Power distribution network grounding fault line selection method based on characteristic frequency band optimization

Abstract

The invention provides a power distribution network grounding fault line selection method based on characteristic frequency band optimization, and the method comprises the following steps: obtaining a transient zero sequence signal of each line, the transient zero sequence signal comprising a normalized transient zero sequence current and a transient zero sequence voltage derivative; decomposing the transient zero-sequence signal by adopting a wavelet packet transformation method; a historical data set is constructed, and frequency bands capable of effectively representing fault features in the historical data set are extracted in an off-line self-adaptive mode through a genetic algorithm; collecting data after the fault, and analyzing and extracting a characteristic frequency band in the fault data by utilizing wavelet packet energy entropy; performing intersection extraction processing on the characteristic frequency bands extracted by the genetic algorithm and the wavelet packet energy entropy to obtain a final optimized characteristic frequency band result; and for a real-time fault, carrying out fault research and judgment by utilizing the transient zero-sequence signal in the optimal characteristic frequency band. According to the invention, waveform data collected by all detection nodes does not need to be uploaded to a power distribution master station, and the position of the single-phase earth fault can be researched and judged on site.

Description

Technical Field

[0001] The invention relates to the field of electric power equipment, and in particular to a method for selecting a ground fault line in a distribution network based on characteristic frequency band optimization. Background Art

[0002] As the link between the transmission network and the load, the stability and reliability of the distribution network operation directly affect the overall performance of the power system and the quality of power supply to users. However, the increasingly complex network structure and diversified power supply operation modes of the distribution system have increased the difficulty of network control and fault handling. Statistics show that most power outages in the daily operation of the power system are caused by distribution network faults, among which single-phase grounding faults account for more than 80% of the total number of distribution network faults. my country's distribution network usually adopts a small current grounding method, including neutral point grounding through arc suppression coils and neutral point ungrounded. When a single-phase grounding fault occurs in a small current grounding system, the fault current is small. At the same time, the number of nonlinear elements and power electronic equipment in the distribution network continues to increase, and the background noise is increasing. The interference of weak fault current and background noise will make it difficult to identify the fault characteristics. If the single-phase grounding fault cannot be isolated in time, the fault range may be expanded, which will have a serious impact on the power supply reliability of the distribution system. The State Grid Corporation of China pointed out in the "Technical Guidelines for Distribution Networks" issued in 2017 that when a permanent single-phase grounding fault occurs in the medium-voltage line of a small current grounding system, it is advisable to handle it according to the principle of rapid isolation of the fault nearby, that is, to achieve rapid line selection and tripping. Therefore, it is of great significance to quickly and accurately identify single-phase grounding faults and quickly remove the faulty line to reduce personal electric shock accidents and ensure the reliability of power supply in the distribution network. Summary of the invention

[0003] The purpose of the present invention is to propose a distribution network grounding fault line selection method based on characteristic frequency band optimization, which can determine the location of single-phase grounding fault on-site without uploading waveform data collected by all detection nodes to the distribution main station.

[0004] To achieve the above object, the technical solution of the present invention is: a method for selecting a ground fault line in a distribution network based on characteristic frequency band optimization, which specifically includes the following steps:

[0005] S1: Acquire transient zero-sequence signals of each line, wherein the transient zero-sequence signals include normalized transient zero-sequence current and transient zero-sequence voltage derivatives;

[0006] S2: Decomposing transient zero-sequence signal by wavelet packet transform method;

[0007] S3: Construct a historical data set, and extract the frequency bands that can effectively characterize the fault characteristics from the historical data set through offline adaptive genetic algorithm; collect post-fault data, and use wavelet packet energy entropy to analyze and extract the characteristic frequency bands in the fault data;

[0008] S4: performing intersection processing on the characteristic frequency bands extracted by the genetic algorithm and the wavelet packet energy entropy respectively, and obtaining the final optimal characteristic frequency band result;

[0009] S5: For real-time faults, use the transient zero-sequence signal in the optimal characteristic frequency band to perform fault analysis;

[0010] S6: The fault characteristic quantity in the preferred characteristic frequency band is used to determine the fault location, and the threshold-free positioning of the ground fault section is achieved by counting the number of determination results of all the preferred characteristic frequency bands.

