A regional protection method for active distribution network based on GFT spectrum coefficient

Abstract

The present invention discloses a method for active distribution network area protection based on GFT spectrum coefficients. According to the topological structure of a feeder of the distribution network, the present invention uses the positive sequence fault components of the current on both sides of each line section to construct an undirected weighted graph, with each line as a node of the graph, and the input signal of each node is the cosine similarity between the positive sequence current fault components on both sides of the line. The weight element in the adjacency matrix of the graph is the average value of the Euclidean distance between the two end nodes of the edge, and a Laplace matrix is ​​constructed. According to the standard orthogonal decomposition, an eigenvector matrix is ​​obtained, and then each spectral coefficient is obtained; the closeness between all spectral coefficients and each row of the eigenvector matrix is ​​calculated to establish a line fault area protection criterion. The method is not affected by the fault location, fault type, distributed power type, and transition resistance, and can tolerate a transition resistance of 100 ohms. By using the current signals of each line in a region, the faulty line can be effectively detected, and it has high reliability.

Description

Technical Field

[0001] The present invention relates to the technical field of active distribution network line protection, and in particular to an active distribution network area protection method based on GFT spectrum coefficients. Background Art

[0002] As my country vigorously develops a low-carbon, safe and efficient energy system, it builds a new power system with new energy as the main body. It is mainly based on distributed energy, connecting small and scattered new energy to the distribution network. The introduction of distributed generation (DG) into the distribution network has changed the structure of the traditional distribution network, from a simple network with single-power radial power supply to a complex network with multiple power supplies and multiple terminals. The fault characteristics of short-circuit current have also changed, which has a great impact on the selectivity, sensitivity and even reliability of traditional three-stage current protection, and will bring disadvantages to the stable operation of the distribution network. For the active distribution network connected to motor type DG (motor type distributed generator, MTDG) and inverter type DG (inverter-interfaced distributed generator, IIDG) fault, their fault characteristics are different, and the output of inverter type DG has intermittent characteristics, which brings difficulties to the judgment and setting of active distribution network line protection.

[0003] Zhu Yuyong, Lv Feipeng, Liao Xiaojun, et al. (Fault line selection in small current grounding system based on GFT spectrum similarity of the road map [J]. Electric Power Automation Equipment, 2023, 43(11): 89-94+116.) proposed a fault line selection method based on the similarity of the graph Fourier transform (GFT) spectrum of the road map. First, the road map GFT analysis of the transient zero-sequence current of each outgoing line was performed, and it was found that the road map GFT spectrum of the non-fault line was similar and significantly different from the fault line; then, the road map GFT spectrum difference between the fault line and the non-fault line was comprehensively characterized by the two dimensions of comprehensive discrete Fréchet distance and comprehensive cosine similarity, so as to screen out the fault line. However, this method can only determine whether a feeder of the distribution network has a fault, and cannot determine the line section fault, and does not consider DG access.

[0004] Liao Xiaojun (Research on Power Grid Protection and Fault Diagnosis Methods Based on Graph Signal Modeling and Analysis [D]. Doctoral Dissertation of Southwest Jiaotong University) proposed a method to use the information of starting components to model graph signals and realize rapid power grid fault diagnosis. First, based on the starting component information of all fault recordings and protection devices of the power grid, the conflict graph model is used to construct the starting sensitivity graph signal based on the starting network information of the main wiring diagram; then, the fault branch has the largest contribution rate to the high-frequency part of the graph spectrum, and the fault branch components of the power grid line or transformer are identified by comparing the contribution rates of each transmission line. However, this method is mainly aimed at transmission lines, and does not consider the impact of DG access on protection.

