Short-circuit fault location method for looped dc microgrid based on line model and euclidean distance
Patent Information
- Application Number
- CN202310664486.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-06-05
- Publication Date
- 2026-09-25
- Estimated Expiration
- 2043-06-05
AI Technical Summary
也有提出利用数学形态法提取电流信号特征检测故障，该方法可应用于高电阻短路故障检测和保护，但未进行故障定位
[0044]综上所述，本发明具有无需额外的信号注入，适用于高电阻故障情况下故障定位等优点。
Smart Images
Figure CN116718872B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of microgrid fault location technology, and in particular to a method for locating short-circuit faults in a ring-shaped DC microgrid based on a line model and Euclidean distance. Background Technology
[0002] In recent years, with the continuous depletion of fossil fuels and the increasing electricity load, people have placed higher and higher demands on the reliability and power quality of power grids. Research has shown that ring-shaped DC microgrids, as a common structure, have higher efficiency than ordinary radial DC microgrids, especially when the distribution lines are short. When a distributed power source or DC bus fails, the entire ring system can still maintain normal operation by quickly isolating the fault area. Therefore, ring-shaped DC microgrids are commonly used in densely populated areas such as large buildings, factories, and data centers, and also have practical value in power distribution systems of ships and spacecraft. When a short-circuit fault occurs in a DC line, due to the large capacitance of the bus and the low line impedance, the fault current will reach its peak within milliseconds. To prevent further damage to system equipment, quickly disconnecting the fault and accurately locating it after it occurs is of great significance for ensuring the normal operation of the microgrid.
[0003] Currently, domestic and international scholars have mainly focused their research on short-circuit fault location methods for DC microgrids from the following three aspects:
[0004] The first type is the active fault location method. The active fault location method is based on auxiliary equipment injecting signals into the faulty line and analyzing the signal characteristics to locate the fault. This method requires additional equipment to be added to the line, which increases the investment cost.
[0005] The second category is passive fault detection and location methods, which are based on the acquisition and analysis of electrical quantity measurements within the system itself. The traveling wave method is a typical passive fault location method, widely used in the fault protection of flexible DC transmission systems. However, this approach requires multiple terminal devices and is not suitable for short-line fault protection. Some researchers have chosen to use line current or voltage signals for analysis and calculation to achieve fault protection, but this method is not compatible with high-resistance grounding faults. Alternatively, the peak value of the fault current can be sampled, and the fault location and other parameters can be calculated using proposed parameter estimation methods. This method requires a high sampling rate to ensure location accuracy.
[0006] With the continuous development of artificial intelligence, third-category methods such as neural networks and image processing techniques have been applied to the electrical field. Some scholars have proposed a method combining multi-criteria systems with neural networks for fault location, which improves fault protection speed compared to traditional differential protection methods; however, this location method has not been verified for high-resistance grounding conditions. Alternatively, a microgrid fault intelligent detection scheme has been proposed combining wavelet transform with deep neural networks; in this scheme, the choice of wavelet family directly affects the accuracy of detection and location. Others have proposed using mathematical morphology to extract current signal features for fault detection; this method can be applied to high-resistance short-circuit fault detection and protection, but fault location has not been performed. Summary of the Invention
[0007] In view of the shortcomings of the prior art, the technical problem to be solved by the present invention is: how to provide a short-circuit fault location method for a ring DC microgrid that does not require additional signal injection and is suitable for fault location under high resistance fault conditions.
[0008] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:
[0009] A method for locating short-circuit faults in a ring-shaped DC microgrid based on a line model and Euclidean distance, characterized by comprising the following steps:
[0010] S1. Establish a system mathematical model for the ring-shaped DC microgrid to be fault located;
[0011] S2. Obtain line current: Obtain the sampled current of the line through intelligent electronic devices or current sensors;
[0012] S3. Fault scheme initialization: Obtain a large number of combination schemes of fault resistors and fault locations through random generation;
[0013] S4. Calculate transient fault current: After the fault scheme is generated, the transient fault current is calculated using the established system mathematical model.
[0014] S5. Calculation of Euclidean distance and correlation coefficient: The correlation between the currents is obtained by calculating the Euclidean distance between the transient fault current and the sampled current.
[0015] S6. Determine if the relevant threshold is met: Set the determination threshold to k1. When the calculated correlation coefficient is greater than k1, determine the fault scheme and output the fault location and transition resistance of the fault scheme; otherwise, use the genetic algorithm to optimize the scheme, find the optimal solution, and repeat steps S4 to S6.
