A power transmission line single-ended traveling wave distance measurement method based on self-adaptive registration of descaling waveform diagram

Abstract

The application relates to a power transmission line single-end traveling wave distance measurement method based on a descaling waveform figure self-adaptive registration and belongs to the technical field of power system protection. First, fault line traveling wave data, line length and topological types are read, and the same kind of topological history samples are searched; current traveling wave data is intercepted to make a normalized waveform figure according to the line length proportion, waveform image migration and sample expansion are realized, a waveform center sequence is extracted, a correlation coefficient is calculated, and the most relevant sample is searched as a reference sample; the area difference calculation region and fault position information of the reference sample are read, and the scaling direction is determined by using disturbance observation; the to-be-measured waveform is finely scaled according to the determined scaling direction, a scale-residual curve is generated, the best matching proportion factor is determined according to the minimum value point position of the scale-residual curve, and finally distance conversion is completed to realize distance measurement.

Description

TECHNICAL FIELD

[0001] The application relates to a power transmission line single-end traveling wave distance measurement method based on descaling waveform figure adaptive registration, and belongs to the technical field of power system protection. BACKGROUND

[0002] With the continuous integration of renewable energy into the power grid, the rapid decarbonization process worldwide has greatly changed the energy pattern of the power system, and also brought great challenges to the operation of the future power system. With the deepening of the double-carbon transformation, in order to handle the relationship between supply and demand of the new power system and promote the orderly construction of the new energy power system, while improving the consumption of renewable energy, the safe and stable operation of the power system needs to be ensured. Among them, it is of great significance to study accurate and reliable fault traveling wave distance measurement methods.

[0003] Considering that the actual waveform is greatly affected by the complex environment at the fault site, it is difficult to detect and accurately calibrate the same reference point of the wave head with the traditional single-end traveling wave algorithm, and there are recurrent faults in the line, the historical fault sample reuse rate is low, the number of fault samples in a single substation is limited, and it is difficult to realize cross-platform joint distance measurement between similar topologies, resulting in poor reliability and unstable precision of automatic distance measurement, and the distance measurement performance cannot be improved with the accumulation of historical data. Therefore, it is of great value to study a method that can adaptively find reference samples and accurately measure the equidistant interval of similar fault waveforms, improve the accuracy and reliability of traveling wave distance measurement, increase the historical sample reuse rate, realize fault distance measurement between similar topologies under the condition of known topology, and enhance the engineering application effect of single-end traveling wave fault distance measurement. SUMMARY

[0004] The technical problem to be solved by the application is to provide a power transmission line single-end traveling wave distance measurement method based on descaling waveform figure adaptive registration, which realizes effective screening of similar samples in a historical sample library, matches the scale of the to-be-measured waveform and the sample waveform, and uses the best matching scale factor between the waveforms to realize distance conversion for fault distance measurement.

[0005] The technical solution of the application is: a power transmission line single-end traveling wave distance measurement method based on descaling waveform figure adaptive registration, and the specific steps are as follows:

[0006] Step 1: read the fault line traveling wave data, line length and topology type, and find similar topology historical samples;

[0007] Step 2: intercept the current traveling wave data to make a normalized waveform figure according to the line length, realize waveform image migration and sample expansion, extract the waveform center sequence and calculate the correlation coefficient, and search for the most relevant sample as the reference sample;

[0008] Step 3: Read the area difference calculation area and fault location information of the benchmark sample, and determine the scaling direction by observing the disturbance;

[0009] Step 4: Finely scale the waveform to be measured in the determined scaling direction to generate a scale-residual curve. Determine the optimal matching scaling factor based on the location of the minimum point of the scale-residual curve, and finally complete the distance conversion to achieve distance measurement.

[0010] Step 1 specifically refers to:

[0011] Step 1.1: Read the traveling wave data, line length and topology type of the fault line. The topology type includes "Category 3-3" bus type where both ends have multiple outgoing lines, and "Category 3-1" bus type where the bus at this end has multiple outgoing lines and the bus at the opposite end has a single outgoing line.

