Double-end traveling wave distance measurement method based on weighted Minkowski distance and cubic spline interpolation

Abstract

The invention relates to a double-end traveling wave fault location method based on weighted Minkowski distance and cubic spline interpolation. The method comprises the following steps: performing smooth fitting expansion processing on fault waveform data W through cubic spline interpolation; training a traveling wave head clipping model, and automatically positioning and clipping a wave head region Wcs; respectively recording the Wcs after the traveling wave processing at the two ends as Wa and Wb, translating one waveform, and calculating a weighted Minkowski distance dist; when dist is minimum, the translation amount is the time difference delta t when the traveling wave reaches the two ends for the first time. Through cubic spline interpolation, under the condition of low sampling rate, smooth fitting expansion is carried out on the signal waveform, the actual error range is obviously reduced, and the positioning precision is improved; according to the method, the weight of the distance near the fault moment is increased, the effect of highlighting the fault point is achieved, for the noisy waveform, noise reduction and other operations which can change the original waveform can be not carried out, the time difference delta t of the wave heads of the two fault traveling waves can be calculated, and the measurement error existing in the practice of the prior art is reduced.

Description

Technical Field

[0001] The invention relates to the field of electric power technology, and in particular to a double-terminal traveling wave ranging method based on weighted Minkowski distance and cubic spline interpolation. Background Art

[0002] Under the background of sustained and rapid economic development, higher requirements are put forward for the safe and stable operation of distribution networks. Various monitoring devices will be disturbed by various factors during actual operation, resulting in the collected traveling wave images not being able to achieve the ideal form. As a result, when using traveling wave positioning, it is impossible to accurately find the traveling wave head time, affecting the precise fault positioning results.

[0003] Over the years, people have conducted extensive research on traveling wave ranging methods. Wave head monitoring specifically includes the following methods:

[0004] (1) Derivative method: The basic idea is to determine whether the traveling wave reaches the busbar by measuring the first-order (or second-order) derivative of the traveling wave at the detection point to see whether it exceeds the set threshold. This method is simple and fast, requires a short data window, is easy to implement, and is still a relatively practical method. However, the derivative method reflects the rate of change of the signal change in discrete signal processing, which will be affected by the change of the signal's own amplitude and is sensitive to noise.

[0005] (2) Wavelet transform method: Wavelet transform is used to identify the wave head. This method has good noise reduction and frequency division characteristics and can accurately capture the arrival time of each traveling wave, thereby improving the accuracy of the ranging results.

[0006] (3) TEO-based method: The Teager energy spectrum of the current traveling wave waveform sequence is calculated through the Teager energy operator, and the wave head position is predicted through singular value search and deep learning to achieve accurate calibration of the wave head position.

[0007] (4) Empirical Mode Decomposition (EMD): The fault traveling wave signal is processed by EMD, the high-frequency IMF component is extracted for wave head detection, and the fault distance is completed by combining the three-terminal traveling wave ranging algorithm.

[0008] At present, there are still many problems in the use of double-terminal traveling wave positioning in distribution networks:

[0009] (1) The equipment waveform sampling rate is limited. At present, most of the network equipment is designed with low traveling wave sampling rate due to power consumption, communication and basic hardware, which leads to large theoretical diagnosis errors and brings great difficulties to maintenance work.

[0010] (2) The distribution network structure is complex. The propagation of traveling waves in power lines is a complex physical process. At present, most distribution networks have a radial structure, with reclosing switches, disconnectors, tie switches and other equipment distributed along the line. At the same time, the distribution network has many branches and the line distance is short, which leads to the superposition and attenuation of waveforms, making it more difficult to locate faults.

[0011] (3) The distribution network operating environment is harsh. The distribution network lines are easily affected by construction and domestic waste. For example, lifting operations, building construction, greening trees, domestic waste, etc. may cause line faults, and the fault traveling waves have different shapes, making it difficult to identify the wave head.

[0012] Due to the above reasons, considering the working environment of on-site operation and maintenance personnel, such errors will bring huge maintenance difficulties. Therefore, in practical applications, in order to accurately calculate the traveling wave speed, it is necessary to comprehensively consider these factors and adopt appropriate mathematical models and calculation methods. At present, the industry is also exploring the use of artificial intelligence to locate faults, but due to the difficulty in sample establishment, complex modeling, and general operating results, it is temporarily impossible to apply it on a large scale in actual distribution network lines. In summary, due to some characteristics of the distribution network, many fault location methods are difficult to implement in the distribution network or the implementation effect is poor. Therefore, there is an urgent need for a fast fault location method using double-end traveling waves. Summary of the invention

[0013] The technical problem to be solved by the present invention is to provide a double-terminal traveling wave ranging method based on weighted Minkowski distance and cubic spline interpolation to overcome the deficiencies in the above-mentioned prior art.

