A method for identifying and locating near-end faults in transmission lines based on time-domain waveform feature recognition

Abstract

This invention relates to a method for identifying and locating near-end faults in transmission lines based on time-domain waveform feature recognition. The method involves collecting traveling wave data of fault current at the end of the transmission line and selecting an appropriate horizontal compression factor based on a near-end fault distance threshold. A moving fixed-area sliding window is used to statistically analyze the pixel density distribution of the pixel stacking feature region of the horizontally compressed waveform image. The correlation coefficient between the pixel density distribution of the pixel stacking feature region and a reference function is calculated, and the correlation coefficient is used to determine whether the fault exhibits near-end fault characteristics. If near-end fault characteristics are present, the peak points of the short-window waveform's outer envelope are extracted, and the variance of the time interval between these peak points is calculated. Based on the correlation coefficient and the variance of the time interval, the method ultimately identifies whether the fault is near-end. After confirming a near-end fault, the estimated period of the test sample, the estimated period of the nearest neighbor sample, and the corresponding fault reference distance are interpolated to calculate the fault distance of the test sample. Simultaneously, a visual comparison result with the reference sample is output.

Description

Technical Field

[0001] This invention relates to a method for identifying and locating near-end faults in transmission lines based on time-domain waveform feature recognition, belonging to the field of power system protection technology. Background Technology

[0002] Transmission lines, operating outdoors for extended periods, are highly susceptible to adverse weather conditions such as lightning, storms, and hail, resulting in a significantly higher probability of failure compared to general power grid equipment. Statistics show that most of these failures are permanent. Timely and accurate fault identification and repair can improve power system stability and ensure reliable and continuous power supply. Transmission lines in complex mountainous areas are particularly challenging due to their long transmission distances, harsh and complex geographical environments, and the difficulty of inspection. Without fault location devices, the workload and working hours of power grid personnel would increase dramatically, significantly impacting efficiency. Therefore, fault identification and location for transmission lines can effectively reduce the workload of inspection operations, quickly locate faults, shorten repair cycles, and reduce economic losses. Technically, this ensures the safety and reliability of the power system, resulting in significant economic and social benefits.

[0003] Traditional abrupt change detection ranging methods, when a near-end fault occurs, exhibit a dense distribution of initial traveling waves and subsequent fault point reflected waves within a short period. However, the abrupt change singularity is weak, making traditional singularity-based wavefront detection and single-end ranging methods unreliable and prone to false alarms for non-near-end faults, easily leading to erroneous output results. Therefore, a certain dead zone exists. However, because near-end faults allow observation of the initial traveling wave followed by multiple fault reflected waves, the overall waveform exhibits a periodicity corresponding to the fault distance, as confirmed by a considerable amount of measured data. Therefore, estimating the periodicity has become a key approach for single-end localization of near-end faults. Researching a near-end fault identification method suitable for traveling wave ranging, screening near-end fault data, and outputting ranging results based on historical near-end samples can effectively reduce the ranging dead zone and significantly improve the success rate of automatic single-end analysis. This is of great value for improving the range and effectiveness of traveling wave ranging results. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method for identifying and locating near-end faults in transmission lines based on time-domain waveform feature cognition, thereby realizing the identification and location of near-end faults in the line, effectively identifying near-end faults and outputting ranging results based on historical near-end faults, reducing the dead zone of traveling wave ranging and improving the reliability of near-end fault ranging results.

[0005] The technical solution of this invention is: a method for identifying and locating near-end faults in transmission lines based on time-domain waveform feature recognition, the specific steps of which are as follows:

[0006] Step 1: Read the period, fault distance, waveform and other information of the historical near-end fault sample library preprocessed by the algorithm, collect the fault phase current traveling wave data at the end of the transmission line, and select the optimal ratio of data length and effective pixel length based on the fault distance threshold to make a waveform diagram, so as to ensure that the initial traveling wave of the fault and the reflected wave group of the fault point within the fault distance threshold can be tightly compressed.