[0011] Preferably, the method of obtaining the transient zero-sequence signal of each line is specifically: using a current transformer and a voltage transformer to obtain the transient zero-sequence current and the transient zero-sequence voltage of each line detection node, differentiating the transient zero-sequence voltage to obtain the transient zero-sequence voltage derivative, and normalizing the transient zero-sequence current and the transient zero-sequence voltage derivative as the transient zero-sequence signal.

[0012] Preferably, the accuracy levels of the current transformer and the voltage transformer are both 0.2.

[0013] Preferably, the wavelet packet transform method is specifically a wavelet packet five-layer transform, and the db10 wavelet is selected as the wavelet basis.

[0014] Preferably, the S3 specifically includes the following steps:

[0015] S3.1 A large amount of fault data of various historical fault scenarios are subjected to wavelet packet five-layer transformation, and five characteristic frequency bands are screened out through genetic algorithm; the fault data is the transient zero-sequence signal of the fault line detection node, and each piece of data obtains 32 sub-frequency bands;

[0016] S3.2 The wavelet packet reconstruction coefficients obtained after five-layer wavelet packet transformation and reconstruction of the transient zero-sequence voltage derivative in the fault data collected within the first half of the power frequency cycle after the fault occurs are used to calculate the wavelet packet energy entropy of each sub-band, and the wavelet packet energy entropy of each sub-band is arranged in order from large to small. Finally, the first five sub-bands with the largest wavelet packet energy entropy are selected as the characteristic band results based on wavelet packet energy entropy screening.

[0017] Preferably, the five characteristic frequency bands are selected by genetic algorithm, specifically:

[0018] Step a: converting the fault information into binary code; the coding principle is: if the polarity of the transient zero-sequence voltage derivative in the transient zero-sequence signal is opposite to that of the transient zero-sequence current within a certain sub-frequency band, it is regarded as a fault section, represented by 1; if the polarity is the same, it is regarded as a healthy section, represented by 0;

[0019] Step b: The population size of the genetic algorithm is set to 10, the individual length is set to 5 according to the number of characteristic frequency bands to be screened out, and the fitness function is defined as the accuracy of segment positioning based on the selected sub-frequency bands:

[0020]

[0021] Wherein, the fitness Fintness is used to measure the segment positioning performance of each combination scheme, Accuracy represents the accuracy of segment positioning, TP represents the number of fault segment samples predicted as fault segment categories, TN represents the number of sound segment samples predicted as sound segment categories, FP represents the number of sound segment samples predicted as fault segment categories, and FN represents the number of fault segment samples predicted as sound segment categories.

[0022] Step c: Select individuals from the population for genetic operations based on the fitness calculation results, and set the crossover probability P c , mutation probability P m and the maximum number of iterations Generations; the genetic algorithm terminates the operation after reaching the maximum number of iterations, and outputs the individual with the maximum fitness as the fault feature frequency band optimized based on the genetic algorithm.

[0023] Preferably, the crossover probability P c Set to 0.8.

[0024] Preferably, the mutation probability P m It is set to decrease as the number of iterations n' increases, expressed as:

[0025] P m =0.9-0.9(n'-1) / Generations.

[0026] Preferably, the S5 specifically includes the following steps:

[0027] S5.1 collects the transient zero-sequence signal of each detection node in the first half of the power frequency cycle after the fault, performs five-layer wavelet packet transformation and reconstructs the wavelet packet reconstruction coefficient, and draws a two-dimensional graph of the transient zero-sequence voltage derivative and the transient zero-sequence current wavelet packet reconstruction coefficient of each sub-band in the preferred characteristic frequency band; in the healthy section, the two have the same polarity and show a positive correlation; in the fault section, the two have opposite polarities and show a negative correlation; thereby judging whether each detection node in the first half of the power frequency cycle after the fault is in the healthy section or the fault section in each sub-band in the preferred characteristic frequency band;

[0028] S5.2 calculates the cumulative sum of the transient zero-sequence power of each sub-band in the preferred characteristic frequency band according to the transient zero-sequence signal of each detection node within the first half of the power frequency cycle after the fault:

[0029]

[0030] Where P sn is the cumulative sum of transient zero-sequence power in the nth sub-band, P nj P sn The transient zero-sequence power of the jth sampling point; the transient zero-sequence power P nj The calculation is performed using the wavelet packet reconstruction coefficient of the nth sub-band of the transient zero-sequence signal:

[0031] P nj =i nj u nj

[0032] Where P nj is the transient zero-sequence power of the jth sampling point in the nth sub-band, i nj and u nj are the wavelet packet reconstruction coefficients of the jth sampling point of the nth sub-band of the transient zero-sequence current and transient zero-sequence voltage derivatives respectively;

[0033] The transient zero-sequence power accumulation sums of the sub-bands in the fault section are all negative values, and the transient zero-sequence power accumulation sums of the sub-bands in the healthy section are all positive values; thus, the positive and negative values ​​of the transient zero-sequence power accumulation sums of each sampling point in the transient zero-sequence power accumulation sums of each sub-band are judged one by one. If, in a certain sub-band, the number of sampling points where the transient zero-sequence power accumulation sum is a positive value is not less than the number of sampling points where it is a negative value, the judgment result of the sub-band is a healthy section; if the number of sampling points where the transient zero-sequence power accumulation sum is a positive value is less than the number of sampling points where it is a negative value, the judgment result of the sub-band is a fault section;

[0034] The accumulated and judged results of the zero-sequence power corresponding to each sub-band are:

[0035]

[0036] In the formula, Z n is the accumulated and judged result of zero-sequence power of the nth sub-band, Z n0 is the number of sampling points where the sum of transient zero-sequence power in the nth sub-band is positive, Z n1 is the number of sampling points where the cumulative sum of transient zero-sequence power is negative in the nth sub-band, "0" indicates that the cumulative sum of zero-sequence power is a healthy section, and "1" indicates that the cumulative sum of zero-sequence power is a faulty section;

[0037] Arrange the zero-sequence power accumulation and judgment results of all sub-bands in the preferred characteristic frequency band in frequency band order, and obtain the zero-sequence power accumulation and judgment result corresponding to the detection node:

[0038] Z=[Z1,Z2,…,Z M ]

[0039] In the formula, Z M is the accumulated and judged result of the zero-sequence power of the Mth sub-band.

[0040] Preferably, S6 specifically includes:

[0041] S6.1: Count the number of sub-frequency bands in healthy sections and faulty sections determined by step S5.1 for each detection node, and take the section type corresponding to the larger number as the first fault analysis result of the detection node;

[0042] S6.2: Count the number of sub-frequency bands in healthy sections and faulty sections determined by step S5.2 for each detection node, and take the section type corresponding to the larger number as the second fault analysis result of the detection node;

[0043] S6.3: When the first fault analysis result and the second fault analysis result of a detection node are both in the fault section, the detection node is determined to be in the fault section, and the ground fault section threshold-free positioning is performed based on the analysis results of all detection nodes.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] 1. In view of the differences in operating parameters of different distribution networks, the traditional threshold method relies on artificially set thresholds for segment location, which has poor adaptability and requires different thresholds to be set according to different line topologies, and has weak universality. The method proposed in the present invention only needs to collect local information of the detection node to achieve segment location, without setting the threshold, and extracts a variety of fault feature quantities within the characteristic frequency band, which has strong adaptability and practicality.

[0047] 2. In the existing single-phase ground fault line selection methods, it is common to use the linear relationship between the transient zero-sequence current and the transient zero-sequence voltage derivative as a criterion, but most of them rely on experience to manually select characteristic frequency bands. The present invention combines adaptive characteristic frequency band extraction based on historical data and characteristic frequency band extraction based on real-time data, does not rely on artificial selection of characteristic frequency bands, and improves generalization ability.

[0048] 3. The method proposed in the present invention only needs to perform fault analysis in the terminal deployed at each detection node and transmit the fault analysis result to the master station, which effectively reduces the amount of data transmission and alleviates the computing burden of the master station. It also extracts multiple fault feature quantities within the characteristic frequency band, has strong adaptability and practicality, and performs well in engineering practicality. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a flow chart of the present invention;

[0050] Figure 2 This is an example diagram of the wavelet packet transform process of the present invention. DETAILED DESCRIPTION

[0051] The following is combined with Figure 1-2 , the technical solution of the present invention is specifically described.

