A longitudinal protection method for transmission lines of large-scale wind farms based on edge detection
By using edge detection technology on the wind farm transmission and outlet line, the Hankel matrix is constructed and the Sobel operator is applied, the problem of the degradation of protection performance of the traditional wind farm longitudinal protection device in the wind power system is solved, and the weak wind farm is achieved Quick fault identification and accurate protection during force output.
Patent Information
- Application Number
- CN202111371391.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-11-18
- Publication Date
- 2024-11-05
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
When the traditional wind farm vertical connection protection device faces fault characteristics such as unstable impedance, weak feeding, frequency deviation and high harmonic content of wind power system, the protection performance is degraded, and the existing methods reduce the sensitivity of the wind farm when the output of the wind farm is weak or Ineffective.
Using edge detection method, Hankel matrix is constructed by the current values on both sides of the wind farm sending line, edge detection is performed through the Sobel operator, average gradient amplitude is calculated, and combined with setting value comparison, to quickly identify faults and regions in the zone. Discrimination of external faults.
This method is not affected by the current frequency offset of the wind farm fault, and is also applicable when the wind farm is weak, and the fault recognition speed is faster. It can accurately identify different types of faults within 1ms, improving the vertical protection performance of the wind farm transmission and outlet line .
Smart Images

Figure CN114219947B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of longitudinal protection of wind farm transmission lines, and in particular to a longitudinal protection method for large-scale wind farm transmission lines based on edge detection. Background Art
[0002] Line longitudinal protection is a protection device that causes the switches on both sides to trip quickly at the same time when a line fault occurs. It is the main protection of the line. It uses the specific relationship between the discrimination quantities on both sides of the line as the judgment criterion, that is, both sides transmit the discrimination quantity to the opposite side through the channel, and then both sides judge whether the fault is inside or outside the area according to the relationship between the discrimination quantity on the opposite side and the local side.
[0003] In recent years, wind power generation has developed rapidly. In order to achieve the goals of carbon peak and carbon neutrality and increase the proportion of wind energy consumption in primary energy consumption, large-scale wind farm grid-connected power generation has become an inevitable trend in China's energy and power sector. The transmission lines of large-scale wind farms are generally at a voltage level of 110kV and above. If the fault characteristics of wind power system impedance instability, weak feed, frequency deviation and high harmonic content are considered, the conventional transmission line protection scheme is used for protection configuration, which will lead to a decrease in the protection performance of traditional protection devices. Therefore, it is particularly important to study the longitudinal protection suitable for wind farm transmission lines.
[0004] When large-scale wind power is connected to a transmission network with a voltage level of 110kV or above, the traditional line longitudinal protection faces the risk of reduced sensitivity or even refusal to operate due to the unstable impedance, weak feedback characteristics, frequency deviation characteristics and fluctuating output of wind turbines compared to traditional power sources.
[0005] There are many literatures on the longitudinal protection of wind farm transmission lines. For example, the different attenuation speeds of the main components of the current on both sides of the line during internal and external faults are used to construct a longitudinal protection criterion based on the attenuation factor difference, but this method is only applicable to the case when the crowbar circuit of the doubly fed wind turbine is put into operation; the frequency difference of the current at both ends of the line after the fault is used to propose a line longitudinal protection method based on the current frequency difference, but this method is only applicable to doubly fed wind farms and not to permanent magnet direct-drive wind farms; the recursive least squares method is used to calculate the impedance value felt by the protection devices on both sides of the line, and a time-domain distance longitudinal direction protection scheme is proposed, but it requires the simultaneous collection of current and voltage values, and has higher requirements on the amount of data; the time-domain current waveform characteristics on both sides of the line are used to propose a line longitudinal protection scheme based on waveform similarity. This method is applicable to all wind farms, but faces the risk of failure when the wind farm has weak output. Summary of the invention
[0006] In view of the problems existing in the existing wind farm transmission line longitudinal protection principle, the present invention provides a large-scale wind farm transmission line longitudinal protection method based on edge detection, which is no longer affected by the frequency offset of the fault current of the wind farm, is also applicable when the wind farm has weak output, and can identify faults more quickly.
