Power distribution network high resistance fault locating method based on feeder segment model mismatch identification
By using a mismatch identification method based on feeder segment model, and by normalizing the low-frequency effective value of zero-sequence current difference and using a first-order accumulation generator to enhance signal characteristics, a comprehensive mismatch index is constructed, which solves the accuracy problem of high-resistivity fault location and achieves reliable segment location in a strong noise environment.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HOHAI UNIV
- Filing Date
- 2026-04-03
- Publication Date
- 2026-07-10
Smart Images
Figure CN122361995A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to a method for locating high-resistivity faults in distribution networks based on feeder segment model mismatch identification, belonging to the field of power system relay protection technology. Background Technology
[0002] Resonant grounding systems are widely used in my country's medium-voltage distribution networks due to their advantages in suppressing single-phase ground fault currents and improving power supply reliability. However, high-impedance faults (HIFs), with their transition resistance reaching thousands of ohms, weak fault currents, and nonlinear arc characteristics, have long been a difficult technical challenge to overcome in the field of distribution network protection. HIFs are often caused by tree branches touching the ground, insulator deterioration, or contact with surface media (such as sand or asphalt). If not located and isolated in time, they may cause fires or escalate into phase-to-phase short-circuit faults, seriously threatening power grid safety. Especially with the increasing penetration of distributed generation, the increasing harmonic content and varied operating modes of the system further exacerbate the complexity of locating high-impedance faults.
[0003] Currently, high-resistivity fault location methods are mainly divided into two categories: steady-state quantity analysis and transient quantity analysis. Steady-state quantity methods utilize power frequency or harmonic components to construct criteria, but the compensation effect of the arc suppression coil significantly weakens the zero-sequence current amplitude, causing a sharp drop in sensitivity when the transition resistance exceeds 1kΩ, and they are easily affected by three-phase imbalance interference. Transient quantity methods achieve location by extracting high-frequency transient signals in the initial stage of the fault, such as the first half-wave method, transient energy method, and wavelet transform. However, the energy of high-resistivity fault transient signals is dispersed and decays rapidly; in strong noise environments or when the initial phase angle of the fault is small, its characteristics are easily submerged, leading to location failure. Furthermore, methods based on waveform similarity comparison (such as correlation coefficient and grey relational analysis) have been widely studied, using the similarity of zero-sequence current waveforms at different measurement points to classify sections. However, these methods ignore the essential differences in the capacitance to ground of each feeder section and only focus on waveform morphology while neglecting the fault intensity information contained in the signal amplitude, making them prone to misjudgment under noise interference. Although multi-criteria fusion technology attempts to integrate multiple feature quantities, the setting of the weights of each criterion often relies on experience, lacks an objective physical basis, and is difficult to adapt to complex and ever-changing fault scenarios.
[0004] Therefore, how to start from the essential differences of the zero-sequence equivalent model, explore the deep characteristics of high-resistivity faults, and achieve stable and interpretable segment location in a strong noise environment has become a key problem that needs to be solved by those skilled in the art. Summary of the Invention
[0005] To address the shortcomings of existing high-resistivity fault feature extraction methods, such as difficulty in effectively utilizing the essential differences in the zero-sequence equivalent models of feeder segments, sensitivity to noise interference, and difficulty in accurately extracting weak fault features in strong noise environments, a new method for high-resistivity fault feature extraction and location in distribution networks based on feeder segment model mismatch identification is proposed. This method leverages the fundamental differences between healthy and faulty segments in the zero-sequence equivalent models, utilizing the energy concentration characteristics of the segment's zero-sequence current difference in the low-frequency band. It aligns the effective signal components through low-frequency RMS value normalization, exhibiting good robustness to measurement noise. Furthermore, a first-order accumulation generator operator is employed to enhance signal features and suppress random noise interference. By constructing a comprehensive mismatch index that integrates model mismatch degree and fault intensity indicators, the dimensions of a single criterion are expanded, giving the location criterion a clear physical meaning. This calculation method is simple, enabling reliable segment location of high-resistivity faults under complex operating conditions such as strong noise, nonlinear arcs, and high-proportion renewable energy integration.
[0006] The present invention adopts the following technical solution.
[0007] The high-resistivity fault location method for distribution networks based on feeder segment model mismatch identification includes:
[0008] Step 1: When the instantaneous calculated value of the zero-sequence voltage of the bus exceeds the preset threshold, the fault location is initiated. The main station recalls and collects the zero-sequence voltage at the bus and the zero-sequence current at each measuring point in the feeder, and calculates the derivative of the zero-sequence voltage and the segment zero-sequence current difference of each feeder segment.