[0005] The methods proposed in the above-mentioned prior art have their own shortcomings. The method of using the three-phase current of the line for protection will be affected by the load current, transition resistance, etc., and the impact of the access of different types of DGs such as IIDG and MTDG on the protection of the distribution network is not fully considered. In order to solve this problem, the present invention proposes an active distribution network regional protection method based on the graph Fourier transform GFT spectrum coefficient. Summary of the invention

[0006] The purpose of the present invention is to propose an active distribution network regional protection method based on GFT spectrum coefficients to solve the problems raised in the background technology.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions:

[0008] A method for protecting an active distribution network area based on GFT spectrum coefficients comprises the following steps:

[0009] S1. Sample the three-phase currents on both sides of each line of a feeder in the distribution network to monitor whether the current mutation exceeds the starting threshold. If it exceeds, the protection is started and the positive sequence fault component of the current on both sides of the line section is calculated;

[0010] S2. Calculate the cosine similarity and Euclidean distance of the current fault components on both sides of each line based on the positive sequence fault components of the current on both sides of the line section obtained in S1, as the cosine similarity and Euclidean distance of each line;

[0011] S3. Construct an undirected weighted graph, with each line of a feeder of the distribution network as each node of the graph. The elements in the adjacency matrix of the graph are the average values ​​of the Euclidean distances between the two end nodes of each edge, i.e., the two lines. The Laplace matrix of the undirected weighted graph is constructed based on this. Then, the eigenvector matrix is ​​obtained by standard orthogonal decomposition.

[0012] S4, taking the normalized value of the cosine similarity of each node as the input signal of each node in the graph, and multiplying it with the eigenvector matrix to obtain each spectral coefficient;

[0013] S5, calculating the closeness between all spectral coefficients and each row of the eigenvector matrix;

[0014] S6. Compare the calculation results of the closeness of each row obtained in S5 with the protection threshold. If the closeness of the row is greater than the protection threshold, it is determined that the node corresponding to the row is faulty, and then it is determined that the line where the node is located is faulty; if the closeness is less than the protection threshold, it is determined to be a normal line.

[0015] Preferably, the S1 specifically includes the following contents:

[0016] Collect the current of each phase on the bus B side of line BC close to the power supply. When the current mutation exceeds the starting threshold, the protection is started and the positive sequence current on the bus B side is calculated. Subtract the normal current 8 cycles ago to get the positive sequence fault component current on the bus B side

[0017] Similarly, the phase currents on the bus C side of line BC close to the load are collected to calculate the positive sequence fault component current on the bus C side:

[0018] Preferably, S2 specifically includes the following contents:

[0019] The positive sequence fault component current on both sides of the line after the fault is obtained Calculate the cosine similarity S between the positive sequence current fault components on both sides of the line in the cycle time window before each sampling time t after the fault. COSBC,t 、Euclidean distance E BC,t , respectively stored in array S COSBC 、E BC ;

[0020] Similarly, for the upstream line AB of line BC on the same feeder line of the distribution network, and the parallel branch line BD on the B side, the cosine similarity S between the positive sequence current fault components on both sides of line AB and line BD is calculated respectively. COSAB , S COSBD , calculate the corresponding Euclidean distance E AB 、E BD ;

[0021] Using the obtained Euclidean distance array E of the four cycles of line AB, line BC, and line BD after the fault BC 、E AB 、E BD , calculate E BC 、E AB 、E BD The average Euclidean distance of ABm 、E BCm 、E BDm .

[0022] Preferably, S3 specifically includes the following contents:

[0023] S3.1. According to the topological structure of each line in a feeder of the distribution network, each line is used as a node and the connection between the lines is used as an edge to construct an undirected weighted graph G, wherein the undirected weighted graph G includes n nodes, and a weighted adjacency matrix W is constructed. A non-diagonal weight element in W is set to be the average value of the average values ​​of the Euclidean distances of the two end nodes of the edge corresponding to the row and column number, i.e., the average value of the average values ​​of the Euclidean distances of the two adjacent lines. The weight element of each node in W to itself is 0, i.e., the diagonal elements of W are 0;

[0024] S3.2. For the undirected weighted graph G constructed in S3.1, calculate its Laplace matrix L, which is defined as:

[0025] L=DW (1)

[0026] Where D is a weighted degree diagonal matrix, and the diagonal elements of D are the degrees of node i. Represents the sum of the weights of the edges connected to node i in the graph; then, the matrix L is decomposed according to the standard orthogonal decomposition to obtain its eigenvector matrix V, where each column in V is an eigenvector and the value range of each element in the eigenvector is [-1,1].

