BFO-SVM-based single-phase earth fault line selection method for small-current grounding system

Abstract

The invention relates to a small current grounding system single-phase grounding fault line selection method based on BFO-SVM, and the method comprises the steps: firstly building a simple 10kV power distribution network model through a PSCAD simulation platform, and collecting a zero-sequence current signal and a zero-sequence voltage signal of each feeder line when a single-phase grounding fault occurs; performing filtering and denoising processing on the zero-sequence transient current signal and the zero-sequence transient voltage signal by adopting a filtering and denoising method combining S transformation threshold filtering and a time-frequency filter based on time-frequency spectrum distribution to respectively obtain an effective zero-sequence transient current signal and an effective zero-sequence transient voltage signal; s transformation is carried out on the obtained effective zero-sequence transient current signal, characteristic current intrinsic mode energy is extracted from the obtained two-dimensional complex time-frequency matrix, and zero-sequence active power is directly calculated according to the obtained effective zero-sequence transient current signal and effective zero-sequence transient voltage signal. Constructing a multi-dimensional fault characteristic vector combining characteristic current intrinsic mode energy and zero-sequence active power; and inputting the constructed multi-dimensional fault feature vector into a BFO-SVM model for training, and then performing line selection so as to obtain a small current grounding system single-phase grounding fault line selection model based on the BFO-SVM.

Description

Technical Field

[0001] The present invention relates to the field of power distribution networks, and in particular to a single-phase grounding fault line selection method for a small current grounding system based on BFO-SVM. Background Art

[0002] As one of the most important secondary energy sources, the reliable production, transmission, and distribution of electricity has become a crucial driving force for the continuous economic development of modern society. Power outages often disrupt normal production and daily life, causing severe economic downturns and even endangering personal safety. As the final link in power transmission, the distribution network serves end-users. Power supply reliability directly impacts economic development and people's lives. Against the backdrop of my country's ongoing energy integration and rapid economic development, people are placing higher demands on power supply reliability. Distribution network failures are the most significant cause of power outages. According to statistics, power outages caused by distribution network failures account for 90% of all power system outages. The majority of these failures are caused by short-circuit or ground faults, with single-phase ground faults accounting for approximately 80% of all power system outages. Therefore, accurately and rapidly locating and troubleshooting faults is crucial for improving distribution network reliability and a crucial step in the development of smart distribution networks. In the distribution network's low-current grounding system, when a single-phase grounding fault occurs, the three-phase line voltage remains symmetrical, and the short-circuit current of the fault line is not large. The system can still operate with the fault for 1-2 hours. Therefore, when a single-phase grounding fault occurs, there is no need to immediately cut off the fault. However, with the continuous expansion of the distribution network scale and the complexity of the operation mode, the number of busbars and feeders continues to increase. When a single-phase grounding fault occurs, the capacitive current to the ground will also continue to increase. Long-term operation with the fault is likely to deteriorate and develop into a two-phase fault or a three-phase fault, expanding the scope of the fault. In addition, the arc grounding overvoltage at the short-circuit point will damage the power equipment and further reduce the stability of the power system. Therefore, when a fault occurs in the distribution network, it is necessary to quickly locate the line where the fault is located and quickly eliminate the fault to prevent the fault from further deteriorating into a more serious accident.

[0003] In summary, for single-phase grounding faults in low-current grounding systems, by analyzing their fault characteristics, the fault line can be quickly identified, and the fault can be quickly eliminated and normal power supply can be quickly restored. This is of great significance to improving the automation and intelligence level of the distribution system, and is of great significance to improving the electricity user satisfaction. Summary of the Invention

[0004] The present invention proposes a single-phase grounding fault line selection method for a small current grounding system based on BFO-SVM, which can effectively improve the line selection classification accuracy of SVM and has good stability and reliability.