[0009] Step 2: Calculate the effective values of the zero-sequence voltage derivative and the segment zero-sequence current difference of each feeder segment in the selected low-frequency band, and normalize the corresponding waveforms using the effective value as a scaling factor.
[0010] Step 3: Apply the first-order accumulation generation operator to the normalized zero-sequence voltage derivative and the segment zero-sequence current difference to enhance the processing, so as to suppress random noise and amplify fault characteristics.
[0011] Step 4: Calculate the root mean square error between the zero-sequence voltage derivative and the zero-sequence current difference of each feeder segment after enhancement treatment, and use it as the model mismatch degree.
[0012] Step 5: Multiply the model mismatch degree of each feeder segment by its corresponding fault intensity index, where the fault intensity index is the effective value of the zero-sequence current difference of the original segment in the selected low-frequency band, to obtain the comprehensive mismatch index.
[0013] Step 6: Compare the overall mismatch index of all feeder segments and determine the feeder segment corresponding to the maximum value as the faulty segment.
[0014] Preferably, step 1 specifically includes:
[0015] Step 1.1: When the instantaneous calculated value of the zero-sequence voltage of the bus exceeds the preset threshold, fault location is initiated. The initiation logic is as follows:
[0016]
[0017] In the formula, This is the zero-sequence voltage of the bus. It is a time series; This refers to a power frequency cycle in the power grid; This is the rated phase voltage of the busbar; and These are preset coefficients;
[0018] Step 1.2: After startup, the main station recalls and collects the zero-sequence voltage at the bus and the zero-sequence current at each measuring point in the feeder, and calculates the derivative of the zero-sequence voltage and the segmental zero-sequence current difference of each feeder segment:
[0019]
[0020] In the formula, The derivative of the zero-sequence voltage of the busbar; For section The zero-sequence current difference in the section; for Zero-sequence current at upstream measuring point of the section; for Downstream of section The zero-sequence current of the branch.
[0021] Preferably, step 2 specifically includes:
[0022] Step 2.1: Select the zero-sequence voltage derivative of the first power frequency cycle of the fault and the segment zero-sequence current difference of each feeder segment, and calculate their effective values in the selected low-frequency band respectively:
[0023]
[0024] In the formula, and These are the zero-sequence voltage derivative and the segment. The effective value of the zero-sequence current difference in the selected low-frequency band; and These are the zero-sequence voltage derivative and the segment. The discrete Fourier transform one-sided amplitude spectrum of the zero-sequence current difference in the segment. For sequence number; These are the weighting coefficients; For the sequence number is The corresponding frequency, and To select the upper and lower limits of the low-frequency band;
[0025] Step 2.2: Normalize the corresponding waveform using this effective value as a scaling factor:
[0026]
[0027] In the formula, and These are the normalized zero-sequence voltage derivative and the segment. The zero-sequence current difference in the section, The sampling point number; For section The zero-sequence current difference in the section; This represents the number of sampling points within one power frequency cycle.
[0028] Preferably, step 3 specifically includes:
[0029] The normalized zero-sequence voltage derivative and the segment zero-sequence current difference are enhanced by applying a first-order accumulation generator to suppress random noise and amplify fault characteristics.
[0030]
[0031] In the formula, for Normalized zero-sequence voltage derivative and segment after enhancement by the first-order accumulation generator operator The zero-sequence current difference in the section.
[0032] Preferably, step 4 specifically includes:
[0033] Calculate the root mean square error between the zero-sequence voltage derivative and the zero-sequence current difference of each feeder segment after enhancement treatment, and use it as the model mismatch degree:
[0034]
[0035] In the formula, For section The model mismatch.
[0036] Preferably, step 5 specifically includes:
[0037] Multiply the model mismatch degree of each feeder segment by its corresponding fault intensity index, where the fault intensity index is the effective value of the zero-sequence current difference of the original segment in the selected low-frequency range, to obtain the comprehensive mismatch index:
[0038]
[0039] In the formula, For section The comprehensive mismatch index.
[0040] Preferably, step 6 specifically includes:
[0041] Compare the overall mismatch index of all feeder segments , will satisfy feeder segment The comprehensive mismatch index The largest segment is identified as the faulty segment.