[0027] Preferably, the S4 specifically includes the following contents:

[0028] Since the cosine similarity S of the positive sequence current fault components on both sides of each line, that is, each node in the undirected weighted graph G at each sampling time t after the fault COS The value range is [-1,1], for S COS Carry out (S COS +1) / 2, and becomes a [0,1] value, which is used as the input signal array S of each node at time t ignal , where each element is S ignal (i), i is the node number, i = 1, 2, ..., n;

[0029] The input signal array S ignal Multiplying with the matrix V, we get the spectral coefficient C of the graph G at time t off An array containing n spectral coefficients, the specific formula is:

[0030]

[0031] In the formula, C off (k) represents the kth spectral coefficient, which is an array S ignal The sum of the products of each element in and each element of the k-th column eigenvector in the matrix V; V ikis the element in the i-th row and k-th column of the matrix V.

[0032] Preferably, the S5 specifically includes the following contents:

[0033] Due to S ignal The value range of is [0,1], and the value range of each element in V is [-1,1]. When node i fails, the corresponding input signal S ignal (i) is close to 1, and the input signals of other non-faulty nodes are close to 0;

[0034] According to formula (2), the kth spectral coefficient C off (k) represents each product S ignal (1) ×V 1k ,…,S ignal (i)×V ik ,…,S ignal (n)×V nk The sum of, where S ignal (i)×V ik It is larger than the value of other products and is dominant. Therefore, C off (k) and S ignal (i)×V ik is numerically close to C off (k) and V ik are close and have the same sign, indicating that when node i fails, the kth spectral coefficient C off (k) and the element V in row i and column k in V ik Similarly, other spectral coefficients such as C off (j) and the element V in row i and column j in V ij Therefore, each spectrum coefficient is close to the corresponding column element of the i-th row in V, and then it is judged that the node i is faulty;

[0035] For each node in the undirected weighted graph G, calculate the n spectral coefficients C within 4 cycles after the fault off The average value C offm ; Through a cycle, the average value of the i-th spectrum coefficient C offm (i) Subtract the element V(1,i) in the ith column of the first row of the matrix V and take the absolute value as the deviation d between the ith spectral coefficient and V(1,i) evia (i) 1-d evia (i) is the closeness between the i-th spectral coefficient and V(1,i), A ppro (i) = 1-d evia (i); In this loop, take i from 1 to n, and get the closeness between each spectral coefficient and the corresponding column element of the first row of matrix V in turn, and put it into array A ppro Calculate array Appro The average value is taken as the closeness A between all spectral coefficients and the first row of matrix V p (1); Repeat the above operation to obtain the closeness between all spectral coefficients and other rows of the matrix V.

[0036] Preferably, the S6 specifically includes the following contents:

[0037] For the closeness array A p Each element A p (i), respectively compare it with the protection threshold A pset Compare and establish the line fault judgment criteria for active distribution network area protection as follows:

[0038] A p (i)>A pset (3)

[0039] In the formula, A pset To protect the threshold, taking into account the influence of noise, take A pset =0.55;

[0040] If A p (i) satisfies formula (3), which means that all spectrum coefficients are close to the i-th row of matrix V, that is, it is judged that node i has a fault and the line where node i is located has a fault.

[0041] Compared with the prior art, the present invention provides an active distribution network regional protection method based on GFT spectrum coefficients, which has the following beneficial effects:

[0042] The present invention proposes a method for regional protection of active distribution network based on GFT spectrum coefficients. According to the topological structure of a feeder of the distribution network, the method uses the positive sequence fault components of the current on both sides of each line section to calculate the cosine similarity and the average value of the Euclidean distance between the positive sequence current fault components on both sides of each line, construct an undirected weighted graph, take each line as a node of the graph, the input signal of each node in the graph is the cosine similarity, the weight of the edge is the average value of the Euclidean distance of the nodes on both sides, construct a Laplace matrix, obtain the eigenvector matrix according to the standard orthogonal decomposition, and then obtain each spectrum coefficient; calculate the closeness between all spectral coefficients and each row of the eigenvector matrix, and establish the regional protection criterion of line fault. The method is not affected by the fault location, fault type, distributed power type, and transition resistance, can tolerate a transition resistance of 100 ohms, and can be adapted to the downstream connected to two types of DGs, IIDG and MTDG, and their mixed conditions. The method can effectively detect the faulty line and has high reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A flow chart of an active distribution network regional protection method based on GFT spectrum coefficients proposed by the present invention;

