A method for locating power grid fault section based on SVMR transition resistance estimation

Abstract

The present invention discloses a method for locating a fault section of a power grid based on SVMR transition resistance estimation. The method simulates a single-phase grounding fault with different transition resistances for each node according to the topological structure of the distribution network, and establishes a database containing fault voltage amplitude and phase angle information; trains a support vector machine regression analyzer with the database to achieve regression estimation of the fault transition resistance; then, based on the transition resistance estimation result of the current fault, traverses and calculates the information distance of all sections of the distribution network, and the section with the shortest distance in the result is judged as the section where the fault is located. The method of the present invention only needs to install a monitoring point at the near-power node of the radial distribution network to complete the positioning of the fault section.

Description

Technical Field

[0001] This invention relates to the field of fault location, and specifically to a method for locating power grid fault sections based on SVMR transition resistance estimation. Background Art

[0002] In my country's power distribution network, the neutral grounding method is mainly ungrounded or resonant grounding (grounded via an arc suppression coil). Two-phase and three-phase faults are the primary focus of relay protection configurations in my country's power distribution network. Protection systems provide sufficient, even redundant, protection measures for two-phase and three-phase faults. For single-phase ground faults, due to their relatively small fault current, the damage to the system is also relatively small; theoretically, the system can operate with the fault for less than 2 hours. However, as the structure of our power distribution network becomes increasingly complex, with more and more connected devices and larger capacities, the risk of a single-phase ground fault evolving into a two-phase or even three-phase fault due to the prolonged maintenance of the single-phase ground fault short-circuit current increases. Therefore, accurate and rapid location of single-phase ground faults in the power distribution network is of significant practical importance for ensuring the safe operation of the power grid.

[0003] Traditional single-phase ground fault location methods mostly require the installation of fault monitoring systems at each node to ensure accurate fault location under the condition that the overall power grid is observable. However, compared to high-voltage power grids, the monitoring conditions of distribution networks are not as ideal. Furthermore, due to the poor automation level of distribution substations in underdeveloped areas of my country, achieving accurate fault location under relatively scarce monitoring conditions is even more difficult. Therefore, it is necessary to study single-phase fault section location methods with relatively lower monitoring requirements to provide richer fault information for distribution networks in underdeveloped areas, thereby improving the operation and maintenance efficiency of power grid companies. Summary of the Invention

[0004] To address or partially address the problems existing in related technologies, this invention provides a power grid fault location method based on SVMR transition resistance estimation. According to the distribution network topology, single-phase grounding faults with different transition resistances are simulated for each node, establishing a database containing fault voltage amplitude and phase angle information. A support vector machine regression analyzer is trained using this database to achieve regression estimation of the fault transition resistance. Then, based on the estimated transition resistance of the current fault, the information distance of all segments in the distribution network is calculated traversally, and the segment with the shortest distance is identified as the fault location. This method only requires installing a monitoring point near the power source node in a radial distribution network to complete the fault location. It should be noted that because the distances between the various segments of the distribution network are relatively short, the fault location can be reflected based on the segment location results. Therefore, the location result of this method is only accurate to the segment position.

[0005] This invention provides a method for locating power grid fault sections based on SVMR transition resistance estimation, comprising:

[0006] Establish a database for estimating transition resistance;

[0007] The fault resistance is estimated using the SVMR method;

[0008] The estimated fault resistance R f The fault section is identified by comparing it with the fault resistance in the transition resistance estimation database.

[0009] Optionally, establishing the transition resistance estimation database specifically includes:

[0010] A corresponding transition resistor is installed at each node of the power distribution system to simulate a single-phase ground fault; the voltage amplitude and angle of the single-phase ground fault after the short circuit are recorded from the measurement node, which are {(U1, sinθ1), (U2, sinθ2), ..., (U n sinθ n )}, where n is the node number;

[0011] In the fault simulation experiment, the transition resistance of a single-phase ground fault is selected in the range of [1, 10], and the selected traversal step size is 1 ohm.

[0012] A transition resistance estimation database for all nodes is established. The transition resistance data for each node is in the form of a 10*3 matrix, where 10 represents the transition resistance value from 1 to 10, and 3 represents the information corresponding to each fault simulation experiment (U). n sinθ n , R f ), where R f This is the transition resistance.