[0016] Furthermore, in step S1, for a ring-shaped DC microgrid with n nodes directly connected to the converter and b transmission lines, its DC system node voltage matrix U nfor:
[0017]
[0018] Converter output current matrix I c for:
[0019]
[0020] The line current matrix I0 is:
[0021]
[0022] The equations for current variation and node voltage variation are as follows:
[0023]
[0024]
[0025] In the formula, L0 is the diagonal matrix of DC line inductance with dimension b×b; R0 is the diagonal matrix of line resistance with dimension b×b; C0 is the diagonal matrix of bus capacitance with dimension n×n; A0 is the node-branch correlation matrix with dimension n×b, and the elements of A0 are defined as follows:
[0026]
[0027] When a node is associated with a branch, if the branch direction leaves the node, then a nb =1, if the branch direction points to the node, then a nb =-1. a = -1 when the node is independent of the branch. nb =0.
[0028] Furthermore, in the event of a fault in the ring-shaped DC microgrid, both the number of nodes n and the number of transmission lines b are incremented by 1; in step S4, the transient fault current can be calculated using the following formula:
[0029]
[0030] Where: the node voltage matrix U under fault conditions n Line current matrix I0, converter output current matrix I c They are respectively:
[0031] U n =[u n1 u n2 u n3 ...u nn u nf ] T
[0032] I0 = [i12 i 23 i 34 ...i bf i f1 ] T
[0033] I c =[i c1 i c2 i c3 ...i cn i f ] T
[0034] L is the diagonal matrix of DC line inductance, with dimensions (b+1)×(b+1); R is the diagonal matrix of line resistance, with dimensions (b+1)×(b+1); C is the diagonal matrix of bus capacitance, with dimensions (n+1)×(n+1); These are represented as follows:
[0035]
[0036]
[0037]
[0038] Where: i f =i f1 -i nf u nf =r f ·(-i f )=E·r f ·i0, E is the coefficient vector, E=[0 0……0 1 -1], r f This is the short-circuit transition resistor.
[0039] Furthermore, in step S5, when the time dimension is fixed, the current sequence in the current space contains k sample points, and the current correlation coefficient based on Euclidean distance is:
[0040]
[0041] In the formula: and They are the sampled current sample sequences I. sam and transient fault current sample sequence I cal Standardize each component and sample the current sample sequence I. sam and transient fault current sample sequence I cal The spatial expressions are as follows:
[0042] I sam =((i sam1 ,t sam1),(i sam2 ,t sam2 ),···,(i samk ,t samk ))
[0043] I cal =((i cal1 ,t cal1 ),(i cal2 ,t cal2 ),···,(i calk ,t calk )).
[0044] In summary, the present invention has advantages such as requiring no additional signal injection and being suitable for fault location in high-resistance fault conditions. Attached Figure Description
[0045] Figure 1 This is a diagram of a four-port ring DC microgrid structure.
[0046] Figure 2 A flowchart for fault location.
[0047] Figure 3 This is a window current correction diagram.
[0048] Figure 4 This is a simulation diagram of a four-port ring DC microgrid.
[0049] Figure 5 This is a schematic diagram of the DC bus voltage.
[0050] Figure 6 This is a schematic diagram of the current in line 1-2 under different fault locations with the same transition resistance.
[0051] Figure 7 This is a schematic diagram of the current in line 1-2 when the transition resistance is different at the same fault location. Detailed Implementation
[0052] The invention will now be described in further detail through modeling and simulation.
[0053] 1. Modeling of a ring-shaped DC microgrid
[0054] Suppose a ring-shaped DC microgrid has n nodes (directly connected to the converter) and b transmission lines (DC lines connecting nodes to each other—branches). Then the DC system node voltage matrix U n It can be represented as:
[0055]
[0056] Converter output current matrix I c It can be represented as:
[0057]
[0058] The line current matrix I0 can be expressed as:
[0059]
[0060] by Figure 1 Taking a four-port loop DC microgrid as an example, we analyze the mathematical relationships between various electrical quantities in a DC system. The system has four nodes directly connected to the converter, and the loop has four DC transmission lines. When the system is operating normally, the DC system node voltage matrix U... n It can be represented as:
[0061] U n =[u n1 u n2 u n3 u n4 ] T (4)
[0062] Converter output current matrix I c It can be represented as:
[0063] I c =[i c1 i c2 i c3 i c4 ] T (5)
[0064] Assuming the clockwise direction of the loop is the positive reference direction for the current, the line current matrix I0 can be expressed as:
[0065] I0 = [i 12 i 23 i 34 i 41 ] T (6)
[0066] In existing research, there are two main types of converter output current models for DC microgrids during short-circuit faults. One type ignores the converter's input current, while the other treats the converter as a constant current source. Therefore, the distributed power source and the converter as a whole can be considered as a constant current source input.