[0007] The technical solution adopted by the present invention is:
[0008] The large-scale wind farm transmission line longitudinal protection method based on edge detection constructs Hankel matrices with the current values collected on both sides of the wind farm transmission line, and performs edge detection through the Sobel operator to obtain the gradient value and edge area corresponding to each element in the current value Hankel matrix; find the current value index and corresponding gradient value corresponding to the edge area; analyze and compare the current value index corresponding to the edge area identified on both sides of the line, calculate the average gradient amplitude in combination with the gradient value, and compare it with the setting value to achieve rapid identification of in-zone and out-of-zone faults of the wind farm transmission line. If the average gradient amplitude is greater than its setting value, the wind farm transmission line has an in-zone fault.
[0009] The large-scale wind farm transmission line longitudinal protection method based on edge detection includes the following steps:
[0010] Step 1: construct the current value Hankel matrix of the current value respectively from the current values collected on both sides of the transmission line;
[0011] Step 2: Use the Sobel operator to perform edge detection on the current value Hankel matrix to obtain the gradient value amplitude matrix D and the binarization matrix E containing edge information;
[0012] Step 3: Find the current value index set Z corresponding to the binary matrix E in the current value Hankel matrix and the gradient value set d corresponding to the index set Z;
[0013] Step 4: partition the current value index set Z according to whether it is continuous, and partition the gradient value set d according to the partition length of the index set Z;
[0014] Step 5: According to the partitioning of the current value index set Z on both sides of the line, calculate the average gradient amplitude S and compare it with its set value. If it is greater than its set value, it is considered that an intra-zone fault has occurred on the transmission line.
[0015] Step 6: Build a wind farm transmission system model, simulate different types of faults occurring at different locations of the transmission lines of different types of wind farms, and identify the faults by calculating the average gradient amplitude.
[0016] In step 1, the sampling frequency of the current signal on the transmission line is 10 kHz, the data window length is 10 ms, and the collected current value is shown in formula (1):
[0017] I={i1,i2,i3,...,i N} (1)
[0018] Where, I is the current value set collected; 1,i2,…,i N is the current value collected; N is the total number of current values collected.
[0019] In step 1, the Hankel matrix refers to a matrix in which the elements on each inverse diagonal are equal. The current value Hankel matrix I h The composition is shown in formula (2):
[0020]
[0021] Among them, I h is the Hankel matrix of current values; N is the total number of current values collected in a data window; i1,i2…,i N-1 ,i N is the collected current value.
[0022] In step 2, the Sobel operator edge detection is an edge detection method based on first-order differentials. First, the elements of the upper, lower, left, and right neighborhoods in the current value Hankel matrix are averaged or weighted averaged, and then first-order differential processing is performed to detect edge points. The basic steps are as follows:
[0023] 1) Calculate the horizontal and vertical gradients.
[0024] Use the Sobel operator in the horizontal and vertical directions and the image f(i, j) to be detected for edge detection (in this paper, the current value Hankel matrix I h ) can be convolved to obtain the horizontal gradient value matrix G x The gradient matrix G in the vertical direction y , and their calculation expressions are shown in formula (3) and formula (4) respectively.
[0025]
[0026]
[0027] Among them, G x is the gradient value matrix in the horizontal direction; f(i,j) represents the detected image (in this method, it is the current value Hankel matrix); G y is the gradient value matrix in the vertical direction.