[0005] The present invention is implemented by the following scheme: a single-phase grounding fault line selection method for a small current grounding system based on BFO-SVM, characterized by comprising the following steps:

[0006] Step 1: Use the PSCAD simulation platform to build a simple 10kV distribution network model and collect the zero-sequence current and voltage signals of each feeder when a single-phase grounding fault occurs;

[0007] Step 2: Using a filtering and denoising method that combines S-transform threshold filtering with a time-frequency filter based on time-frequency spectrum distribution, the zero-sequence transient current signal and the zero-sequence transient voltage signal are filtered and denoised to obtain the effective zero-sequence transient current signal and the effective zero-sequence transient voltage signal respectively;

[0008] Step 3: Perform S-transform on the obtained effective zero-sequence transient current signal, extract the characteristic current intrinsic modal energy from the obtained two-dimensional complex time-frequency matrix, and directly calculate the zero-sequence active power based on the obtained effective zero-sequence transient current signal and effective zero-sequence transient voltage signal, and construct a multi-dimensional fault feature vector that combines the characteristic current intrinsic modal energy and zero-sequence active power;

[0009] Step 4: Input the constructed multi-dimensional fault feature vector into the BFO-SVM model for training and then use it for line selection.

[0010] The above-mentioned single-phase grounding fault line selection method for a small current grounding system based on BFO-SVM, the model building and simulation described in step 1 include: building a 10kV distribution network model on the PSCAD simulation platform, simulating single-phase grounding faults under various operating conditions (such as changing the fault initial phase angle, transition resistance, fault location, etc.), and collecting the zero-sequence current signal and zero-sequence voltage signal of each feeder at the time of the fault.

[0011] The above-mentioned single-phase grounding fault line selection method for a small current grounding system based on BFO-SVM, the filtering and denoising method described in step 2 is a combination of S-transform threshold filtering and a time-frequency filter based on time-frequency spectrum distribution, and the effective zero-sequence transient current signal and the effective zero-sequence transient voltage signal are obtained by filtering and denoising the zero-sequence transient current signal and the zero-sequence transient voltage signal respectively, including: performing S-transform on the collected zero-sequence transient current and zero-sequence transient voltage discrete sequences to obtain a two-dimensional time-frequency complex matrix S respectively. i (k,n) and S u (k,n), and select the appropriate filter threshold parameter γ i and γ u , and the filter factors H are obtained respectively. 1i (k,n) and H 1u(k, n), respectively used to perform the first layer of filtering and noise reduction on the collected zero-sequence transient current and zero-sequence transient voltage signals to obtain the effective signal S 1i and S 1u Filter factor H 1i (k,n) and H 1u The (k,n) expressions are

[0012]

[0013] After analyzing the time-frequency distribution of the signal and distinguishing the time-frequency range of noise and effective signal, the noise time-frequency range is removed and the effective signal part is retained. The expression of the time-frequency filter function H2(k,n) based on the time-frequency spectrum distribution is:

[0014]

[0015] Among them, [t k-1 ,t k ] and [f n-1 ,f n ] represents the time-frequency range of the effective signal.

[0016] The effective zero-sequence transient current signal S obtained 1i and effective zero-sequence transient voltage signal S 1u After being processed by the time-frequency filter method, the final effective signal S is extracted 2i and S 2u , respectively, perform S inverse transform to obtain the denoised one-dimensional time domain zero-sequence transient current signal and one-dimensional time domain zero-sequence transient voltage signal x' i (t) and x' u (t), its expression is:

[0017] x' i (t) = S -1 [S 2i (k,n)]

[0018] x' u (t) = S -1 [S 2u (k,n)]

[0019] The above-mentioned single-phase grounding fault line selection method for a small current grounding system based on BFO-SVM, in step 3, the effective zero-sequence transient current signal obtained is subjected to S transformation, and the characteristic current inherent modal energy is extracted from the obtained two-dimensional complex time-frequency matrix, and the zero-sequence active power is directly calculated based on the obtained effective zero-sequence transient current signal and the effective zero-sequence transient voltage signal to construct a multi-dimensional fault characteristic vector, including: performing S transformation on the denoised one-dimensional time-frequency zero-sequence transient current signal obtained in step 2, and calculating the corresponding characteristic current inherent modal energy of each feeder for the IMF component with a frequency of 0 to 20000 Hz after S transformation, and directly calculating the zero-sequence active power based on the obtained effective zero-sequence transient current signal and the effective zero-sequence transient voltage signal to construct a multi-dimensional fault characteristic vector, wherein the expressions of the characteristic current inherent modal energy and the zero-sequence active power are respectively

[0020]

[0021] Among them, n feeders on a bus are taken as research objects, x j (j=1,2,…,N) is the amplitude of the discrete points of the two-dimensional complex time-frequency matrix after the S-transformation of the effective zero-sequence transient current signal, E i (i=1,2,…,n) is the characteristic current natural modal energy of the i-th line, W i The energy normalized form, the characteristic current inherent modal energy of all lines constitutes a set of column vectors [W1,…,W n ];

[0022]

[0023] Where P is the zero-sequence active power, V(t) is the zero-sequence transient voltage at the bus, I(t) is the zero-sequence transient current, T includes 3 / 4 power frequency cycle before the fault and 1 / 4 power frequency cycle after the fault, and the zero-sequence transient voltage at the bus measured in PSCAD is 3U0; the zero-sequence active power of all lines constitutes another set of vectors [P1,…,P n ].