[0052] like Figure 1 The present invention proposes a method for selecting a ground fault line in a distribution network based on characteristic frequency band optimization, which specifically includes the following steps:

[0053] S1: Acquire transient zero-sequence signals of each line, wherein the transient zero-sequence signals include normalized transient zero-sequence current and transient zero-sequence voltage derivatives;

[0054] S2: Decomposing transient zero-sequence signal by wavelet packet transform method;

[0055] S3: Construct a historical data set, and extract the frequency bands that can effectively characterize the fault characteristics from the historical data set through offline adaptive genetic algorithm; collect post-fault data, and use wavelet packet energy entropy to analyze and extract the characteristic frequency bands in the fault data;

[0056] S4: performing intersection processing on the characteristic frequency bands extracted by the genetic algorithm and the wavelet packet energy entropy respectively, and obtaining the final optimal characteristic frequency band result;

[0057] S5: For real-time faults, use the transient zero-sequence signal in the optimal characteristic frequency band to perform fault analysis;

[0058] S6: The fault characteristic quantity in the preferred characteristic frequency band is used to determine the fault location, and the threshold-free positioning of the ground fault section is achieved by counting the number of determination results of all the preferred characteristic frequency bands.

[0059] In this embodiment, the method of obtaining the transient zero-sequence signal of each line is specifically as follows: using a current transformer and a voltage transformer to obtain the transient zero-sequence current and the transient zero-sequence voltage of each line detection node, differentiating the transient zero-sequence voltage to obtain the transient zero-sequence voltage derivative, and normalizing the transient zero-sequence current and the transient zero-sequence voltage derivative as the transient zero-sequence signal.

[0060] In this embodiment, the accuracy level of the current transformer and the voltage transformer are both 0.2.

[0061] like Figure 2 In this embodiment, it is assumed that the original signal is The wavelet packet transform method is specifically a five-layer wavelet packet transform, and the db10 wavelet is selected as the wavelet basis. and are the approximate component and detail component of the t-th scale, t>0, H k and G k are the low-pass and high-pass filter coefficients respectively.

[0062] In this embodiment, S3 specifically includes the following steps:

[0063] S3.1 A large amount of fault data of various historical fault scenarios are subjected to wavelet packet five-layer transformation, and five characteristic frequency bands are screened out through genetic algorithm; the fault data is the transient zero-sequence signal of the fault line detection node, and each piece of data obtains 32 sub-frequency bands;

[0064] S3.2 The wavelet packet reconstruction coefficients obtained after five-layer wavelet packet transformation and reconstruction of the transient zero-sequence voltage derivative in the fault data collected within the first half of the power frequency cycle after the fault occurs are used to calculate the wavelet packet energy entropy of each sub-band, and the wavelet packet energy entropy of each sub-band is arranged in order from large to small. Finally, the first five sub-bands with the largest wavelet packet energy entropy are selected as the characteristic band results based on wavelet packet energy entropy screening.

[0065] In this embodiment, the five characteristic frequency bands are selected by genetic algorithm, specifically:

[0066] Step a: converting the fault information into binary code; the coding principle is: if the polarity of the transient zero-sequence voltage derivative in the transient zero-sequence signal is opposite to that of the transient zero-sequence current within a certain sub-frequency band, it is regarded as a fault section, represented by 1; if the polarity is the same, it is regarded as a healthy section, represented by 0;

[0067] Step b: The population size of the genetic algorithm is set to 10, and the individual length is set to 5 according to the number of characteristic frequency bands to be screened. In order to verify the quality of each randomly selected sub-band, the method uses the selected sub-band for segment positioning. The fitness function is defined according to the accuracy of segment positioning as follows:

[0068]

[0069] Wherein, the fitness Fintness is used to measure the segment positioning performance of each combination scheme, Accuracy represents the accuracy of segment positioning, TP represents the number of fault segment samples predicted as fault segment categories, TN represents the number of sound segment samples predicted as sound segment categories, FP represents the number of sound segment samples predicted as fault segment categories, and FN represents the number of fault segment samples predicted as sound segment categories.

[0070] Step c: Select individuals from the population for genetic operations based on the fitness calculation results, and set the crossover probability P c , mutation probability P mand the maximum number of iterations Generations; the genetic algorithm terminates the operation after reaching the maximum number of iterations, and outputs the individual with the maximum fitness as the fault characteristic frequency bands optimized based on the genetic algorithm. These characteristic frequency bands will be considered as the frequency bands that best characterize the fault characteristics of the historical data.

[0071] In this embodiment, the crossover probability P c Setting it to 0.8 means that there will be a greater probability of crossover steps. The maximum number of iterations is set to 50.

