Single-phase grounding fault line selection method for small current grounding system

Abstract

The small current grounding system single-phase grounding fault line selection method solves the problem of how to improve the line selection accuracy, and belongs to the field of small current fault line selection. The two power frequency period data before the fault are used as the reference, and the data after the fault are compared by difference, so that the inherent interference such as load imbalance, harmonic background and line parameter asymmetry during normal operation of the system is eliminated, and the fault characteristics are more prominent. Secondly, four criteria (differential steady-state phase, wavelet packet differential energy spectrum, differential instantaneous phase consistency and envelope weighted cumulative polarity) are used to describe the fault from different physical aspects. Each criterion independently outputs a 0~1 confidence level, forming a multi-dimensional complementary verification mechanism. Even if a certain criterion fails under certain conditions, other criteria can still provide reliable evidence. Finally, the prior probability and the four criterion likelihood values corrected by dynamic weight are fused through the Bayes formula to obtain the posterior probability of each branch and bus grounding, and the high and low threshold values are used for hierarchical decision-making.

Description

Technical Field

[0001] This invention relates to a method for locating single-phase grounding faults in low-current grounding systems, belonging to the field of low-current fault location. Background Technology

[0002] current, Distribution networks at certain voltage levels primarily employ a neutral point non-effective grounding method, often referred to as a low-current grounding system. Based on specific differences in the neutral point grounding method, this system can be categorized into ungrounded neutral points, grounded via arc suppression coils, and other types. Statistics show that in this type of system, over 80% of faults are single-phase grounding faults. While this fault theoretically allows the system to maintain short-term operation without being disconnected (thanks to the relatively small fault current and line voltage symmetry), this traditional operating mode is facing serious challenges.

[0003] When a single-phase ground fault occurs, the fault current is relatively small, and the line voltage on the load side remains symmetrical. Therefore, the system can usually continue to operate with the fault for a period of time, generally up to several hours. However, with rapid economic development, the demand for electricity for production and daily life continues to grow, and the upgrading and transformation of urban and rural power distribution networks are constantly advancing. The scale of the power distribution network is expanding rapidly, and the usage and length of cable lines are increasing significantly, resulting in a significant increase in the system's distributed capacitance to ground and an increase in capacitive leakage current. Under these circumstances, the impact of single-phase ground faults becomes more severe: on the one hand, a large grounding current may lead to excessively high local temperatures; on the other hand, if the fault is not cleared for a long time, it may further develop into more serious faults such as phase-to-phase short circuits, posing a potential risk to the safe and stable operation of the power grid. In summary, although the initial harm of a single-phase ground fault is relatively limited, its potential danger cannot be ignored under the current increasingly complex power distribution network architecture. Therefore, given the current complex network architecture, even if the direct destructive force of a single-phase ground fault is small in its initial stage, a strategy of "shifting from passive monitoring to active and rapid clearing" must be adopted to curb the spread of potential risks.

[0004] After extensive research, a wealth of methods for fault location in low-current grounding systems under single-phase grounding faults have been developed. However, existing fault location technologies are still insufficient to adapt to the increasing complexity and scale of distribution networks, and their accuracy in identifying certain fault types, such as high-resistance grounding, remains low. Existing fault location methods still have shortcomings in terms of accuracy and universality, and there is room for improvement in their technical aspects. In general, the main technical challenges are as follows:

[0005] (1) Insufficient steady-state fault signal strength

[0006] In low-current grounding systems, the overall amplitude of the grounding current is relatively low under single-phase grounding fault conditions. Especially when the fault occurs near the zero-crossing point of the phase voltage, the inductive attenuation component is significant, leading to blurred electrical characteristics of the fault. If arcing occurs, the already limited current signal is further affected by arc fluctuations, increasing the difficulty of fault line identification. Furthermore, the compensation effect of the arc suppression coil further reduces the characteristic differences between faulty and healthy lines.

[0007] (2) Transient fault signals are highly variable and random.

[0008] In cases of arc grounding and intermittent grounding, transient fault signals exhibit high randomness and significant fluctuations. Furthermore, these signals contain numerous harmonic components, with different frequency bands intertwined, making accurate extraction of their spectral energy quite difficult.

[0009] (3) Interference factors affect signal acquisition

[0010] In addition to the aforementioned issues of weak fault characteristics and complex transient signals, the complex operating conditions on site can also interfere with the normal operation of the signal acquisition device. When a single-phase ground fault occurs, factors such as the amplification effect of the zero-sequence circuit on higher harmonic components, external electromagnetic interference, and background noise can all make it difficult for the acquisition device to obtain a sufficiently clear fault waveform, thereby affecting the accurate identification of the faulty line.

