Resonant grounding system fault line selection method based on DTW-Hilbert and improved K-means

Abstract

The invention discloses a DTW-Hilbert and improved K-means-based resonant grounding system fault line selection method, and belongs to the field of medium-voltage power distribution system fault line selection. The method comprises the following steps: building a power distribution network simulation model, setting different types of single-phase earth faults in the built model, and collecting transient zero-sequence current data of each line; a DTW distance algorithm is adopted to quantitatively describe the waveform similarity degree between the transient zero-sequence currents of all the lines, and Hilbert envelope energy is adopted to measure the high-frequency component amplitude in transient zero-sequence current signals; and finally, carrying out classification processing on the fault feature data set by adopting an improved K-means clustering algorithm, and judging a fault line according to clustering analysis.

Description

Technical Field

[0001] The present invention relates to the field of fault line selection in a medium voltage distribution network, and in particular to a fault line selection method for a resonant grounding system based on DTW-Hilbert and improved K-means. Background Art

[0002] The resonant grounding system can distinguish between transient grounding faults and permanent grounding faults and has high power supply reliability, so it has been widely used in my country's medium-voltage distribution network. Single-phase grounding faults are the most common type of faults in distribution networks, accounting for more than 80% of total faults. In the past, the operating specifications required that the system could operate with the fault for 1 to 2 hours after a single-phase grounding fault occurred, and the requirements for the fault handling capability of the protection were relatively low. After 2017, my country's two major power supply companies successively issued new guidelines for distribution network fault handling, requiring rapid isolation of permanent faults and safe arc extinguishing of transient faults, and put forward higher requirements for the fault handling capability of protection. Affected by factors such as arc extinguishing coils, transition resistance, environmental noise, and operating mode, the line selection of single-phase grounding faults in resonant grounding systems is particularly difficult. Therefore, how to accurately and reliably identify the fault line after a fault occurs in the resonant grounding system is of great significance to the promotion of intelligent distribution networks. Summary of the invention

[0003] In order to solve the problems mentioned in the above background technology, the present invention proposes a fault line selection method for a resonant grounding system based on DTW-Hilbert and improved K-means.

[0004] The present invention achieves the above technical objectives and technical effects through the following technical solutions: a fault line selection method for a resonant grounding system based on DTW-Hilbert and improved K-means, which comprises the following steps:

[0005] Step 1: Build a distribution network simulation model with N outgoing lines, set different types of single-phase grounding faults in the built model, collect transient zero-sequence current data of each line, and select the signal sequence in a specific section as the original current data;

[0006] Step 2: input the original current data obtained in step 1 into the DTW distance algorithm, and use the DTW distance algorithm to calculate the distance coefficient of each line, thereby measuring the waveform similarity between the zero-sequence currents of each line;

[0007] Step 3, inputting the original current data obtained in step 1 into the VMD algorithm, the VMD algorithm decomposes the original current data into high-frequency components and low-frequency components, and calculates the Hilbert envelope energy of the high-frequency components;

[0008] Step 4: Use the improved K-means clustering algorithm to classify the fault feature data set and determine the fault line based on cluster analysis.

[0009] Furthermore, the specific section in step 1 is: selecting 1 / 4 power frequency cycle before the fault occurs to 3 / 4 power frequency cycle after the fault occurs as the fault feature extraction section, and only using the transient zero-sequence current data in this section as the original current data to complete the subsequent steps.

[0010] Furthermore, the specific steps of calculating the distance coefficient in step 2 are:

[0011] 1) Normalize the original current data according to the following formula:

[0012]

[0013] Among them, l i is the i-th data point in the original current data, L i is the i-th data point after normalization.

[0014] 2) Calculate the DTW distance between the normalized original current data of N lines to form a DTW distance matrix D:

[0015]

[0016] Among them, d(L i ,L j ) represents line L i and line L j DTW distance between raw current data;

[0017] 3) The elements of each column in the D matrix are summed up to obtain the dynamic time bending distance vector d, which can reduce the order of the D matrix:

[0018]

[0019] Among them, d i is the algebraic sum of the elements in the i-th column of the D matrix. The branch with the largest element value in the vector d is the fault branch.

[0020] The distance coefficient ρ is calculated according to the following formula:

[0021]

[0022] Among them, d max is the largest element in the vector d. After normalization, the distance coefficient of the faulty line is ρ=1, and is greater than the distance coefficients of all sound lines.