[0042] Compared with the prior art, the beneficial effects of the present invention are:
[0043] In this invention, based on the essential difference between healthy and faulty feeder segments in the zero-sequence equivalent model, the model mismatch characteristics of the faulty segment are revealed by defining an error current. Utilizing the energy concentration characteristics of the segment's zero-sequence current difference in the low-frequency band, the low-frequency RMS value is used as a scaling factor to normalize the segment's zero-sequence current difference and zero-sequence voltage derivative. This effectively suppresses strong noise interference, improves the alignment accuracy of effective signal components, and avoids the problem of traditional maximum absolute value normalization methods being susceptible to outliers in noisy environments. Furthermore, a first-order accumulation generator operator is used to enhance the normalized signal, amplifying fault characteristics while suppressing random noise, making the characteristic differences between the healthy and faulty segments more significant. Based on this, the root mean square error representing the model mismatch degree and the low-frequency effective value representing the fault intensity are multiplied and fused to construct a comprehensive mismatch index with clear physical meaning as a location criterion. This criterion reflects both the degree of deviation of the feeder segment from the ideal capacitor model and the intensity of the fault current, avoiding the problem of lack of objective basis for weight allocation in multi-criterion fusion, and making the location results more interpretable and robust.
[0044] The constructed comprehensive mismatch index This invention achieves a physical multiplication fusion of model mismatch and fault intensity, preserving waveform morphology and amplitude energy information without subjective weight allocation. The physical meaning is clear—it directly quantifies the fault current components in the segment that cannot be explained by the self-capacitance model after first-order accumulation generation. This invention requires no prior line parameters, is computationally simple, and has strong real-time performance. It significantly outperforms existing methods in terms of noise immunity, principle universality, criterion interpretability, and engineering applicability. This invention only requires the acquisition of bus zero-sequence voltage and zero-sequence current at various measurement points along the line, making data acquisition convenient. It is applicable to complex conditions such as nonlinear arc faults, high-proportion renewable energy integration, and strong noise interference, achieving reliable segment location of high-resistance faults. It has the advantages of strong anti-interference capability, wide applicability, and high engineering practical value.
[0045] The above overview is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments, and features described above, further aspects, embodiments, and features of the invention will become readily apparent from the accompanying drawings and the following detailed description. Attached Figure Description
[0046] Figure 1 The flowchart shows the algorithm for locating faulty sections.
[0047] Figure 2 This is a schematic diagram of a 10kV radial resonant grounding system.
[0048] Figure 3 Visualization of the impact of strong noise on this method: (a) Normalized results; (b) 1-AGO enhancement results; (c) Proportional distribution of each index. Detailed Implementation
[0049] Next, the technical solutions of the embodiments of the present invention will be clearly and comprehensively described with reference to the accompanying drawings. It is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of them. It should be noted that the accompanying drawings are schematic and not drawn to scale. For clarity and convenience, the relative sizes and proportions of the parts shown in the drawings have been exaggerated or reduced in size; any size is merely exemplary and not restrictive.
[0050] Example 1
[0051] Step 1.1: When the instantaneous calculated value of the zero-sequence voltage of the bus exceeds the preset threshold, fault location is initiated. The initiation logic is as follows:
[0052]
[0053] In the formula, This is the zero-sequence voltage of the bus. It is a time series; This refers to a power frequency cycle in the power grid; This is the rated phase voltage of the busbar; and These are preset coefficients;
[0054] Step 1.2: After startup, the main station recalls and collects the zero-sequence voltage at the bus and the zero-sequence current at each measuring point in the feeder, and calculates the derivative of the zero-sequence voltage and the segmental zero-sequence current difference of each feeder segment:
[0055]
[0056] In the formula, The derivative of the zero-sequence voltage of the busbar; For section The zero-sequence current difference in the section; for Zero-sequence current at upstream measuring point of the section; for Downstream of section The zero-sequence current of the branch.
[0057] Step 2.1: Select the zero-sequence voltage derivative of the first power frequency cycle of the fault and the segment zero-sequence current difference of each feeder segment, and calculate their effective values in the selected low-frequency band respectively:
[0058]
[0059] In the formula, and These are the zero-sequence voltage derivative and the segment. The effective value of the zero-sequence current difference in the selected low-frequency band; and These are the zero-sequence voltage derivative and the segment. The discrete Fourier transform one-sided amplitude spectrum of the zero-sequence current difference in the segment. For sequence number; These are the weighting coefficients; For the sequence number is The corresponding frequency, and In order to select the upper and lower limits of the low frequency band, the frequency band selected in this embodiment is 0~100 Hz;
[0060] Step 2.2: Normalize the corresponding waveform using this effective value as a scaling factor:
[0061]
[0062] In the formula, and These are the normalized zero-sequence voltage derivative and the segment. The zero-sequence current difference in the section, The sampling point number; For section The zero-sequence current difference in the section; This represents the number of sampling points within one power frequency cycle.