[0067] When the system is running normally, taking line 1-2 between node 1 and node 2 as an example, the line equations are written according to Kirchhoff's voltage law:
[0068]
[0069] Among them, l 12 ,r 12 These represent the line inductance and line resistance of line 1, respectively.
[0070] Extending this to every transmission line, the current variation equation can be expressed in matrix form as follows:
[0071]
[0072] Where L0 is the DC line inductance diagonal matrix with dimension b×b. R0 is the line resistance diagonal matrix with dimension b×b. A0 is the node-branch correlation matrix with dimension n×b, and the elements of A0 are defined as follows:
[0073]
[0074] When a node is associated with a branch, if the branch direction leaves the node, then a nb =1, if the branch direction points to the node, then a nb =-1. a = -1 when the node is independent of the branch. nb =0.
[0075] Similarly, according to Kirchhoff's current law, the equation for the node voltage change matrix is as follows:
[0076]
[0077] Where C0 is the diagonal matrix of bus capacitors, and its dimension is n×n.
[0078] by Figure 1 Taking the network shown as an example, if a short-circuit fault occurs on line 4-1, the original four-port network is equivalent to adding a faulty node between nodes 1 and 4. At this time, the number of nodes and the number of branches both increase by 1. The system node voltage matrix, line current matrix, and converter output current matrix are then as follows:
[0079] U n =[u n1 u n2 u n3 u n4 u nf ] T (11)
[0080] I0 = [i 12 i 23 i 34 i 4f i f1 ] T (12)
[0081] I c =[i c1 i c2 i c3 i c4 i f ] T(13)
[0082] In equation (13),
[0083] i f =i f1 -i 4f (14)
[0084] u nf =r f ·(-i f )=E·r f ·I0 (15)
[0085] Where E is the coefficient vector, E = [0 0 0 1 -1], r f This is the short-circuit transition resistor.
[0086] Since the original DC line can be considered as being cut off into two lines by the fault point when a fault occurs, the inductance and resistance matrices in the line current change equation (8) are respectively corrected as follows:
[0087]
[0088]
[0089] The capacitance matrix in the node voltage change equation (10) is corrected as follows:
[0090]
[0091] The simultaneous voltage and current differential equations are as follows:
[0092]
[0093] The dimension of A0 is corrected to (n+1)×(b+1), and the fault current when a short circuit fault occurs in a DC microgrid line can be quickly calculated through equation (19).
[0094] 2. Fault Location Scheme
[0095] Euclidean distance, as one of the important analytical methods in discriminant analysis, is usually used to describe the true distance between two points in N-dimensional space. Its purpose is to analyze and calculate the distance between samples to characterize the similarity between two samples. The smaller the distance, the more related the samples are; the larger the distance, the less related the samples are.
[0096] Define two sets of sample sequences X and Y, whose spatial representations are as follows:
[0097] X = (x1, x2, ..., x n (20)
[0098] Y = (y1, y2, ..., yn ) (twenty one)
[0099] The Euclidean distance between the two sets of sequences can be defined as:
[0100]
[0101] Since the distribution of components varies significantly across dimensions, to eliminate the influence of units, it is necessary to first standardize each component. In this embodiment, the component mean is chosen for standardization, and the formula is as follows:
[0102]
[0103] In the formula: The mean of the X sequence samples; The mean of the Y sequence samples.
[0104] If the Euclidean distance of the samples remains close to 0 throughout the entire space, it can be determined that the selected samples have a high degree of similarity and correlation, and it can also be said that the sample curves have the same trend. In this embodiment, Euclidean distance is used to represent the similarity and amplitude difference of the fault current. Since the Euclidean distance between the sampled current and the transient fault current is different in actual conditions, in order to better represent the correlation of the fault current, this embodiment uses formula (24) to convert the Euclidean distance into a correlation coefficient.