[0028] 2) Calculate the gradient modulus matrix and "normalize"
[0029] The gradient modulus matrix G corresponding to the current value Hankel matrix is calculated and "standardized" to obtain the standardized gradient modulus matrix M. The calculation formula is shown in formula (5) and formula (6):
[0030]
[0031] M=G / max(G) (6)
[0032] Where G is the gradient modulus matrix corresponding to the current value Hankel matrix; G x is the gradient value matrix in the horizontal direction; G y is the gradient value matrix in the vertical direction; M is the gradient modulus value matrix after "standardization"; max(G) represents the maximum value in the matrix G.
[0033] Since the gradient modulus at this time is obtained by summing the squares of the horizontal gradient and the vertical gradient and taking the square root, it can no longer represent the positive or negative direction of the gradient. Considering the need to calculate the average gradient amplitude later, the gradient amplitude matrix D(i, j) with direction information is determined by equation (7).
[0034] D(i,j)=G x +G y (7)
[0035] Where D(i,j) is the gradient magnitude matrix; G x is the gradient value matrix in the horizontal direction; G y is the gradient value matrix in the vertical direction.
[0036] 3) Binarization to find edges
[0037] The standardized gradient value matrix M is binarized by the threshold value K to obtain the binarized matrix E, as shown in formula (8):
[0038]
[0039] Among them, E(i,j) is the binarization matrix; M(i,j) is the gradient modulus matrix after "standardization"; K is the threshold value, which is 0.5.
[0040] The part of the binary matrix E that is 1 represents the edge position of the image, corresponding to the part of the current value collected where the current value changes rapidly; the part of E that is 0 represents the non-edge position.
[0041] In step 2, the binary matrix refers to a matrix in which elements have only two values, 0 or 1.
[0042] In step 2, the binarization matrix E is obtained by binarizing the gradient value matrix M calculated during edge detection through a threshold value K, as shown in formula (8). The part of the binarization matrix E that is 1 represents the position of the edge, and the part that is 0 represents the non-edge position.
[0043] In step 3, the current value index refers to the arrangement sequence number corresponding to a single current value in its data window.
[0044] In step 3, the current value index set Z is composed of non-repeated current value indexes corresponding to the edge positions in the binary matrix E in the current value Hankel matrix, as shown in formula (9):
[0045] Z=unique(i-1+j),E(i,j)=1 (9)
[0046] Where Z represents the current value index set, unique represents taking non-repeated elements in the array, and E(i,j) is the element in the i-th row and j-th column of the matrix E.
[0047] In step 3, the gradient value set d is a set of gradient values of the sampled current value corresponding to each element in the current value index set Z that is closest to the main diagonal in the current value Hankel matrix, as shown in formula (10):
[0048] d(i)=D(round(Z(i) / 2),round((Z(i)+1) / 2))i=1,2,…,n Z (10)
[0049] Where d represents the gradient value set, D represents the gradient amplitude matrix corresponding to the current value Hankel matrix, round represents rounding, Z represents the current value index set, and n Z Indicates the length of the current value index set.
[0050] In step 4, the current value index set Z and the gradient value set d are partitioned according to whether the elements in the current value index set Z are continuous. When there are discontinuous elements, they are respectively shown in equations (11) and (12):
[0051]
[0052]
[0053] Wherein, z1 represents the first partition of the current value index set Z, z2 represents the second partition of the current value index set Z, Z represents the current value index set, n Z represents the length of the current value index set Z, d1 represents the first partition of the gradient value set d, d2 represents the second partition of the gradient set d, and d represents the gradient value set.
[0054] When the elements in the current value index set Z are continuous, its partitioning method and the gradient value set d partitioning method are shown in equations (13) and (14), respectively:
[0055]
[0056]
[0057] Wherein, z1 represents the first partition of the current value index set Z, z2 represents the second partition of the current value index set Z, and Z represents the current value index set. represents the empty set, d1 represents the first partition of the gradient value set d, d2 represents the second partition of the gradient set d, and d represents the gradient value set.