[0011] Existing fault location methods can be broadly categorized into three types: steady-state fault information-based fault location, fault transient information-based fault location, and comprehensive fault location methods. Each method has its own shortcomings and limitations. For steady-state fault information-based fault location methods, the most commonly used approach is phase and amplitude comparison. However, in distribution networks containing arc suppression coils, due to the compensation effect of inductive current, the amplitude and phase characteristics of the fault current become less apparent, making it difficult to accurately identify the faulty line. Furthermore, if the grounding fault is severe, leading to arcing, the influence of current fluctuations further complicates fault location. For fault transient information-based fault location methods, the most common approach is also phase and amplitude comparison. However, fault transient signals exhibit strong randomness and large fluctuations. In addition, transient fault signals also contain a large number of harmonic signals, with components of different frequencies mixed within, making spectral energy extraction difficult. Both of these methods are implemented using hardware circuits, relying excessively on the accuracy of circuit signal acquisition, resulting in low accuracy. For the comprehensive route selection method, existing methods are mostly combined with artificial intelligence and deep learning algorithms. However, artificial intelligence algorithms rely excessively on the programming of host computer chips, which makes the host computer extremely expensive to manufacture. At the same time, the overly complex overall system is also prone to equipment failure and increases the failure rate. Summary of the Invention

[0012] To address the issue of improving fault location accuracy, this invention provides a fault location method for single-phase grounding faults in low-current grounding systems based on a comparison of the total current before and after the fault.

[0013] The present invention provides a method for selecting the fault location of a single-phase grounding system in a low-current grounding system, comprising:

[0014] Save the current sampling data of each branch before the fault occurs and after the fault occurs, and process the sampling data to obtain the steady-state component and high-frequency transient component of the power frequency.

[0015] Based on the extracted power frequency steady-state components and high-frequency transient components, the following four types of confidence are calculated in parallel for each branch k: confidence of differential steady-state power frequency phase change, confidence of differential energy spectrum, confidence of differential instantaneous phase accumulation consistency, and confidence of envelope weighted accumulation.

[0016] For each confidence level, the dynamic weight of the confidence level is calculated and normalized based on the difference between the maximum and the second largest confidence level of the confidence level in all branches. The normalized result is used as the confidence weight of the confidence level of the confidence level.

[0017] The prior probability of bus grounding and branch k is calculated by using the deviation between the power frequency current amplitude of branch k before the fault and the average power frequency current amplitude of all branches.

[0018] Based on the fundamental phase difference, differential energy ratio, differential phase consistency coefficient, and weighted cumulative energy of branch k, the likelihood value under the fault assumption is obtained.

[0019] The likelihood value under the fault assumption is corrected by using a confidence weight. The posterior probability of bus grounding and branch k is calculated based on the corrected likelihood value and the prior probability of branch k. The decision is made based on the maximum value of the posterior probability.

[0020] The beneficial effects of this invention are that it systematically improves the accuracy of single-phase grounding fault location in low-current grounding systems by comparing the total current before and after a fault, performing parallel calculations based on multiple criteria, dynamically adjusting weights adaptively, and using Bayesian probability fusion. First, by using pre-fault power frequency period data as a benchmark and performing differential comparison with post-fault data, it eliminates inherent interferences such as load imbalance, harmonic background, and asymmetric line parameters during normal system operation, making the fault characteristics more prominent. Second, four criteria characterizing the fault from different physical perspectives (differential steady-state phase, wavelet packet differential energy spectrum, differential instantaneous phase consistency, and envelope weighted cumulative activity) are used, with each criterion output independently. A multi-dimensional complementary verification mechanism is formed based on confidence levels, ensuring that even if one criterion fails under specific operating conditions, other criteria can still provide reliable evidence. On this basis, dynamic weights are calculated in real-time based on the ratio of the maximum to the second-largest confidence level of each criterion for all branches. Criteria with strong discriminative power receive higher weights, while the weights of failed criteria are automatically reduced, thus avoiding misleading information in Bayesian fusion. Finally, the prior probabilities and the likelihood values ​​of the four criteria after dynamic weight correction are fused using a Bayesian formula to obtain the posterior probability of each branch and bus grounding, and decisions are made in a graded manner based on high and low thresholds. These methods comprehensively solve the problem of low accuracy in traditional line selection methods under complex conditions such as weak steady-state signals, variable transient characteristics, high-resistance grounding, and arc suppression coil compensation, significantly improving the line selection success rate. Attached Figure Description

[0021] Figure 1 This is a schematic diagram illustrating the principle of the invention of this application;

[0022] Figure 2 This is a timing diagram of the data sampling window before and after the fault;

[0023] Figure 3 Comparison of waveforms before and after zero-phase digital filtering;

[0024] Figure 4 The flowchart shows the sub-processes for calculating the confidence level of the four criteria;

[0025] Figure 5 A comparison chart of the accuracy of line selection using dynamic weights and fixed weights;

[0026] Figure 6 This is a schematic diagram illustrating the adaptive changes in the dynamic weights of each criterion.