[0072] In this embodiment, in order to optimize the parameters and adapt to the changes in demand during the algorithm evolution process, the mutation probability P m It is set to decrease as the number of iterations n' increases, expressed as:

[0073] P m =0.9-0.9(n'-1) / Generations.

[0074] In this embodiment, S5 specifically includes the following steps:

[0075] S5.1 collects the transient zero-sequence signal of each detection node in the first half of the power frequency cycle after the fault, performs five-layer wavelet packet transformation and reconstructs the wavelet packet reconstruction coefficient, and draws a two-dimensional graph of the transient zero-sequence voltage derivative and the transient zero-sequence current wavelet packet reconstruction coefficient of each sub-band in the preferred characteristic frequency band; in the healthy section, the two have the same polarity and show a positive correlation; in the fault section, the two have opposite polarities and show a negative correlation; thereby judging whether each detection node in the first half of the power frequency cycle after the fault is in the healthy section or the fault section in each sub-band in the preferred characteristic frequency band;

[0076] S5.2 calculates the cumulative sum of the transient zero-sequence power of each sub-band in the preferred characteristic frequency band according to the transient zero-sequence signal of each detection node within the first half of the power frequency cycle after the fault:

[0077]

[0078] Where P sn is the cumulative sum of transient zero-sequence power in the nth sub-band, P nj P sn The transient zero-sequence power of the jth sampling point; the transient zero-sequence power P nj The calculation is performed using the wavelet packet reconstruction coefficient of the nth sub-band of the transient zero-sequence signal:

[0079] P nj =i nj u nj

[0080] Where P njis the transient zero-sequence power of the jth sampling point in the nth sub-band, i nj and u nj are the wavelet packet reconstruction coefficients of the jth sampling point of the nth sub-band of the transient zero-sequence current and transient zero-sequence voltage derivatives respectively;

[0081] The transient zero-sequence power accumulation sums of the sub-bands in the fault section are all negative values, and the transient zero-sequence power accumulation sums of the sub-bands in the healthy section are all positive values; thus, the positive and negative values ​​of the transient zero-sequence power accumulation sums of each sampling point in the transient zero-sequence power accumulation sums of each sub-band are judged one by one. If, in a certain sub-band, the number of sampling points where the transient zero-sequence power accumulation sum is a positive value is not less than the number of sampling points where it is a negative value, the judgment result of the sub-band is a healthy section; if the number of sampling points where the transient zero-sequence power accumulation sum is a positive value is less than the number of sampling points where it is a negative value, the judgment result of the sub-band is a fault section;

[0082] The accumulated and judged results of the zero-sequence power corresponding to each sub-band are:

[0083]

[0084] In the formula, Z n is the accumulated and judged result of zero-sequence power of the nth sub-band, Z n0 is the number of sampling points where the sum of transient zero-sequence power in the nth sub-band is positive, Z n1 is the number of sampling points where the cumulative sum of transient zero-sequence power is negative in the nth sub-band, "0" indicates that the cumulative sum of zero-sequence power is a healthy section, and "1" indicates that the cumulative sum of zero-sequence power is a faulty section;

[0085] The zero-sequence power accumulation and judgment results of all sub-bands in the preferred characteristic frequency band are arranged in frequency band order, and the zero-sequence power accumulation and judgment result corresponding to the detection node is obtained as follows:

[0086] Z=[Z1,Z2,…,Z M ]

[0087] In the formula, Z M is the accumulated and judged result of the zero-sequence power of the Mth sub-band.

[0088] In this embodiment, S6 is specifically:

[0089] S6.1: Count the number of sub-frequency bands in healthy sections and faulty sections determined by step S5.1 for each detection node, and take the section type corresponding to the larger number as the first fault analysis result of the detection node;

[0090] S6.2: Count the number of sub-frequency bands in healthy sections and faulty sections determined by step S5.2 for each detection node, and take the section type corresponding to the larger number as the second fault analysis result of the detection node;

[0091] S6.3: When the first fault analysis result and the second fault analysis result of a detection node are both in the fault section, the detection node is determined to be in the fault section, and the ground fault section threshold-free positioning is performed based on the analysis results of all detection nodes.

[0092] The above are preferred embodiments of the present invention. Any changes made according to the technical solution of the present invention, as long as the resulting functions do not exceed the scope of the technical solution of the present invention, belong to the protection scope of the present invention.