[0058] In step 5, according to the different partitioning conditions of the current value index set Z identified by edge detection using the Sobel operator on the current values on both sides of the lines M and N, the calculation method of the average gradient amplitude S is shown in formula (15):
[0059]
[0060] Among them, S is the average gradient amplitude, d m1 and d m2 are the gradient values corresponding to the Z1 and Z2 partitions identified on the M side of the line, d n1 and d n2 are the gradient values corresponding to the Z1 and Z2 partitions identified on the N side of the line, mean means finding the average value of the array, max means finding the maximum value in the array, and z m1 ∩z n1 Represents array z m1 With array z n1 The intersection of .
[0061] In step 5, the set value is also called the set value, which means that in the automatic control system, when a certain physical quantity reaches a certain value, a certain action will occur. In this method, if the average gradient amplitude is greater than its set value, it means that an internal fault has occurred in the line. Considering that there may be transmission errors in the current information on both sides of the line and the influence of noise on the average gradient amplitude, the set value of the average gradient amplitude in this method is 0.2.
[0062] In step 5, the in-zone fault refers to a fault occurring within the protection range of the protection device. The protection proposed in this method is line longitudinal protection, and the protection range is the entire length of the line.
[0063] In step 5, when a fault occurs in the transmission line, the judgment method is as shown in formula (16):
[0064] S>Sset (16)
[0065] Among them, S is the average gradient amplitude; S set The average gradient amplitude setting value is 0.2.
[0066] In step 6, the wind farm transmission system model consists of a wind turbine, a collection system, a wind farm main step-up transformer, a transmission line and a system equivalent power supply. In the model, the types of wind farms are double-fed wind farms and permanent magnet direct-drive wind turbine wind farms, and the rated power of both types of wind farms is 400MW. The voltage level of the wind farm transmission line is 220kV, the length is 40km, the positive sequence impedance and zero sequence impedance per unit length are (0.076+j0.338)Ω / km and (0.284+j0.824)Ω / km, respectively, and the positive sequence capacitance and zero sequence capacitance per unit length are 0.0086μF / km and 0.0061μF / km, respectively. The capacity of the main transformer of the wind farm is 450MVA, YNd connection, and the voltage level is 220kV / 35kV.
[0067] In step 6, faults are set at the wind farm side exit of the wind farm transmission system, at the locations where the transmission line is 10km, 20km, and 30km away from the wind farm, and at the external exit of the system side, which are respectively denoted as K1, K2, K3, K4, and K5.
[0068] In step 6, the fault types are phase A ground short circuit, AB two-phase short circuit, AB two-phase ground short circuit, and three-phase short circuit, which are respectively recorded as AG, AB, ABG, and ABC.
[0069] The present invention provides a large-scale wind farm transmission line longitudinal protection method based on edge detection, and the technical effects are as follows:
[0070] 1) The present invention constructs the current values collected on both sides of the line into Hankel matrix forms and performs edge detection, and then calculates the average gradient amplitude S. By comparing it with its set value, the in-area fault of the wind farm transmission line can be identified.
[0071] 2) The present invention builds a large-scale wind farm transmission system model. The method of the present invention can accurately identify the occurrence of faults at different locations of the transmission lines of different types of wind farms when single-phase grounding faults, phase-to-phase short circuit faults, two-phase grounding faults and three-phase short circuit faults occur, and the action time is less than 1ms.
[0072] 3) The method of the present invention adopts time domain quantity, will not be affected by the frequency offset of the wind farm, and can still be applied when the wind farm has weak output. BRIEF DESCRIPTION OF THE DRAWINGS
[0073] Figure 1 The figure is a flow chart of the method of the present invention.
[0074] Figure 2 Transmission system model diagram for large-scale wind farms.