[0013] Optionally, the estimation of fault resistance using the SVMR method specifically includes:

[0014] Let the training data samples be:

[0015] S={(x1, y1), (x2, y2),..., (x n ,y n )} x i ∈R n ,y i ∈R n (1)

[0016] Where x i Represents the input vector (U) in the sample space i sinθ i );y i The output vector R in the sample spacef ; i is the number of samples;

[0017]

[0018] In the formula, ω is the weight vector; b is the mapping function; b is the bias.

[0019] The core problem of SVMR training is minimizing the objective function of the following expression:

[0020]

[0021] In the formula, C is the penalty parameter; ω is the weight vector in the above formula; ε is the upper limit of error, that is, when the predicted value f(x) is equal to the upper limit of error, the upper limit of error is calculated. i ) and actual value y i If the error does not exceed the given error precision ε, it indicates that the regression function fits the point.

[0022] Incorporating Lagrange multipliers and duality theory, the final mathematical description is as follows:

[0023]

[0024]

[0025] Where, α i , α′ i α j , α′ j K(x) is the Lagrange multiplier; ε is the upper limit of error in the above equation; K(x) i x j ) is the inner product kernel function;

[0026] The expression for RBF as a kernel function is:

[0027] K(x i x j )=exp(-γ||x i -x j || 2 (6)

[0028] Once SVMR training is complete, the fault information (U) measured at the monitoring point for the current fault will be used. f θ f As the input to SVMR, the corresponding output is the fault resistance transition R. f .

[0029] Optionally, the identification of faulty sections specifically includes:

[0030] The estimated fault resistance R fCompare with the fault resistance in the database; if the estimated fault resistance is between two adjacent fault resistances in the database, R f (x) and R f (x+1), as shown in the following formula:

[0031] R f (x)<R f <R f (x+1) (7)

[0032] Fault section identification is performed using voltage data; x represents the multiple of the fault transition resistance step size in the database. If the step size is 5 when the database is established, then the single-phase ground fault simulation for each node is divided into 3 times, corresponding to transition resistances of 1 ohm, 5 ohms, and 10 ohms. If x = 1, then x + 1 = 5; if the step size is 1 ohm, then each node needs to perform 10 fault simulations; if x = 1, then x + 1 = 2; then, the adjacent fault resistance R... f (x) and R f (x+1) Compare the voltage amplitude and angle information corresponding to the database with the actual voltage amplitude and angle of the fault to determine the fault section;

[0033] Taking the fault segment between nodes i and j as an example, there are no other nodes between nodes i and j, and the line between nodes i and j can be represented by the ij segment;

[0034] With a transition resistance of R f When (x), the voltage amplitude at node i is Node j is At node i, when the transition resistance is R f At (x+1), the corresponding voltage amplitude and phase angle are Node j is

[0035] The location of the fault section can be determined by information on the range of fault voltage and phase angle; the corresponding range information is as follows:

[0036]

[0037]

[0038] Optionally, after an actual short-circuit fault occurs, the fault information point (U) is calculated. f θ f The shortest distance from the line connecting the fault voltage and phase angle data of a certain section can be used to determine the section where the fault occurred.

[0039] Let segment 1 be ij, and segment 2 be mn, where m represents the starting node of the segment, and n represents the ending node of the segment. There are no other nodes between the two segments. The voltage and phase angle data for segment 1 are as follows: and The voltage and phase angle data for section 2 are as follows: and d m-n and d i-j Indicates the fault point (U) f θ f The shortest distance between data points and segments 1 and 2;

[0040] The overall algorithm process involves iterating through all distance results and then selecting the shortest distance as the segment where the fault occurs.

[0041] The technical solution provided by this invention may include the following beneficial effects:

[0042] This invention simulates single-phase grounding faults with different transition resistances at each node based on the distribution network topology, establishing a database containing fault voltage amplitude and phase angle information. A support vector machine regression analyzer is trained using this database to estimate the fault transition resistance. Then, based on the estimated transition resistance for the current fault, the information distances to all segments of the distribution network are calculated traversally. The segment with the shortest distance is identified as the faulty segment. This method only requires installing a monitoring point near the power source node in a radial distribution network to locate the faulty segment. It should be noted that because the distances between the various segments of the distribution network are relatively short, the fault location can be determined based on the segment location results. Therefore, the location result of this method is only accurate to the segment level.