[0023] Furthermore, the implementation method of calculating the Hilbert envelope energy of the high-frequency component in step 3 is:

[0024] 1) The original current data is decomposed into high-frequency components and low-frequency components using the VMD algorithm, and the Hilbert envelope energy of the high-frequency component is calculated;

[0025] 2) Calculate the high frequency component c according to the following formula: i Hilbert transform of (t):

[0026]

[0027] in, c i (t) Hilbert transform, t is time, τ is a certain moment;

[0028] 3) Calculate the Hilbert envelope energy according to the following formula:

[0029]

[0030] in, Z is the number of sampling points;

[0031] 4) Normalize the envelope energy and calculate the energy coefficient according to the following formula:

[0032]

[0033] Where N is the total number of lines.

[0034] 5. The method for fault line selection in a resonant grounding system based on DTW-Hilbert and improved K-means according to claim 1, characterized in that the fault information (ρ i ,h i ) is organized into a fault data set as the input of the improved K-means clustering algorithm. The number of clusters K is set to 2, and all data are clustered. The clustering algorithm outputs the clustering label of each line, and the faulty line is determined based on the clustering label.

[0035] After adopting the above method, the present invention has the following advantages compared with the existing resonant grounding fault line selection technology:

[0036] 1. The dual line selection criterion combined with the transient zero-sequence current waveform similarity and the transient zero-sequence current high-frequency component amplitude characteristics adopted by the present invention has better line selection effect and superior performance than the single line selection criterion; the transient process after the fault is affected by various factors, and it is difficult to achieve complete line selection using a single fault criterion. The use of dual characteristic quantities for fault line selection can achieve complementary advantages and improve the accuracy of line selection.

[0037] 2. Compared with several clustering algorithms currently used, the improved K-means clustering algorithm has significantly improved the performance in line selection accuracy and clustering quality; compared with the fifth harmonic line selection method based on steady-state information and the fundamental amplitude method and transient energy method based on transient information, the criteria used in the present invention have a wider range of application and are more compatible with different fault conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and specific embodiments.

[0039] Figure 1 Framework flow chart of the line selection method for resonant ground fault

[0040] Figure 2 Distribution network simulation structure diagram

[0041] Figure 3 Clustering results

[0042] Figure 4 The first group of original zero-sequence current and filtered current

[0043] Figure 5 The second group of original zero-sequence current and filtered current DETAILED DESCRIPTION

[0044] In order to better understand the present invention, the content of the present invention is further described below in conjunction with the accompanying drawings and examples.

[0045] A fault line selection method for resonant grounding system based on DTW-Hilbert and improved K-means, such as Figure 1 As shown, the specific process includes the following steps:

[0046] Step 1: Build a distribution network simulation model with N outgoing lines, set different types of single-phase grounding faults in the built model, collect transient zero-sequence current data of each line, and select the signal sequence in a specific section as the original current data;

[0047] Step 2: input the original current data obtained in step 1 into the DTW distance algorithm, and use the DTW distance algorithm to calculate the distance coefficient of each line, thereby measuring the waveform similarity between the zero-sequence currents of each line;

[0048] Step 3, inputting the original current data obtained in step 1 into the VMD algorithm, the VMD algorithm decomposes the original current data into high-frequency components and low-frequency components, and calculates the Hilbert envelope energy of the high-frequency components;

[0049] Step 4: Use the improved K-means clustering algorithm to classify the fault feature data set and determine the fault line based on cluster analysis.

[0050] Furthermore, the specific section in step 1 is: selecting 1 / 4 power frequency cycle before the fault occurs to 3 / 4 power frequency cycle after the fault occurs as the fault feature extraction section, and only using the transient zero-sequence current data in this section as the original current data to complete the subsequent steps.

[0051] Furthermore, the specific steps of calculating the distance coefficient in step 2 are:

[0052] 1) Normalize the original current data according to the following formula:

[0053]

[0054] Among them, l i is the i-th data point in the original current data, L i is the i-th data point after normalization.

[0055] 2) Calculate the DTW distance between the normalized original current data of N lines to form a DTW distance matrix D:

[0056]

[0057] Among them, d(L i ,L j ) represents line L i and line L j DTW distance between raw current data;

[0058] 3) The elements of each column in the D matrix are summed up to obtain the dynamic time bending distance vector d, which can reduce the order of the D matrix:

[0059]

[0060] Among them, d i is the algebraic sum of the elements in the i-th column of the D matrix. The branch with the largest element value in the vector d is the fault branch.

[0061] The distance coefficient ρ is calculated according to the following formula:

[0062]

[0063] Among them, d max is the largest element in the vector d. After normalization, the distance coefficient of the faulty line is ρ=1, and is greater than the distance coefficients of all sound lines.