[0075] Figure 3(a) is a dynamic performance diagram of the protection algorithm when a fault occurs in the transmission line of a doubly-fed wind farm;
[0076] Figure 3(b) is a dynamic performance diagram of the protection algorithm when a fault occurs in the transmission line of a permanent magnet direct-drive wind farm. DETAILED DESCRIPTION
[0077] like Figure 1 As shown, the large-scale wind farm transmission line longitudinal protection method based on edge detection includes the following steps:
[0078] Step 1: construct the current value Hankel matrix of the current value respectively from the current values collected on both sides of the transmission line;
[0079] Step 2: Use the Sobel operator to perform edge detection on the current value Hankel matrix to obtain the gradient value amplitude matrix D and the binarization matrix E containing edge information;
[0080] Step 3: Find the current value index set Z corresponding to the binary matrix E in the current value Hankel matrix and the gradient value set d corresponding to the index set Z;
[0081] Step 4: partition the current value index set Z according to whether it is continuous, and partition the gradient value set d according to the partition length of the index set Z;
[0082] Step 5: According to the partitioning of the current value index set Z on both sides of the line, calculate the average gradient amplitude S and compare it with its set value. If it is greater than its set value, it is considered that an intra-zone fault has occurred on the transmission line.
[0083] Step 6: Build a wind farm transmission system model, simulate different types of faults occurring at different locations of the transmission lines of different types of wind farms, and identify the faults by calculating the average gradient amplitude.
[0084] Below in conjunction with the accompanying drawings, the preferred embodiment is described in detail:
[0085] 1. The current signal of the transmission line is collected, the sampling frequency is 10kHz, the data window length is 10ms, and the collected current value is shown in formula (1):
[0086] I={i1,i2,…,i N} (1)
[0087] Where, I is the current value set collected; 1,i2,…,i Nis the current value collected; N is the total number of current values collected.
[0088] 2. The collected current value is constructed into a Hankel matrix form. The current value Hankel matrix is shown in formula (2):
[0089]
[0090] Among them, I h is the Hankel matrix of current values; i1,i2,…,i N is the current value collected; N is the total number of current values collected.
[0091] 3. Use the Sobel operator to perform edge detection on the current value Hankel matrix. The steps are as follows:
[0092] 1) Calculate the horizontal and vertical gradients.
[0093] The Sobel operator in the horizontal and vertical directions is used to detect the image f(i, j) to be edge detected, which is the current value Hankel matrix I in the present invention. h Convolution can obtain the horizontal gradient value matrix G x The gradient matrix G in the vertical direction y , and their calculation expressions are shown in formula (3) and formula (4) respectively.
[0094]
[0095]
[0096] Among them, G x is the gradient value matrix in the horizontal direction; f(i,j) represents the detected image, which is the current value Hankel matrix in the present invention; G y is the gradient value matrix in the vertical direction.
[0097] 2) Calculate the gradient magnitude matrix and "normalize" it:
[0098] The gradient modulus matrix G corresponding to the current value Hankel matrix is calculated and "standardized" to obtain the standardized gradient modulus matrix M. The calculation formula is shown in formula (5) and formula (6):
[0099]
[0100] M=G / max(G) (6)
[0101] Where G is the gradient modulus matrix corresponding to the current value Hankel matrix; G x is the gradient value matrix in the horizontal direction; G yis the gradient value matrix in the vertical direction; M is the gradient modulus value matrix after "standardization"; max(G) represents the maximum value in the matrix G.
[0102] Since the gradient modulus at this time is obtained by summing the squares of the horizontal gradient and the vertical gradient and taking the square root, it can no longer represent the positive or negative direction of the gradient. Considering the need to calculate the average gradient amplitude later, the gradient amplitude matrix with directional information is determined by equation (7).
[0103] D(i,j)=G x +G y (7)
[0104] Where D(i,j) is the gradient magnitude matrix; G x is the gradient value matrix in the horizontal direction; G y is the gradient value matrix in the vertical direction.
[0105] 3) Binarization to find edges:
[0106] The standardized gradient value matrix M is binarized by the threshold value K to obtain the binarized matrix E, as shown in formula (8):
[0107]
[0108] Among them, E(i,j) is the binarization matrix; M(i,j) is the standardized gradient value matrix; K is the threshold value, which is 0.5.
