Power distribution network high resistance fault locating method and locating system based on multi-stage joint filtering

Abstract

The application discloses a power distribution network high-resistance fault positioning method and positioning system based on multi-stage joint filtering, which comprises collecting power data of a power distribution network in real time, including a zero sequence voltage signal and a zero sequence current signal; using a multi-stage joint filtering technology, a signal flow graph strategy is constructed to process the power data, and output data is generated; signal disturbance is removed while fault feature extraction is completed; a dynamic threshold is established, and comparison and analysis are carried out based on the output data and the dynamic threshold to determine the fault occurrence time; based on the fault occurrence time, a trial method and characteristic polarity accumulation are used to generate a fault measure; based on the amplitude and polarity of the fault measure two-dimensional characteristics, and according to the distribution characteristics of the fault measure of each measuring point relative to the fault point, a recursive positioning algorithm with an adaptive threshold is used to position the fault section, the fault feeder is quickly positioned, and the adaptive threshold is used for recursive positioning to eliminate abnormal values caused by interference, so that accurate positioning under complex working conditions is realized.

Description

Technical Field

[0001] This invention relates to the field of power system relay protection technology, specifically to a method and system for locating high-resistivity faults in distribution networks based on multi-level joint filtering. Background Technology

[0002] Resonant grounding systems are widely used in distribution networks to suppress single-phase ground fault currents. However, high-impedance faults (HIFs) remain a technical challenge for detection and location due to their large transition resistance (hundreds to thousands of ohms) and weak fault current (several amperes to a few tenths of an ampere). Traditional methods rely on the amplitude or phase characteristics of the zero-sequence current, but the signal amplitude under high-impedance faults is close to the background noise level, making direct identification difficult. With the large-scale integration of distributed generation (DG), the complexity of distribution network topologies has increased, exacerbating problems such as line parameter asymmetry, harmonic pollution, and load fluctuations, further amplifying the challenge of high-impedance fault location.

[0003] Current positioning technologies are mostly based on transient signal analysis, such as wavelet transform or time-frequency analysis, which requires multi-scale decomposition and frequency domain feature extraction of the signal, resulting in computational complexity as high as [missing information]. The computational requirements for edge devices (such as feeder terminals, FTUs) are stringent, making it difficult to achieve millisecond-level real-time response. While deep learning-based models perform well in the laboratory, their large number of parameters and reliance on high-performance computing resources result in high deployment costs and limited generalization capabilities in practice. Some optimization algorithms (such as least squares parameter identification) improve accuracy through iteration, but this significantly increases latency, failing to meet the requirements for rapid fault isolation. Insufficient anti-interference capability is another core bottleneck. For example, the zero-sequence mutation amplitude comparison method is susceptible to steady-state imbalance components caused by three-phase asymmetry in line parameters, especially in DG access scenarios, where inverter harmonics and power fluctuations exacerbate background noise, leading to a higher misjudgment rate. Existing algorithms are typically designed for single interferences and lack comprehensive suppression capabilities against composite interferences (harmonics, noise, and imbalance superposition), limiting their engineering practicality.

[0004] Existing high-resistivity fault location methods suffer from limitations in computational complexity and anti-interference capabilities, hindering their practical application in complex power distribution networks. To address the demands of modern power systems for rapid fault isolation and high reliability, a lightweight and robust fault location method and system are urgently needed to solve core issues such as weak signal extraction, multi-source interference suppression, and real-time response. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for locating high-resistance faults in distribution networks based on multi-level joint filtering. By jointly filtering the zero-sequence voltage and zero-sequence current measurements, using a multi-level filtering architecture to suppress interference and highlight fault characteristics, a lightweight algorithm to reduce computational load, and an adaptive recursive algorithm, the invention achieves rapid and accurate location of high-resistance faults in complex operating conditions through the collaboration of the master station and distributed measurement devices.

[0006] This invention adopts the following technical solution: a high-resistivity fault location method for distribution networks based on multi-level joint filtering, comprising the following steps:

[0007] Real-time acquisition of power data from the distribution network, including zero-sequence voltage and zero-sequence current signals;

[0008] By utilizing multi-level joint filtering technology, a signal flow graph strategy is constructed to process power data and generate output data.