[0044] Figure 2 A topological diagram of a distribution network containing two types of DGs mentioned in Embodiment 2 of the present invention;

[0045] Figure 3 It is a spectrum coefficient curve diagram of the ABG short circuit fault in the occurrence area at 95% of the downstream IIDG line BC mentioned in Example 2 of the present invention and the transition resistance is 100 ohms;

[0046] Figure 4 It is a spectrum coefficient curve diagram of the ABG short circuit fault in the occurrence area at 95% of the downstream IIDG line BC mentioned in Example 2 of the present invention, with a transition resistance of 100 ohms and 20 dB noise on the line AB;

[0047] Figure 5 It is a spectrum coefficient curve diagram of the ABG short circuit fault in the occurrence area at 95% of the downstream MTDG line BD mentioned in Example 2 of the present invention and the transition resistance is 100 ohms;

[0048] Figure 6 It is a spectrum coefficient curve diagram of the ABG short circuit fault in the occurrence area at 95% of the downstream MTDG line BD mentioned in Example 2 of the present invention, with a transition resistance of 100 ohms and 20 dB noise on line AB. DETAILED DESCRIPTION

[0049] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0050] See also Figure 1 The present invention proposes a method for protecting an active distribution network area based on GFT spectrum coefficients, which specifically includes the following steps:

[0051] Step 1: Collect the current of each phase on the bus B side of line BC close to the power supply. When the current mutation exceeds the starting threshold, the protection is started and the positive sequence current on the bus B side is calculated. Subtract the normal current 8 cycles ago to get the positive sequence fault component current on the bus B side Similarly, the phase currents on the bus C side of line BC close to the load are collected to calculate the positive sequence fault component current on the bus C side:

[0052] Step 2: Use the positive sequence fault component current on both sides of the line after the fault obtained in step 1 Calculate the cosine similarity S between the positive sequence current fault components on both sides of the line in the cycle time window before each sampling time t after the fault. COSBC,t 、Euclidean distance E BC,t, respectively stored in array S COSBC 、E BC ;

[0053] Similarly, for the upstream line AB of line BC on the same feeder line of the distribution network, and the parallel branch line BD on the B side, the cosine similarity S between the positive sequence current fault components on both sides of line AB and line BD is calculated respectively. COSAB , S COSBD , calculate the corresponding Euclidean distance E AB 、E BD ;

[0054] Using the obtained Euclidean distance array E of the four cycles of line AB, line BC, and line BD after the fault BC 、E AB 、E BD , calculate their average Euclidean distance E ABm 、E BCm 、E BDm ;

[0055] Step 3: According to the topological structure of each line in a feeder of the distribution network, with each line as a node and the connection between the lines as an edge, an undirected weighted graph G is constructed, which includes n nodes, and a weighted adjacency matrix W is constructed. A non-diagonal weight element in W is set to be the average value of the average value of the Euclidean distances of the two end nodes of the edge corresponding to the row and column number, that is, the two adjacent lines. The weight element of each node in W to itself is 0, that is, the diagonal elements of W are 0;

[0056] Step 4: For graph G, calculate its Laplace matrix L, defined as L = DW, where D is the degree diagonal matrix and the diagonal elements of D are the degrees of node i. Represents the sum of the weights of the edges connected to node i in the graph; then, for the matrix L, according to the standard orthogonal decomposition, find its eigenvector matrix V, each column in V is an eigenvector, and the value range of each element in the eigenvector is [-1,1];

[0057] Step 5: Since the cosine similarity S of the positive sequence current fault component of each line at each sampling time t after the fault, that is, each node in the graph G on both sides COS The range is [-1,1], and they are (S COS +1) / 2, and becomes a [0,1] value, which is used as the input signal array S of each node at time t ignal , where each element is S ignal (i), i is the node number, i = 1, 2, ..., n;

[0058] The input signal array S ignal Multiplying with the matrix V, we get the spectral coefficient C of the graph G at time toff An array containing n spectral coefficients, the kth spectral coefficient C off (k) is the array S ignal The sum of the products of each element in and each element of the k-th column eigenvector in V, that is,

[0059]

[0060] In the formula, C off (k) represents the kth spectral coefficient, which is an array S ignal The sum of the products of each element in and each element of the k-th column eigenvector in the matrix V; V ik is the element in the i-th row and k-th column of the matrix V.