[0024] The BFO optimization algorithm described in step 4 of the BFO optimization algorithm for single-phase ground fault line selection in low-current grounded systems based on BFO-SVM includes the following: The BFO algorithm uses multiple search strategies, such as minnows occupying oysters, escaping from unoccupied oysters, breeding in oysters, and capturing offspring. Furthermore, it introduces a distributed search mechanism, also known as a dual strategy, which significantly enhances the ability to find the global optimal solution.

[0025] In the above-mentioned single-phase grounding fault line selection method for a small current grounding system based on BFO-SVM, the mathematical model of the minnow occupying the oyster is as follows:

[0026]

[0027] Among them, F i t and F i t+1 F represents the current position and new position of the i-th fish (or i-th solution) at the t-th iteration and t+1-th iteration respectively. * represents the best oyster (i.e. the optimal solution), F + is a high-quality oyster randomly selected from the population. δ and r are random numbers between 0 and 1. J represents the step length, or movement rate, of a fish escaping from or approaching an oyster. This parameter decreases with each algorithm iteration. The reason for the decreasing J parameter is that after a male fish successfully mates, its activity decreases over time. This decrease in J gradually shifts the global search to a local search.

[0028] In the above-mentioned single-phase grounding fault line selection method for a small current grounding system based on BFO-SVM, the mathematical model of the minnow not occupying the oyster and escaping is as follows:

[0029]

[0030] Here, the value of M is equal to the average position of the school. In this sense, after escaping, a minnow can search the space between the average and the optimal value, or it can explore random space problems. The formula for M is as follows:

[0031]

[0032] In the above-mentioned single-phase grounding fault line selection method for a small current grounding system based on BFO-SVM, the mathematical model of the minnow breeding in oysters is as follows:

[0033] F i t+1 =F i t +R*rand(0,1)

[0034] Where R is the radius of the fish population around its host oyster. This parameter is initially set to [0, 2], with a typical default value of 2. It decreases over time as the BFO algorithm iterates.

[0035] In the above-mentioned single-phase grounding fault line selection method for a small current grounding system based on BFO-SVM, the mathematical model of the minnow offspring being captured is as follows:

[0036]

[0037] Among them, d(F it ) is the elimination solution F i t The probability of f(F i t ) represents the objective function value of the solution.

[0038] In the above-mentioned BFO-SVM-based single-phase grounding fault line selection method for a low-current grounding system, the distributed search mechanism, i.e., the dual strategy, is as follows: a global search is performed in the initial iteration stage to avoid the local optimal solution; and a local search is performed in the final iteration stage to obtain a more accurate solution.

[0039] In the above-mentioned BFO-SVM-based single-phase grounding fault line selection method for a small current grounding system, the optimization of the SVM using the BFO optimization algorithm in step 4 includes: optimizing the penalty parameter C and the kernel parameter γ in the SVM using the BFO optimization algorithm.

[0040] In the above-mentioned single-phase grounding fault line selection method for a small current grounding system based on BFO-SVM, the establishment of the BFO-SVM fault line selection model in step 4 includes: dividing the multi-dimensional fault feature vector constructed previously into a test set and a training set, inputting the training set into the BFO-SVM model, training the model, and then using the test set to verify the training results of the model, thereby obtaining the BFO-SVM fault line selection model. The model has good accuracy and strong robustness, and has a faster line selection speed, thereby shortening the line selection time. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 Flowchart of a single-phase grounding fault line selection method for a small current grounding system based on BFO-SVM

[0042] Figure 2 10kV low-current grounding system simulation model diagram

[0043] Figure 3 BFO algorithm optimization SVM flow chart

[0044] Figure 4 Comparison of SVM prediction results for the training set

[0045] Figure 5 Comparison of SVM prediction results for the test set

[0046] Figure 6 Comparison of BFO-SVM prediction results for the training set

[0047] Figure 7 Comparison of BFO-SVM prediction results for the test set

[0048] Figure 8Comparison of prediction results of BFO-SVM on the training set based on 50% training set

[0049] Figure 9 Comparison of prediction results of BFO-SVM on the test set based on 50% training set DETAILED DESCRIPTION

[0050] In order to better understand the present invention, the present invention is further described below with reference to the accompanying drawings and examples.