[0043] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention. Attached Figure Description

[0044] To more clearly illustrate the technical solutions of the embodiments of this invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 This is a flowchart of the positioning method in an embodiment of the present invention;

[0046] Figure 2 This is a schematic diagram illustrating the relationship between fault point information in an embodiment of the present invention.

[0047] Figure 3 This is a schematic diagram of the segment positioning principle in an embodiment of the present invention;

[0048] Figure 4 This is the IEEE-33 node network topology in an embodiment of the present invention;

[0049] Figure 5 This is the result of the segment number distance calculation in this embodiment of the invention. Detailed Implementation

[0050] Embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that the invention will be more thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

[0051] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The singular forms “a,” “the,” and “the” used in this invention and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0052] It should be understood that although the terms "first," "second," "third," etc., may be used in this invention to describe various information, this information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this invention, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Thus, features defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0053] This invention provides a method for locating power grid fault sections based on SVMR transition resistance estimation, which can...

[0054] The technical solutions of the embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0055] Step A: Establish a transition resistance estimation database

[0056] Step A01: Install corresponding transition resistors at each node of the power distribution system to simulate a single-phase ground fault. Record the voltage amplitude and angle of the single-phase ground fault after the short circuit from the measurement node, which are {(U1, sinθ1), (U2, sinθ2), ..., (U... n sinθ n)}, where n is the node number.

[0057] Step A02: In the above fault simulation experiment, the transition resistance of a single-phase ground fault is selected in the range of [1, 10]. The traversal step size selected in this patent is 1 ohm.

[0058] Step A03: Establish a transition resistance estimation database for all nodes. The transition resistance data for each node is in the form of a 10*3 matrix, where 10 represents the transition resistance value from 1 to 10, and 3 represents the information corresponding to each fault simulation experiment (U). n sinθ n , R f ), where R f This represents the transition resistance. The establishment of the transition resistance database lays the data foundation for the subsequent training of machine learning algorithms.

[0059] Step B: Fault Transition Resistance Estimation

[0060] This patent employs the Support Vector Machine Regression (SVMR) method to estimate fault resistance. Let the training data sample be:

[0061]

[0062] Where x i Represents the input vector (U) in the sample space i sinθ i );y i R represents the output vector in the sample space. f ; i represents the number of samples. Based on statistical theory, the optimal regression function is constructed in the feature space:

[0063]

[0064] In the formula, ω is the weight vector; is the mapping function; b is the bias.

[0065] The core problem in SVMR training is minimizing the objective function of the following expression:

[0066]

[0067] In the formula, C is the penalty parameter; ω is the weight vector in the above formula; ε is the upper limit of error, that is, when the predicted value f(x) is equal to the upper limit of error, the upper limit of error is calculated. i ) and actual value y i If the error does not exceed the given error precision ε, it indicates that the regression function fits the point.

[0068] Incorporating Lagrange multipliers and duality theory, the final mathematical description is as follows:

[0069]

[0070]

[0071] In equations (4) and (5), α i ,a′ i α j , α′ j K(x) is the Lagrange multiplier; ε is the upper limit of error in the above equation; K(x) i x j γ is the inner product kernel function. There are four types of kernel functions. Since the radial basis function (RBF) has a wide range of applications and only has one parameter γ to be optimized, this paper chooses RBF as the kernel function. Its expression is:

[0072] K(x i x j )=exp(-γ||x i -x j || 2 (6)

[0073] The transition resistance estimation database established in step A is used as a sample for SVMR training. Once SVMR training is complete, the fault information (U) measured at the monitoring point for the current fault is used. f θ f This is used as the input to the SVMR. The corresponding output is the fault resistance transition R. f This method only discusses the case where the fault transition resistance is purely resistive, therefore R f It contains only the real part.