[0027] Figure 7 The intention is to construct and represent the likelihood function lookup table;

[0028] Figure 8 This is a schematic diagram of hierarchical decision-making based on posterior probability. Detailed Implementation

[0029] 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. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0031] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but this is not intended to limit the scope of the invention.

[0032] This specific embodiment provides a method for selecting the fault location of a single-phase grounding system in a low-current grounding system, which solves the problem of low accuracy in existing methods under complex conditions such as weak steady-state signals, variable transient characteristics, high-resistance grounding, and arc suppression coil compensation. This method collects branch current and bus zero-sequence voltage data for two power frequency cycles before and eight power frequency cycles after a fault. First, it performs zero-phase digital filtering to separate the power frequency steady-state component from the high-frequency transient component. Then, it calculates the confidence scores of four independent criteria for each branch in parallel (differential steady-state power frequency phase change confidence score, differential energy spectrum confidence score, differential instantaneous phase cumulative consistency confidence score, and envelope weighted cumulative positivity confidence score). Next, it calculates and normalizes the dynamic weight of each criterion based on the difference between the maximum and second-largest confidence scores for all branches. Simultaneously, it calculates the prior probability of bus grounding and each branch based on the power frequency current amplitude deviation in the second power frequency cycle before the fault. Finally, it converts the characteristic quantities of the four criteria into likelihood values ​​using a preset likelihood function, corrects them with dynamic weights, and performs Bayesian fusion with the prior probabilities to obtain the posterior probability of each branch and bus grounding. Finally, it makes hierarchical decisions based on the maximum posterior probability and a preset threshold. The single-phase grounding fault location method for low-current grounding systems in this embodiment includes:

[0033] Step 1: Save the current sampling data of each branch before the fault occurred and after the fault occurred, and process the sampling data to obtain the steady-state component and the high-frequency transient component; the steady-state component and the high-frequency transient component are specifically the steady-state current and the high-frequency transient current.

[0034] In this implementation, the current sampling data of each branch is saved for the two power frequency cycles before the fault occurs and the eight power frequency cycles after the fault occurs.

[0035] Specifically, the current of each branch of phases A, B, and C and the zero-sequence voltage of the bus are synchronously sampled at a sampling rate of 20kHz. A circular buffer is used to store the sampling data of the most recent 20 power frequency cycles to ensure that even if there is a delay in fault detection, the data segment from the two cycles before the fault to the eight cycles after the fault can still be completely extracted.

[0036] Real-time calculation of the effective value of zero-sequence voltage and the periodic effective value change rate of each phase current. When >0.15 , The rate of change of the effective value of phase voltage or the effective value of a certain phase current is greater than When the rated current is applied, determine that a single-phase ground fault has occurred in that phase, and record the time of the fault. ;

[0037] Extract the raw sampled data from the two complete power frequency cycles before the fault and the eight complete power frequency cycles after the fault from the circular buffer, and use them as the fault data segment;

[0038] Zero-phase digital filtering is performed on the faulty data segment: After forward filtering with an FIR low-pass filter (cutoff frequency 100Hz), the sequence is reversed and filtered again to eliminate phase delay, thus obtaining the power frequency steady-state component. Using an FIR high-pass filter (cutoff frequency 300Hz) in a zero-phase manner, high-frequency transient components were obtained. .

[0039] Traditional fault location methods directly use absolute electrical quantities after a fault, which are easily affected by inherent three-phase imbalance, harmonic background, and load fluctuations in the system. This implementation introduces a pre-fault reference. By performing differential calculations (such as phase difference and energy difference) between the data from the two cycles before the fault and the data after the fault, a high-pass filter is essentially constructed. This operation filters out DC offset and slowly changing steady-state background, retaining only the transient and steady-state incremental characteristics caused by the sudden change in fault, thus highlighting fault characteristics in environments with extremely low signal-to-noise ratios.

[0040] Step 2: Based on the extracted components, calculate the following four types of confidence scores in parallel for each branch k:

[0041] Four confidence levels are used as criteria to characterize fault features from different physical perspectives. The output is a confidence level of 0 to 1, indicating the degree of certainty that the criterion indicates that branch k is faulty. The four criteria are calculated in parallel and are independent of each other.

[0042] Criterion 1: Utilize the fundamental phase difference of the steady-state power frequency component of branch k in the second cycle before the fault and the first cycle after the fault. Calculate the confidence level of the differential steady-state power frequency phase change of branch k. ;

[0043] When a single-phase ground fault occurs in branch k, the phase of its zero-sequence current changes significantly before and after the fault, while the phase change in a healthy branch is minimal. By utilizing the fundamental phase difference between the second cycle before the fault and the first cycle after the fault, the influence of the arc suppression coil compensation current can be effectively eliminated. Specifically:

[0044] The steady-state current at power frequency of branch k during the second power frequency cycle before the fault Perform Discrete Fourier Transform (DFT) to extract the fundamental phase. .

