CN114966324A - Single-phase earth fault positioning method based on improved variational modal decomposition - Google Patents

Abstract

The invention discloses a single-phase earth fault positioning method based on improved variational modal decomposition, which comprises the following steps: s1, collecting transient zero sequence current waveform i of fault line _d (t) and an initial fault angle phi; s2, for transient zero sequence current waveform i _d (t) processing and obtaining zero sequence current waveform f for filtering power frequency component _d (t); s3, determining variation modal decomposition parameters K and alpha by using a slime algorithm; s4, filtering out zero sequence current waveform f of power frequency component _d (t) carrying out metamorphic modal decomposition, and removing high-frequency and white noise components to obtain a main resonant frequency component u' _d，k (g) (ii) a S5, calculating a correlation coefficient between main resonance frequency components of transient zero-sequence currents of adjacent detection points and a set threshold value rho _T And comparing to judge the fault section. The method of the invention removes the influence of noise and high-frequency components by utilizing variational modal decomposition, has strong anti-noise interference capability and greatly improves the positioning accuracy.

Description

A Single-Phase Earth Fault Location Method Based on Improved Variational Mode Decomposition

technical field

The invention relates to the field of single-phase grounding fault location of distribution network, in particular to a single-phase grounding fault location method based on improved variational mode decomposition.

Background technique

Most of my country's power systems use the neutral point ungrounded or grounded through arc suppression coils. The faults that occur in this grounding method during operation are mainly single-phase grounding faults. After the fault occurs, the system can continue to run for 1-2 hours. However, the rising voltage of the non-faulty phase may break down the weak link of the insulation, thereby causing a short-circuit fault between phases, causing serious losses to the power system. Therefore, it is necessary to conduct in-depth research on the single-phase grounding fault location of the grounding system in order to quickly and accurately find the fault location and solve the fault problem.

In recent years, many scholars at home and abroad have done a lot of research on the location of small current ground faults, but various methods have more or less defects. The active positioning method has high cost and is easily affected by the detection equipment. Although the passive positioning method overcomes some disadvantages of the former to a certain extent, it also has the problem of inaccurate positioning under the influence of the noise environment, and the existing In the method of judging the fault section by using the similarity of the transient zero-sequence current waveform upstream and downstream of the fault point, the method of extracting the main resonant frequency component of the zero-sequence current is too cumbersome, and the parameters K and α in the variational mode decomposition algorithm need artificial Therefore, the decomposition effect is unstable, and the slime mold algorithm can be used to optimize it. Therefore, in order to solve the above problems, the present invention proposes a single-phase grounding fault location method based on improved variational mode decomposition.

"A method for locating single-phase grounding fault based on data processing of distribution network" disclosed in Chinese patent documents, whose publication number is CN106990332B, discloses a method for locating single-phase grounding fault based on data processing of power distribution network, which is Including S01: Obtain massive real-time data of distribution network through distribution terminals, fault indicators, smart meters and other equipment; S02: Establish a real-time stream data analysis platform for distribution network based on Storm cluster; S03: Design and integrate multiple single-phase grounding faults Stream data processing topology of the location technology; S04: output and store the results according to the criteria of different single-phase-to-ground fault location technologies. However, the Chinese patent with publication number CN106990332B has a simple process and does not involve specific algorithms.

SUMMARY OF THE INVENTION

The invention solves the problem of inaccurate section positioning when the zero-sequence current has noise and power frequency and high frequency components in the small current grounding system of the distribution network after a metallic single-phase grounding fault occurs. The single-phase grounding fault location method based on modal decomposition, the method of the present invention uses variational modal decomposition to remove the influence of noise and high-frequency components, has strong anti-noise interference ability, and greatly improves the positioning accuracy.

In order to achieve the above object, the present invention adopts the following technical solutions: a single-phase grounding fault location method based on improved variational modal decomposition, comprising the following steps:

S1, collect the transient zero-sequence current waveform id ( _t ) and the initial fault angle φ of the faulty line;

S2, process the transient zero-sequence current waveform id (t) and obtain the zero-sequence current waveform f _d (t) with the power frequency component removed _;

S3, use the slime mold algorithm to determine the variational mode decomposition parameters K and α;

S4, perform variational modal decomposition on the zero-sequence current waveform f _d (t) filtered out of the power frequency component, and obtain the main resonance frequency component u′ _{d, k} (g) after removing the high frequency and white noise components;

S5: Calculate the correlation coefficient between the main resonant frequency components of the transient zero-sequence current at adjacent detection points, and compare it with the set threshold ρ _T to determine the faulty section.