[0007] Step 2: Select the first wave head and the area with a suitable pixel length in the waveform, construct a sliding window to calculate the image density of the area, and calculate the correlation coefficient between the pixel density of the sample to be tested and the benchmark function for further confirmation.

[0008] Step 3: Select a window length related to the fault distance threshold and plot the waveform, ensuring that the window can observe the initial traveling wave of the fault and subsequent reflected wave groups from the fault point. Extract the intersection of the waveform peak points and the outer envelope of the waveform within the window as the outer envelope peak point set, and calculate the variance of the distance between the first 5 outer envelope peak points. Check whether the correlation coefficient is greater than the threshold and whether the variance is less than the threshold. If both conditions are met, proceed to Step 4; otherwise, the locking distance measurement algorithm is handed over to manual processing.

[0009] Step 4: Extract the distance between the two points preceding the outer envelope peak as the estimated period, and search for the nearest neighbor period sample and the corresponding reference fault distance sample in the historical near-end fault sample library. Calculate the fault distance of the test sample based on historical fault indications using the estimated period and nearest neighbor sample interpolation, and output a visual waveform comparison of the test sample and the nearest neighbor sample.

[0010] Step 1 specifically refers to:

[0011] Step 1.1: Read the fault current traveling wave data at the end of the transmission line and read the historical near-end fault sample library pre-processed by the algorithm, including the fault waveform library, the fault estimation period library, and the fault reference distance library.

[0012] Step 1.2: Based on the near-end fault distance threshold x TH and plotting line width w L Calculate the ratio k of the optimal data length to the effective pixel length, and use this ratio to plot a waveform, ensuring that near-end fault waveforms within the threshold can be sufficiently compressed. The formula for calculating the ratio k is:

[0013]

[0014] Among them, the near-end fault threshold can be taken as 8km, the line width can be taken as 4.167 pixels, and v is the traveling wave velocity.

[0015] Step 2 specifically includes:

[0016] Step 2.1: Take the initial traveling wave and the subsequent P-pixel long image from the waveform obtained in Step 1 as the pixel stacking feature region. P is calculated by the following formula:

[0017]

[0018] Where m is the reliability coefficient, which can be taken as 1.13, and k is the ratio of the optimal data length to the effective pixel length.

[0019] Step 2.2: Construct a sliding window with a fixed area sw, where the length of the sliding window is sw. L , width sw W The sliding window is moved along the horizontal pixel increment direction to calculate the pixel density η of the pixel stacking feature region, where η is calculated as follows:

[0020]

[0021] In the formula, px is the waveform pixel value, p(c i g i ) is the point (c i g i The pixel value of ).

[0022] Step 2.3: Obtain the empirical function by fitting the waveform density distribution of multiple measured and simulated near-end data. In the formula, the decreasing constant γ is 0.08, and c i Given a horizontal pixel sequence, calculate the correlation coefficient ρ between the pixel density η in the neighborhood of the first wave head of the sample under test and the baseline function F. The calculation formula is as follows:

[0023]

[0024] Step 3 specifically refers to:

[0025] Step 3.1: After cutting the first wave head The data is plotted as waveforms to ensure that the initial traveling wave of the near-end fault and the subsequent reflected wave group at the fault point can be observed within the window.

[0026] Step 3.2: Extract the intersection of the waveform peak points and the waveform outer envelope within the window. The outer envelope contour point set S is obtained by the Alpha-Shape algorithm, as shown in equations (4) and (5): Take any two points p and q, and calculate the center (x) of the circle passing through points p and q with radius α using equation (4). c ,y c ), determine whether there are other points inside the circle. If not, then p and q belong to the point set S.

[0027]

[0028]

[0029] Extract the first 5 points from the intersection of the outer envelope contour points and the peak points, and calculate the variance of their distance, as shown in the following formula:

[0030]

[0031] In the formula, n represents the length of the sequence in the intersection of the peak points and the outer envelope of the waveform under test. t pi+1 t pi This represents the time series of the outer envelope peak points.