[0045] The steady-state current at power frequency during the first power frequency cycle after the fault Perform DFT to extract the fundamental phase. .

[0046] Calculate phase difference and adjust to scope.

[0047] Calculate the confidence level of differential instantaneous phase cumulative consistency:

[0048]

[0049] As the phase difference threshold, it can be... Within the scope of this embodiment ;

[0050] Based on Kirchhoff's current law, the zero-sequence current phase of a faulty line is opposite to that of a healthy line. Criterion 1 eliminates the masking effect of the arc suppression coil's inductive compensation current on the amplitude characteristics by comparing the change in the fundamental phase before and after the fault, utilizing the rigidity of the phase dimension.

[0051] Criterion 2: Perform wavelet packet decomposition on the high-frequency transient components of branch k during the first power frequency cycle before and after the fault, extract the characteristic frequency band energy of the preset characteristic frequency band, and calculate the differential energy ratio of the two extracted characteristic frequency band energies. According to this differential energy ratio Calculate the confidence level of the differential energy spectrum of branch k. ;

[0052] Single-phase ground faults generate high-frequency transient currents with energy far exceeding the high-frequency background energy during normal operation. By comparing energy changes within characteristic frequency bands before and after the fault, the faulty branch can be identified. Wavelet packet decomposition can precisely extract the energy of a specified frequency band. Specifically:

[0053] Take the high-frequency transient current of branch k in the first power frequency cycle before the fault. and the high-frequency transient current in the first power frequency cycle after the fault .

[0054] A three-level wavelet packet decomposition is performed using the Db4 wavelet basis. The third level has eight sub-bands, each with a bandwidth of [missing value]. Select a sub-band that covers a preset characteristic frequency band;

[0055] The preset characteristic frequency band in this application can be a fixed frequency band, or it can automatically select the sub-band with the most concentrated energy based on the spectral distribution of the transient signal after the fault, achieving adaptability to different fault resistances and different system parameters, and always extracting the strongest fault characteristic frequency band. Specifically:

[0056] Perform a Fast Fourier Transform on the high-frequency transient current during the first power frequency cycle after the fault, calculate the power spectral density, and find the frequency corresponding to the maximum value of the power spectral density. , choose to Centered on a frequency band with a bandwidth of B, this is the preset characteristic frequency band.

[0057] ;

[0058] This is the proportionality coefficient. In this way, the bandwidth automatically widens when the main frequency is high and narrows when the main frequency is low, avoiding the inclusion of too much noise or the omission of energy-concentrated sidelobes. For branches... k The high-frequency transient components of the first power frequency cycle after the fault are subjected to wavelet packet decomposition of a preset number of layers to extract the reconstruction coefficients of a preset characteristic frequency band. The sum of the squares of each reconstruction coefficient is... For the high-frequency transient components of branch k in the first power frequency cycle before the fault, the same wavelet packet decomposition as after the fault is performed to extract the reconstruction coefficients of the same preset characteristic frequency bands, and the sum of squares of each coefficient is calculated. ;

[0059] Calculate the differential energy ratio:

[0060]

[0061] in, Prevent division by zero;

[0062] Differential energy spectrum confidence for:

[0063]

[0064] This represents the differential energy ratio reference, adjusted according to the background noise level. The range of values This implementation method You can choose 5;

[0065] The high-frequency transient traveling wave generated by the instantaneous voltage change during a fault is mainly concentrated in a specific frequency band. Criterion 2 extracts the differential energy of this frequency band through wavelet packet decomposition, capturing the strong transient characteristics in the early stage of the fault.

[0066] Criterion 3: Calculate the differential phase trajectory of the instantaneous phase of the high-frequency transient component of branch k at the last preset duration of the first power frequency cycle before the fault and at the preset duration before the first power frequency cycle after the fault, and calculate the differential phase consistency coefficient of the current branch with all other branches. According to the differential phase consistency coefficient Calculate the differential instantaneous phase cumulative consistency confidence of branch k ;

[0067] The phase of the high-frequency transient current in the faulted branch is essentially opposite to that in the healthy branch. By calculating the differential phase trajectories before and after the fault, the inherent phase shift of the line can be eliminated, allowing for a more accurate comparison of the phase relationships between branches. A consistency coefficient is used to measure the degree of phase similarity. Specifically:

[0068] Take the high-frequency transient current at the last preset duration (5ms) of the first power frequency cycle before the fault and the first preset duration (5ms) of the first power frequency cycle after the fault;

[0069] Perform a Hilbert transform on each signal segment to obtain an analytic signal, and extract the instantaneous phase. and ;

[0070] The differential phase trajectory is:

[0071]

[0072] Phase unwinding is then performed to ensure the trajectory is continuous.