In the present invention, firstly collect the d-th transient zero-sequence current waveform id (t), initialize _d =1, filter out the power frequency component in _{id (t), obtain f d (} t), and then use slime mold The algorithm determines the variational modal decomposition parameters K and α, and uses the numbers K and α to perform variational modal decomposition on f _d (t) to obtain K zero-sequence current components ud _{, k} (t), remove ud _{, k} ( t) Middle and high frequency components and noise components, obtain the main resonant frequency component waveform u' _{d, k} (g), calculate the correlation coefficient between u' _{d, k} (g) and u' _d+1 , k (g), Compared with the set threshold ρ _T , the fault section is judged; the method of the present invention uses the improved variational mode algorithm and the fundamental wave migration method to remove the influence of high-frequency components, power frequency components and noise in the zero-sequence current, and uses the extraction method. The obtained main resonant frequency is compared with the correlation coefficient, which improves the accuracy of fault section location.

Preferably, the step S1 is specifically: when a metallic single-phase ground fault occurs in the distribution network, the feeder terminal device is used to collect the transient zero-sequence current waveform id ( _t ) and the initial fault angle φ, where id ( _t ) is The d-th zero-sequence current waveform starting from the bus terminal, 0<t<t _b , t _b has the following relationship with the initial fault angle φ:

t _b =4T _b -2|sinφ|

In the formula: T _b is the power frequency period, and t _b is the acquisition duration. In the present invention, when a fault occurs, the transient zero-sequence current waveform id ( _t ) and the initial fault angle φ are collected according to the feeder terminal devices on all detection points of the faulty line.

Preferably, the step S2 includes the following steps:

S21, first perform Hilbert transform on the transient zero-sequence current waveform id ( _t ), specifically:

对

取实部得到滤除工频的零序电流波形f_d(t)。本发明中，首先对i_d(t)进行希尔伯特变换，随后构建信号，并滤除i_d(t)的工频分量，得到滤除工频的零序电流波形f_d(t)。S22, build signal

right

Take the real part to get the zero-sequence current waveform f _d (t) with the power frequency filtered out. In the present invention, firstly perform Hilbert transform on id ( _t ), then construct the signal, and filter out the power frequency component of _{id (t) to obtain the zero-sequence current waveform f d (} t) with filtered power frequency .

Preferably, the step S3 includes the following steps:

S31, initialize the number of components K and the accuracy factor α after the variational modal decomposition f _d (t), take K and α as the abscissa and ordinate, (2, 100) as the coordinate origin to establish a coordinate system, and the node coordinate is X _m,n (K,α), initialization m=1, n=1, m is the number of updates, M is the number of update terminations, n is the coordinate ordinal, and N is the number of coordinate points to be updated;

更新W_m(n)，其中，g为离散化点序列，G为离散化后的总点数，a_d，k(g)为u_d，k(g)的包络信号，

E_m(n)为第m次更新时第n个节点的适应度值，W_m(n)为第m次更新时第n个节点的权重；S32, perform variational modal decomposition on f _d (t) using the values of X _{m, n} coordinates K and α to obtain K zero-sequence current components ud _{, k} (t), k=1, 2,...,K ; and perform time discretization to obtain u _{d, k} (g), using the fitness function

Update W _m (n), where g is the discretized point sequence, G is the total number of points after discretization, a _{d, k} (g) is the envelope signal of ud _{, k} (g),

E _m (n) is the fitness value of the n-th node in the m-th update, and W _m (n) is the weight of the n-th node in the m-th update;

S33, judge whether n>=N is established, if yes, execute step S34, if not, n=n+1, return to step S32,

S34: Calculate the fitness values Em of all _N nodes updated for the _mth time, and sort them; and represents the smaller half of the nodes in Em, dis represents the larger half of the nodes in _Em , and E _b is _Em In the minimum value, E _w is the maximum value in E _m , and then let n=1;

更新S35, using

renew

In the formula, rand is a random number uniformly distributed between 0 and 1, x _min is (2, 10), x _max is (100, 20000), rand1, rand2 are random numbers between 0 and 1, X _b are the node coordinates with fitness value E _b , X _rand1 and X _rand2 are two random node coordinates, Z is 0.3, p=tanh(E _b -E _m (n)), randA is the sum of [-a, a] A random number between , randB is a random number between [-b, b], a=artanh(1-m/M), b=1-m/M;

