMS) uses a database intermediate table or message queue method to achieve automatic dispatching of fault work orders and tracking of processing progress; and integration with dispatch automation systems uses E language files or CIM / XML format to achieve synchronous updates of distribution network models and topology data.

[0093] System integration follows the principle of loose coupling, using standardized interfaces to shield differences between heterogeneous systems, thereby reducing integration complexity and maintenance costs. In one embodiment of this invention, the system provides a visual integration configuration interface, allowing operations and maintenance personnel to complete interface mapping configurations via drag-and-drop, achieving data interoperability with third-party systems without writing code.

[0094] The embodiments of the present invention are not limited to the specific embodiments described above. Those skilled in the art can make various equivalent changes or substitutions based on the technical solutions of the present invention, and all such changes or substitutions should be included within the protection scope of the present invention.

Claims

1. A method for locating single-phase grounding fault sections in a distribution network, characterized in that, include: When a single-phase ground fault is detected in a distribution network with a non-effectively grounded neutral point, the transient zero-sequence current waveform within a preset time window after the fault occurs is collected by the transient zero-sequence current transformer installed at each section switch of the distribution line. Perform wavelet packet decomposition on the transient zero-sequence current waveform and extract the energy spectrum within a preset frequency band as the transient feature vector; Section determination is based on the transient energy ratio method of adjacent monitoring points. The transient energy ratio and current direction factor between two adjacent monitoring points are calculated. In the same fault section, the transient zero-sequence current direction of the two adjacent monitoring points is consistent and the energy ratio is within the preset range. In non-fault sections, the transient current direction of adjacent monitoring points is opposite or the energy ratio exceeds the preset range. The amplitude and phase information of the steady-state zero-sequence current are extracted as steady-state characteristic quantities; A Bayesian fusion framework is used to integrate the transient feature vector and the steady-state feature quantity. The fault confidence based on the transient criterion and the fault confidence based on the steady-state criterion are calculated respectively. The two types of confidence are probabilistically fused to obtain the fault probability of each segment and then sorted. An isolation command is issued for the section with the highest probability of failure, and the corresponding sectionalizing switch is activated to automatically disconnect the faulty section and restore power to the non-faulty section.

2. The method according to claim 1, characterized in that, The sampling frequency of the transient zero-sequence current transformer is 100kHz, the length of the preset time window is 20ms, and the fault initiation criterion adopts the zero-sequence voltage change detection method. When the zero-sequence voltage change exceeds 15% of the rated phase voltage, a single-phase grounding fault is determined to have occurred and transient data acquisition is triggered.

3. The method according to claim 1, characterized in that, The wavelet packet decomposition uses the db4 wavelet basis function for four-level decomposition. The preset frequency band range is the third to the eighth frequency band, corresponding to a frequency range of 6.25kHz to 18.75kHz. The transient feature vector is a six-dimensional normalized energy spectrum vector, with each dimension corresponding to the proportion of energy in the third to the eighth frequency bands to the total energy.

4. The method according to claim 1, characterized in that, The lower threshold of the preset range is 0.3, and the upper threshold is 3.

0. The energy ratio threshold supports adaptive adjustment. Based on historical data, the mean and standard deviation of the transient energy ratio of adjacent monitoring points in each segment are statistically analyzed, and the threshold is set to the range of the mean plus or minus twice the standard deviation.

5. The method according to claim 1, characterized in that, The current direction factor is calculated using the first half-wave polarity method. The polarity of the first peak point of the zero-sequence current waveform within the transient time window is taken as the current direction indicator. When the polarity of the first peak point of two adjacent monitoring points is the same, the direction factor is positive one, indicating that the directions are consistent. When the polarities are opposite, the direction factor is negative one, indicating that the directions are opposite.

6. The method according to claim 1, characterized in that, The fault confidence based on transient criteria is calculated using a Gaussian kernel function. When the current directions are consistent, the confidence is equal to the Gaussian mapping value of the degree to which the energy ratio deviates from one. When the current directions are opposite, the confidence is taken as the low reference value. The Gaussian kernel bandwidth parameter is set to 0.5, and the low reference value is set to 0.

1.

7. The method according to claim 1, characterized in that, The steady-state zero-sequence current is extracted in the time period from 60ms to 100ms after the fault occurs. The amplitude and phase of the fundamental component are calculated using fast Fourier transform, and the change is processed to eliminate the influence of the unbalanced current before the fault. The fault confidence based on the steady-state criterion uses the Sigmoid function to map the amplitude ratio and phase difference of adjacent monitoring points to the interval between zero and one.

8. The method according to claim 1, characterized in that, The Bayesian fusion framework introduces a criterion reliability weighting mechanism. When a high-resistance grounding fault is detected, the transient criterion weight index is set to 1.5 and the steady-state criterion weight index is set to 0.

8. When the transient signal-to-noise ratio is lower than the preset signal-to-noise ratio threshold, the transient criterion weight index is set to 0.8 and the steady-state criterion weight index is set to 1.

2. The preset signal-to-noise ratio threshold is set to 20dB.

9. The method according to claim 1, characterized in that, The isolation command is issued when the probability value of the section with the highest fault probability exceeds the isolation action threshold of 0.

7. The power supply restoration adopts a segmented step-by-step closing method, closing the segment switches of each healthy section in sequence from the substation outgoing line according to the line topology until the upstream boundary of the fault section, and simultaneously closing the segment switches of each healthy section downstream of the fault section in sequence from the end of the line.

10. A single-phase grounding fault location system for a distribution network, used to implement the method described in any one of claims 1-9, characterized in that, include: The transient signal acquisition module is used to acquire the transient zero-sequence current waveform within a preset time window after the fault occurs by using transient zero-sequence current transformers installed at each section switch of the distribution line when a single-phase ground fault is detected in a distribution network with a non-effectively grounded neutral point. The feature extraction module is used to perform wavelet packet decomposition on the transient zero-sequence current waveform and extract the energy spectrum within a preset frequency band as the transient feature vector; The section determination module is used to determine the section based on the transient energy ratio method of adjacent monitoring points. It calculates the transient energy ratio and current direction factor between two adjacent monitoring points. In the same fault section, the transient zero-sequence current direction of the two adjacent monitoring points is consistent and the energy ratio is within the preset range. In non-fault sections, the transient current direction of adjacent monitoring points is opposite or the energy ratio exceeds the preset range. The steady-state feature extraction module is used to extract the amplitude and phase information of the steady-state zero-sequence current as steady-state feature quantities. The Bayesian fusion module is used to integrate the transient feature vector and the steady-state feature quantity using a Bayesian fusion framework, calculate the fault confidence based on the transient criterion and the fault confidence based on the steady-state criterion respectively, and perform probability fusion of the two types of confidence to obtain the fault probability of each segment and sort them. The isolation execution module is used to issue isolation commands to the section with the highest probability of failure, and to link the corresponding sectionalizing switch to automatically disconnect the faulty section and restore power supply to the non-faulty section.

Images