[0064] Furthermore, the implementation method of calculating the Hilbert envelope energy of the high-frequency component in step 3 is:

[0065] 1) The original current data is decomposed into high-frequency components and low-frequency components using the VMD algorithm, and the Hilbert envelope energy of the high-frequency component is calculated;

[0066] 2) Calculate the high frequency component c according to the following formula: i Hilbert transform of (t):

[0067]

[0068] in, c i (t) Hilbert transform, t is time, τ is a certain moment;

[0069] 3) Calculate the Hilbert envelope energy according to the following formula:

[0070]

[0071] in, Z is the number of sampling points;

[0072] 4) Normalize the envelope energy and calculate the energy coefficient according to the following formula:

[0073]

[0074] Where N is the total number of lines.

[0075] 5. The method for fault line selection in a resonant grounding system based on DTW-Hilbert and improved K-means according to claim 1, characterized in that the fault information (ρ i ,h i ) is organized into a fault data set as the input of the improved K-means clustering algorithm. The number of clusters K is set to 2, and all data are clustered. The clustering algorithm outputs the clustering label of each line, and the faulty line is determined based on the clustering label.

[0076] Simulation Verification

[0077] exist Figure 2 A large number of simulation experiments are carried out in the simulation model shown in the figure to simulate different fault conditions, obtain transient zero-sequence current signals and extract fault characteristics from them. The case simulation results are shown in Table 1 and Figure 3 shown.

[0078] Table 1 Fault characteristics and line selection results of the case

[0079]

[0080] from Figure 3As can be seen from Table 1, the clustering algorithm successfully divides the fault data into two categories, one is the "fault cluster" and the other is the "healthy cluster". According to the clustering analysis, the fault line can be determined, which shows that it is feasible to use this method to select the fault line.

[0081] It is assumed that a single-phase grounding fault with different resistance values ​​occurs at L1 3 km away from the busbar. The initial fault angle is 90°, and the transition resistances are 10Ω, 100Ω, 500Ω, 1000Ω, and 1500Ω respectively. The line selection results are shown in Table 2.

[0082] Table 2 Line selection effect with different transition resistance

[0083]

[0084]

[0085] It can be seen from Table 2 that when the transition resistance increases from 0 to 1.5 kΩ, the result obtained by cluster analysis is consistent with the actual fault line, indicating that the line selection method of the present invention has a high transition resistance tolerance and accuracy.

[0086] A single-phase grounding fault with a fault transition resistance of 150Ω is set on L4, the initial fault angle is 45°, and the fault locations are set to be 2km, 4km, 6km, 8km, and 11km away from the busbar. The line selection results are shown in Table 3.

[0087] Table 3 Line selection results at different fault distances

[0088]

[0089] It can be seen from Table 3 that when the fault location is changed, the line selection method of the present invention can still correctly select the faulty line. It is also found that since line L3 contains a cable section, the capacitive current flowing through the line is very large, and the transient zero-sequence current of L3 is much larger than the zero-sequence current on the two overhead lines, resulting in a small difference between the transient zero-sequence current amplitudes on lines L3 and L4. If the fault line is determined only by the amplitude difference of the high-frequency component, misselection is likely to occur. However, the difference between the similarity of the zero-sequence currents of L3 and L4 still exists at this time. It can be seen that when a single-phase grounding fault occurs on different types of lines or at different positions of the lines, the line selection method of the present invention is still effective and has good adaptability.

[0090] It is assumed that a single-phase grounding fault with different initial fault angles occurs at L2, which is 5 km away from the busbar. The fault transition resistance is 75Ω, and the initial fault angles are 0°, 30°, 45°, and 90° respectively. The specific line selection results are shown in Table 4.

[0091] Table 4 Line selection results at different fault initial phase angles

[0092]

[0093] It can be seen from Table 4 that when a single-phase ground fault with different initial fault angles occurs in the system, the result obtained by cluster analysis is consistent with the actual fault line, and the line selection method of the present invention can correctly determine the fault line. Therefore, when the fault occurs near the phase voltage passing through 0, near the peak value, or in other situations, the line selection method of the present invention can accurately determine the fault line and has good adaptability.

[0094] In summary, in the simulation model, different fault conditions are simulated, the energy coefficient of the high-frequency component of the transient zero-sequence current of each line and the distance coefficient characterizing the similarity characteristics of the transient zero-sequence current waveforms between different lines are calculated, and the line selection results obtained by cluster analysis are consistent with the actual fault line. The line selection results of the line selection method proposed in the present invention are still correct under different fault conditions.

[0095] On-site data verification

[0096] In order to further verify the practical value of the present invention, actual measured data was obtained using a low-voltage equivalent experimental platform built by a company to verify the practicality of the present invention.

[0097] The experimental platform consists of a 0.4 / 0.4kV power transformer, a three-phase five-column arc suppression coil, and four lines. The experimental operation is completed at a voltage level of 0.4kV. In order to fit the actual distribution network system fault characteristics as closely as possible, the three-phase conductor is wound 25 turns on the primary side of the zero-sequence current transformer to achieve consistency with the primary side current in the 10kV system.