[0032] Step 3.3: Based on the magnitude of the correlation coefficient ρ and the variance δ of the time series intervals, select those with ρ ≥ ρ TH And δ < δ TH Using samples as near-end faults, and combining extensive experimental data, simulation waveforms, and domain knowledge, δ can be taken as... TH =8、ρ TH =0.9. Determine whether the time interval variance of the test sample meets the threshold. If it does, proceed to Step 4. If it does not, close the ranging algorithm and hand the test sample over to manual processing.

[0033] Step 4 specifically refers to:

[0034] Step 4.1: Based on the sample to be tested meeting the variance threshold, the distance between the first two points of the outer envelope peak point sequence is taken as the estimated period T1 of the sample. The historical near-end fault sample library pre-processed by the algorithm is read, including the fault waveform library, the fault estimation period library, and the fault reference distance library.

[0035] Step 4.2: Determine the nearest neighbor sample T of the estimated period T1 of this sample in the historical fault estimated period database. L T R And retrieve the corresponding historical fault reference distance x L x R The nearest neighbor period should satisfy the following formula:

[0036] T L <T1<T R (8)

[0037] Step 4.3: Calculate the fault distance of the test sample based on historical fault indications using the estimated period and nearest neighbor sample interpolation. To ensure the validity of the interpolation results, add a left boundary (T0,x0) = (0,0) and a right boundary threshold sample (T0,x0) to the historical samples. max ,x max = (57, 7.9). The reference fault distance x1 corresponding to T1 can be obtained by interpolation according to equation (9), and the visualization comparison results are output.

[0038]

[0039] The final output is the ranging result x1 and a visual comparison of the waveform of the reference sample.

[0040] The beneficial effects of this invention are: by combining historical data and domain knowledge, this invention can automatically identify traveling wave data of near-end faults, reduce the inherent error of ranging during near-end faults, make the results more in line with engineering needs, significantly improve the success rate of single-end automatic analysis, the effectiveness and interpretability of ranging results, and overcome the shortcomings of traditional single-end traveling wave methods such as difficulty in identifying near-end faults, low reliability of automatic analysis, high deviation of ranging results, and poor stability. It is widely applicable to overhead transmission lines, plant-station outgoing lines and direct distribution lines of various voltage levels. Attached Figure Description

[0041] Figure 1 This is a measured current traveling wave waveform diagram of Embodiment 1 of the present invention;

[0042] Figure 2 This is a diagram illustrating the specific implementation steps included in Step 1 of the present invention;

[0043] Figure 3 This is a waveform diagram that is 12 times the effective pixel length in Embodiment 1 of the present invention;

[0044] Figure 4 This is a diagram illustrating the specific implementation steps included in Step 2 of the present invention;

[0045] Figure 5 This is a waveform pixel density distribution and reference function diagram of the 22-pixel length region following the first wave head in Embodiment 1 of the present invention;

[0046] Figure 6 This is a diagram illustrating the specific implementation steps included in Step 3 of the present invention;

[0047] Figure 7 These are the waveform diagram and outer envelope peak point diagram of the 300μs following the first wave head in Embodiment 1 of the present invention;

[0048] Figure 8 This is a diagram illustrating the specific implementation steps included in Step 4 of the present invention;

[0049] Figure 9 This is a comparison of the visualized waveforms of the sample to be tested and its nearest neighbor samples in Embodiment 1 of the present invention;

[0050] Figure 10 This is a flowchart of the steps of the present invention. Detailed Implementation

[0051] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0052] Example 1: Measured waveforms acquired by a 1MHz sampling rate waveform recorder in a 220kV transmission line are as follows. Figure 1 As shown, the faulty line is 93.11km long, and the reference fault distance obtained by the subsequent line inspection is 5km. The fault is a C-phase ground fault.