[0074] Step C: Identifying the faulty section

[0075] Step C01: Estimate the fault resistance R f Compare the estimated fault resistance with the fault resistance in the database. If the estimated fault resistance is between two adjacent fault resistances in the database, R... f (x) and R f (x+1). As shown in formula (7):

[0076] R f (x)<R f <R f (x+1) (7)

[0077] The voltage data is then used to identify the fault section. Here, x represents the multiple of the fault transition resistance step size in the database. If the step size is 5 when the database is established, then the single-phase ground fault simulation for each node is divided into 3 times, for transition resistances of 1 ohm (approximately metallic grounding), 5 ohms, and 10 ohms. If x = 1, then x + 1 = 5. As mentioned above, the step size in this patent is 1 ohm, so each node needs to undergo 10 fault simulations. If x = 1, then x + 1 = 2. Then, the adjacent fault resistance R... f (x) and R f (x+1) Compare the voltage amplitude and angle information corresponding to the database with the actual voltage amplitude and angle of the fault to determine the fault section.

[0078] Taking the fault segment between nodes i and j as an example, there are no other nodes between nodes i and j, therefore the line between nodes i and j can be represented by segment ij. After regression calculation, the resulting transition resistance is R. f Then R f With R f (x) and R f The relationship between (x+1) is shown in the figure.

[0079] With a transition resistance of R f When (x), the voltage amplitude at node i is Node j is At node i, when the transition resistance is R f At (x+1), the corresponding voltage amplitude and phase angle are Node j is

[0080] like Figure 2 As shown, the shaded area represents the resistance value R. f (x) and R f The search boundary is at (x+1). It is worth noting that the fault voltage and phase angle corresponding to the actual fault (U) f θ f The fault section is not within the shaded area. The location of the fault section can be determined using information about the fault voltage and phase angle range. The corresponding range information is as follows:

[0081]

[0082]

[0083] Step C02: Determining the faulty section. After an actual short-circuit fault occurs, the voltage and phase angle (U) measured in the faulty section are determined. f θ fThe data point should be the closest to the corresponding section information connection in the database, therefore the fault information point (U) is calculated. f θ f The shortest distance from the line connecting the fault voltage and phase angle data of a certain section (selected from the database) can be used to determine the section where the fault occurred.

[0084] like Figure 3 As shown, let segment 1 be ij and segment 2 be mn. Consistent with the previous description, m represents the starting node of the segment, and n represents the ending node. There are no other nodes between the two segments. The voltage and phase angle data for segment 1 are as follows: and Similarly, the voltage and phase angle data for segment 2 are as follows: and d m-n and d i-j Indicates the fault point (U) f θ f The shortest distance between data points and segments 1 and 2.

[0085] As can be seen from the figure, d m-n Significantly less than d i-j Therefore, compared to segment ij, segment mn is more likely to be the segment where the fault is located. The overall algorithm process involves iterating through all distance results and selecting the shortest distance as the segment where the fault is located.

[0086] To verify the universality of the algorithm proposed in this patent, the algorithm is validated using a typical radial distribution network IEEE-33 node model in the example discussion section. This network contains 33 nodes and 32 branches, with a system rated voltage of 12.66kV. Relevant network data are shown in Tables 1 and 2. Node 1 is selected as the monitoring point. It is assumed that a single-phase ground fault occurs in section 28-29, and the preset fault transition resistance is 4.5 ohms.

[0087] Table 1 IEEE-33 Line Parameters

[0088]

[0089]

[0090]

[0091] Table 2 IEEE 33-node load parameters

[0092]

[0093]

[0094] Based on the above steps, the short-circuit fault information at monitoring point 1 (3118.59, 0.2994*π) is used as the monitoring location. This patent uses radians to represent the angle value, and the estimated transition resistance is 4.8 ohms obtained from SVMR regression calculation. The distance calculation results for all sections of IEEE-33 are shown in Table 3 below.

[0095] Table 3 Distance Calculation Results

[0096]

[0097] As shown in red in the table, the distance to the fault is shortest in section 28-29, and the fault can be accurately located in section 28-29. The corresponding distance bar chart is shown below:

[0098] The results show that segment 28 starts at node 28 and ends at node 29, which is the segment where the fault occurs. Figure 5 As shown, d 28-29 The result is much smaller than that of other sections, making it easy to determine the location of the faulty section.

[0099] The above description is merely an embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and scope of the present invention are included within the scope of protection of the present invention.