[0009] Establish a dynamic threshold and compare the output data with the dynamic threshold to determine the time when the fault occurs;

[0010] Based on the time of failure occurrence, a fault measure is generated using a trial-and-error method and feature polarity accumulation.

[0011] Based on the dual-dimensional features of fault measurement amplitude and polarity, and according to the distribution characteristics of fault measurement at each measurement point relative to the fault point, a recursive localization algorithm with adaptive threshold is used to locate the fault section.

[0012] Preferably, a fixed sampling frequency is used at various detection points distributed throughout the distribution network. The zero-sequence voltage and zero-sequence current signals at the detection point are acquired in real time.

[0013] Preferably, at least the latest two power grid frequency cycles are stored based on rolling storage technology. zero-sequence voltage With zero-sequence current signal k is the detection point number.

[0014] Preferably, the signal flow graph strategy includes:

[0015] For zero-sequence voltage Use window length is The differential filter is used to process the voltage, resulting in the processed zero-sequence voltage output value. ;

[0016] For zero-sequence current Use window length is The zero-sequence current output value is obtained by processing the signal through a sliding accumulator filter. ;

[0017] Zero-sequence voltage output value and zero-sequence current output value Multiply and fuse to obtain the fused output value. ;

[0018] For output values Use window length is The output value is obtained by processing the sliding accumulator filter. .

[0019] Preferably, Equal to one power frequency cycle Internal sampling frequency The number of sampling points; .

[0020] Preferably, the process of determining the time of fault occurrence includes:

[0021] Using zero-sequence voltage output value Constructing difference values :

[0022] ;

[0023] Based on sliding window Calculate dynamic threshold :

[0024] ;

[0025] in, The mean of the differences within the window. The standard deviation of the differences within the window. This is the sensitivity coefficient;

[0026] when When the number of points continuously exceeds a threshold, it is marked as a candidate mutation point; and when the number of points continuously exceeds a threshold, it is marked as a candidate mutation point. The sampling point was determined to be the time when the fault started. .

[0027] Preferably, the fault measurement generation includes:

[0028] judge The polarity at the initial moment of the fault is determined using a trial-and-error method to obtain a trial value. :

[0029] ;

[0030] Based on the trial value The polarity of the fixed window Fault measures are generated by accumulating eigenpolarities. :

[0031] If the trial value Then calculate the time within one power frequency cycle after the fault. The sum of all positive values ​​yields the fault measure. :

[0032] ;

[0033] If the trial value Then calculate the time within one power frequency cycle after the fault. The sum of all negative values ​​yields the fault measure. :

[0034] .

[0035] Preferably, the fault segment is located using a recursive localization algorithm with an adaptive threshold, including:

[0036] Determine the fault measurement at the beginning of each feeder If all signs are positive, the fault is determined to be a bus fault; if there are negative values, the fault is determined to be a feeder fault.

[0037] Select The feeder where the measuring point is located is directly identified as a faulty feeder, and is compared sequentially with the measuring points downstream of this measuring point:

[0038] If there exists a measurement point that satisfies And all adjacent measuring points downstream of this measuring point satisfy the following conditions. If so, the fault location is determined to be in the area between the measuring point and its downstream adjacent measuring point;

[0039] If the measuring point at the end of the feeder exists If so, the fault location is determined to be in the area downstream of the measuring point and at the end of the line.

[0040] This invention also discloses a high-resistivity fault location system for distribution networks based on multi-level joint filtering, comprising:

[0041] Signal acquisition module: Real-time acquisition of power data from the distribution network, including zero-sequence voltage signal and zero-sequence current signal;

[0042] Signal processing module: It incorporates a signal flow graph strategy and combines multi-level joint filtering technology to process power data and generate output data, including zero-sequence voltage output values. and zero-sequence current output value and the output value after merging and processing the two. ;

[0043] Timing determination module: Utilizing zero-sequence voltage output value Calculate the difference and dynamic threshold And compare the difference values and dynamic threshold Determine and output the sampling point at the start time of the fault. ;

[0044] Measurement generation module: Uses trial and error and characteristic polarity accumulation to generate fault measures, and uploads the fault measures to the main station;

[0045] The main station includes a fault location module, which incorporates a recursive location algorithm with adaptive thresholds to locate faulty sections based on fault metrics.