[0012] Step 1.2: Based on the fault line topology type, find historical samples of the same type of topology as a sample library;

[0013] Step 2 specifically refers to:

[0014] Step 2.1: Using the detected initial mutation t0 as a baseline, take t backwards. pre Take t backwards post The amount of data is controlled, and the fault data is extracted and normalized into a waveform by maintaining a proportional relationship between the data volume before and after the fault data. pre / t post =const, at [t0-t pre ,t0+t post This function extracts data within a specified interval. It fixes the relative position of the first wave leader in the waveform diagram, facilitating subsequent alignment of the first wave leader and waveform scaling. post The extraction formula is:

[0015] t post =CL / v(1)

[0016] In the formula, C is a factor that controls the display scale of the waveform in the image, L is the total length of the faulty line, and v is the traveling wave velocity.

[0017] Since the most similar matching historical sample may not occur on the current faulty line under test, the length of the historical line may differ from the length of the current faulty sample line. To increase the success rate of historical sample matching and enhance its reusability, historical samples need to be migrated across lines according to the current faulty line under test. Historical samples are then migrated according to their respective line lengths L. m As t post After calculating and extracting data at a fixed ratio, and then descaling and plotting the data, the system can retain the percentage information of each fault distance (x%), ignoring its relationship with the total length of the line, thereby enabling cross-line migration of different line lengths.

[0018] On the basis of cross-line migration, considering the sparsity of historical samples, to make full use of the existing historical samples and increase the matching probability of benchmark samples, x% is used as the basis to expand the samples. When the historical sample fault distance percentage is less than half the length of the line, the amount of data is reduced to enlarge the sample as an expansion sample; when the historical sample fault distance percentage is greater than half the length of the line, the amount of data is increased to reduce the sample as an expansion sample, and the remaining case is not expanded. The sample expansion formula can be expressed as:

[0019] t′ post =k ex CL m / v(2)

[0020] In the formula, t′ post is the initial mutation point of the expansion sample, k ex is the sample expansion coefficient, and L m is the length of the historical sample line. A large number of measured data show that C=4 can fully display the fault wave head information and realize cross-line migration of different lines.

[0021] The sample expansion coefficient satisfies:

[0022]

[0023] In the formula, x m % is the historical sample fault distance percentage.

[0024] Step 2.2: Extract the waveform center sequence using the gray gravity center algorithm, and the gray gravity coordinates (x k ,y i ) are calculated as follows:

[0025]

[0026] In the formula, k is the horizontal pixel width of the waveform image, h(x k ,y i ) is the gray value corresponding to the kth column coordinates of the waveform image, and M is the column pixel width corresponding to the waveform curve. As can be seen from formula (4), the waveform curve can be approximately represented by the center line sequence connected by (x k ,y k ) point set from low to high horizontal coordinates. Since the horizontal coordinates x are arranged continuously, the equalization of different length data sequences can be realized, and the shortcoming that the correlation coefficient cannot be directly used for different sampling rates and unequal length sequences is overcome.

[0027] The equalized sequence can be used to represent the similarity of two waveforms through the Pearson correlation coefficient. The Pearson correlation coefficient is the most commonly used index to measure the similarity of two sequences. The Pearson correlation coefficient of two waveform sequences is calculated as follows:

[0028]

[0029] wherein y k1 , y k2 represent the waveform to be matched and the reference sample center line coordinates, respectively; represent the corresponding average values, respectively.

[0030] Step 2.3: Select the sample with the highest correlation coefficient as the reference sample, and determine whether the correlation coefficient of the reference sample is greater than the threshold value. If yes, it means that there is a similar historical sample in the sample library as the sample to be measured, and the selected reference sample meets the requirements, then go to Step 2.4; if no, it means that there is no similar historical sample in the sample library as the sample to be measured, and the sample to be measured is a new sample, then use another method or send it to the artificial ranging to supplement into the sample library. The reference sample screening formula is:

[0031] ρ > p Th (6)

[0032] wherein p Th is the reference sample screening threshold value, which can be temporarily taken as 0.9, and can be dynamically adjusted during operation.

[0033] Step 2.4: Determine whether the selected reference sample is a non-proximal fault sample. If yes, it means that the sample to be measured is also a non-proximal fault sample, then go to Step 3; if no, it means that the sample to be measured is a proximal fault sample, then the two waveform graphs of the reference sample and the sample to be measured are both cut according to the proximal window and drawn to go to Step 3.