[0053] like Figure 10 As shown, the specific implementation steps are as follows:

[0054] Step 1: Collect traveling wave data of the fault phase current at the end of the transmission line and select an appropriate horizontal compression factor based on the near-end fault distance threshold. Display the data with the appropriate horizontal compression factor in the image. The specific implementation steps are as follows: Figure 2 As shown.

[0055] Step 1.1: Preprocess historical near-end fault samples to obtain fault waveform library, fault cycle library and corresponding fault distance library.

[0056] Step 1.2: Collect fault traveling wave data from the transmission line and calculate the ratio of optimal data length to effective pixel length, k = 12.

[0057]

[0058] Plot the current traveling wave data with a length 12 times the effective pixel length to create a waveform, as shown below. Figure 3 This ensures that the fault waveform is fully compressed.

[0059] Step 2: Using a sliding window with a fixed area, statistically analyze the pixel density distribution of the pixel stacking feature region. Calculate the correlation coefficient between the pixel density distribution of the pixel stacking feature region and the baseline function. Use the correlation coefficient to determine whether a near-end fault characteristic exists. The specific implementation steps are as follows: Figure 4 As shown:

[0060] Step 2.1: Take the initial traveling wave and subsequent images of a specific pixel length from the waveform obtained in Step 1 as the pixel stacking feature region. Calculate the pixel length P = 25 of the pixel stacking feature region according to equation (3):

[0061]

[0062] Step 2.2: Construct a fixed-area sliding window, translate the sliding window, and according to formula (4), statistically analyze the pixel density distribution along the positive horizontal direction of the pixel stacking feature area η=[479, 909, 1336, 1363, 1363, 1361, 1181, 1001, 821, 763, 676, 589, 505, 464, 423, 454, 427, 402, 369, 344, 312, 239, 209, 160, 130].

[0063] Step 2.3: The baseline function F = e is obtained by fitting the shape density distribution of multiple measured and simulated near-end fault data. -γx γ is taken as 0.06. According to equation (4), the correlation coefficient ρ = 0.98 between the pixel density of the sample to be tested and the reference function is calculated. The pixel density distribution and the reference function are as follows: Figure 5 As shown.

[0064] Step 3. Based on the presence of near-end fault characteristics, extract the peak points of the outer envelope of the short-window waveform and calculate the variance of the time interval between these peak points. Finally, determine whether it is a near-end fault based on the variance and similarity coefficient. The specific implementation steps are as follows: Figure 6 As shown:

[0065] Step 3.1: Plot the data from the first wavefront and the subsequent 300μs to ensure that the first wavefront of the fault waveform and its subsequent reflected wave groups can be observed, such as... Figure 7 As shown.

[0066] Step 3.2: Take the intersection of the outer envelope boundary points of the first 5 fault waveforms and the peak points, that is, take the first 5 outer envelope peak points as shown below. Figure 7 As shown, the variance of the spacing δ = 0.67 is calculated according to equation (7).

[0067] Step 3.3: Based on the correlation coefficient threshold ρ = 0.98 > 0.9 and variance δ = 0.67 < 8, the near-end fault criterion is met, and the case is determined to be a near-end fault, proceeding to Step 4.

[0068] Step 4: After confirming a near-end fault, calculate the fault distance of the test sample by interpolating the estimated period of the sample under test, the estimated period of the nearest neighbor samples, and the corresponding fault reference distance. Simultaneously, output a visual comparison result with the reference sample. The specific implementation steps are as follows: Figure 8 As shown:

[0069] Step 4.1: Estimate the waveform period of the near-end fault, T1 = t p2 -t p1 =41μs. The nearest sample period and sample reference distance were searched in the historical data. The estimated period and nearest sample parameters are shown in the table below:

[0070] Table 1: Samples to be tested and neighboring samples

[0071]

[0072] Step 4.2: Select the nearest neighbor historical periodic sample and the fault distance sample as the periodic constraint and fault distance interpolation boundary of the sample to be tested.