Claims

1. A method for selecting a ground fault line in a distribution network based on characteristic frequency band optimization, characterized in that: The specific steps include: S1: Acquire transient zero-sequence signals of each line, wherein the transient zero-sequence signals include normalized transient zero-sequence current and transient zero-sequence voltage derivatives; S2: Decomposing transient zero-sequence signal by wavelet packet transform method; S3: Construct a historical data set and extract the frequency bands that can effectively characterize the fault characteristics from the historical data set through offline adaptive genetic algorithm; Collect post-fault data, and use wavelet packet energy entropy to analyze and extract characteristic frequency bands in the fault data; S4: performing intersection processing on the characteristic frequency bands extracted by the genetic algorithm and the wavelet packet energy entropy respectively, and obtaining the final optimal characteristic frequency band result; S5: For real-time faults, use the transient zero-sequence signal in the optimal characteristic frequency band to perform fault analysis; S6: The fault characteristic quantity in the preferred characteristic frequency band is used to determine the fault location, and the threshold-free positioning of the ground fault section is achieved by counting the number of determination results of all the preferred characteristic frequency bands.

2. A method for selecting a ground fault line in a distribution network based on characteristic frequency band optimization according to claim 1, characterized in that: The method of obtaining the transient zero-sequence signal of each line is specifically as follows: using a current transformer and a voltage transformer to obtain the transient zero-sequence current and the transient zero-sequence voltage of each line detection node, differentiating the transient zero-sequence voltage to obtain the transient zero-sequence voltage derivative, and normalizing the transient zero-sequence current and the transient zero-sequence voltage derivative as the transient zero-sequence signal.

3. A method for selecting a ground fault line in a distribution network based on characteristic frequency band optimization according to claim 2, characterized in that: The accuracy level of the current transformer and the voltage transformer is 0.

2.

4. The method for selecting a ground fault line in a distribution network based on characteristic frequency band optimization according to claim 1, characterized in that: The wavelet packet transform method is specifically a wavelet packet five-layer transform, and the db10 wavelet is selected as the wavelet basis.

5. A method for selecting a ground fault line in a distribution network based on characteristic frequency band optimization according to claim 4, characterized in that: The S3 specifically includes the following steps: S3.1 A large amount of fault data of various historical fault scenarios are subjected to wavelet packet five-layer transformation, and five characteristic frequency bands are screened out through genetic algorithm; the fault data is the transient zero-sequence signal of the fault line detection node, and each piece of data obtains 32 sub-frequency bands; S3.2 The wavelet packet reconstruction coefficients obtained after five-layer wavelet packet transformation and reconstruction of the transient zero-sequence voltage derivative in the fault data collected within the first half of the power frequency cycle after the fault occurs are used to calculate the wavelet packet energy entropy of each sub-band, and the wavelet packet energy entropy of each sub-band is arranged in order from large to small. Finally, the first five sub-bands with the largest wavelet packet energy entropy are selected as the characteristic band results based on wavelet packet energy entropy screening.

6. A method for selecting a ground fault line in a distribution network based on characteristic frequency band optimization according to claim 5, characterized in that: The five characteristic frequency bands are screened out by the genetic algorithm, specifically: Step a: converting the fault information into binary code; the coding principle is: if the polarity of the transient zero-sequence voltage derivative in the transient zero-sequence signal is opposite to that of the transient zero-sequence current within a certain sub-frequency band, it is regarded as a fault section, represented by 1; if the polarity is the same, it is regarded as a healthy section, represented by 0; Step b: The population size of the genetic algorithm is set to 10, the individual length is set to 5 according to the number of characteristic frequency bands to be screened out, and the fitness function is defined as the accuracy of segment positioning based on the selected sub-bands: Wherein, the fitness Fintness is used to measure the segment positioning performance of each combination scheme, Accuracy represents the accuracy of segment positioning, TP represents the number of fault segment samples predicted as fault segment categories, TN represents the number of sound segment samples predicted as sound segment categories, FP represents the number of sound segment samples predicted as fault segment categories, and FN represents the number of fault segment samples predicted as sound segment categories. Step c: Select individuals from the population for genetic operations based on the fitness calculation results, and set the crossover probability P c , mutation probability P m and the maximum number of iterations Generations; The genetic algorithm terminates the operation after reaching the maximum number of iterations, and outputs the individual with the maximum fitness as the fault feature frequency band optimized based on the genetic algorithm.