[0105]
[0106] When analyzing current, Euclidean space can be considered as a two-dimensional space consisting of time and current; therefore, the sampled current sequence can be represented as I. sam The transient fault current sample sequence is represented as I cal The spatial expressions are as follows:
[0107] I sam =((i sam1 ,t sam1 ),(i sam2 ,t sam2 ),···,(i samk ,t samk (25)
[0108] I cal =((i cal1 ,t cal1 ),(i cal2 ,t cal2 ),···,(i calk ,t calk (26)
[0109] The current correlation coefficient based on Euclidean distance is then expressed as:
[0110]
[0111] When the time dimension is fixed, and the current sequence in the current space contains k sample points, the current correlation coefficient simplifies to:
[0112]
[0113] 3. Fault location methods
[0114] DC microgrid lines typically have intelligent electronic devices (IEDs) installed at both ends for monitoring and protection. When a short-circuit fault occurs, the system monitors and acquires current and voltage information through the IEDs. This section presents a fault location method by calculating the Euclidean distance between the sampled current and the transient fault current, combined with genetic algorithm analysis.
[0115] Fault location process as follows Figure 2 As shown, the specific implementation process of the positioning method is as follows:
[0116] 1) Obtaining line current: The sampling line current is obtained through intelligent electronic devices, current sensors, etc.
[0117] 2) Fault scheme initialization: A large number of combination schemes of fault resistors and fault locations are obtained by random generation.
[0118] 3) Calculate transient fault current: After the fault scheme is generated, the transient fault current is calculated using the established system mathematical model.
[0119] 4) Calculation of Euclidean distance and correlation coefficient: The correlation between the currents is obtained by calculating the Euclidean distance between the transient fault current and the sampled current.
[0120] 5) Determining if the relevant threshold is met: After a fault occurs, the closer the Euclidean distance between the sampled current and the transient fault current, the more accurate the fault diagnosis. Therefore, a threshold of k1 is set. When the calculated correlation coefficient is greater than k1, the fault diagnosis is determined. After extensive simulation verification, k1 = 0.99 was selected.
[0121] 6) Genetic algorithm iterative update: When the generated fault solution does not meet the Euclidean distance correlation coefficient threshold requirement, the genetic algorithm is used to optimize the solution and find the optimal solution.
[0122] 4. Algorithm data window current correction
[0123] When analyzing current, Euclidean space can be considered a two-dimensional space composed of time and current. The sampled current and the transient fault current constitute two sets of sample sequences, respectively. The number of samples is determined by the sampling frequency and the calculation window. A higher sampling frequency results in more samples; a larger calculation window also results in more samples. The selection of the data window plays a crucial role in accurately assessing the Euclidean distance between the sampled current and the transient fault current.
[0124] This embodiment employs a sliding window method, sliding a fixed-length actual current sampling window along the time axis to ensure alignment between the actual current sampling point and the transient fault current point. Due to the discrete nature of the sampled current and the influence of noise, the exact moment of fault occurrence is difficult to determine. However, after a fault occurs, as the current rises to its maximum value and eventually stabilizes, the sampling window will slide and sample the maximum current value. Therefore, the maximum current value can be selected as the reference time for window current correction. Figure 3 As shown.
[0125] Based on simulation data analysis, the data window in this embodiment is T = 1 × 10⁻⁶. -4 s.
[0126] When current correction is required, there are two relationships between the actual sampled current and the transient fault current:
[0127] (1) Maximum value of sampling current I s_max Greater than or equal to the maximum fault current I c_max At this point, the transient fault current is moved until the maximum value of the sampled current is located on the transient fault current.
[0128] (2) Maximum sampling current I s_max Less than the maximum value of the transient fault current I c_max At this point, the transient fault current is moved until the maximum value of the sampled current is aligned with the maximum value of the transient fault current on the time axis.
[0129] 5. Simulation Verification
[0130] To preliminarily verify the proposed fault location scheme, this embodiment constructs a four-port ring DC microgrid based on the Matlab / Simulink simulation platform to simulate a low-voltage 400V DC microgrid system. The simulation structure model is shown below. Figure 4 As shown, the system parameters are detailed in Table 1 below.
[0131] Table 1 Fault Simulation Parameters
[0132]
[0133] The simulation is set to a short-circuit fault occurring at t=1s, with an initial grounding transition resistance R. f =1Ω. According to Figure 5As shown in the bus voltage waveform, when a short-circuit fault occurs, the bus voltage drops rapidly and gradually stabilizes after a brief period of high-frequency oscillation.
[0134] Figure 6 and Figure 7 The changes in line current of line 1-2 when a ground fault occurs in line 1-4 were simulated. As the fault location d increases, the line current I... 21 The oscillation frequency gradually increases, but the current rise rate gradually decreases. When the fault location is fixed, the transition resistance directly affects the amplitude of the line current. Based on the above current variation characteristics, the proposed fault location algorithm can accurately locate the fault and determine the magnitude of the transition resistance.
[0135] The errors in fault location and fault resistance under different conditions are given in Tables 2 and 3 below.