S36, judge whether n>=N is established, if yes, execute step S37, if not, n=n+1, return to step S35;

S37, determine whether m+1>=M is established or whether more than 99% of the nodes are located at the same coordinate, if so, the node coordinate with the smallest E _m is the optimal solution of parameters K and α, if not, m=m+1, n =1, return to step S32. The method of the invention optimizes the selection of K and α values in the variational modal algorithm by using the slime mold algorithm, and can accurately decompose the transient zero-sequence current with the power frequency removed, so as to accurately locate the fault.

Preferably, the step S4 includes the following steps:

S41, perform variational modal decomposition on f _d (t) to obtain K zero-sequence current components ud _{, k} (t), k=1, 2, . . . , K, ud _{, k} (t) represent f _d ( t) The kth zero-sequence current component obtained by decomposition, perform time discretization on ud _,k (t) to obtain ud _,k (g), and calculate the kth discretized current component ud _,k (g) energy of

_{S42 ,} take the maximum value max(Q ₁ , Q ₂ , _. (g), removal of high frequency and white noise components. In the present invention, K and α are used to perform variational modal decomposition on f _d (t) to obtain K zero-sequence current components ud _,k (t), and the energy of ud _,k (t) is used to distinguish each component, and Remove the high frequency components and noise components, and obtain the main resonance frequency component waveform u' _{d, k} (g).

Preferably, the step S5 includes the following steps:

式中：u′_d，_k(g)和u′_d+1，k(g)分别为第d个与第d+1个检测点的暂态零序电流主谐振频率分量，d代表线路上第d个与第d+1个检测点之间的区段；S51, calculate the correlation coefficient between u' _{d, k} (g) and u' _{d+1, k} (g)

In the formula: u′ _d , _k (g) and u′ _{d+1, k} (g) are the main resonant frequency components of the transient zero-sequence current at the dth and d+1th detection points, respectively, and d represents the The section between the dth and d+1th detection points;

S52, if the correlation coefficient ρ _d is less than the set threshold ρ _T , determine that the d-th section is a faulty section, otherwise, go to step S53; S53, carry out d=d+1, and judge whether d>=D is established, if so , the system determines that the section after d+1 is a faulty section, and exits; if not, returns to step S2. In the present invention, starting from the busbar section of the faulty line, the main resonance frequency components u' _{d, k} (g) and u' _{d+1, k} (g) of the transient zero-sequence current waveform of two adjacent detection points are calculated. The correlation coefficient ρ _d of , and finally according to the correlation coefficient ρ _d , the fault section is judged.

The beneficial effects of the present invention are as follows: the method of the present invention utilizes the initial fault angle to select the optimal waveform duration, which reduces the time required for subsequent similarity judgment; also utilizes power frequency filtering to remove the power frequency, eliminating the power frequency pair similarity It also uses the slime mold algorithm to optimize the selection of K and α values in the variational modal algorithm, which can accurately decompose the transient zero-sequence current that removes the power frequency, so as to accurately locate the fault. Variational modal decomposition removes the influence of noise and high-frequency components, has strong anti-noise interference ability, and greatly improves positioning accuracy.

Description of drawings

1 is a flowchart of a single-phase-to-ground fault location method based on improved variational modal decomposition of the present invention;

2 is a schematic diagram of a distribution network of a single-phase-to-ground fault location method based on improved variational modal decomposition of the present invention;

3 is a schematic diagram of the original zero-sequence current waveform and the waveform after power frequency filtering in a single-phase-to-ground fault location method based on improved variational modal decomposition of the present invention;

4 is a schematic diagram of each component waveform of the zero-sequence current after decomposition and filtering using variational modal algorithm when determining K and α values in a single-phase grounding fault location method based on improved variational modal decomposition of the present invention.

Detailed ways

Example 1:

This embodiment proposes a single-phase-to-ground fault location method based on improved variational modal decomposition. Referring to FIG. 1 , the method mainly includes the following steps.