[0061] Step 6: Due to S ignal The value range of is [0,1], and the value range of each element in V is [-1,1]. When node i fails, the corresponding input signal S ignal (i) is large, close to 1, and the input signals of other non-faulty nodes are small, close to 0. From formula (1), we can see that the kth spectrum coefficient C off (k) is the product S ignal (1) ×V 1k ,…,S ignal (i)×V ik ,…,S ignal (n)×V nk The sum of, where S ignal (i)×V ik It is larger than the value of other products and is dominant. Therefore, C off (k) and S ignal (i)×V ik is numerically close to C off (k) and V ik are close and have the same sign, indicating that when node i fails, the kth spectral coefficient C off (k) and the element V in row i and column k in V ik Similarly, other spectrum coefficients such as C off (j) and the element V in row i and column j in V ij Therefore, each spectrum coefficient is close to the corresponding column element of the i-th row in V, from which it can be judged that node i has a fault.

[0062] For each node in graph G, calculate the n spectral coefficients C within 4 cycles after the fault off The average value C offm ; Through a cycle, the average value of the i-th spectrum coefficient C offm(i) Subtract the element V(1,i) in the ith column of the first row of the matrix V and take the absolute value as the deviation d between the ith spectral coefficient and V(1,i) evia (i) 1-d evia (i) as the closeness A between the i-th spectral coefficient and V(1,i) ppro (i) = 1-d evia (i); In this loop, take i from 1 to n, and get the closeness between each spectral coefficient and the corresponding column element of the first row of matrix V in turn, and put it into array A ppro Calculate array A ppro The average value is taken as the closeness A between all spectral coefficients and the first row of matrix V p (1);

[0063] Similarly, the closeness between all spectral coefficients and other rows of matrix V is obtained, such as A p (2), …, A p (n);

[0064] Step 7: For the closeness array A p Each element A p (i), respectively compare it with the protection threshold A pset Compare and establish the line fault judgment criteria for active distribution network area protection as follows:

[0065] A p (i)>A pset (3)

[0066] In the formula, A pset To protect the threshold, taking into account the influence of noise, take A pset =0.55;

[0067] If A p (i) satisfies formula (3), which means that all spectrum coefficients are close to the i-th row of matrix V, that is, if node i fails, then the line where node i is located fails.

[0068] Based on the above content, the specific implementation is as follows.

[0069] Embodiment 1:

[0070] In order to verify the effectiveness of this protection method, the PSCAD / EMTDC simulation software is used to build the following Figure 2 The 10kV active distribution network model is shown in Figure 1. The reference voltage is 10.5kV, the transformer capacity is 50MVA, the line parameters are (0.17+j0.34)Ω / km, the length is 4km, the IIDG capacity is 2.5MW, the MTDG capacity is 5MW, and the load active and reactive power are 0.8MW and 0.4MVA respectively. It is assumed that various faults occur in each line at 2.4s.

[0071] Construct an undirected weighted graph G, let Figure 2 Line AB is node 1 of graph G, line BC is node 2 of graph G, and line BD is node 3 of graph G. Graph G has 3 nodes. The weight of edge 12 between node 1 and node 2 is w 12 is the average Euclidean distance E ABm 、E BCm The effective value of w 12 =(E ABm +E BCm ) / 2; Similarly, the weight w of edge 13 between node 1 and node 3 13 =(E ABm +E BDm ) / 2, the weight w of edge 23 between nodes 2 and 3 23 =(E BCm +E BDm ) / 2; adjacency matrix W = [0w 12 w 13 ;w 21 0w 23 ;w 31 w 32 0], where w 21 =w 12 , w 31 =w 13 , w 32 =w 23 .