[0073] Calculate the differential phase consistency coefficient between branch k and another branch j:

[0074]

[0075] The overall consistency coefficient of branch k is the average of that of all other branches. for:

[0076]

[0077] Calculate the confidence level of differential instantaneous phase cumulative consistency. for:

[0078]

[0079] Criterion 4: Extract the envelope of the high-frequency transient current of branch k within 8 power frequency cycles after the fault, and calculate the weighted cumulative current. Based on weighted cumulative motivation Calculate the envelope-weighted cumulative positivity confidence of branch k. ;

[0080] During transient decay, the high-frequency current of the faulty branch has the opposite instantaneous polarity to the total current of all branches, and its contribution is greater with larger amplitude. Envelope-weighted integration enhances noise immunity.

[0081] The high-frequency transient current of branch k during 8 power frequency cycles after the fault Perform a Hilbert transform to obtain the analytic signal, and then take the modulus to obtain the envelope. .

[0082] Constructing a reference current ;

[0083] Determine the transient decay time From the moment of failure Initially, the envelope maximum value decays to 10% of its peak value;

[0084] Calculate the weighted cumulative motivation:

[0085]

[0086] Envelope-weighted cumulative positivity confidence level for:

[0087]

[0088] The weighted cumulative positivity of branch z.

[0089] Criteria 3 and 4 are based on the propagation polarity of the fault transient current. The direction of the transient current flowing through the faulty branch is opposite to the transient direction of the total current in all branches. The instantaneous phase and envelope polarity are extracted by Hilbert transform, and the identification is performed using the principle of vector superposition.

[0090] Step 3: For each type of confidence, calculate and normalize the dynamic weight of the confidence based on the difference between the maximum and second largest values ​​of the confidence in all branches. The normalization result is used as the confidence weight of the confidence in that type.

[0091] The reliability of each criterion varies with fault conditions. By analyzing the confidence distribution of the criterion's output across all branches, if the maximum value is significantly greater than the second-largest value, it indicates strong discriminative ability and should be assigned a higher weight; if the maximum and second-largest values ​​are close or the maximum value is very small, it indicates criterion failure and its weight should be reduced. Specifically:

[0092] Confidence level for all branches Find the maximum value and the second largest value , ;

[0093] Set basic weights Criterion 1 is 0.4, and criteria 2, 3, and 4 are each 0.2;

[0094] Existing technologies mostly employ fixed-weight fusion, which cannot adapt to complex and variable power grid conditions (such as weak transient components during high-resistance grounding and abundant transient components during overvoltage). This implementation introduces a dynamic weight calculation model based on the range between the maximum and second-largest values ​​to calculate the dynamic weights. :

[0095]

[0096] in, Prevent division by zero;

[0097] For dynamic weights Normalization is performed to obtain the confidence weight of branch k. :

[0098] .

[0099] is the dynamic weight of the confidence level of the j-th class.

[0100] When a certain criterion has a high confidence level for all branches (i.e., the maximum value is much greater than the second largest value), it indicates that the criterion has high identification ability in the current fault environment, and the algorithm automatically assigns it a high weight; conversely, if the outputs of all branches are similar (criterion fails), the weight automatically approaches zero, thereby realizing the algorithm's self-diagnosis and self-correction.

[0101] Furthermore, in route selection practice, a single criterion may output an incorrect high confidence level due to specific interference (e.g., harmonic contamination causing a phase criterion to misjudge a healthy branch's phase change as large). If a fixed weight is used, this erroneous evidence will continue to participate in the fusion, leading to posterior probability bias. However, in the dynamic weight calculation model of this invention:

[0102] If the difference between the maximum and second-largest confidence values ​​of a certain criterion for all branch outputs is not significant, it indicates that the criterion cannot distinguish between faulty and non-faulty branches. Even if the confidence values ​​of individual branches are high, the overall distribution is flat, and its dynamic weights are automatically reduced, thus weakening the influence of likelihood values ​​in Bayesian fusion.

[0103] If a certain criterion outputs an abnormality but still has discriminative power (for example, the maximum value truly corresponds to the faulty branch, but the second largest value is also relatively high), the range can still reflect its relative advantage, and the weight will not be mistakenly reduced.

[0104] This application eliminates the need for manual preset weight switching logic for different fault types (metallic grounding, high-resistance grounding, arcing grounding, intermittent grounding). It autonomously perceives the real-time effectiveness of each criterion, achieves adaptive weighting under operating conditions, avoids a single failure criterion from vetoing or misleading the final result, and significantly improves the success rate of line selection under harsh conditions such as high-resistance grounding and strong noise.

[0105] Step 4: Calculate the prior probability of bus grounding and branch k by using the deviation between the power frequency current amplitude of branch k in the second power frequency cycle before the fault and the average power frequency current amplitude of all branches.