[0014] The technical solution of the present invention to solve the above technical problems is as follows: a double-terminal traveling wave ranging method based on weighted Minkowski distance and cubic spline interpolation, comprising the following steps:

[0015] Step S01: Collecting double-end fault waveform data W;

[0016] Step S02: performing smooth fitting and expansion processing on the fault waveform data W by cubic spline interpolation;

[0017] Step S03: training the traveling wave head cutting model, using the processed fault waveform data W to train the traveling wave head cutting model, automatically locating and cutting out the wave head area W cs ;

[0018] Step S04: W after traveling wave processing at both ends cs Denoted as Wa and Wb respectively, one of the waveforms is translated, and the waveform with a fixed length of L in the overlapping part is taken. The weighted Minkowski distance dist is calculated according to the weighting function corresponding to the fault type. When dist is the smallest, the translation amount is the time difference Δt when the traveling wave first arrives at the two ends.

[0019] Step S05: Using the time difference Δt as a parameter of the dual-end positioning method to calculate the fault point location.

[0020] The beneficial effects of the present invention are as follows: the present invention uses cubic spline interpolation to perform smooth fitting and expansion on the signal waveform under low sampling rate conditions, significantly reducing the actual error range and improving positioning accuracy; based on the weighted Minkowski distance, the weight of the distance near the fault moment is increased to achieve the effect of highlighting the fault point, and for noisy waveforms, operations such as noise reduction that will change the original waveform do not need to be performed, and the time difference Δt between the two fault traveling wave heads is calculated, thereby reducing the measurement error existing in the prior art practice.

[0021] Based on the above technical solution, the present invention can also be improved as follows.

[0022] Furthermore, the training wave head cropping model in step S03 is specifically:

[0023] From the processed fault waveform data, the wave head area is manually selected and marked to obtain the traveling wave waveform dataset, which is then used to train the traveling wave head cropping model in the YOLOv8 algorithm.

[0024] Further, the step S04 of calculating the weighted Minkowski distance dist according to the fault type specifically includes the following steps:

[0025] The wave head area W cut out based on the traveling wave waveform data set and the traveling wave head cutting model cs , through the LASSO algorithm, the correlation coefficient a is calculated by sparsely calculating each segment i , and construct the weighting function w i ;

[0026] The weighted Minkowski distance formula is as follows:

[0027]

[0028] Among the fault types, the weighting function of short circuit fault is:

[0029]

[0030] The weighting function of the ground fault type is:

[0031]

[0032] Further, the calculation formula of the dual-end positioning method in step S05 is:

[0033]

[0034] Where, T S-T R or T R -T S The absolute value of is Δt; T S is the time when the fault traveling wave first reaches the measuring end S; T R is the time when the fault traveling wave first reaches the measuring end R; X s is the distance from the measuring end S to the fault point F, X R is the distance from the measuring end R to the fault point F; r is the length of the transmission line, F is the fault point location, and V is the traveling wave propagation velocity. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a schematic diagram of the overall solution process structure of the present invention;

[0036] Figure 2 It is a schematic diagram of the structure of the double-end traveling wave positioning principle of the present invention;

[0037] Figure 3 It is a schematic diagram of the local cubic spline interpolation of the present invention;

[0038] Figure 4 It is a schematic diagram of the traveling wave cutting of the present invention;

[0039] Figure 5 It is a short-circuit fault traveling wave current waveform diagram of the present invention;

[0040] Figure 6 A traveling wave current waveform diagram of a single-phase grounding fault of the present invention;

[0041] Figure 7 This is a waveform diagram of the time difference Δt of the maximum overlap of the present invention. DETAILED DESCRIPTION

[0042] The principles and features of the present invention are described below in conjunction with the accompanying drawings. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0043] like Figure 1 to Figure 7 As shown, embodiment 1, a double-terminal traveling wave ranging method based on weighted Minkowski distance and cubic spline interpolation, comprises the following steps:

[0044] Step S01: Collecting double-end fault waveform data W;

[0045] Step S02: performing smooth fitting and expansion processing on the fault waveform data W by cubic spline interpolation;