7. A method for selecting a ground fault line in a distribution network based on characteristic frequency band optimization according to claim 6, characterized in that: The crossover probability P c Set to 0.

8.

8. The method for selecting a ground fault line in a distribution network based on characteristic frequency band optimization according to claim 6, characterized in that: The mutation probability P m It is set to decrease as the number of iterations n' increases, expressed as: P m =0.9-0.9(n'-1) / Generations。 9. The method for selecting a ground fault line in a distribution network based on characteristic frequency band optimization according to claim 4, characterized in that: The S5 specifically includes the following steps: S5.1 collects the transient zero-sequence signal of each detection node in the first half of the power frequency cycle after the fault, performs five-layer wavelet packet transformation and reconstructs the wavelet packet reconstruction coefficient, and draws a two-dimensional graph of the transient zero-sequence voltage derivative and the transient zero-sequence current wavelet packet reconstruction coefficient of each sub-band in the preferred characteristic frequency band; in the healthy section, the two have the same polarity and show a positive correlation; in the fault section, the two have opposite polarities and show a negative correlation; thereby judging whether each detection node in the first half of the power frequency cycle after the fault is in the healthy section or the fault section in each sub-band in the preferred characteristic frequency band; S5.2 calculates the cumulative sum of the transient zero-sequence power of each sub-band in the preferred characteristic frequency band according to the transient zero-sequence signal of each detection node within the first half of the power frequency cycle after the fault: Where P sn is the cumulative sum of transient zero-sequence power in the nth sub-band, P nj P sn The transient zero-sequence power of the jth sampling point; the transient zero-sequence power P nj The calculation is performed using the wavelet packet reconstruction coefficient of the nth sub-band of the transient zero-sequence signal: P nj =and nj in nj Where P nj is the transient zero-sequence power of the jth sampling point in the nth sub-band, i nj and u nj are the wavelet packet reconstruction coefficients of the jth sampling point of the nth sub-band of the transient zero-sequence current and transient zero-sequence voltage derivatives respectively; The transient zero-sequence power accumulation sums of the sub-bands in the fault section are all negative values, and the transient zero-sequence power accumulation sums of the sub-bands in the healthy section are all positive values; thus, the positive and negative values ​​of the transient zero-sequence power accumulation sums of each sampling point in the transient zero-sequence power accumulation sums of each sub-band are judged one by one. If, in a certain sub-band, the number of sampling points where the transient zero-sequence power accumulation sum is a positive value is not less than the number of sampling points where it is a negative value, the judgment result of the sub-band is a healthy section; if the number of sampling points where the transient zero-sequence power accumulation sum is a positive value is less than the number of sampling points where it is a negative value, the judgment result of the sub-band is a fault section; The accumulated and judged results of the zero-sequence power corresponding to each sub-band are: In the formula, Z n is the accumulated and judged result of zero-sequence power of the nth sub-band, Z n0 is the number of sampling points in the nth sub-band where the sum of transient zero-sequence power is positive, Z n1 is the number of sampling points where the cumulative sum of transient zero-sequence power is negative in the nth sub-band, "0" indicates that the cumulative sum of zero-sequence power is a healthy section, and "1" indicates that the cumulative sum of zero-sequence power is a faulty section; Arrange the zero-sequence power accumulation and judgment results of all sub-bands in the preferred characteristic frequency band in frequency band order, and obtain the zero-sequence power accumulation and judgment result corresponding to the detection node: Z=[Z1,Z2,…,Z M ] In the formula, Z M is the accumulated and judged result of the zero-sequence power of the Mth sub-band.

10. A method for selecting a ground fault line in a distribution network based on characteristic frequency band optimization according to claim 9, characterized in that: The S6 is specifically: S6.1: Count the number of sub-frequency bands in healthy sections and faulty sections determined by step S5.1 for each detection node, and take the section type corresponding to the larger number as the first fault analysis result of the detection node; S6.2: Count the number of sub-frequency bands in healthy sections and faulty sections determined by step S5.2 for each detection node, and take the section type corresponding to the larger number as the second fault analysis result of the detection node; S6.3: When the first fault analysis result and the second fault analysis result of a detection node are both in the fault section, the detection node is determined to be in the fault section, and the ground fault section threshold-free positioning is performed based on the analysis results of all detection nodes.

Images