[0051] Step 1: Use the PSCAD simulation platform to build a simple 10kV distribution network model and collect the zero-sequence current and voltage signals of each feeder when a single-phase grounding fault occurs;

[0052] Step 2: Using a filtering and denoising method that combines S-transform threshold filtering with a time-frequency filter based on time-frequency spectrum distribution, the zero-sequence transient current signal and the zero-sequence transient voltage signal are filtered and denoised to obtain the effective zero-sequence transient current signal and the effective zero-sequence transient voltage signal respectively;

[0053] Step 3: Perform S-transform on the obtained effective zero-sequence transient current signal, extract the characteristic current intrinsic modal energy from the obtained two-dimensional complex time-frequency matrix, and directly calculate the zero-sequence active power based on the obtained effective zero-sequence transient current signal and effective zero-sequence transient voltage signal, and construct a multi-dimensional fault feature vector that combines the characteristic current intrinsic modal energy and zero-sequence active power;

[0054] Step 4: Input the constructed multi-dimensional fault feature vector into the BFO-SVM model for training and then use it for line selection.

[0055] Furthermore, in step 2, the collected zero-sequence transient current and zero-sequence transient voltage discrete sequences are subjected to S transformation to obtain the two-dimensional time-frequency complex matrix S i (k,n) and S u (k,n), and select the appropriate filter threshold parameter γ i and γ u , and the filter factors H are obtained respectively. 1i (k,n) and H 1u (k, n), respectively used to perform the first layer of filtering and noise reduction on the collected zero-sequence transient current and zero-sequence transient voltage signals to obtain the effective signal S 1i and S 1u Filter factor H 1i (k,n) and H 1u The (k,n) expressions are

[0056]

[0057] Furthermore, in step 2, the time-frequency distribution of the zero-sequence transient current signal and the zero-sequence transient voltage signal is analyzed. After distinguishing the time-frequency range of noise and effective signal, the noise time-frequency range is eliminated and the effective signal part is retained. The expression of the time-frequency filter function H2(k,n) based on the time-frequency spectrum distribution is:

[0058]

[0059] Among them, [t k-1 ,t k ] and [f n-1 ,f n ] represents the time-frequency range of the effective signal.

[0060] Furthermore, in step 2, the effective zero-sequence transient current signal S 1i and effective zero-sequence transient voltage signal S 1u After being processed by the time-frequency filter method, the final effective signal S is extracted 2i and S 2u , respectively, perform S inverse transform to obtain the denoised one-dimensional time domain zero-sequence transient current signal and one-dimensional time domain zero-sequence transient voltage signal x' i (t) and x' u (t), its expression is:

[0061] x' i (t) = S -1 [S 2i (k,n)]

[0062] x' u (t) = S -1 [S 2u (k,n)]

[0063] Furthermore, the characteristic current intrinsic modal energy in step 3 is extracted as follows:

[0064]

[0065] Among them, n feeders on a bus are taken as research objects, x j (j=1,2,…,N) is the amplitude of the discrete points of the two-dimensional complex time-frequency matrix after the S-transformation of the effective zero-sequence transient current signal, E i (i=1,2,…,n) is the characteristic current natural modal energy of the i-th line, W i The energy normalized form, the characteristic current inherent modal energy of all lines constitutes a set of column vectors [W1,…,W n ];

[0066] Furthermore, the zero-sequence active power in step 3 is:

[0067]

[0068] Where P is the zero-sequence active power, V(t) is the zero-sequence transient voltage at the bus, I(t) is the zero-sequence transient current, T includes 3 / 4 power frequency cycle before the fault and 1 / 4 power frequency cycle after the fault, and the zero-sequence transient voltage at the bus measured in PSCAD is 3U0; the zero-sequence active power of all lines constitutes another set of vectors [P1,…,P n ].