[0072] set up Figure 2 At 95% of the line BD containing MTDG in the middle and downstream, the transition resistance is 100 ohms and an ABG short circuit grounding fault occurs in the area. At this time, the following matrices D, W, L, and V are calculated:

[0073]

[0074]

[0075]

[0076]

[0077] The calculated input signal of each node is [0,0,0.9190];

[0078] The calculated average values ​​of the spectrum coefficients after the first cycle after the fault are 0.5351, 0, and 0.7567 respectively;

[0079] The calculated closeness of the average values ​​of each spectral coefficient to the 1st, 2nd, and 3rd rows of the matrix V is 0.064, 0.064, and 0.970 respectively;

[0080] According to formula (3), the closeness between the average value of each spectrum coefficient and the third row of V is 0.970, which is greater than the protection threshold. It can be judged that node 3, that is, line BD, is faulty, and the judgment is correct.

[0081] The spectral coefficient curve of the fault is as follows: Figure 3 shown. Figure 3 The three horizontal lines represent the values ​​of the elements in the third row of the matrix V, and the three curves represent the three spectral coefficient curves at each moment. It can be seen that the three spectral coefficient curves are relatively close to the elements in the third row of the matrix V, which verifies the effectiveness of this method.

[0082] Other fault conditions are as follows: 95% of the downstream IIDG line BC, the transition resistance is 100 ohms, the ABG short circuit fault occurs in the area, and the line AB has 20dB noise; 95% of the downstream MTDG line BD, the transition resistance is 100 ohms, the ABG short circuit fault occurs in the area, and the spectrum coefficient curve of the downstream MTDG line BD is 95% and the transition resistance is 100 ohms. Figures 4 to 6 shown.

[0083] Embodiment 2:

[0084] Based on Example 1, but different in that a specific experiment is designed to illustrate the effectiveness of the active distribution network area protection method based on GFT spectrum coefficients proposed by the present invention, and the specific content is as follows.

[0085] 1) Short circuit fault test

[0086] Figure 2 When the middle lines AB, BC, and BD have short-circuit faults in the AG, AB, ABG, and ABC regions at 5%, 50%, and 95% with transition resistances of 0.01Ω, 50Ω, and 100Ω, respectively, the corresponding simulation results are shown in Tables 1 to 9. AG, AB, ABG, and ABC represent phase A grounding fault, phase A and phase B short-circuit fault, phase A and phase B short-circuit grounding fault, and three-phase short-circuit grounding fault, respectively.

[0087] Table 1 Fault simulation results of line AB at f1 with 0.01Ω transition resistance

[0088]

[0089]

[0090] Table 2 Fault simulation results of line AB at f1 with 50Ω transition resistance

[0091]

[0092] Table 3 Fault simulation results of line AB at f1 with 100Ω transition resistance

[0093]

[0094]

[0095] Table 4 Fault simulation results of line BC at f2 with 0.01Ω transition resistance

[0096]

[0097] Table 5 Fault simulation results of line BC at f2 with 50Ω transition resistance

[0098]

[0099]

[0100] Table 6 Fault simulation results of line BC at f2 with 100Ω transition resistance

[0101]

[0102] Table 7 Fault simulation results of line BD at f3 with 0.01Ω transition resistance

[0103]

[0104]

[0105] Table 8 Fault simulation results of line BD at f3 with 50Ω transition resistance

[0106]

[0107] Table 9 Fault simulation results of line BD at f3 with 100Ω transition resistance

[0108]

[0109]

[0110] It can be seen from Tables 1 to 9 that this protection method is not affected by the fault location, fault type, distributed power type, and transition resistance of each line. Whether the line BC is connected to IIDG downstream, the line BD is connected to MTDG downstream, or the line AB is connected to IIDG and MTDG downstream at the same time, this protection method can accurately detect the fault line, and different transition resistances have little effect on this protection method. This verifies the effectiveness of this protection method.