[0034] The Step 3 is specifically:

[0035] Step 3.1: Read the area difference calculation region and fault position information of the reference sample selected in Step 2, and mark the corresponding area difference calculation region in the reference sample. In order to facilitate computer discretization implementation, the area difference is realized by counting the number of pixels of the area surrounded by the two curves in the statistical calculation region, which can be expressed by the formula:

[0036] ΔS = S A -S B = (S A ∪ S B ) - (S A ∩ S B ) (7)

[0037] wherein S A , S B are the areas surrounded by the rising edge of the fault point reflected wave of the waveform to be measured and the reference sample and the information window boundary, respectively.

[0038] Step3.2: Considering the poor reliability of single wave front matching, non-continuous accidental local mismatch may occur. However, the actual measured traveling wave can often observe more than one continuous subsequent fault traveling wave, so the area difference ΔS1 calculated by the first fault point reflection wave and the area difference ΔS2 calculated by the second fault point reflection wave are combined and weighted to obtain the overall area difference ΔS w , as follows:

[0039] ΔS w = a x ΔS1+ (1-a) x ΔS2 (8)

[0040] In the formula, ΔS j (j = 1, 2) represents the area difference of the calculation region of the first fault point reflection wave and the second fault point reflection wave of the reference sample, respectively; a is the weight of the area difference of the information window of the first fault point reflection wave, and considering that the energy attenuation of the first fault point reflection wave is small and the calibration is more accurate, the weight of the first fault point reflection wave region is higher than that of the second fault point reflection wave region, so a = 0.8 is taken.

[0041] In order to determine the scaling direction of the to-be-measured waveform relative to the historical nearest neighbor reference sample, the positive and negative disturbance average area differences are defined to represent the overall trend of the area difference when the waveform graph is scaled down and enlarged, respectively. The overall trend of the area difference can be characterized by the average area difference , and the formula is as follows:

[0042]

[0043] In the formula, ΔS w is the overall area difference; n s is the number of ΔS w sought in the scaling interval; ΔS w (j s )(j s = 1, 2, …, n s ) is the j s th overall area difference sought in the scaling interval.

[0044] Based on the idea of disturbance observation, the un-scaled waveform is taken as the original scale, which is scaled in the [1, 0.95] and [1, 1.05] interval ranges by 0.01 steps, and the scaling direction of the to-be-measured waveform is determined according to the scaling area difference trend, that is, if the to-be-measured waveform is contracted to reduce the area difference, it is determined that the to-be-measured waveform should increase the data amount to monotonically contract; if the to-be-measured waveform is enlarged to reduce the area difference, it is determined that the to-be-measured waveform should reduce the data amount to monotonically enlarge; if the changes in both directions increase the area difference, it is determined that the to-be-measured waveform should be locally and continuously scaled in a small range. The scaling direction criterion can be abstracted as:

[0045]

[0046] In the formula, flag = -1, 1, 0 represent that the waveform to be measured should be magnified, reduced, and scaled, respectively; ε represents the average area difference between the positive and negative perturbations, calculated in the intervals [1, 1.05] and [1, 0.95], respectively; S The threshold value for the difference between the average areas of the positive and negative disturbances can be taken as...

[0047] Step 4 specifically refers to:

[0048] Step 4.1: Finely scale the waveform under test according to the scaling direction determined in Step 3. For continuous scaling and matching in a small local range, since the required wavefront proximity is high, it is only necessary to scale k in the small range of [0.95, 1.05]. When the waveform under test needs to reduce the amount of data, the data in the range of [1, 0.6] is truncated to enlarge the waveform under test. When the waveform under test needs to increase the amount of data, the data in the range of [1, 1.4] is truncated to shrink the waveform under test.

[0049] Based on determining the scaling direction of the waveform under test, the original waveform is finely scaled with small steps, using the weighted area difference ΔS. w As a measure, the waveform under test is scaled horizontally to match the reference sample, with a step size of 0.01, so as to approximate the waveform under test to the reference sample and form a scale-residual curve.