[0069] Furthermore, the advantage of the BFO algorithm in step 4 lies in its ability to leverage both the current optimal solution and the location information of high-quality fish within the population. This means that it not only focuses on searching for the optimal solution but also considers other suitable solutions, making it less susceptible to local optima. Furthermore, it employs a dual strategy: a global search in the initial iteration phase to avoid local optima; and a local search in the final iteration phase to obtain a more accurate solution. The introduction of this distributed search mechanism significantly enhances the ability to find the global optimal solution.

[0070] Furthermore, in step 4, a BFO-SVM fault line selection model is established, the collected fault feature quantities are divided into a test set and a training set, and the BFO-SVM model is trained using the data of the training set to obtain the BFO-SVM fault line selection model.

[0071] Simulation verification:

[0072] A 10kV low-current grounding system simulation model was built in the commonly used power system electromagnetic transient (EMT) simulation platform PSCAD. Figure 2 As shown, the line parameters of the feeder L1~L4 used are shown in Table 1. The simulation model consists of power supply G, step-down transformer, grounding transformer T z , arc suppression coil (L, r L ), distribution transformer, 10kV busbar and four outgoing lines.

[0073] Table 1 Parameters of overhead lines and cable lines

[0074]

[0075] When obtaining single-phase grounding fault signals, in order to better simulate the types of faults that may occur in actual power systems and improve the accuracy of algorithm classification, single-phase grounding faults with grounding resistances of 0Ω (metallic grounding fault), 5Ω, 100Ω, 500Ω, 1000Ω, 3000Ω, 5000Ω, 7000Ω, 9000Ω, and 10000Ω occur at 20%, 50%, and 80% of the line length of feeders L1 to L4, respectively. The fault angles corresponding to the fault moments are 0°, 45°, 90°, and 100°. Five fault conditions, namely, 135°, 180°, and 600 fault conditions in total, were selected. Under each fault condition, the zero-sequence transient current and zero-sequence transient voltage of each feeder were measured using a zero-sequence transient current transformer and a zero-sequence transient voltage transformer. The zero-sequence transient current signal and zero-sequence transient voltage signal of the 3 / 4 cycle before the fault and the 1 / 4 cycle after the fault were selected as research targets. The natural modal energy and zero-sequence active power of each feeder were calculated under the 600 fault conditions. Some of the calculation results are shown in Tables 2 and 3.

[0076] Table 2 Natural mode energy

[0077]

[0078] Analysis of the data in Table 2 shows that when single-phase grounding faults of 0Ω (metallic grounding fault), 5Ω, 100Ω, 500Ω, 1000Ω, 3000Ω, 5000Ω, 7000Ω, 9000Ω, and 10000Ω occur at 80% of the length of feeder L1, the natural modal energy of the fault line L1 is higher than the natural modal energy of other normal lines; when single-phase grounding faults of 0Ω (metallic grounding fault), 5Ω, 100Ω, 500Ω, 1000Ω, 3000Ω, 5000Ω, 7000Ω, 9000Ω, and 10000Ω occur at 80% of the length of feeder L2, the natural modal energy of the fault line L2 is higher than that of other normal lines. Intrinsic modal energy; when single-phase grounding faults of 0Ω (metallic grounding fault), 5Ω, 100Ω, 500Ω, 1000Ω, 3000Ω, 5000Ω, 7000Ω, 9000Ω, and 10000Ω occur at 80% of the length of feeder L3, the inherent modal energy of the faulty line L3 is higher than the inherent modal energy of other normal lines; when single-phase grounding faults of 0Ω (metallic grounding fault), 5Ω, 100Ω, 500Ω, 1000Ω, 3000Ω, 5000Ω, 7000Ω, 9000Ω, and 10000Ω occur at 80% of the length of feeder L4, the inherent modal energy of the faulty line L4 is higher than the inherent modal energy of other normal lines.

[0079] From the above experiments, we can see that the natural mode energy can distinguish normal lines from faulty lines and can be used as one of the fault characteristic quantities.