[0046] Step S03: training the traveling wave head cutting model, using the processed fault waveform data W to train the traveling wave head cutting model, automatically locating and cutting out the wave head area W cs ;

[0047] Step S04: W after traveling wave processing at both ends cs W a , W b , translate one of the waveforms, take the waveform with a fixed length of L in the overlapping part, and calculate the weighted Minkowski distance dist according to the weighting function corresponding to the fault type; when dist is the smallest, the translation amount is the time difference Δt when the traveling wave first arrives at both ends;

[0048] Step S05: Using the time difference Δt as a parameter of the dual-end positioning method to calculate the fault point location.

[0049] The present invention uses cubic spline interpolation to smoothly fit and expand the signal waveform under low sampling rate conditions, significantly reducing the actual error range and improving positioning accuracy. Based on the weighted Minkowski distance, the weight of the distance near the fault moment is increased to achieve the effect of highlighting the fault point. For noisy waveforms, operations such as noise reduction that will change the original waveform do not need to be performed, and the time difference Δt between the two fault traveling wave heads is calculated, thereby reducing the measurement error existing in the prior art practice.

[0050] The cubic spline interpolation in step S02 is as follows:

[0051] Step S21: There are n+1 known data points on the interval [a, b], and [a, b] is divided into n sub-intervals:

[0052] [x0,x1],[x1,x2],...,[x n-1 , x n ];

[0053]

[0054] Step S22: Construct f(x) with respect to nodes x0, x1, ..., x n The cubic spline function S(x) of , S(x) satisfies the following conditions:

[0055] a. In each subinterval, there is a polynomial of degree not more than 3, that is:

[0056] S i (x i )=a i +b i (xx i )+c i (xx i ) 2 +d i (xx i ) 3 ;

[0057] b. Continuous and smooth over the entire interval [a, b], that is, the first-order derivative S′(x i ) and the second-order derivative S″(x i ) exists and is continuous;

[0058] c. Satisfy the interpolation conditions, namely:

[0059] S i (x i )=y i , i=0, 1, ..., n.

[0060] Taking the traveling wave sampling rate of 1M / s as an example, the sampling interval is 1us. Under the ideal state of traveling wave, the wave speed is 300m / us. According to the relevant calculation formula, it can be obtained that the positioning error is 1us. In order to compensate for this part of the information, cubic spline interpolation is performed on the two adjacent points. According to the function of each interval, 99 points are inserted at equal distances between the two points, reducing the original sampling interval from 1us to 0.01us.

[0061] Embodiment 2: This embodiment is a further improvement on the basis of embodiment 1, and its details are as follows:

[0062] The training wave head cropping model in step S03 is specifically:

[0063] From the processed fault waveform data, the wave head area is manually selected and marked to obtain the traveling wave waveform dataset, which is then used to train the traveling wave head cropping model in the YOLOv8 algorithm.

[0064] For the collected fault traveling wave data, the main key information is concentrated near the wave head; in the specific implementation, the present invention uses the YOLOv8 algorithm to train the traveling wave head clipping model, performs target detection on the overall waveform data, automatically locates the wave head area and completes precise clipping, thereby significantly reducing the processing amount of irrelevant data. This method has high real-time performance and high precision, and is suitable for the rapid analysis and processing of power system fault traveling waves;

[0065] The specific steps are as follows: collect and annotate a traveling wave waveform data set containing wave head area annotations, and train it using the YOLOv8 algorithm; use the trained YOLOv8 algorithm to perform target detection on the input fault traveling wave waveform data and identify the area where the wave head is located; crop the detected wave head area to extract high-value information from the target area; the cropped data is used for subsequent analysis, significantly improving computational efficiency and accuracy.

[0066] Embodiment 3, this embodiment is a further improvement on the basis of embodiment 2, and its details are as follows:

[0067] Calculating the weighted Minkowski distance dist according to the fault type in step S04 specifically includes the following steps:

[0068] The wave head area W cut out based on the traveling wave waveform data set and the traveling wave head cutting model cs , through the LASSO algorithm, the correlation coefficient a is calculated by sparsely calculating each segment i , and construct the weighting function w i ; The weighting function constructed by LASSO algorithm training samples can have a stronger effect of highlighting the main features of different fault traveling waves when calculating distances compared to ordinary weighting functions;