[0106] Branches with a larger load current imbalance before a fault have a slightly higher probability of weak insulation points, which can be used as prior information to improve the reliability of line selection. Specifically,

[0107] For each branch k, extract the power frequency current amplitude during the second power frequency cycle before the fault. ;

[0108] Calculate the average amplitude of all branches ;

[0109] The unnormalized prior probability of busbar grounding is:

[0110]

[0111] The unnormalized prior probability of branch k is:

[0112]

[0113] The unbalance coefficient preset for the branch is... The prior base coefficient of the busbar;

[0114] Normalized prior probability of bus grounding Prior probability of branch k They are respectively:

[0115]

[0116] .

[0117] Step 5: Based on the fundamental phase difference, differential energy ratio, differential phase consistency coefficient, and weighted cumulative probability of branch k, obtain the likelihood value under the fault assumption; correct the likelihood value under the fault assumption using confidence weights; calculate the posterior probability of bus grounding and branch k based on the corrected likelihood value and the prior probability of branch k; and make a decision based on the maximum value of the posterior probability.

[0118] The original features of the four criteria (phase difference, energy ratio, consistency coefficient, and cumulative positivity) are converted into probability density values ​​under the fault hypothesis through a preset likelihood function, then corrected by dynamic weights, and finally combined with the prior probability to obtain the posterior probability, thus realizing multi-evidence fusion decision-making.

[0119] This implementation method obtains the probability distributions of fundamental phase difference, differential energy ratio, differential phase consistency coefficient, and weighted cumulative energy under bus grounding, bus ungrounding, faulty branch, and healthy branch conditions through simulation and statistics. Parameter fitting is then performed, and the fitted probability density function is discretized into a lookup table. Specifically, the probability distributions of fundamental phase difference, differential energy ratio, differential phase consistency coefficient, and weighted cumulative energy are obtained through simulation and statistics for two scenarios: branch k fault and bus grounding.

[0120] Fault conditions of branch k: Statistically analyze the distribution of characteristic quantities of the faulty branch itself. Criterion 1 and 3 conform to normal distribution, criterion 2 conforms to log-normal distribution, and criterion 4 is negative and has a large absolute value and skewed distribution.

[0121] Bus grounding scenario: When the bus fault occurs, the characteristic quantity distribution of all branches is statistically analyzed. The fundamental phase difference is close to 0°, the differential energy ratio is small, the differential phase consistency coefficient is close to +1, and the weighted cumulative energy is positive or close to 0.

[0122] For the two scenarios described above, parameter fitting is performed separately, and the fitted probability density function is discretized into two independent lookup tables: a branch fault lookup table and a bus grounding lookup table. Each lookup table is constructed as follows: the possible value range of the feature quantity is uniformly divided into M intervals; the likelihood value corresponding to the midpoint of each interval is calculated; and the interval boundary array and the likelihood value array are stored.

[0123] During real-time calculation, for each branch k:

[0124] Based on the fundamental phase difference, differential energy ratio, differential phase consistency coefficient, and weighted cumulative likelihood calculated for this branch, a binary search is used to determine the interval in the branch fault lookup table, and the corresponding likelihood value is read as... ;

[0125] Meanwhile, for the bus grounding assumption, the same characteristic quantities are used. However, it should be noted that when the bus is grounded, it is necessary to evaluate whether the characteristic quantities of all branches match the bus grounding pattern. In actual implementation, the characteristic quantities of all branches can be pre-substituted into the bus grounding lookup table to obtain the likelihood value of each branch under the bus grounding assumption. Since all branches behave consistently when the bus is grounded, the average of the likelihood values ​​of each branch can be taken, or the value can be applied directly to each branch individually. For simplification, the bus grounding table is usually looked up independently for each branch k to obtain the value. The branch components are then processed collectively. Alternatively, a unified characteristic quantity can be used to characterize the bus grounding assumption (e.g., the average value of all branch characteristic quantities), and then substituted into the bus grounding lookup table to obtain the result. Alternatively, the arithmetic mean can be taken by looking up the table for each branch separately.

[0126] The posterior probability of busbar grounding is:

[0127] ;

[0128] in, The prior probability of busbar grounding. Let k be the prior probability of branch k. For the number of branch roads, ;

[0129] The posterior probability of branch k is:

[0130] ;

[0131] in, and These are the likelihood values ​​for bus grounding and branch k, respectively. The likelihood value after correction for bus grounding is... The likelihood value after correction for branch k is .

[0132] By combining prior probabilities with the likelihood function, the posterior probability is output. This method transforms the route selection problem from traditional hard logic judgment to soft probability estimation, effectively handling measurement uncertainty and random interference, and improving the scientific nature of the decision-making.

[0133] Make a decision based on the maximum posterior probability: if If the value is greater than 0.85, then output the corresponding branch number or bus grounding. ;

[0134] If 0.65 < If the value is ≤0.85, output the corresponding branch and issue a low confidence alarm;

[0135] like If the value is ≤0.65, the line selection is considered a failure and the system switches to an auxiliary line selection method.