[0080] Table 3 Zero sequence active power

[0081]

[0082] Analysis of the data in Table 3 shows that when single-phase grounding faults of 0Ω (metallic grounding fault), 5Ω, 100Ω, 500Ω, 1000Ω, 3000Ω, 5000Ω, 7000Ω, 9000Ω, and 10000Ω occur at 80% of the length of feeder L1, the zero-sequence active power of the fault line L1 is negative, while the zero-sequence active power of other normal lines is positive; when single-phase grounding faults of 0Ω (metallic grounding fault), 5Ω, 100Ω, 500Ω, 1000Ω, 3000Ω, 5000Ω, 7000Ω, 9000Ω, and 10000Ω occur at 80% of the length of feeder L2, the zero-sequence active power of the fault line L2 is negative, while the zero-sequence active power of other normal lines is positive. The active power is positive; when single-phase grounding faults of 0Ω (metallic grounding fault), 5Ω, 100Ω, 500Ω, 1000Ω, 3000Ω, 5000Ω, 7000Ω, 9000Ω, and 10000Ω occur at 80% of the length of feeder L3, the zero-sequence active power of the fault line L3 is negative, while the zero-sequence active power of other normal lines is positive; when single-phase grounding faults of 0Ω (metallic grounding fault), 5Ω, 100Ω, 500Ω, 1000Ω, 3000Ω, 5000Ω, 7000Ω, 9000Ω, and 10000Ω occur at 80% of the length of feeder L4, the zero-sequence active power of the fault line L4 is negative, while the zero-sequence active power of other normal lines is positive.

[0083] From the above experiments, we can see that zero-sequence active power can be used to distinguish normal lines from fault lines and can be used as one of the fault characteristic quantities.

[0084] The processed natural mode energy and zero-sequence active power are used as fault feature quantities and divided into training set and test set. The training set is input into SVM and BFO-SVM for training respectively, and then used for classification and line selection. The flowchart of the BFO optimization algorithm optimizing SVM is shown in the following figure. Figure 3 As shown, the SVM training and testing results are as follows Figure 4 and Figure 5 As shown, the BFO-SVM training and testing results are as follows Figure 6 and Figure 7 As shown, the horizontal axis is the sample number, and the vertical axes 1, 2, 3, and 4 represent single-phase grounding faults occurring in feeders L1, L2, L3, and L4, respectively.

[0085] According to the analysis of the prediction results of the SVM model on the training set samples, it can be seen that the SVM model made prediction errors in predicting multiple samples in the training set. After MATLAB analysis, the prediction accuracy was only 76.17%, while the BFO-SVM model made no mistakes in the prediction results of all samples in the training set, with a prediction accuracy of 100%. In addition, in the comparison chart of the prediction results of the SVM model on the test set, it can be seen that the SVM model still made prediction errors at multiple points when predicting the test set, and the prediction accuracy was only 80.56%, while the BFO-SVM model still made no mistakes in predicting the test set. By comparing the prediction results of the SVM model and the BFO-SVM model, it can be concluded that the prediction accuracy of the SVM model after optimization by the BFO optimization algorithm has been significantly improved.

[0086] The method of reducing the number of samples in the training set is used to test the prediction accuracy of the BFO-SVM model under the condition of reducing the number of sample training. 50% of the original training set is used to train the BFO-SVM, that is, 300 data samples are used as the training set, and then the trained model is used to predict the training set and the test set respectively. The results are as follows: Figure 8 and Figure 9 As shown in the figure, the analysis shows that the prediction accuracy of the 300 samples in the training set can still reach 100%. When predicting the 180 samples in the test set, the prediction accuracy can still reach 100%, and the correct line selection with high accuracy can still be achieved. Therefore, the model still maintains good line selection accuracy when reducing the number of training samples, and has good robustness.

[0087] The description of the above embodiments is intended to illustrate the technical solutions of this application and does not constitute a limitation thereof. Although this application has been fully described, those skilled in the art may modify these embodiments or substitute equivalent features for certain technical features. Such modifications or substitutions will not cause the relevant technical solutions to deviate from the core spirit and scope of the technical solutions of this application.

Claims

1. A single-phase grounding fault line selection method for a small current grounding system based on BFO-SVM, characterized in that: The following steps are involved: Step 1: Use the PSCAD simulation platform to build a simple 10kV distribution network model and collect the zero-sequence current and voltage signals of each feeder when a single-phase grounding fault occurs; Step 2: Using a filtering and denoising method that combines S-transform threshold filtering with a time-frequency filter based on time-frequency spectrum distribution, the zero-sequence transient current signal and the zero-sequence transient voltage signal are filtered and denoised to obtain the effective zero-sequence transient current signal and the effective zero-sequence transient voltage signal respectively; Step 3: Perform S-transform on the obtained effective zero-sequence transient current signal, extract the characteristic current intrinsic modal energy from the obtained two-dimensional complex time-frequency matrix, and directly calculate the zero-sequence active power based on the obtained effective zero-sequence transient current signal and effective zero-sequence transient voltage signal, and construct a multi-dimensional fault feature vector that combines the characteristic current intrinsic modal energy and zero-sequence active power; Step 4: Input the constructed multi-dimensional fault feature vector into the BFO-SVM model for training and then use it for line selection.