[0046] Beneficial Effects: This invention utilizes multi-level joint filtering technology and a signal flow graph strategy to process power data, effectively removing the influence of measurement noise and zero-sequence imbalance components, achieving fault feature fusion and enhancement. It can adapt to more complex environments, exhibits better algorithm robustness, and completes fault feature extraction simultaneously with signal descrambling, compressing the process steps and improving algorithm efficiency. Specifically, the signal flow graph strategy achieves stream processing of sampled signals point-by-point; the computational load is proportional to the data size, with each sample point requiring only 5 additions / subtractions and 1 multiplication, resulting in an algorithm complexity of only [value missing]. Therefore, it can run in real time on edge devices with limited computing power, resulting in lower computational complexity.

[0047] In this invention, only the fault measurement can be uploaded to the main station, and the fault measurement is only uploaded to the main station as a signed number, which makes feature extraction more comprehensive. At the same time, the amount of data transmitted with the main station is reduced to a minimum, which greatly improves the efficiency of the localization process.

[0048] In the fault location process, this invention adopts an adaptive threshold recursive location algorithm. It utilizes the two-dimensional features of the amplitude and polarity of the fault measurement, and quickly locates the fault feeder based on the distribution characteristics of each measurement point relative to the fault point. It then uses an adaptive threshold for recursive location to eliminate abnormal values ​​caused by interference, thereby achieving accurate location under complex working conditions. Attached Figure Description

[0049] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0050] In the attached diagram:

[0051] Figure 1 This is a flowchart of the high-resistance fault location method and system for power distribution networks of the present invention;

[0052] Figure 2 This is the multi-level joint filtering signal flow graph of the present invention;

[0053] Figure 3 This is a schematic diagram of the 10kV radial resonant grounding system of the present invention;

[0054] Figure 4 This is a diagram illustrating the effect of the invention under complex working conditions. Detailed Implementation

[0055] The embodiments of the present invention will now be described with reference to the accompanying drawings. The terminology used in the embodiments section is for illustrative purposes only and is not intended to limit the scope of the invention. The embodiments of this application will now be described with reference to the accompanying drawings.

[0056] Example 1: A method for locating high-resistivity faults in distribution networks based on multi-stage joint filtering, referenced Figure 1 As shown: It includes the following steps:

[0057] Step 1: Collect power data from the distribution network in real time, including zero-sequence voltage signals and zero-sequence current signals;

[0058] By using a fixed sampling frequency at various detection points distributed throughout the power distribution network Real-time acquisition of zero-sequence voltage and zero-sequence current signals at the detection point location;

[0059] And based on rolling storage technology, at least the latest two power grid frequency cycles are stored. zero-sequence voltage With zero-sequence current signal k is the detection point number.

[0060] Step 2: Utilize multi-level joint filtering technology to construct a signal flow graph strategy to process power data and generate output data; effectively remove unbalanced components, suppress noise interference, and achieve the fusion and enhancement of fault characteristics;

[0061] The signal flow graph strategy includes multiple related processing modules, such as... Figure 2 As shown, each module is configured to perform the following functions:

[0062] 1) Data filtering and feature extraction module:

[0063] For zero-sequence voltage Use window length is The differential filter is used to process the voltage, resulting in the processed zero-sequence voltage output value. :

[0064] ;

[0065] The amplitude-frequency response curve of the differential filter is as follows:

[0066] ;

[0067] in, For frequency, This is the system power frequency.

[0068] therefore, exist ( The existence of a zero at point () means that the frequency is The signal is completely suppressed, thus eliminating power frequency unbalanced components and high-order harmonic interference;

[0069] For zero-sequence current Use window length is The zero-sequence current output value is obtained by processing the signal through a sliding accumulator filter. :

[0070] ;

[0071] It has a recursive form that reduces computational complexity:

[0072] ;

[0073] The amplitude-frequency response curve of the sliding accumulator filter is as follows:

[0074] ;

[0075] Similar to differential filters, Also there The presence of a zero point at this location can remove power frequency unbalanced components and higher harmonic interference. However, unlike other filters, for low-frequency signals, the sliding accumulator filter has a near-zero value. The gain is increased by a factor of 1, but decreases rapidly as the frequency increases, thereby achieving the purpose of extracting low-frequency signal characteristics, eliminating the influence of fixed frequency, and suppressing high-frequency noise.