Claims

1. A method for locating a power grid fault section based on SVMR transition resistance estimation, characterized in that: include: Establishing a transition resistance estimation database; The SVMR method is used to measure the fault resistance R f Make estimates; The estimated fault resistance R f Compare the fault resistance with the transition resistance estimation database to identify the fault section; The identifying of the fault section specifically includes: The estimated fault resistance R f Compare with the fault resistance in the database; If the estimated fault resistance is between two adjacent fault resistances in the database R f (x) and R f (x+1), as shown in the following formula: R f (x)<R f <R f (x+1) (7) The voltage data is used to identify the fault section; x represents the multiple of the fault transition resistance step in the database. If the step selected when establishing the database is 5, the single-phase grounding fault simulation performed at each node is divided into 3 times, namely, the transition resistance is 1 ohm, 5 ohms, and 10 ohms. If x=1, then x+1=5; if the step is 1 ohm, each node needs to perform 10 fault simulations; if x=1, then x+1=2; then, the adjacent fault resistance R f (x) and R f (x+1) Compare the voltage amplitude and angle information corresponding to the database with the voltage amplitude and angle of the actual fault to determine the fault section; Taking the fault segment between nodes i and j as an example, there is no other node between nodes i and j, so the line between nodes i and j can be represented by the ij segment; The transition resistance is R f (x), the voltage amplitude at node i is Node j is At node i, when the transition resistance is R f When (x+1), the corresponding voltage amplitude and phase angle are Node j is The location of the fault section can be determined by the information of the range of fault voltage and phase angle; The corresponding scope information is as follows: After the actual short-circuit fault occurs, calculate the fault information point (U f ,θ f ) to the shortest distance obtained by connecting the fault voltage and phase angle data of a certain section, the section where the fault occurs can be determined; Segment 1 is ij, segment 2 is mn, m represents the starting node of the segment, n represents the ending node of the segment, and there are no other nodes between the two nodes; the voltage and phase angle data of segment 1 are and The voltage and phase angle data of segment 2 are and d m-n and d i-j Indicates the fault point (U f ,θ f )The shortest distance between the data point and segments 1,2; The overall process of the algorithm is to traverse all distance results and take the shortest distance as the section where the fault is located.

2. The method for locating a fault section of a power grid based on SVMR transition resistance estimation according to claim 1, characterized in that: The establishment of the transition resistance estimation database specifically includes: setting a corresponding transition resistance at each node of the distribution system to simulate a single-phase grounding fault; recording the voltage amplitude and angle of the single-phase grounding fault after short circuit from the measurement node, that is, {(U1, sinθ1), (U2, sinθ2), ..., (U n ,sinθ n )}, where n is the node number; In the fault simulation experiment, the transition resistance of the single-phase grounding fault is selected in the range of [1,10], and the traversal step size is selected as 1 ohm; Establish a transition resistance estimation database for all nodes. The transition resistance data of each node is in the form of a 10*3 matrix, where 10 represents the transition resistance value of 1 to 10, and 3 represents the corresponding information of each fault simulation experiment (U n ,sinθ n ,R f ), where R f is the transition resistor.

3. The method for locating a fault section of a power grid based on SVMR transition resistance estimation according to claim 2, characterized in that: The SVMR method is used to estimate the fault resistance, specifically including: Assume that the data training sample S is: S={(x1,y1),(x2,y2),...,(x n ,y n )}x i ∈R n ,y i ∈R n (1) In the formula, x i is the input vector in the sample space (U i ,sinθ i );y i is the output vector R in the sample space f ; i is the number of samples; Where ω is the weight vector; is the mapping function; b is the bias; The core problem of SVMR training is to minimize the objective function of the following formula: In the formula, C is the penalty parameter; ω is the weight vector; ε is the upper limit of error; y i is the output vector R in the sample space f ; Introducing Lagrange multipliers and duality theory, the final expression is as follows: In the formula, α i , α i ′,α j , α' j is the Lagrange multiplier; ε is the upper limit of error; K(x i , x j ) is the inner product kernel function; The expression of RBF as kernel function is: K(x i ,x j )=exp(-γ||x i -x j || 2 ) (6) When the SVMR training is completed, the fault information (U f ,θ f ) as the input of SVMR, and the corresponding output is the fault resistance transition R f .

Images