2. The method for selecting a single-phase grounding fault line in a low-current grounding system based on BFO-SVM according to claim 1, characterized in that: In step 2, the collected zero-sequence transient current and zero-sequence transient voltage discrete sequences are subjected to S transformation to obtain two-dimensional time-frequency complex matrices S and S, respectively. i (k,n) and S u (k,n), and select the appropriate filter threshold parameter γ i and γ u , and the filter factors H are obtained respectively. 1i (k,n) and H 1u (k, n), respectively used to perform the first layer of filtering and noise reduction on the collected zero-sequence transient current signal and zero-sequence transient voltage signal to obtain the effective signal S 1i and S 1u Filter factor H 1i (k,n) and H 1u The (k,n) expressions are 3. The method for selecting a single-phase grounding fault line in a low-current grounding system based on BFO-SVM according to claim 1, characterized in that: In step 2, the time-frequency distribution of the zero-sequence transient current signal and the zero-sequence transient voltage signal is analyzed, and after distinguishing the time-frequency range of noise and effective signal, the noise time-frequency range is eliminated and the effective signal part is retained. The time-frequency filter function H2(k,n) based on the time-frequency spectrum distribution is expressed as follows: Among them, [t k-1 ,t k ] and [f n-1 ,f n ] represents the time-frequency range of the effective signal.

4. The method for selecting a single-phase grounding fault line in a low-current grounding system based on BFO-SVM according to claim 1, characterized in that: The effective signal S obtained in step 2 1i and S 1u After being processed by the time-frequency filter method, the final effective signal S is extracted 2i and S 2u , respectively, perform S inverse transform to obtain the denoised one-dimensional time domain zero-sequence transient current signal and one-dimensional time domain zero-sequence transient voltage signal x i '(t) and x' u (t), its expression is: x i '(t)=S -1 [S 2i (k,n)] x' u (t)=S -1 [S 2u (k,n)]。 5. The method for selecting a single-phase grounding fault line in a low-current grounding system based on BFO-SVM according to claim 1, characterized in that: In step 3, the obtained effective zero-sequence transient current signal is subjected to S transformation, the characteristic current inherent modal energy is extracted from the obtained two-dimensional complex time-frequency matrix, and the zero-sequence active power is directly calculated based on the obtained effective zero-sequence transient current signal and the effective zero-sequence transient voltage signal, wherein the expressions of the characteristic current inherent modal energy and the zero-sequence active power are respectively Among them, n feeders on a bus are taken as research objects, x j (j=1,2,…,N) is the amplitude of the discrete points of the two-dimensional complex time-frequency matrix after the S-transformation of the effective zero-sequence transient current signal, E i (i=1,2,…,n) is the characteristic current natural modal energy of the i-th line, W i The energy normalized form, the characteristic current inherent modal energy of all lines constitutes a set of column vectors [W1,…,W n ]; Where P is the zero-sequence active power, V(t) is the zero-sequence transient voltage at the bus, I(t) is the zero-sequence transient current, T includes 3 / 4 power frequency cycle before the fault and 1 / 4 power frequency cycle after the fault, and the zero-sequence transient voltage at the bus measured in PSCAD is 3U0; the zero-sequence active power of all lines constitutes another set of vectors [P1,…,P n ].

6. The method for selecting a single-phase grounding fault line in a low-current grounding system based on BFO-SVM according to claim 1, characterized in that: The advantage of the BFO algorithm in step 4 is that it leverages the current optimal solution while also incorporating information about the locations of high-quality fish within the population. This means it not only focuses on searching for the optimal solution but also considers other suitable solutions, making it less susceptible to local optima. Furthermore, it employs a dual strategy: a global search in the initial iteration phase to avoid local optima; and a local search in the final iteration phase to obtain a more accurate solution. The introduction of this distributed search mechanism significantly enhances the ability to find the global optimal solution.

Images