In step S1, the transient zero-sequence current waveform id ( _t ) and the initial fault angle φ of the faulty line are collected; specifically, in this step, when a metallic single-phase grounding fault occurs in the distribution network, all detection points of the faulty line are used. The feeder terminal device collects the transient zero-sequence current waveform id ( _t ) and the initial fault angle φ, id (t) is the _d -th zero-sequence current waveform from the bus terminal, initialized d=1, D is the fault to be diagnosed The number of line detections, 0<t<t _b , t _b has the following relationship with the initial fault angle φ: t _b =4T _b -2|sinφ|

Among them: T _b represents the power frequency period, and t _b represents the acquisition duration.

Step S2, process the transient zero-sequence current waveform id (t) and obtain the zero-sequence current waveform f _d (t) with the power frequency component filtered out _; specifically, this step includes two sub-steps, step S21 and step S22 .

In step S21, Hilbert transform is performed on the transient zero-sequence current waveform id ( _t ). Specifically, there is the following formula:

对

取实部得到滤除工频的零序电流波形f_d(t)。具体的，本实施例中，本发明中，首先对i_d(t)进行希尔伯特变换，随后构建信号，并滤除i_d(t)的工频分量，得到滤除工频的零序电流波形f_d(t)。Step S22, build a signal

right

Take the real part to get the zero-sequence current waveform f _d (t) with the power frequency filtered out. Specifically, in this embodiment, in the present invention, the Hilbert transform is first performed on id ( _t ), then a signal is constructed, and the power frequency component of id ( _t ) is filtered out to obtain zero filtered power frequency. sequence current waveform f _d (t).

Step S3, using the slime mold algorithm to determine the variational modal decomposition parameters K and α; specifically, the following sub-steps are included.

Step S31, initialize the number of components K and the accuracy factor α after the variational modal decomposition f _d (t), take K and α as the abscissa and ordinate, (2, 100) as the coordinate origin to establish a coordinate system, and the node coordinates are X _m,n (K,α), initialization m=1, n=1, m represents the number of updates, M represents the number of update terminations, n represents the coordinate number, and N represents the number of coordinate points to be updated; specifically, K is 2 A random natural number between 10 and 10; α is a natural number between 100 and 20000.

更新W_m(n)，上式中，g表示离散化点序列，G表示离散化后的总点数，a_d，k(g)表示u_d，k(g)的包络信号，其中

E_m(n)表示第m次更新时第n个节点的适应度值，W_m(n)表示第m次更新时第n个节点的权重。Step S32, perform variational modal decomposition on f _d (t) by the values of X _{m, n} coordinates K and α to obtain K zero-sequence current components ud _{, k} (t), k=1, 2,..., K; and perform time discretization to obtain u _{d, k} (g), and use the fitness function

Update W _m (n), in the above formula, g represents the discretized point sequence, G represents the total number of points after discretization, a _{d, k} (g) represents the envelope signal of ud _{, k} (g), where

E _m (n) represents the fitness value of the n-th node in the m-th update, and W _m (n) represents the weight of the n-th node in the m-th update.

In step S33, it is judged whether n>=N is established, if yes, execute step S34, if not, n=n+1, and return to step S32.

Step S34: Calculate the fitness value Em of all _N nodes updated for the _mth time, and sort them; and is the smaller half of the nodes in Em, dis is the larger half of the nodes in _Em , and E _b represents E The minimum value in _m , E _w represents the maximum value in E _m , and let n=1.

更新Step S35, use

renew

Among them, rand represents random numbers evenly distributed between 0 and 1, x _min represents (2,10), x _max represents (100, 20000), rand1, rand2 represent random numbers between 0 and 1, and X _b represents The fitness value is the node coordinates of E _b , X _rand1 and X _rand2 represent two random node coordinates, in this embodiment, Z is specifically 0.3, p=tanh(|E _b -E _m (n)|), randA represents A random number between [-a, a], randB represents a random number between [-b, b], a=artanh(1-m/M), b=1-m/M;

In step S36, it is judged whether n>=N is established, if yes, go to step S37, if not, n=n+1, and return to step S35.

Step S37, it is judged whether m+1>=M is established or whether more than 99% of the nodes are located at the same coordinate, if so, the coordinate of the node with the smallest E _m is the optimal solution of parameters K and α, if not, m=m+1, n=1, and the process returns to step S32. The method of the invention optimizes the selection of K and α values in the variational modal algorithm by using the slime mold algorithm, and can accurately decompose the transient zero-sequence current with the power frequency removed, so as to accurately locate the fault.