[0073] Step 4.3: Based on the boundary conditions in Step 4.2, calculate the fault distance x1 = 5.43 km of the sample under test using Equation (9) through interpolation, and output the visual waveform comparison between the sample under test and its nearest neighbor samples. Figure 9 As shown.

[0074] The specific 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. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.

Claims

1. A method for identifying and locating near-end faults in transmission lines based on time-domain waveform feature recognition, characterized in that: Step 1: Collect fault current traveling wave data at the end of the transmission line and select the horizontal compression factor of the waveform based on the near-end fault distance threshold, and display the data of the horizontal compression factor in the image; Step 2: Move a fixed-area sliding window to statistically analyze the pixel density distribution of the pixel stacking feature region, calculate the correlation coefficient between the pixel density distribution of the pixel stacking feature region and the baseline function, and use the correlation coefficient to determine whether the fault has near-end fault characteristics. Step 3: Based on the presence of near-end fault characteristics, extract the peak points of the outer envelope of the short-window waveform and calculate the variance of the timing interval of the peak points. Finally, determine whether it is a near-end fault based on whether the variance is less than the threshold. Step 4: After confirming that it is a near-end fault, take the estimated period of the sample to be tested, the estimated period of the historical neighboring samples, and the corresponding fault reference distance to interpolate and calculate the fault distance of the sample to be tested. At the same time, output the visual comparison results of the reference samples.

2. The method for identifying and locating near-end faults in transmission lines based on time-domain waveform feature recognition according to claim 1, characterized in that, Step 1 specifically refers to: Step 1.1: Preprocess historical near-end fault samples to obtain their waveform library, fault cycle library and corresponding fault distance library; Step 1.2: Collect fault traveling wave data from the transmission line, select the optimal ratio of data length to effective pixel length to plot the waveform, and ensure that the fault waveform is fully compressed.

3. The method for identifying and locating near-end faults in transmission lines based on time-domain waveform feature recognition as described in claim 1, characterized in that, Step 2 specifically refers to: Step 2.1: Take the initial traveling wave and subsequent images of a specific pixel length from the waveform obtained in Step 1 as the pixel stacking feature region; Step 2.2: Construct a fixed-area sliding window, translate the sliding window, and statistically analyze the pixel density distribution in the positive direction of the horizontal pixels; Step 2.3: Fit the waveform density distribution of multiple measured and simulated data to obtain the benchmark function, and calculate the correlation coefficient between the pixel density of the sample to be tested and the benchmark function.

4. The method for identifying and locating near-end faults in transmission lines based on time-domain waveform feature recognition as described in claim 1, characterized in that, Step 3 specifically refers to: Step 3.1: Based on Step 2, select an appropriate window for plotting to ensure that the first wave head of the fault waveform and its subsequent reflected wave groups can be observed; Step 3.2: Take the first 5 intersection points of the intersection between the outer envelope boundary point and the peak point of the fault waveform, that is, the outer envelope peak points, and calculate the variance of the distance between the outer envelope peak points; Step 3.3: Based on the correlation coefficient threshold and variance threshold, determine the near-end fault. When the threshold conditions are met, determine it as a near-end fault and proceed to Step 4.

5. The method for identifying and locating near-end faults in transmission lines based on time-domain waveform feature recognition according to claim 1, characterized in that, Step 4 specifically refers to: Step 4.1: Estimate the period of the near-end fault waveform by searching for the nearest sample period and sample reference distance in historical data; Step 4.2: Select the nearest neighbor historical periodic sample and the fault distance sample as the interpolation boundary value of the period and fault distance of the sample to be tested; Step 4.3: Calculate the fault distance of the sample under test based on the interpolation boundary conditions in Step 4.2 and output a visual waveform comparison between the sample under test and its nearest neighbor samples.

Images