[0136] Table 2. Error conditions for the same transition resistance at different fault locations.
[0137]
[0138]
[0139] To verify the accuracy of the fault location method, this embodiment defines the error between the fault location and the transition resistance as follows:
[0140]
[0141] Among them, L e and L a It is to detect the location of the fault and the actual location of the fault, L T R is the total length of the faulty line. e and R a These are the detected fault resistance and the actual fault resistance, R. T This is the reference fault resistor, which is set to 10Ω in this embodiment.
[0142] Table 3 Error details for different transition resistances at the same location.
[0143]
[0144] When a short-circuit fault occurs, the transient fault current in the fault location algorithm does not perfectly match the simulated current. The fault location accuracy is highest when the fault location is in the middle of the line. Its accuracy gradually decreases as the fault location moves closer to both ends of the line. As the fault transition resistance increases, the difficulty of fault location increases, and its accuracy decreases. When the transition resistance is at its maximum, R... f=10Ω, the fault detection error also reaches its maximum. Among them, the average error of fault location is 2.68%, and the average error of fault resistance is 2.77%. It can be seen that the proposed method can accurately locate the fault and determine the transition resistance.
[0145] 6. Conclusion
[0146] To address the issue of inaccurate fault location in ring-shaped DC microgrids, this embodiment establishes a system model of the ring-shaped DC microgrid for ground faults and proposes a fault location scheme based on the Euclidean distance correlation coefficient. The following conclusions are drawn through theoretical analysis and simulation verification:
[0147] 1) By establishing a model of the ring microgrid during a fault, the transient fault current can be calculated quickly and accurately.
[0148] 2) The proposed method can accurately detect the fault location and transition resistance, and is applicable in both low-resistance and high-resistance short-circuit conditions, meeting the reliability requirements of DC microgrids.
[0149] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for locating short-circuit faults in a ring-shaped DC microgrid based on a line model and Euclidean distance, characterized in that, Includes the following steps: S1. Establish a system mathematical model for the ring-shaped DC microgrid to be fault located; S2. Obtain line current: Obtain the sampled current of the line through intelligent electronic devices or current sensors; S3. Fault scheme initialization: Obtain a large number of combination schemes of fault resistors and fault locations through random generation; S4. Calculate transient fault current: After the fault scheme is generated, the transient fault current is calculated using the established system mathematical model. S5. Calculation of Euclidean distance and correlation coefficient: The correlation between the currents is obtained by analyzing the Euclidean distance between the calculated transient fault current and the sampled current. S6. Determine if the relevant threshold is met: Set the threshold as follows: When the calculated correlation coefficient is greater than If the fault is found, determine the fault scheme and output the fault location and transition resistance of the fault scheme; otherwise, use the genetic algorithm to optimize the scheme, find the optimal solution, and repeat steps S4 to S6. In step S5, when the time dimension is fixed, the current sequence in the current space respectively includes For each sample point, the current correlation coefficient based on Euclidean distance is: In the formula: and These are the sampled current sample sequences. and transient fault current sample sequence Standardize each component and sample current sample sequence and transient fault current sample sequence The spatial expressions are as follows: 。 2. The short-circuit fault location method for a ring-shaped DC microgrid based on a line model and Euclidean distance as described in claim 1, characterized in that, In step S1, for those having A node directly connected to the converter and A ring-shaped DC microgrid with a transmission line, its DC system node voltage matrix for: Converter output current matrix for: Line current matrix for: The equations for current variation and node voltage variation are as follows: In the formula, is a diagonal matrix of DC line inductance with a dimension of ; is a diagonal matrix of line resistance with a dimension of ; is a diagonal matrix of bus capacitance with a dimension of ; is a node-branch incidence matrix with a dimension of , elements are defined as follows: When a node is associated with a branch, if the branch direction leaves the node, then If the branch direction points to the node, then When the node is unrelated to the branch .
3. The short-circuit fault location method for a ring-shaped DC microgrid based on a line model and Euclidean distance as described in claim 2, characterized in that, In a fault-prone ring-shaped DC microgrid, the number of nodes and the number of transmission lines Add 1 to all; in step S4, the transient fault current can be calculated using the following formula: Where: Node voltage matrix under fault conditions Line current matrix Converter output current matrix They are respectively: The inductance of a DC line is a diagonal matrix with dimensions of . ; The line resistance is a diagonal matrix with dimensions of . ; The bus capacitor is a diagonal matrix with dimension O(n). ; respectively represented as: in: , , For the coefficient vector, , This is the short-circuit transition resistor.