[0109] The part of the binary matrix E that is 1 represents the edge position of the image, corresponding to the part of the current value collected where the current value changes rapidly; the part of E that is 0 represents the non-edge position.
[0110] 4. Find the edge points in the binary matrix E, that is, the points with a value of 1. The corresponding non-repeated current value index set Z is shown in formula (9):
[0111] Z=unique(i-1+j),E(i,j)=1 (9)
[0112] Where Z represents the current value index set, unique means taking non-repeated elements in the array, and E(i,j) is the element in the i-th row and j-th column of the matrix E.
[0113] 5. Find the gradient value set d corresponding to the index set Z, as shown in formula (10):
[0114] d(i)=D(round(Z(i) / 2),round((Z(i)+1) / 2)) i=1,2,…,n Z (10)
[0115] Where d represents the gradient value set, D represents the gradient amplitude matrix corresponding to the current value Hankel matrix, round represents rounding, Z represents the current value index set, and n Z Indicates the length of the current value index set.
[0116] 6. Partition the current value index set Z and the gradient value set d. When there are discontinuous elements in the current value index set Z, the calculation formulas are shown in formulas (11) and (12) respectively:
[0117]
[0118]
[0119] Wherein, z1 represents the first partition of the current value index set Z; z2 represents the second partition of the current value index set Z; Z represents the current value index set; n Z represents the length of the current value index set Z; d1 represents the first partition of the gradient value set d; d2 represents the second partition of the gradient set d; d represents the gradient value set.
[0120] If the elements in the current value index set Z are continuous, its partitioning method and the gradient value set d partitioning method are shown in equations (13) and (14) respectively:
[0121]
[0122]
[0123] Wherein, z1 represents the first partition of the current value index set Z, z2 represents the second partition of the current value index set Z, and Z represents the current value index set. represents the empty set, d1 represents the first partition of the gradient value set d, d2 represents the second partition of the gradient set d, and d represents the gradient value set.
[0124] 7. According to the different partitioning conditions of the current value index set Z identified by edge detection using the Sobel operator on the current values on both sides of the lines M and N, the calculation method of the average gradient amplitude S is shown in formula (15):
[0125]
[0126] Among them, S is the average gradient amplitude, d m1 and d m2 are the gradient values corresponding to the Z1 and Z2 partitions identified on the M side of the line; d n1 and d n2 are the gradient values corresponding to the Z1 and Z2 partitions identified on the N side of the line; |d m1| means take d m1 The absolute value of the array; mean means to find the average value of the array, max means to find the maximum value in the array, z m1 ∩z n1 Represents array z m1 With array z n1 The intersection of .
[0127] 8. Compare the calculated average gradient amplitude S with its set value. If it is greater than the set value, it is considered that an in-zone fault has occurred.
[0128] The technical effects of the present invention will be further illustrated by examples below:
[0129] Build a wind farm transmission system model and conduct simulation analysis. The model diagram is as follows: Figure 2 As shown. The wind farm transmission system model consists of wind turbines, collection systems, wind farm main step-up transformers, transmission lines and system equivalent power supplies. In the model, the types of wind farms are double-fed wind farms and permanent magnet direct-drive wind turbine wind farms, and the rated power of both types of wind farms is 400MW. The voltage level of the wind farm transmission line is 220kV, the length is 40km, the positive sequence impedance and zero sequence impedance per unit length are (0.076+j0.338)Ω / km and (0.284+j0.824)Ω / km, respectively, and the positive sequence capacitance and zero sequence capacitance per unit length are 0.0086μF / km and 0.0061μF / km, respectively. The capacity of the wind farm main transformer is 450MVA, YNd connection, and the voltage level is 220kV / 35kV.