[0076] 2) Feature fusion module:

[0077] Zero-sequence voltage output value and zero-sequence current output value Multiply and fuse to obtain the fused output value. :

[0078] ;

[0079] 3) Feature enhancement and result output module:

[0080] For the fused output value Use window length is The output value is obtained by processing the sliding accumulator filter. :

[0081] ;

[0082] It has a recursive form that reduces computational complexity:

[0083] ;

[0084] Through the Utilizing the high-frequency attenuation characteristics of a sliding accumulator filter to suppress the zero-sequence voltage output value The noise interference introduced, while the integral characteristic of the sliding accumulator filter can enhance the initial fault moment. The characteristics of the sliding accumulator filter also delay the flow of the zero-sequence current through the fault path (the path through which the arc suppression coil zero-sequence current flows). The zero-crossing point allows the trial-and-error method in step 4 to extract the feature polarity more accurately;

[0085] In this embodiment, Equal to one power frequency cycle Internal sampling frequency The number of sampling points; ;

[0086] In this embodiment, a multi-stage filtering module is combined with... Figure 2 The signal flow graph shown is executed using a sample-by-sample processing method. and As input, As output, it enables streaming processing of sampled data, thus it is not limited by the data size and can run in real time on terminal devices.

[0087] Step 3: Establish a dynamic threshold and compare the output data with the dynamic threshold to determine the time of fault occurrence. Specifically:

[0088] Using zero-sequence voltage output value Constructing difference values :

[0089] ;

[0090] Based on sliding window Calculate dynamic threshold :

[0091] ;

[0092] in, The mean of the differences within the window. The standard deviation of the differences within the window. The sensitivity coefficient is used in this embodiment. The value is between 3 and 5;

[0093] when When the number of points continuously exceeds a threshold, it is marked as a candidate mutation point; and when the number of points continuously exceeds a threshold, it is marked as a candidate mutation point. The sampling point was determined to be the time when the fault started.

[0094] Step 4: Based on the time of fault occurrence, generate a fault measure using a trial-and-error method and feature polarity accumulation, specifically:

[0095] judge The polarity at the initial moment of the fault is determined using a trial-and-error method to obtain a trial value. :

[0096] ;

[0097] Based on the trial value The polarity of the fixed window Fault measures are generated by accumulating eigenpolarities. :

[0098] If the trial value Then calculate the time within one power frequency cycle after the fault. The sum of all positive values ​​yields the fault measure. :

[0099] ;

[0100] If the trial value Then calculate the time within one power frequency cycle after the fault. The sum of all negative values ​​yields the fault measure. :

[0101] .

[0102] The generated fault metrics can be uploaded to the main station for comprehensive analysis.

[0103] Step 5: Based on the dual-dimensional features of fault measurement amplitude and polarity, and according to the distribution characteristics of fault measurement at each measurement point relative to the fault point, the fault segment is located using a recursive localization algorithm with adaptive thresholds. Specifically:

[0104] Determine the fault measurement at the beginning of each feeder If all signs are positive, the fault is determined to be a bus fault; if there are negative values, the fault is determined to be a feeder fault.

[0105] Select The feeder where the measuring point is located is directly identified as a faulty feeder, and is compared sequentially with the measuring points downstream of this measuring point:

[0106] If there exists a measurement point that satisfies And all adjacent measuring points downstream of this measuring point satisfy the following conditions. If so, the fault location is determined to be in the area between the measuring point and its downstream adjacent measuring point;

[0107] If the measuring point at the end of the feeder exists If so, the fault location is determined to be in the area downstream of the measuring point and at the end of the line.