[0050] Step 4.2: Based on the minimum point of the scaling-residual curve obtained in Step 4.1, calculate the optimal matching scaling factor. The optimal matching scaling factor k can be determined from the coordinates of the minimum point of the curve. x The point z with the minimum area difference min The conversion relationships are as follows:

[0051]

[0052] In the formula, k x The optimal matching ratio factor; N x N b These represent the time window lengths for the waveform under test before scaling and the reference sample, respectively. Since the faulty line topology and the fault information of the reference sample are known in practice, N... x and N b All are known quantities; f x f b These represent the sampling rates of the test sample and the reference sample, respectively, in MHz; flag is the scaling direction flag; step is the scaling step size; z min The x-coordinate is the point corresponding to the minimum value of the area difference curve.

[0053] Step4.3: Directly convert the fault distance of the to-be-tested waveform from the historical fault location by using the proportional transformation relationship between the to-be-tested waveform mode and the most matched historical reference waveform mode, to realize distance measurement. The fault distance conversion formula is:

[0054] x=k x x b (12)

[0055] In the formula, x is the distance measurement result of the to-be-tested waveform; x b is the actual fault distance of the reference sample; k x is the best matching scale factor.

[0056] The beneficial effects of the present application are: the present application can realize the migration and adaptive matching of fault traveling wave data of different sampling rates and different line lengths, equivalent scaling waveform graph to proportional transformation by changing the data window length, can select the most similar historical sample as the reference sample, can determine the scale change direction by using the perturbation observation method, can realize the matching of the to-be-tested waveform and the reference waveform and determine the scaling scale based on the minimum criterion of the waveform weighted area difference, can realize the accurate and reliable calculation of the to-be-tested fault distance according to the scaling of the fault distance of the reference sample and the scaling scale, can fully learn from the prior information of the historical sample, greatly avoid the difficulty of identifying and calibrating the fault point reflection wave in the traditional single-end traveling wave method, can efficiently, automatically and visually identify and match the waveform, significantly improve the distance measurement reliability, can effectively identify the low matching degree sample and update it to the historical sample library, promote the continuous improvement of the performance of the traveling wave analysis and distance measurement, enhance the heuristic and intelligent nature of the single-end traveling wave analysis and distance measurement, and have very important value for improving the engineering application effect of the single-end traveling wave fault distance measurement. BRIEF DESCRIPTION OF DRAWINGS

[0057] Figure 1 is a 220kV transmission line model schematic diagram of the present application;

[0058] Figure 2 is a line conductor and tower arrangement diagram of the present application;

[0059] Figure 3 is a to-be-tested sample fault current traveling wave waveform of the present application;

[0060] Figure 4 is a specific implementation step of Step1 of the present application;

[0061] Figure 5 is a specific implementation step of Step2 of the present application;

[0062] Figure 6 is a to-be-tested sample normalized waveform graph of the present application;

[0063] Figure 7 is a to-be-tested waveform center sequence graph of the present application;

[0064] Figure 8 is the selected reference sample normalized waveform chart of the present application;

[0065] Figure 9 is the selected reference sample de-normalized waveform chart of the present application;

[0066] Figure 10 is the specific implementation step of Step 3 of the present application;

[0067] Figure 11 is the reference sample area difference calculation region calibration chart of the present application;

[0068] Figure 12 is the specific implementation step of Step 4 of the present application;

[0069] Figure 13 is the scale-residual curve chart of the present application. DETAILED DESCRIPTION

[0070] The present application will be further described below in combination with the drawings and specific embodiments.

[0071] Example 1: Based on PSCAD / EMTDC, a 220kV transmission line model as shown in Figure 1 is established, the line adopts frequency-dependent model, the line conductor and tower arrangement is as shown in Figure 2 , the m-side busbar has 4 outgoing lines except the fault line itself, the n-side busbar has 3 outgoing lines except the fault line itself, the fault line is 100km long, the simulation calculation step is 1μs, the scale transformation step is 0.01, the A-phase ground fault occurs at the observation point 45km away from the m-end, wherein the transition resistance is 50Ω, the fault angle is 85°, the m-side busbar current traveling wave sampling frequency is 1MHz, 40dB Gaussian white noise is added to the fault current simulation data obtained at the end of the line, and the fault current waveform is as shown in Figure 3 . The specific implementation steps are as follows:

[0072] Step 1: Read the fault line traveling wave data, line length and topology type, find the same type of topology historical sample, the specific implementation steps are as shown in Figure 4 ;

[0073] Step 1.1: Read the fault current traveling wave data of the m-end of the transmission line, the line length L d =100km, and the busbar types at both ends are the "three-three type" busbars with multi-outgoing lines at the observation end and the opposite end;

[0074] Step 1.2: According to the fault line topology type, find the same type of topology historical sample as the sample library, and select the historical sample of the busbar topology with multi-outgoing lines at both ends as the sample library;

[0075] Step2: intercept current traveling wave data to make normalized waveform graph according to line length proportion, realize waveform image migration and sample expansion, extract waveform center sequence and calculate correlation coefficient, search the most relevant sample as reference sample, the specific implementation steps are as shown in Figure 5 The advantage of this step is to realize cross-line migration and normalization of different line traveling wave data, expand the success rate of historical sample matching, increase its reusability, realize data serialization, and overcome the shortcoming that correlation coefficient cannot be directly used for different sampling rates and unequal length sequences.

[0076] Step2.1: take the detected initial mutation t0 as the reference, take t pre = 336 μs forward and t post = 1344 μs backward data amount, and keep the front and back proportional t pre / t post = 1:4, intercept data in [t0-t pre , t0+t post ] interval, as shown in Figure 6 . Realize fixing the relative position of the first wave head in the waveform graph, which is convenient for subsequent first wave head alignment and waveform graph stretching matching;

[0077] Calculate t post for each line full length L m of the waveform in the sample library selected in Step1 and intercept data amount according to fixed proportion, descale and make graph display to realize cross-line migration of different line full length, and realize sample expansion according to respective fault distance percentage information x% and formulas (2), (3);

[0078] Step2.2: use formula (4) gray center algorithm to extract each waveform center sequence, wherein the center sequence graph of the to-be-tested waveform is shown in Figure 7 .

[0079] Calculate the correlation coefficient of the to-be-tested waveform and each waveform center sequence in the sample library according to formula (5);

[0080] Step2.3: select the most relevant sample with ρ max = 0.9982 as the reference sample according to the correlation coefficient result, as shown in Figure 8 . Its descaled waveform graph is shown in Figure 9 . Since ρ max > 0.9, it means that there is a similar historical sample in the sample library as the to-be-tested sample;

[0081] Step2.4: since the selected reference sample is a non-near-end fault sample, it means that the to-be-tested sample is also a non-near-end fault sample, so the two waveform graphs are directly entered into Step3;

[0082] Step3.1: Read the reference sample fault distance x selected in Step2 b = 40km, L m = 100km, sampling rate f b = 1MHz, according to fault distance x b The first fault point reflection wave mutation point distance is calculated, and the initial wave head mutation point time interval At = 2x b / v = 2x40 / 0.298 = 268us, so mark the first fault point reflection wave and the second fault point reflection wave area difference calculation area in the reference sample waveform graph, as shown in Figure 11 ;

[0083] Step3.2: Use the perturbation observation method to scale the measured waveform by 0.01 steps in the [1, 0.95] and [1, 1.05] directions, and the perturbation observation method scaling direction determination result is shown in Table 1, and finally determine that the measured waveform should increase the data amount to reduce the waveform graph;

[0084]

[0085] Table 1: Perturbation observation method scaling direction determination result

[0086] Step4: Fine scale the measured waveform according to the determined scaling direction, generate a scale-residual curve, determine the best matching scale factor according to the minimum point position of the scale-residual curve, and finally complete the distance conversion to realize distance measurement, and the specific implementation steps are shown in Figure 12 ;

[0087] Step4.1: According to the scaling direction determined in Step3, fine scale the measured waveform in the [1, 1.4] interval, and match with the reference sample to generate a scale-residual curve, as shown in Figure 13 .