[0098] The first set of parameters is set as follows: fault line L1, fault transition resistance 500Ω, sampling frequency 12.5kHz. Since the collected zero-sequence current contains a large noise signal, a zero-sequence current filtering link is added. The actual measured zero-sequence current of each line and the zero-sequence current waveform after filtering are as follows: Figure 4 shown.

[0099] The second set of parameters is set as follows: fault line L1, fault transition resistance 1000Ω, the actual measured zero-sequence current of each line and the zero-sequence current waveform after filtering are as follows: Figure 5 shown.

[0100] By changing the transition resistance value, the zero-sequence current of each line under different transition resistances is collected, and the line selection criterion proposed in the present invention is used to extract the fault characteristic quantity of each line. The original data set is composed of the simulation data as the input of the clustering algorithm, and the clustering label of each line is obtained as the output. The calculated fault characteristic quantity and clustering results are shown in Table 5.

[0101] Table 5 Fault characteristics extracted from experimental data and line selection results

[0102]

[0103] It can be seen from Table 5 that when the transition resistance increases from 500Ω to 5000Ω, the fault characteristics of the field data can still be measured by the energy coefficient and the distance coefficient. The line selection results prove that the line selection method of the present invention has certain practicality.

[0104] The above description and embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A fault line selection method for a resonant grounding system based on DTW-Hilbert and improved K-means, characterized in that: The steps include: Step 1: Build a distribution network simulation model with N outgoing lines, set different types of single-phase grounding faults in the built model, collect transient zero-sequence current data of each line, and select the signal sequence in a specific section as the original current data; Step 2: input the original current data obtained in step 1 into the DTW distance algorithm, and use the DTW distance algorithm to calculate the distance coefficient of each line, thereby measuring the waveform similarity between the zero-sequence currents of each line; Step 3, inputting the original current data obtained in step 1 into the VMD algorithm, the VMD algorithm decomposes the original current data into high-frequency components and low-frequency components, and calculates the Hilbert envelope energy of the high-frequency components; Step 4: Use the improved K-means clustering algorithm to classify the fault feature data set and determine the fault line based on cluster analysis.

2. The method for fault line selection in a resonant grounding system based on DTW-Hilbert and improved K-means according to claim 1, characterized in that: The specific section in step 1 is: selecting 1 / 4 power frequency cycle before the fault occurs to 3 / 4 power frequency cycle after the fault occurs as the fault feature extraction section, and only using the transient zero-sequence current data in this section as the original current data to complete the subsequent steps.

3. The method for fault line selection in a resonant grounding system based on DTW-Hilbert and improved K-means according to claim 1, characterized in that: The specific steps of calculating the distance coefficient in step 2 are: 1) Normalize the original current data according to the following formula: Among them, l i is the i-th data point in the original current data, L i is the i-th data point after normalization. 2) Calculate the DTW distance between the normalized original current data of N lines to form a DTW distance matrix D: Among them, d(L i ,L j ) represents line L i and line L j DTW distance between raw current data; 3) The elements of each column in the D matrix are summed up to obtain the dynamic time bending distance vector d, which can reduce the order of the D matrix: Among them, d i is the algebraic sum of the elements in the i-th column of the D matrix. The branch with the largest element value in the vector d is the fault branch. The distance coefficient ρ is calculated according to the following formula: Among them, d max is the largest element in the vector d. After normalization, the distance coefficient of the faulty line is ρ=1, and is greater than the distance coefficients of all sound lines.

4. The method for fault line selection in a resonant grounding system based on DTW-Hilbert and improved K-means according to claim 1, characterized in that: The implementation method of calculating the Hilbert envelope energy of the high-frequency component in step 3 is: 1) The original current data is decomposed into high-frequency components and low-frequency components using the VMD algorithm, and the Hilbert envelope energy of the high-frequency component is calculated; 2) Calculate the high frequency component c according to the following formula: i Hilbert transform of (t): in, c i (t) Hilbert transform, t is time, τ is a certain moment; 3) Calculate the Hilbert envelope energy according to the following formula: in, Z is the number of sampling points; 4) Normalize the envelope energy and calculate the energy coefficient according to the following formula: Where N is the total number of lines.

5. The method for fault line selection in a resonant grounding system based on DTW-Hilbert and improved K-means according to claim 1, characterized in that: The fault information of each line (ρ i ,h i ) is organized into a fault data set as the input of the improved K-means clustering algorithm. The number of clusters K is set to 2, and all data are clustered. The clustering algorithm outputs the clustering label of each line, and the faulty line is determined based on the clustering label.

Images