[0130] Faults are set at the wind farm side exit of the wind farm transmission system, at the 10km, 20km, and 30km distance of the transmission line from the wind farm, and at the external exit of the system side, which are respectively recorded as K1, K2, K3, K4, and K5. The fault types simulated are A-phase ground short circuit, AB two-phase short circuit, AB two-phase ground short circuit, and three-phase short circuit, which are respectively recorded as AG, AB, ABG, and ABC.
[0131] At the five fault points K1, K2, K3, K4 and K5, four fault types, namely, A-phase ground short circuit, AB two-phase phase short circuit, AB two-phase ground short circuit and three-phase short circuit, are simulated to verify the effectiveness of the method of the present invention. The data window is 10ms after the fault occurs, and the sampling frequency is 10kHz. The average gradient amplitude S of different wind farms at different fault locations and fault types is shown in Table 1.
[0132] Table 1 Protection action performance under different fault scenarios
[0133]
[0134]
[0135] It can be seen from Table 1 that under various fault conditions, whether it is a doubly fed wind farm or a permanent magnet direct drive wind farm, the average gradient amplitude of the fault phase when an in-zone fault occurs is greater than the set value of 0.2; the average gradient amplitude of the non-fault phase when an in-zone fault occurs and the average gradient amplitude of an out-of-zone fault are less than 0.2. The protection algorithm based on edge detection can reliably and accurately identify the faulty line and has good action performance.
[0136] Figure 3(a) and Figure 3(b) are the dynamic performance diagrams of the method of the present invention when a phase A ground fault occurs at point K4 of the wind farm transmission line at t=0s. As can be seen from Figure 3(a) and Figure 3(b), for different wind farms, the proposed protection algorithm can exceed the set value of 0.2 within 1ms, so that the line longitudinal protection can act quickly and reliably, and because the time domain current value is used, it is no longer affected by the frequency offset of the short-circuit current of the wind farm. The simulation results show that the large-scale wind farm transmission line longitudinal protection method based on edge detection of the present invention can accurately identify the faults occurring on the transmission line, thereby verifying the feasibility of the proposed method.
Claims
1. A large-scale wind farm transmission line longitudinal protection method based on edge detection, characterized in that: The current values collected on both sides of the wind farm transmission line are used to construct Hankel matrices respectively, and edge detection is performed using the Sobel operator to obtain the gradient value and edge area corresponding to each element in the current value Hankel matrix; the current value index and the corresponding gradient value corresponding to the edge area are found; the current value indexes corresponding to the edge areas identified on both sides of the line are analyzed and compared, the average gradient amplitude is calculated in combination with the gradient value, and compared with the set value, to achieve rapid identification of in-zone and out-of-zone faults of the wind farm transmission line.
2. A large-scale wind farm transmission line longitudinal protection method based on edge detection, characterized in that The following steps are involved: Step 1: construct the current value Hankel matrix of the current value respectively from the current values collected on both sides of the transmission line; Step 2: Use the Sobel operator to perform edge detection on the current value Hankel matrix to obtain the gradient value amplitude matrix D and the binarization matrix E containing edge information; Step 3: Find the current value index set Z corresponding to the binary matrix E in the current value Hankel matrix and the gradient value set d corresponding to the index set Z; Step 4: partition the current value index set Z according to whether it is continuous, and partition the gradient value set d according to the partition length of the index set Z; Step 5: According to the partitioning of the current value index set Z on both sides of the line, calculate the average gradient amplitude S and compare it with its set value. If it is greater than its set value, it is considered that an intra-zone fault has occurred on the transmission line.
3. The large-scale wind farm transmission line longitudinal protection method based on edge detection according to claim 2 is characterized by: In step 1, the current value Hankel matrix I h The composition is shown in formula (2): Among them, I h is the Hankel matrix of current values; N is the total number of current values collected in a data window; i1,i2…,i N-1 ,i N is the collected current value.