In step S4, variational modal decomposition is performed on the zero-sequence current waveform f _d (t) from which the power frequency component is filtered out, and the high frequency and white noise components are removed to obtain the main resonance frequency component u' _d,k (g). Specifically, the following sub-steps are included.

Step S41, perform variational modal decomposition on f _d (t) to obtain K zero-sequence current components ud _{, k} (t), where k=1, 2, . . . , K, ud _{, k} (t) represent The k-th zero-sequence current component obtained by decomposing f _d (t) is time discretized to ud _{, k} (t) to obtain ud _{, k} (g), and the k-th discretized current component ud _{, k} is calculated (g) energy

Step _{S42 ,} take the maximum value max (Q ₁ , Q ₂ , _{. k} (g), removing high frequency and white noise components. In this embodiment, K and α are used to perform variational modal decomposition on f _d (t) to obtain K zero-sequence current components ud _{, k} (t), and the energy of ud _{, k} (t) is used to distinguish each component, And remove the high frequency components and noise components, get the main resonance frequency component waveform ud _,k (g).

其中：u′_d，k(g)和u′_d+1，k(g)分别表示第d个和第d+1个检测点的暂态零序电流主谐振频率分量，d为线路上第d个与第d+1个检测点之间的区段。Step S5: Calculate the correlation coefficient between the main resonance frequency components of the transient zero-sequence current at adjacent detection points, and compare it with the set threshold ρ _T to determine the faulty section. Specifically, the correlation coefficient is obtained first, and then the comparison and judgment are performed. In step S51, the correlation coefficient between u'd _,k (g) and u'd _+1,k (g) is calculated

Among them: u' _{d, k} (g) and u' _{d+1, k} (g) represent the transient zero-sequence current main resonant frequency components of the d-th and d+1-th detection points, respectively, and d is the first The segment between the d and d+1th detection points.

Step S52, if the correlation coefficient ρ _d is smaller than the set threshold ρ _T , then determine that the d-th section is a fault section, otherwise, go to step S53 .

In step S53, d=d+1 is performed, and it is judged whether d>=D is established, if so, the system determines that the section after d+1 is a faulty section, and exits; if not, returns to step S2. In this embodiment, starting from the busbar section of the faulty line, calculate the sum of the main resonance frequency components u'd _,k (g) and u'd _+1,k (g) of the transient zero-sequence current waveform of two adjacent detection points The correlation coefficient ρ _d between them is finally determined according to the correlation coefficient ρ _d .

In the present invention, firstly collect the d-th transient zero-sequence current waveform id (t), initialize _d =1, filter out the power frequency component in _{id (t), obtain f d (} t), and then use slime mold The algorithm determines the variational modal decomposition parameters K and α, and uses the numbers K and α to perform variational modal decomposition on f _d (t) to obtain K zero-sequence current components ud _{, k} (t), remove ud _{, k} ( t) Middle and high frequency components and noise components, obtain the main resonant frequency component waveform u' _{d, k} (g), calculate the correlation coefficient between u' _{d, k} (g) and u' _{d+1, k} (g), Compared with the set threshold ρ _T , the fault section is judged; the method of the present invention uses the improved variational mode algorithm and the fundamental wave migration method to remove the influence of high-frequency components, power frequency components and noise in the zero-sequence current, and uses the extraction method. The obtained main resonant frequency is compared with the correlation coefficient, which improves the accuracy of fault section location.

Referring to Fig. 1, the working principle of the present invention is as follows: when a metallic single-phase grounding fault occurs in the low-current grounding system, the transient zero-sequence current is collected by the feeder terminal device that comes with the distribution network circuit, and the demand is calculated by the initial fault angle. Then use the fundamental wave migration method to filter the acquired transient zero-sequence current waveform; then use the slime mold algorithm to determine the K and α values of the variational modal decomposition, and then use the variational The modal algorithm decomposes the filtered zero-sequence current waveform, removes the noise and high-frequency components, and extracts the main resonant frequency component; finally, the faulty section is determined by using the correlation coefficient of the main resonant frequency components at both ends of each section.

Example 2:

On the basis of Example 1, the simulation experiment is used to describe in more detail. Referring to Figure 2, the system in Figure 2 is a cable-overhead line hybrid line, and the system capacitance current is 73A; the arc suppression coil adopts a compensation degree of 8%. Compensation, the equivalent inductance is 0.244H, and the series resistance is 1.53Ω; the specific parameters of the line are shown in Table 1; the detection points M, N, P, and Q are set on the faulty line, where M is the detection point of the bus end, and the fault point is O. It can be divided into 4 sections MN (4km), NP (7km), PQ (2km) and the last section (with H to represent its distance). Other parameters are shown in Figure 2 and Table 1.