[0136] This application systematically improves the accuracy of single-phase grounding fault location in low-current grounding systems by comparing the total current before and after the fault, performing parallel calculations based on multiple criteria, dynamically adjusting weights adaptively, and using Bayesian probabilistic fusion. First, it uses data from two power frequency cycles before the fault as a benchmark and performs differential comparison with data after the fault, eliminating inherent interferences such as load imbalance, harmonic background, and asymmetric line parameters during normal system operation, thus making the fault characteristics more prominent. Second, it designs four criteria to characterize the fault from different physical perspectives (differential steady-state phase, wavelet packet differential energy spectrum, differential instantaneous phase consistency, and envelope weighted cumulative energy), with each criterion outputting independently. A multi-dimensional complementary verification mechanism is formed based on confidence levels, ensuring that even if a criterion fails under specific operating conditions, other criteria can still provide reliable evidence. On this basis, dynamic weights are calculated in real-time based on the ratio of the maximum to the second-largest confidence levels for each criterion across all branches. Criteria with strong discriminative power receive higher weights, while the weights of failed criteria are automatically reduced, thus avoiding misleading information in Bayesian fusion. Finally, the prior probability (based on the pre-fault load current imbalance) and the likelihood values ​​of the four criteria, corrected by dynamic weights, are fused using a Bayesian formula to obtain the posterior probability of each branch and busbar grounding, and decisions are made based on high and low thresholds. The entire process eliminates the rapid early termination of any single criterion, forcing the completion of all criterion calculations and fusions, ensuring the reliability and robustness of the decision. These methods comprehensively address the low accuracy issues of traditional line selection methods under complex conditions such as weak steady-state signals, variable transient characteristics, high-resistance grounding, and arc suppression coil compensation, significantly improving the line selection success rate.

[0137] This specific implementation significantly improves the accuracy and robustness of single-phase grounding fault location in low-current grounding systems by comparing the full current before and after the fault, quantifying confidence based on multiple criteria, adaptive dynamic weighting, and Bayesian fusion. It is particularly suitable for complex operating conditions such as high-resistance grounding, arcing grounding, and overcompensation of arc suppression coils.

[0138] While the invention has been described herein with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the invention. Therefore, it should be understood that many modifications can be made to the exemplary embodiments, and other arrangements can be designed without departing from the spirit and scope of the invention as defined by the appended claims. It should be understood that different dependent claims and features herein can be combined in ways different from those described in the original claims. It is also understood that features described in conjunction with individual embodiments can be used in other embodiments.

Claims

1. A method for selecting the fault location of a single-phase grounding fault in a low-current grounding system, characterized in that, include: Save the current sampling data of each branch before the fault occurs and after the fault occurs, and process the current sampling data to obtain the steady-state component and the high-frequency transient component. Based on the extracted power frequency steady-state components and high-frequency transient components, the following four types of confidence are calculated in parallel for each branch k: confidence of differential steady-state power frequency phase change, confidence of differential energy spectrum, confidence of differential instantaneous phase accumulation consistency, and confidence of envelope weighted accumulation. For each confidence level, the dynamic weight of the confidence level is calculated and normalized based on the difference between the maximum and the second largest confidence level of the confidence level in all branches. The normalized result is used as the confidence weight of the confidence level of the confidence level. The prior probability of bus grounding and branch k is calculated by using the deviation between the power frequency current amplitude of branch k before the fault and the average power frequency current amplitude of all branches. Based on the fundamental phase difference, differential energy ratio, differential phase consistency coefficient, and weighted cumulative energy of branch k, the likelihood value under the fault assumption is obtained. The likelihood value under the fault assumption is corrected using the aforementioned confidence weight. The posterior probability of bus grounding and branch k is calculated based on the corrected likelihood value and the prior probability of branch k. The decision is made based on the maximum value of the posterior probability.

2. The method for selecting the fault location of a single-phase grounding system in a low-current grounding system according to claim 1, characterized in that, Save the current sampling data of each branch for 2 power frequency cycles before the fault occurs and 8 power frequency cycles after the fault occurs, and process the current sampling data to obtain the power frequency steady-state component and the high-frequency transient component. Using the fundamental phase difference of the steady-state power frequency component of branch k in the second cycle before the fault and the first cycle after the fault, the confidence level of the differential steady-state power frequency phase change of branch k is calculated. ; Wavelet packet decomposition is performed on the high-frequency transient components of branch k during the first power frequency cycle before and after the fault. The characteristic frequency band energy of the preset characteristic frequency bands is extracted, and the differential energy ratio of the two extracted characteristic frequency band energies is calculated. Based on this differential energy ratio, the confidence level of the differential energy spectrum of branch k is calculated. ; Calculate the differential phase trajectory of the instantaneous phase of the high-frequency transient component of branch k at the last preset duration of the first power frequency cycle before the fault and at the preset duration before the first power frequency cycle after the fault. Calculate the differential phase consistency coefficient between the current branch and all other branches. Based on the differential phase consistency coefficient, calculate the cumulative consistency confidence level of the differential instantaneous phase of branch k. ; Extract the envelope of the high-frequency transient current of branch k within 8 power frequency cycles after the fault, calculate the weighted cumulative positivity, and then calculate the confidence level of the envelope weighted cumulative positivity of branch k based on the weighted cumulative positivity. .