[0108] Example 2: A high-resistivity fault location system for distribution networks based on multi-stage joint filtering, referencing... Figure 1 As shown, the positioning method based on Embodiment 1 includes:

[0109] Signal acquisition module: Real-time acquisition of power data from the distribution network, including zero-sequence voltage signal and zero-sequence current signal;

[0110] Signal processing module: It incorporates a signal flow graph strategy and combines multi-level joint filtering technology to process power data and generate output data, including zero-sequence voltage output values. and zero-sequence current output value and the output value after merging and processing the two. ;

[0111] Timing determination module: Utilizing zero-sequence voltage output value Calculate the difference and dynamic threshold And compare the difference values and dynamic threshold Determine and output the sampling point at the start time of the fault. ;

[0112] Measurement generation module: Uses trial and error and characteristic polarity accumulation to generate fault measures, and uploads the fault measures to the main station;

[0113] The main station includes a fault location module, which has a built-in recursive location algorithm with adaptive thresholds to locate faulty sections based on fault metrics.

[0114] In a specific case, based on the method of Example 1 or the system of Example 2, a simulation model of a 10.5 kV resonant grounding system is built using the electromagnetic transient simulation software PSCAD / EMTDC, such as... Figure 3 As shown. The cable line uses model YJV22-3*400, and the overhead line uses model JKLYJ-150. The arc suppression coil is overcompensated by 5%, and the inductance of the arc suppression coil is L=0.58H. ~ For different single-phase ground fault points. On the feeder and Measuring points are set at the beginning and end respectively. and Used for fault location; in feeders Measuring points are set at the beginning of the branch feeder. and This is used for fault location. The simulation sampling frequency is 10 kHz, therefore the window length N = 200.

[0115] To investigate the robustness of this method under conditions of both unbalanced components and measurement noise, the feeder was adjusted. The three-phase parameters are artificially created to induce an imbalance in the three-phase parameters. Figure 3 of At point A, a single-phase-to-ground short-circuit fault with a transition resistance of 3500Ω and an initial phase angle of 288° occurred in phase A, with the fault occurring at 3.016s. White noise with a signal-to-noise ratio of 53.7 dB within one power frequency cycle at the initial fault time was added to the zero-sequence voltage, such as... Figure 4 As shown in (a); white noise with an average energy of 1.05 A² is added to the zero-sequence current at each sampling point, as shown in Figure (a). Figure 4 As shown in (b), the signal-to-noise ratios of each measuring point within one power frequency cycle at the initial moment of the fault are respectively 1.99 dB -83.29 dB -22.75 dB -7.73 dB. According to Figure 4 Fault measurement at each measuring point in section (f), measuring point upstream of the fault point The fault measure is negative, while the fault measures at the downstream and intact feeder measuring points are both positive. Therefore, the final location result is determined to be at the measuring point. and The positioning is correct.

[0116] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. For those skilled in the art, after learning the contents described in the present invention, several equivalent changes and substitutions can be made without departing from the principle of the present invention. These equivalent changes and substitutions should also be considered to fall within the protection scope of the present invention.

Claims

1. A method for locating high-resistivity faults in distribution networks based on multi-stage joint filtering, characterized in that: Includes the following steps: Real-time acquisition of power data from the distribution network, including zero-sequence voltage and zero-sequence current signals; By utilizing multi-level joint filtering technology, a signal flow graph strategy is constructed to process power data and generate output data. Establish a dynamic threshold and compare the output data with the dynamic threshold to determine the time of fault occurrence; Based on the time of failure occurrence, a fault measure is generated using a trial-and-error method and feature polarity accumulation. Based on the dual-dimensional features of fault measurement amplitude and polarity, and according to the distribution characteristics of fault measurement at each measurement point relative to the fault point, a recursive localization algorithm with adaptive threshold is used to locate the fault section. The signal flow graph strategy includes: For zero-sequence voltage Use window length is The differential filter is used to process the voltage, resulting in the processed zero-sequence voltage output value. ; For zero-sequence current Use window length is The zero-sequence current output value is obtained by processing the signal through a sliding accumulator filter. ; Zero-sequence voltage output value and zero-sequence current output value Multiply and fuse to obtain the fused output value. ; For output values Use window length is The output value is obtained by processing the sliding accumulator filter. ; The process of determining the time of the fault includes: Using zero-sequence voltage output value Constructing difference values : ； Based on window length Calculate dynamic threshold : ； in, The mean of the differences within the window. The standard deviation of the differences within the window. This is the sensitivity coefficient; when When the number of points continuously exceeds a threshold, it is marked as a candidate mutation point; and when the number of points continuously exceeds a threshold, it is marked as a candidate mutation point. The sampling point was determined to be the time when the fault started. .