Table 1 Line Model Parameters

Line parameters Overhead lines cable Positive sequence resistance/(Ω·km^-1) 0.17 0.27 Zero sequence resistance/(Ω·km^-1) 0.32 2.7 Positive sequence inductance/(mH·km^-1) 1.017 0.255 Zero-sequence inductance/(mH·km^-1) 3.56 1.109 Positive sequence capacitance/(μF·km^-1 0.115 0.376 Zero-sequence capacitance/(μF·km^-1 0.0062 0.276

When a single-phase ground fault occurs at point O, the system is A-phase ground fault, the fault line is L ₅ , the fault distance MO=9km, H=2km, the fault initial phase angle φ is 90°, and the grounding resistance R _g is 0.5 Ω and NP are fault sections.

Referring to Figure 3 and Figure 4, collect the transient zero-sequence current waveform of each detection point M, N, P, Q, and use the initial fault angle to select the waveform length as 2 power frequency cycles (the power frequency cycle is 0.02s, in Figure 3 , the waveform between 0.025s-0.065s in Figure 4), inject Gaussian white noise with signal-to-noise ratios of 10dB, 5dB, 15dB, and 10dB into the collected M, N, P, and Q signals, respectively, and then get Decompose the required zero-sequence current waveform; use the fundamental wave offset method to filter it at the power frequency to obtain the waveform shown in Figure 3; use the slime mold algorithm to determine the K and α values of the variational modal decomposition, and determine M , N point K and α values are 3 and 2058, P, Q point K and α values are 3 and 1721, and the transient zero-sequence current waveforms of the four detection points are the highest energy of NMF1 component, the main resonant frequency, the specific waveform The components are shown in Figure 4. Finally, the correlation coefficient between the main resonant frequency component waveform of the zero-sequence current waveform at each point is used to compare the similarity with the set threshold (the threshold is set to 0.8 in this simulation experiment), so as to locate the fault section, and the obtained simulation data As shown in table 2.

Table 2 Positioning results

It can be seen from Table 2 that the system will make a misjudgment when the similarity of zero-sequence current containing noise, power frequency and high frequency components is judged. Fault section; when the transient zero-sequence current is further injected with lower signal-to-noise ratio noise (3dB, 3dB, 5dB, and 5dB noise are injected at M, N, P, and Q points respectively), the method of the present invention can still be accurate to determine the faulty section.

To sum up, in the noise environment, the transient zero-sequence current waveform similarity comparison will be affected by noise, power frequency and high frequency components, resulting in misjudgment of fault location; and the present invention uses the initial fault angle to select the best The waveform duration reduces the time required for subsequent similarity judgment; and the variational modal decomposition algorithm after the fundamental wave migration method and the slime mold algorithm are used to optimize parameters to filter out power frequency components, noise components and high frequency components. After the components are extracted, the similarity is judged by using the extracted main resonance frequency components, which greatly improves the accuracy of segment location.

The above-mentioned embodiments are further elaboration and description of the present invention for the convenience of understanding, and are not intended to limit the present invention. Any modification, equivalent replacement and improvement made within the spirit and principle of the present invention shall be included in within the scope of protection of the invention.

Claims (6)

1. a single-phase grounding fault location method based on improved variational modal decomposition, is characterized in that, comprises the following steps: S1, collect the transient zero-sequence current waveform id ( _t ) and the initial fault angle φ of the faulty line; S2, process the transient zero-sequence current waveform id (t) and obtain the zero-sequence current waveform f _d (t) with the power frequency component removed _; S3, use the slime mold algorithm to determine the variational mode decomposition parameters K and α; S4, perform variational modal decomposition on the zero-sequence current waveform f _d (t) filtered out of the power frequency component, and obtain the main resonance frequency component u′ _{d, k} (g) after removing the high frequency and white noise components; S5: Calculate the correlation coefficient between the main resonant frequency components of the transient zero-sequence current at adjacent detection points, and compare it with the set threshold ρ _T to determine the faulty section. 2. A single-phase grounding fault location method based on improved variational modal decomposition according to claim 1, wherein the step S1 is specifically: when a metallic single-phase grounding fault occurs in the distribution network, use The feeder terminal device collects the transient zero-sequence current waveform id ( _t ) and the initial fault angle φ, id (t) is the _d -th zero-sequence current waveform from the bus terminal, 0<t<t _b , t _b is the same as the initial The fault angle φ has the following relationship: t _b =4T _b -2|sinφ| In the formula: T _b is the power frequency period, and t _b is the acquisition duration. 3. The single-phase grounding fault location method based on improved variational modal decomposition according to claim 1 or 2, wherein the step S2 comprises the following steps: S21, first perform Hilbert transform on the transient zero-sequence current waveform id ( _t ), specifically:

S22, build signal

right

Take the real part to get the zero-sequence current waveform f _d (t) with the power frequency filtered out. 4. A single-phase grounding fault location method based on improved variational modal decomposition according to claim 1 or 2, wherein the step S3 comprises the following steps: S31, initialize the number of components K and the accuracy factor α after the variational modal decomposition f _d (t), take K and α as the abscissa and ordinate, (2, 100) as the coordinate origin to establish a coordinate system, and the node coordinate is X _m,n (K,α), initialization m=1, n=1, m is the number of updates, M is the number of update terminations, n is the coordinate ordinal, and N is the number of coordinate points to be updated; S32, perform variational modal decomposition on f _d (t) using the values of X _{m, n} coordinates K and α to obtain K zero-sequence current components ud _{, k} (t), k=1, 2,...,K ; and perform time discretization to obtain u _{d, k} (g), using the fitness function

Update W _m (n), where g is the discretized point sequence, G is the total number of points after discretization, a _{d, k} (g) is the envelope signal of ud _{, k} (g),

E _m (n) is the fitness value of the n-th node in the m-th update, and W _m (n) is the weight of the n-th node in the m-th update; S33, judge whether n>=N is established, if yes, execute step S34, if not, n=n+1, return to step S32, S34: Calculate the fitness values Em of all _N nodes updated for the _mth time, and sort them; and represents the smaller half of the nodes in Em, dis represents the larger half of the nodes in _Em , and E _b is _Em In the minimum value, E _w is the maximum value in E _m , and then let n=1; S35, using

renew

In the formula, rand is a random number uniformly distributed between 0 and 1, x _min is (2, 10), x _max is (100, 20000), rand1, rand2 are random numbers between 0 and 1, X _b are the node coordinates with fitness value E _b , X _rand1 and X _rand2 are two random node coordinates, Z is 0.3, p=tanh(|E _b -E _m (n)|), randA is [-a, a ], randB is a random number between [-b,b], a=artanh(1-m/M), b=1-m/M; S36, judge whether n>=N is established, if yes, execute step S37, if not, n=n+1, return to step S35; S37, determine whether m+1>=M is established or whether more than 99% of the nodes are located at the same coordinate, if so, the node coordinate with the smallest E _m is the optimal solution of parameters K and α, if not, m=m+1, n =1, return to step S32. 5. A single-phase-to-ground fault location method based on improved variational modal decomposition according to claim 4, wherein the step S4 comprises the following steps: S41, perform variational modal decomposition on f _d (t) to obtain K zero-sequence current components ud _{, k} (t), k=1, 2, . . . , K, ud _{, k} (t) represent f _d ( t) The kth zero-sequence current component obtained by decomposing, perform time discretization processing on ud _,k (t) to obtain ud _,k (g), and calculate the kth discretized current component ud _,k (g) energy of

_{S42 ,} take the maximum value max(Q ₁ , Q ₂ , _. (g), removal of high frequency and white noise components. 6. The single-phase-to-ground fault location method based on improved variational modal decomposition according to claim 5, wherein the step S5 comprises the following steps: S51, calculate the correlation coefficient between u' _{d, k} (g) and u' _{d+1, k} (g)

In the formula: u′ _{d, k} (g) and u′ _{d+1, k} (g) are the main resonant frequency components of the transient zero-sequence current at the dth and d+1th detection points, respectively, and d represents the The section between the dth and d+1th detection points; S52, if the correlation coefficient ρ _d is less than the set threshold ρ _T , then determine that the d-th section is a fault section, otherwise step S53 is performed; S53, perform d=d+1, and determine whether d>=D is established, if yes, the system determines that the segment after d+1 is a faulty segment, and exits; if not, returns to step S2.

Images