[0111] 2) Noise resistance test

[0112] Set various faults at 95% of line BC and 100Ω transition resistance, and add 30dB and 20dB white noise to line AB respectively. The simulation results are shown in Tables 10 and 11. It can be seen that the noise added to the normal line AB increases from 30dB to 20dB, and the closeness of each spectrum coefficient to the second row of matrix V is slightly reduced, but still greater than the protection threshold, and node 2 can be accurately judged as a faulty node, that is, line BC is a faulty line; due to the addition of noise to the normal line AB, the closeness of each spectrum coefficient to the first and third rows of matrix V is slightly increased, but nodes 1 and 3 will not be misjudged as faulty lines, ensuring the reliability of this method. It is verified that this method has good noise tolerance and can tolerate 20dB noise.

[0113] Table 10 Fault simulation results at 95% of line BC f2, 100Ω transition resistance and 30dB noise on line AB

[0114]

[0115] Table 11 Fault simulation results at 95% of line BC f2, 100Ω transition resistance and 20dB noise on line AB

[0116]

[0117]

[0118] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for protecting an active distribution network area based on GFT spectrum coefficients, characterized in that: The following steps are involved: S1. Sample the three-phase currents on both sides of each line of a feeder in the distribution network to monitor whether the current mutation exceeds the starting threshold. If it exceeds, the protection is started and the positive sequence fault component of the current on both sides of the line section is calculated; S2. Calculate the cosine similarity and Euclidean distance of the current fault components on both sides of each line based on the positive sequence fault components of the current on both sides of the line section obtained in S1, as the cosine similarity and Euclidean distance of each line; S3. Construct an undirected weighted graph, with each line of a feeder of the distribution network as each node of the graph. The elements in the adjacency matrix of the graph are the average values ​​of the Euclidean distances between the two end nodes of each edge, i.e., the two lines. The Laplace matrix of the undirected weighted graph is constructed based on this. Then, the eigenvector matrix is ​​obtained by standard orthogonal decomposition. S4, taking the normalized value of the cosine similarity of each node as the input signal of each node in the graph, and multiplying it with the eigenvector matrix to obtain each spectral coefficient; S5, calculating the closeness between all spectral coefficients and each row of the eigenvector matrix; S6. Compare the calculation results of the closeness of each row obtained in S5 with the protection threshold. If the closeness of the row is greater than the protection threshold, it is determined that the node corresponding to the row is faulty, and then it is determined that the line where the node is located is faulty; if the closeness is less than the protection threshold, it is determined to be a normal line.

2. The method for protecting an active distribution network area based on GFT spectrum coefficients according to claim 1, characterized in that: The S1 specifically includes the following contents: Collect the current of each phase on the bus B side of line BC close to the power supply. When the current mutation exceeds the starting threshold, the protection is started and the positive sequence current on the bus B side is calculated. , minus the normal current 8 cycles ago, to obtain the positive sequence fault component current on the bus B side ; Similarly, the phase currents on the bus C side of line BC close to the load are collected to calculate the positive sequence fault component current on the bus C side: .

3. The method for protecting an active distribution network area based on GFT spectrum coefficients according to claim 2, characterized in that: The S2 specifically includes the following contents: The positive sequence fault component current on both sides of the line after the fault is obtained , , calculate each sampling time after the fault t Cosine similarity between the positive sequence current fault components on both sides of the line within the first cycle time window S COSBC,t , Euclidean distance E BC,t , stored in arrays S COSBC , E BC ; Similarly, for the upstream line AB of line BC on the same feeder line of the distribution network, and the parallel branch line BD on the B side, the cosine similarity between the positive sequence current fault components on both sides of line AB and line BD is calculated respectively. S COSAB , S COSBD , calculate the corresponding Euclidean distance E AB , E BD ; Using the obtained Euclidean distance array of the four cycles of line AB, line BC, and line BD after the fault E BC , E AB , E BD , calculated E BC , E AB , E BD The average Euclidean distance E ABm , E BCm , E BDm .

4. The method for protecting an active distribution network area based on GFT spectrum coefficients according to claim 3 is characterized in that: The S3 specifically includes the following contents: S3.