[0063] Step 3: Apply a first-order accumulation generator to the normalized zero-sequence voltage derivative and the segment zero-sequence current difference to enhance them, thereby suppressing random noise and amplifying fault characteristics.
[0064]
[0065] In the formula, for Normalized zero-sequence voltage derivative and segment after enhancement by the first-order accumulation generator operator The zero-sequence current difference in the section.
[0066] Step 4: Calculate the root mean square error between the zero-sequence voltage derivative and the zero-sequence current difference of each feeder segment after enhancement treatment, as the model mismatch degree.
[0067]
[0068] In the formula, For section The model mismatch.
[0069] Step 5: Multiply the model mismatch degree of each feeder segment by its corresponding fault intensity index, where the fault intensity index is the effective value of the zero-sequence current difference of the original segment in the selected low-frequency range, to obtain the comprehensive mismatch index:
[0070]
[0071] In the formula, For section The comprehensive mismatch index.
[0072] Step 6: Compare the overall mismatch index of all feeder segments. , will satisfy feeder segment The comprehensive mismatch index The largest segment is identified as the faulty segment.
[0073] Example 2
[0074] To verify the effectiveness of the proposed method, a typical radial power distribution network simulation model was built in PSCAD / EMTDC, as follows: Figure 2 As shown. The arc suppression coil is set to an overcompensation degree of 8%, corresponding to... The cable lines are of type YJV22-3*400, with a total length of 34 km; the overhead lines are of type JKLYJ-240, with a total length of 28 km; detailed parameters are listed in Table 1. DG1 and DG2 represent a wind farm and a photovoltaic power station, respectively, with capacities of 3.5 MW and 0.8 MW. Six zero-sequence current transformers M1~M6 are distributed throughout the network, dividing the network into six sections, as detailed in Table 2. One zero-sequence voltage transformer measures the zero-sequence voltage of the busbar. The sampling frequency of all transformers is 10 kHz, corresponding to N=200 sampling points per cycle.
[0075] Table 1 Parameters of Overhead Cable Lines
[0076] Table 2 Feeder Section Division
[0077] exist Figure 2 In the model shown, different fault sections and grounding resistances are considered. and fault initial phase angle Tests were conducted, and the results are summarized in Table 3. In all cases, due to the fault section deviating from the capacitance model, its... The values were all significantly higher than those in other segments. Furthermore, due to... Since the faulty section is typically large, after restoring the error to its original dimensions, the faulty section and the healthy section are... The differences in values are more obvious, thus enabling accurate fault location.
[0078] Table 3 Location Results
[0079] Taking the fourth case in Table 3 as an example, the signal-to-noise ratio of the zero-sequence current difference in the segment reached -3 dB. Figure 3 As shown in (a), the proposed low-frequency RMS normalization method reduces noise interference, making... and The effective components are well aligned. However, due to excessive noise, the distinction between faulty and healthy sections remains unclear. Figure 3 (b) shows that after processing by the first-order accumulation generator operator, random noise is suppressed and low-frequency signal characteristics are enhanced, thereby reducing the impact of noise on the model mismatch calculation. Nevertheless, due to strong noise interference, a large model mismatch still occurs in segment 2. Figure 3 (c) shows the proportional distribution of each indicator. Since section 2 is an overhead line, its zero-sequence capacitance is relatively small, therefore its... The value is relatively low. Ultimately, the corresponding The value remained small, and the fault was correctly located in segment 1.
[0080] The applicant of this invention has provided a detailed explanation and description of embodiments of the invention based on the accompanying drawings. However, those skilled in the art should recognize that the above embodiments are merely preferred embodiments of the invention. The detailed explanation is intended to assist the reader in gaining a deeper understanding of the spirit of the invention, rather than to limit the scope of protection of the invention. On the contrary, any improvements or variations made based on the spirit of the invention should be included within the scope of protection of the invention.