4. The large-scale wind farm transmission line longitudinal protection method based on edge detection according to claim 2 is characterized by: In step 2, the Sobel operator edge detection is an edge detection method based on first-order differentials. The elements of the upper, lower, left, and right neighborhoods in the current value Hankel matrix are first averaged or weighted averaged, and then first-order differential processing is performed to detect edge points.
5. The large-scale wind farm transmission line longitudinal protection method based on edge detection according to claim 2 is characterized by: In step 2, the binarization matrix E is obtained by binarizing the gradient value matrix M calculated during edge detection through a threshold value K, as shown in formula (8); the part of the binarization matrix E that is 1 represents the position of the edge, and the part that is 0 represents the non-edge position; Among them, E is the binarization matrix, M is the gradient modulus matrix calculated and normalized by the Sobel operator, K is the threshold value, E(i,j) is the element in the i-th row and j-th column of the matrix E, and M(i,j) is the element in the i-th row and j-th column of the matrix M.
6. The large-scale wind farm transmission line longitudinal protection method based on edge detection according to claim 2 is characterized by: In step 3, the current value index set Z is composed of non-repeated current value indexes corresponding to the edge positions in the binary matrix E in the current value Hankel matrix, as shown in formula (9): Z=unique(i-1+j),E(i,j)=1 (9) Where Z represents the current value index set, unique means taking non-repeated elements in the array, and E(i,j) is the element in the i-th row and j-th column of the matrix E.
7. The large-scale wind farm transmission line longitudinal protection method based on edge detection according to claim 2 is characterized by: In step 3, the gradient value set d is a set of gradient values of the sampled current value corresponding to each element in the current value index set Z that is closest to the main diagonal in the current value Hankel matrix, as shown in formula (10): d(i)=D(round(Z(i) / 2),round((Z(i)+1) / 2))i=1,2,…,n Z (10) Where d represents the gradient value set, D represents the gradient amplitude matrix corresponding to the current value Hankel matrix, round represents rounding, Z represents the current value index set, and n Z Indicates the length of the current value index set.
8. The large-scale wind farm transmission line longitudinal protection method based on edge detection according to claim 2 is characterized by: In step 4, the current value index set Z and the gradient value set d are partitioned according to whether the elements in the current value index set Z are continuous. When there are discontinuous elements, they are respectively shown in equations (11) and (12): Wherein, z1 represents the first partition of the current value index set Z, z2 represents the second partition of the current value index set Z, Z represents the current value index set, n Z represents the length of the current value index set Z, d1 represents the first partition of the gradient value set d, d2 represents the second partition of the gradient set d, and d represents the gradient value set; When the elements in the current value index set Z are continuous, its partitioning method and the gradient value set d partitioning method are shown in equations (13) and (14), respectively: Wherein, z1 represents the first partition of the current value index set Z, z2 represents the second partition of the current value index set Z, and Z represents the current value index set. represents the empty set, d1 represents the first partition of the gradient value set d, d2 represents the second partition of the gradient set d, and d represents the gradient value set.
9. The large-scale wind farm transmission line longitudinal protection method based on edge detection according to claim 2 is characterized by: In step 5, according to the different partitioning conditions of the current value index set Z identified by edge detection using the Sobel operator on the current values on both sides of the lines M and N, the calculation method of the average gradient amplitude S is shown in formula (15): Among them, S is the average gradient amplitude, d m1 and d m2 are the gradient values corresponding to the Z1 and Z2 partitions identified on the M side of the line, d n1 and d n2 are the gradient values corresponding to the Z1 and Z2 partitions identified on the N side of the line, mean means finding the average value of the array, max means finding the maximum value in the array, and z m1 ∩z n1 Represents array z m1 With array z n1 The intersection of .
10. The large-scale wind farm transmission line longitudinal protection method based on edge detection according to claim 1 is characterized by: In step 5, when a fault occurs in the transmission line, the judgment method is as shown in formula (16): S>S set (16) Among them, S is the average gradient amplitude, S set Average gradient amplitude setting value.