1. According to the topological structure of each line in a feeder of the distribution network, an undirected weighted graph is constructed with each line as a node and the connection between the lines as an edge. G , the undirected weighted graph G include n nodes, construct an adjacency matrix with weights W ,set up W The weight element of a non-diagonal line in is the average of the average Euclidean distances of the two end nodes of the edge corresponding to the row and column number, i.e., the average of the average Euclidean distances of the two adjacent lines. W The weight element of each node to itself is 0, that is W The diagonal elements of are 0; S3.

2. For the undirected weighted graph constructed in S3.1 G , calculate its Laplacian matrix L , defined as: L = D – W (1) in, D is the weighted degree diagonal matrix, D The diagonal elements are nodes i Degree , which means that the graph is related to the node i The sum of the weights of the edge connections; Then, for the matrix L According to the standard orthogonal decomposition, find its eigenvector matrix V , V Each column in is a eigenvector, and the value range of each element in the eigenvector is [-1, 1].

5. The method for protecting an active distribution network area based on GFT spectrum coefficients according to claim 4 is characterized in that: The S4 specifically includes the following contents: Each sampling moment after the fault t Each path is an undirected weighted graph G The cosine similarity of the positive sequence current fault components on both sides of each node S COS The value range of is [-1,1]. S COS conduct( S COS +1) / 2, and becomes a value of [0,1] as the moment t The input signal array of each node S ignal , where each element is S ignal ( i ), i is the node number, i =1, 2,… , n ; The input signal array S ignal With the matrix V Multiply them to get the sampling time t Below G The spectral coefficient C off Array, the C off Array contains n Spectral coefficients, the specific formula is expressed as: (2) In the formula, C off ( k ) indicates the k spectral coefficients, which is an array S ignal Each element and matrix V Middle k The sum of the products of the elements of the column eigenvector; V ik Representation Matrix V Middle i Line k Column elements.

6. The method for protecting an active distribution network area based on GFT spectrum coefficients according to claim 5, characterized in that: The S5 specifically includes the following contents: S ignal The value range of is [0,1], V The value range of each element in is [-1,1]. i When a fault occurs, the corresponding input signal S ignal ( i ) is close to 1, and the input signals of other non-faulty nodes are close to 0; According to formula (2), k Spectral coefficients C off ( k ) is the product of S ignal (1)× V 1k , … , S ignal ( i )× V ik , …, S ignal ( n )× V nk of and, in, S ignal ( i )× V ik It is larger than the value of other products and is dominant. Therefore, C off ( k )and S ignal ( i )× V ik is numerically close to C off ( k )and V ik Close and with the same sign, indicating nodes i When a fault occurs, k Spectral coefficients C off ( k )and V Middle i Line k Column Elements V ik Similarly, other spectral coefficients are analyzed as follows C off ( j )and V Middle i Line j Column Elements V ij close; therefore, the spectral coefficients are respectively V Middle i The row and column elements are close, and then the node is determined. i A malfunction occurs; For undirected weighted graphs G For each node in the n Spectral coefficients Coff The average Coffm ; Through a cycle, the first i The average value of the spectral coefficient Coffm ( i ) and the matrix V In row 1 i Column Elements V (1, i ) and then take the absolute value as the i The spectral coefficients and V (1, i ) devia ( i ), 1- devia ( i ) as the first i The spectral coefficients and V (1, i ), Appro ( i )=1- devia ( i ); In this loop, each spectrum coefficient and matrix are obtained in turn V No. i The closeness between the elements of each row and column is placed in a temporary array Appro Calculate the array Appro The average value of all spectral coefficients and the matrix V No. i Proximity between rows Ap ( i );in, i Take 1 to n , repeat the above operation to obtain all spectral coefficients and matrices V The closeness between the rows.

7. The method for protecting an active distribution network area based on GFT spectrum coefficients according to claim 6, characterized in that: The S6 specifically includes the following contents: For the proximity array Ap Elements Ap ( i ), and compare it with the protection threshold Apset Compare and establish the line fault judgment criteria for active distribution network area protection as follows: Ap ( i ) > Apset (3) In the formula, Apset To protect the threshold, considering the influence of noise, take Apset =0.55; if A p ( i ) satisfies formula (3), which means that all spectral coefficients and matrix V No. i The rows are close, that is, the judgment node i A failure occurred, the node i The line where the fault occurs.

Images