3. The method for selecting the fault location of a single-phase grounding system in a low-current grounding system according to claim 2, characterized in that, For each confidence level, the dynamic weight of that confidence level is calculated based on the difference between the maximum and second-largest confidence levels across all branches: ； Subscript , The dynamic weights for the corresponding class confidence. As the preset base weights, Small positive number, This represents the maximum confidence level for the corresponding class. This is the second largest value of the corresponding class confidence score; The credibility weight is: ； is the dynamic weight of the confidence level of the j-th class.

4. The method for selecting the fault location of a single-phase grounding system in a low-current grounding system according to claim 3, characterized in that, The posterior probability of busbar grounding is: ； in, The prior probability of busbar grounding. Let k be the prior probability of branch k. For the number of branch roads, ; The posterior probability of branch k is: ； in, and These are the likelihood values ​​for bus grounding and branch k, respectively. The likelihood value after correction for bus grounding is... The likelihood value after correction for branch k is .

5. The method for selecting the fault location of a single-phase grounding system in a low-current grounding system according to claim 4, characterized in that, Prior probability of bus grounding Prior probability of branch k They are respectively: ； ； Among them, the unnormalized prior probability of bus grounding is The unnormalized prior probability of branch k is: ; The amplitude of the power frequency current in branch k during the second power frequency cycle before the fault; This represents the average power frequency current amplitude of all branches during the second power frequency cycle prior to the fault. The unbalance coefficient preset for the branch is... Let be the prior basic coefficient of the busbar, and .

6. The method for selecting the fault location of a single-phase grounding system in a low-current grounding system according to claim 4, characterized in that, The probability distributions of fundamental phase difference, differential energy ratio, differential phase consistency coefficient, and weighted cumulative energy are obtained through simulation statistics under bus grounding, bus ungrounding, faulty branch, and healthy branch conditions. The parameters are then fitted, and the fitted probability density function is discretized into a lookup table. Based on the calculated fundamental phase difference, differential energy ratio, differential phase consistency coefficient, and weighted cumulative likelihood, the interval is determined in the lookup table through binary search, and the corresponding likelihood value is read.

7. The method for selecting the fault location of a single-phase grounding system in a low-current grounding system according to claim 2, characterized in that, The confidence level of the differential steady-state power frequency phase change for: ； in, For the fundamental phase difference, This is the phase difference threshold.

8. The method for selecting the fault location of a single-phase grounding system in a low-current grounding system according to claim 2, characterized in that, Differential energy spectrum confidence for: ； Among them, differential energy ratio The high-frequency transient components of branch k in the first power frequency cycle after the fault are decomposed into wavelet packet decompositions of a preset number of layers to extract the reconstruction coefficients of a preset characteristic frequency band. The sum of the squares of each reconstruction coefficient is... For the high-frequency transient components of branch k in the first power frequency cycle before the fault, the same wavelet packet decomposition as after the fault is performed to extract the reconstruction coefficients of the same preset characteristic frequency bands, and the sum of squares of each coefficient is calculated. , It is a small positive number; This represents the differential energy ratio reference, adjusted according to the background noise level. The method for obtaining the preset feature frequency band is as follows: Perform a Fast Fourier Transform on the high-frequency transient component of the first power frequency cycle after the fault, calculate the power spectral density, and find the frequency corresponding to the maximum value of the power spectral density. , choose to Centered on a frequency band with a bandwidth of B, this is the preset characteristic frequency band. ； This is the proportionality coefficient.

9. The method for selecting the fault location of a single-phase grounding system in a low-current grounding system according to claim 2, characterized in that, Differential instantaneous phase cumulative consistency confidence for: ； Differential phase consistency coefficient of branch k with all other branches j for: ； The overall consistency coefficient of branch k is the average of that of all other branches: ； For preset duration, For the number of branch roads, This is the differential phase trajectory.

10. The method for selecting the fault location of a single-phase grounding system in a low-current grounding system according to claim 2, characterized in that, Envelope-weighted cumulative positivity confidence level for: ； Among them, weighted cumulative motivation Reference current , This refers to the high-frequency transient component of branch k, specifically the high-frequency transient current. Indicates the envelope. Indicates the transient decay time. The time when the fault occurred. The weighted cumulative positivity of branch z.

Images