2. The method for locating high-resistivity faults in distribution networks based on multi-stage joint filtering according to claim 1, characterized in that: By using a fixed sampling frequency at various detection points distributed throughout the power distribution network The zero-sequence voltage and zero-sequence current signals at the detection point are acquired in real time.

3. The method for locating high-resistivity faults in distribution networks based on multi-stage joint filtering according to claim 1, characterized in that: Equal to one power frequency cycle Internal sampling frequency The number of sampling points, i.e.: .

4. The method for locating high-resistivity faults in distribution networks based on multi-stage joint filtering according to claim 3, characterized in that: The fault metric generation includes: judge The polarity at the initial moment of the fault is determined using a trial-and-error method to obtain a trial value. : ； Based on the trial value The polarity of the fixed window Fault measures are generated by accumulating eigenpolarities. : If the trial value Then calculate the time within one power frequency cycle after the fault. The sum of all positive values ​​yields the fault measure. : ； If the trial value Then calculate the time within one power frequency cycle after the fault. The sum of all negative values ​​yields the fault measure. : 。 5. The method for locating high-resistivity faults in distribution networks based on multi-stage joint filtering according to claim 4, characterized in that: The faulty section is located using a recursive localization algorithm with adaptive thresholds, including: Determine the fault measurement at the beginning of each feeder If all signs are positive, the fault is determined to be a bus fault; if there are negative values, the fault is determined to be a feeder fault. Select The feeder where the measuring point is located is directly identified as a faulty feeder, and is compared sequentially with the measuring points downstream of this measuring point: If there exists a measurement point that satisfies And all adjacent measuring points downstream of this measuring point satisfy the following conditions. If so, the fault location is determined to be in the area between the measuring point and its downstream adjacent measuring point; If the measuring point at the end of the feeder exists If the fault location is determined to be in the area downstream of the measuring point and at the end of the line, then the fault location can be determined.

6. A high-resistivity fault location system for distribution networks based on multi-stage joint filtering, characterized in that: include: Signal acquisition module: Real-time acquisition of power data from the distribution network, including zero-sequence voltage signal and zero-sequence current signal; Signal processing module: It incorporates a signal flow graph strategy and combines multi-level joint filtering technology to process power data and generate output data, including zero-sequence voltage output values. and zero-sequence current output value and the output value after merging and processing the two. ; Timing determination module: Utilizing zero-sequence voltage output value Calculate the difference and dynamic threshold And compare the difference values and dynamic threshold Determine and output the sampling point at the start time of the fault. ; Measurement generation module: Uses trial and error and characteristic polarity accumulation to generate fault measures, and uploads the fault measures to the main station; The main station includes a fault location module, which has a built-in recursive location algorithm with adaptive thresholds to locate faulty sections based on fault measurement. The signal flow graph strategy includes: For zero-sequence voltage Use window length is The differential filter is used to process the voltage, resulting in the processed zero-sequence voltage output value. ; For zero-sequence current Use window length is The zero-sequence current output value is obtained by processing the signal through a sliding accumulator filter. ; Zero-sequence voltage output value and zero-sequence current output value Multiply and fuse to obtain the fused output value. ; For output values Use window length is The output value is obtained by processing the sliding accumulator filter. ; The process of determining the time of the fault includes: Using zero-sequence voltage output value Constructing difference values : ； Based on window length Calculate dynamic threshold : ； in, The mean of the differences within the window. The standard deviation of the differences within the window. This is the sensitivity coefficient; when When the number of points continuously exceeds a threshold, it is marked as a candidate mutation point; and when the number of points continuously exceeds a threshold, it is marked as a candidate mutation point. The sampling point was determined to be the time when the fault started. .

Images