Claims
1. A method for locating high-resistivity faults in distribution networks based on feeder segment model mismatch identification, characterized in that, The method includes the following steps: Step 1: When the instantaneous calculated value of the zero-sequence voltage of the bus exceeds the preset threshold, the fault location is initiated. The main station recalls and collects the zero-sequence voltage at the bus and the zero-sequence current at each measuring point in the feeder, and calculates the derivative of the zero-sequence voltage and the segment zero-sequence current difference of each feeder segment. Step 2: Calculate the effective values of the zero-sequence voltage derivative and the segment zero-sequence current difference of each feeder segment in the selected low-frequency band, and normalize the corresponding waveforms using the effective value as a scaling factor. Step 3: Apply the first-order accumulation generation operator to the normalized zero-sequence voltage derivative and the segment zero-sequence current difference to enhance the processing, so as to suppress random noise and amplify fault characteristics. Step 4: Calculate the root mean square error between the zero-sequence voltage derivative and the zero-sequence current difference of each feeder segment after enhancement treatment, and use it as the model mismatch degree. Step 5: Multiply the model mismatch degree of each feeder segment by its corresponding fault intensity index, where the fault intensity index is the effective value of the zero-sequence current difference of the original segment in the selected low-frequency band, to obtain the comprehensive mismatch index. Step 6: Compare the overall mismatch index of all feeder segments and determine the feeder segment corresponding to the maximum value as the faulty segment.
2. The method for locating high-resistivity faults in distribution networks based on feeder segment model mismatch identification as described in claim 1, characterized in that, Step 1 specifically includes: Step 1.1: When the instantaneous calculated value of the zero-sequence voltage of the bus exceeds the preset threshold, fault location is initiated. The initiation logic is as follows: In the formula, Let t be the zero-sequence voltage of the bus, and t be the time series. This refers to a power frequency cycle in the power grid; This is the rated phase voltage of the busbar; These are preset coefficients; Step 1.2: After startup, the main station recalls and collects the zero-sequence voltage at the busbar and the zero-sequence current at each measuring point in the feeder, and calculates the derivative of the zero-sequence voltage and the segmental zero-sequence current difference of each feeder segment: In the formula, The derivative of the zero-sequence voltage of the busbar; The zero-sequence current difference in segment k; The zero-sequence current at the upstream measuring point of section k; For the downstream of section k The zero-sequence current of the branch.
3. The method for locating high-resistivity faults in distribution networks based on feeder segment model mismatch identification according to claim 1, characterized in that, Step 2 specifically includes: Step 2.1: Select the zero-sequence voltage derivative of the first power frequency cycle of the fault and the segment zero-sequence current difference of each feeder segment, and calculate their effective values in the selected low-frequency band respectively: In the formula, and These are the effective values of the zero-sequence voltage derivative and the segment zero-sequence current difference in segment k within the selected low-frequency band, respectively. and , i, are the discrete Fourier transform one-sided amplitude spectra of the zero-sequence voltage derivative and the segment zero-sequence current difference of segment k, respectively, where i is the sequence number; These are the weighting coefficients; The frequency corresponding to sequence number i. and To select the upper and lower limits of the low-frequency band; Step 2.2: Normalize the corresponding waveform using this effective value as a scaling factor: In the formula, and They are the normalized zero-sequence voltage derivative and the segment zero-sequence current difference in segment k, respectively, where n is the sampling point number; denoted as segment zero-sequence current difference in segment k; n is the number of sampling points within one power frequency cycle.
4. The method for locating high-resistivity faults in distribution networks based on feeder segment model mismatch identification according to claim 1, characterized in that, Step 3 specifically includes: The normalized zero-sequence voltage derivative and the segment zero-sequence current difference are enhanced by applying a first-order accumulation generator to suppress random noise and amplify fault characteristics. In the formula, for The normalized zero-sequence voltage derivative and the segment zero-sequence current difference of segment k after enhancement by the first-order accumulation generator operator.
5. The method for locating high-resistivity faults in distribution networks based on feeder segment model mismatch identification according to claim 1, characterized in that, Step 4 specifically includes: Calculate the root mean square error between the zero-sequence voltage derivative and the zero-sequence current difference of each feeder segment after enhancement treatment, and use it as the model mismatch degree: In the formula, Let k be the model mismatch degree for segment k.
6. The method for locating high-resistivity faults in distribution networks based on feeder segment model mismatch identification according to claim 1, characterized in that, Step 5 specifically includes: Multiply the model mismatch degree of each feeder segment by its corresponding fault intensity index, where the fault intensity index is the effective value of the zero-sequence current difference of the original segment in the selected low-frequency range, to obtain the comprehensive mismatch index: In the formula, Let be the comprehensive mismatch index for segment k.
7. The method for locating high-resistivity faults in distribution networks based on feeder segment model mismatch identification according to claim 1, characterized in that, Step 6 specifically includes: Compare the overall mismatch index of all feeder segments , will satisfy feeder segment The comprehensive mismatch index The largest segment is identified as the faulty segment.