[0088] Step4.2: Based on the minimum point z min = 12 of the scale-residual curve obtained in Step4.1, since N x = N b = 1680, f x = f b = 1MHz, flag = 1, and the best matching scale factor k x is converted to:

[0089] k x = 1680x1x[1+1x0.01x12] / (1680x1) = 1.12

[0090] Step4.3: using the proportional transformation relationship between the to-be-tested sample waveform mode and the most matched historical reference waveform mode, the fault distance of the to-be-tested waveform is directly converted from the historical fault position, and the to-be-tested sample ranging result is:

[0091] x=k x x x b =1.12*40=44.8km

[0092] The specific embodiments of the application are described in detail above with reference to the drawings, but the application is not limited to the above embodiments, and various changes can be made within the knowledge of those skilled in the art without departing from the purpose of the application.

Claims

1. A power line single-ended traveling wave distance measurement method based on self-adaptive registration of descaling waveform diagram, characterized in that: Step1: reading fault line traveling wave data, line length and topology type, finding the same type of topology historical sample; Step2: intercepting current traveling wave data to make normalized waveform diagram according to line length ratio, realizing waveform image migration and sample expansion, extracting waveform center sequence and calculating correlation coefficient, searching for the most relevant sample as the reference sample; Step3: reading the area difference calculation area and fault location information of the reference sample, and determining the scaling direction by disturbance observation; Step4: fine scaling the to-be-measured waveform according to the determined scaling direction, generating a scale-residual curve, determining the best matching scale factor according to the minimum point position of the scale-residual curve, and finally completing distance conversion to realize distance measurement; Step4.1: fine scaling the to-be-measured waveform according to the scaling direction determined in Step3, matching with the reference sample, calculating the area difference change in the area difference calculation area during scaling, and generating a scale-residual curve; Step4.2: calculating the best matching scale factor based on the minimum point of the scale-residual curve obtained in Step4.1; Step4.3: directly converting the fault distance of the to-be-measured waveform from the historical fault location by using the proportional transformation relationship between the to-be-measured waveform mode and the most matched historical reference waveform mode, to realize distance measurement; flag wherein the best matching scale factor k x is: ； In the formula, N x , N b respectively are the window length of the to-be-tested waveform and the reference sample before scaling, f x , f b respectively are the sampling rate of the to-be-tested sample and the reference sample, and bit is MHz, step is a scaling direction identification bit; The fault distance conversion formula is: is a scaling step, z min is the horizontal coordinate corresponding to the minimum value point of the area difference curve. The Step1 is specifically: ； wherein, x is the measured waveform ranging result, x b is the reference sample actual fault distance, k x is the best matching scale factor.

2. The power line single-ended traveling wave distance measurement method based on self-adapting registration of de-scaling waveform graph according to claim 1, characterized in that, Step1.1: reading the fault line current traveling wave data, line length and topology type, and the topology type includes "three-three type" bus type with both ends of the bus being multi-outlet, and "three-one type" bus type with the bus at one end being multi-outlet and the bus at the other end being single-outlet; Step1.2: finding the same type of topology historical sample as the sample library according to the fault line topology type. The Step2 is specifically:

3. The power line single-ended traveling wave distance measurement method based on self-adapting registration of de-scaled waveform graph according to claim 1, characterized in that, Step2.1: intercepting current traveling wave data to make normalized waveform diagram according to line length ratio, realizing waveform image migration and sample expansion; Step2.2: extracting the waveform center sequence using the gray center algorithm, and calculating the correlation coefficient of the to-be-measured waveform and the waveform in the sample library according to the waveform center sequence; Step2.3: selecting the most relevant sample as the reference sample according to the correlation coefficient result, and judging whether the correlation coefficient of the reference sample is greater than the threshold value; if yes, go to Step2.4; if no, use another method or report manual distance measurement and supplement into the sample library; Step2.4: judging whether the selected reference sample is a non-near-end fault sample; if yes, go to Step3 with the current two waveform diagrams; if no, intercept and draw the two waveform diagrams of the reference sample and the to-be-measured sample according to the near-end window, and then go to Step3. The Step3 is specifically:

4. The power line single-ended traveling wave distance measurement method based on self-adapting registration of de-scaled waveform graph according to claim 1, characterized in that, Step3.1: reading the area difference calculation area and fault location information of the reference sample selected in Step2, and marking the corresponding area difference calculation area in the reference sample; ​ Step3.2: The test waveform is scaled in two directions of waveform reduction and amplification respectively by equal step length ratio using the perturbation observation method, and the scaling direction of the test waveform is determined according to the change trend of the scaling area difference.

Images