[0069] In the general method, the time difference Δt when the traveling wave first arrives at the two ends is calculated by the wave head moment of the two traveling waves. S -T R or T R -T S . In the distribution network, the wave head shape is not ideal, especially the traveling wave under high-resistance fault. Therefore, the distance or similarity of the overlapping part is calculated by translating the two traveling waves to achieve the effect of aligning the wave head. The amount of translation time is the time difference Δt at both ends. At the same time, because there are many branches in the distribution network and the line distance is short, the wave shape of different fault types is different. At the same time, there is a big gap in the quality of the traveling wave waveforms at both ends, and the translation will have a large error. Therefore, the weighted Minkowski distance is considered to highlight the distance of the wave head part by weighting;

[0070] The weighted Minkowski distance formula is as follows:

[0071]

[0072] For short-circuit faults among the fault types, since the amplitude of the traveling wave is large and changes significantly during the fault, its weighting function is:

[0073]

[0074] Among the fault types, the ground fault is mostly a small current fault with a weak wave head, and its weighting function is:

[0075]

[0076] Embodiment 4: This embodiment is a further improvement on the basis of any one of Embodiments 1 to 3, and its details are as follows:

[0077] The calculation formula of the dual-end positioning method in step S05 is:

[0078]

[0079] Where, T S -TR or T R -T S The absolute value of is Δt; T S is the time when the fault traveling wave first reaches the measuring end S; T R is the time when the fault traveling wave first reaches the measuring end R; X s is the distance from the measuring end S to the fault point F, X R is the distance from the measuring end R to the fault point F; r is the length of the transmission line, F is the fault point location, and V is the traveling wave propagation velocity.

[0080] See also Figure 2 In the two-terminal traveling wave ranging principle formula, it can be seen that it is necessary to accurately obtain the traveling wave propagation speed V and the time difference T when the traveling wave first arrives at both ends. S -T R or T R -T S , and the total length of the line r, these three key parameters, can be used to accurately determine the location of the fault point. By using the optimized time difference Δt for calculation, the measurement error existing in the existing technical practice is reduced.

[0081] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A double-terminal traveling wave ranging method based on weighted Minkowski distance and cubic spline interpolation, characterized in that: The steps include: Step S01: Collecting double-end fault waveform data W; Step S02: performing smooth fitting and expansion processing on the fault waveform data W by cubic spline interpolation; Step S03: training a traveling wave head cutting model, using the processed fault waveform data W for the training traveling wave head cutting model, automatically locating and cutting out the wave head area W cs ; Step S04: W after traveling wave processing at both ends cs W a , W b , translate one of the waveforms, take the waveform with a fixed length of L in the overlapping part, and calculate the weighted Minkowski distance dist according to the weighting function corresponding to the fault type; when dist is the smallest, the translation amount is the time difference Δt when the traveling wave first arrives at both ends; Step S05: using the time difference Δt as a parameter of a dual-end positioning method to calculate the fault point location.

2. The double-terminal traveling wave ranging method based on weighted Minkowski distance and cubic spline interpolation according to claim 1, characterized in that: The training wave head cutting model in step S03 is specifically: From the processed fault waveform data, the wave head area is manually selected and marked to obtain a traveling wave waveform data set, and the traveling wave waveform data set is used to train a traveling wave head cropping model in the YOLOv8 algorithm.

3. The double-terminal traveling wave ranging method based on weighted Minkowski distance and cubic spline interpolation according to claim 2, characterized in that: The step S04 of calculating the weighted Minkowski distance dist according to the fault type specifically includes the following steps: The wave head area W cut out according to the traveling wave waveform data set and the traveling wave wave head cutting model cs , through the LASSO algorithm, the correlation coefficient a is calculated by sparsely calculating each segment i , and construct the weighting function w i ; The weighted Minkowski distance formula is as follows: The weighting function of the short circuit fault in the fault type is: The weighting function of the ground fault in the fault type is:

4. A two-terminal traveling wave ranging method based on weighted Minkowski distance and cubic spline interpolation according to any one of claims 1 to 3, characterized in that: The calculation formula of the dual-end positioning method in step S05 is: Where, T S -T R or T R -T S The absolute value of is Δt; T s is the time when the fault traveling wave first reaches the measuring end S; T R is the time when the fault traveling wave first reaches the measuring end R; X s is the distance from the measuring end S to the fault point F, X R is the distance from the measuring end R to the fault point F; r is the length of the transmission line, F is the fault point location, and V is the traveling wave propagation velocity.

Images