Medium-voltage power distribution single-phase earth fault fusion identification model

Abstract

The invention discloses a medium-voltage power distribution single-phase earth fault fusion identification model, and relates to the field of power distribution network earth fault identification, and the model specifically comprises the following steps: building an identification rule and model, and extracting effective multi-domain features reflecting different insulation degradation types; processing is carried out by adopting multivariate variance analysis MANOVA, and waveform features are sorted according to an F statistical result; constructing an insulation degradation type identification model based on ELM; a plurality of support vector machines (SVM) are combined, and based on a Libsvm software package, a single-phase earth fault insulation degradation type identification model is established; the Dempster-Shafer evidence theory is used for fusing the recognition results of the two recognition models, and a recognition fusion model is established; and testing and analyzing the fusion model through the selected data set to obtain a correct recognition result. According to the method, the fault detection and maintenance efficiency is improved, and SPGF development caused by insulation degradation is prevented.

Description

Technical Field

[0001] The present invention relates to the field of grounding fault identification in distribution networks, and in particular to a fusion identification model for single-phase grounding faults in medium-voltage distribution networks. Background Art

[0002] The research on fault type identification of distribution lines has attracted much attention. Generally, the fault type between short circuit and ground fault as well as the fault phase should be distinguished first. Information about the three line currents is used to identify the phases involved in all 10 types of parallel faults in the distribution system under different fault resistances, starting angles, system topologies, and load levels. The energy information of the current signal is extracted to determine whether the fault is a three-phase fault, a two-phase fault, or a single-phase fault. The characteristics of the waveform data collected from the indicators of various fault types are analyzed, and a multi-level fault type identification method is proposed. Since the treatment scheme for short circuit faults is relatively mature and SPGF accounts for the majority of faults occurring in the distribution system, the identification of SPGF is more meaningful.

[0003] Identifying progressive SPGF can not only improve maintenance efficiency, but also enable predictive maintenance and prevent failures; however, it faces many challenges. Compared with other classification standards, progressive SPGF is usually related to insulation degradation and exhibits more complex characteristics in waveform data; therefore, the extraction and selection of corresponding fault features are more difficult.

[0004] Therefore, a medium voltage distribution single-phase grounding fault fusion identification model is provided to solve the above problems. Summary of the invention

[0005] The purpose of the present invention is to provide a fusion identification model for single-phase grounding faults in medium-voltage power distribution. In order to reflect the fault characteristics of the progressive SPGF type, a fusion identification tool is developed; by extracting multiple features in the time domain, frequency domain and time-frequency domain from the fault waveform data, the statistical analysis based on multivariate variance analysis MANOVA shows the effectiveness of the extracted features; then a type identification tool based on extreme learning machine ELM and support vector machine SVM is established, and a fusion tool based on Dempster–Shafer (DS) evidence theory is developed to improve the efficiency of fault detection and maintenance, and prevent the development of SPGF caused by insulation degradation.

[0006] To achieve the above object, the present invention provides a medium voltage power distribution single-phase grounding fault fusion identification model, comprising the following steps: Step S1: Establish identification rules and models to extract effective multi-domain features reflecting different insulation degradation types; Step S2: Multivariate analysis of variance (MANOVA) was used for processing, and the waveform features were ranked according to the F statistical results; Step S3: constructing an insulation degradation type identification model based on ELM; Step S4: by combining multiple support vector machines (SVMs), a recognition model for insulation degradation type of single-phase grounding fault is established based on the Libsvm software package; Step S5: Using the Dempster-Shafer evidence theory to fuse the recognition results of the two recognition models in step S3 and step S4, and establishing a recognition fusion model; Step S6: Test and analyze the fusion model using the selected data set to obtain the correct recognition result.

[0007] Preferably, in step S2, the multi-waveform feature quantity selected is: C1-After Fault and Ratio: ; ; C2-Distortion degree: The distortion of a waveform compared to a sine wave is expressed by total harmonic distortion THD: ; in represents the number of harmonics, Indicates The RMS value of the subharmonic current, M Represents the highest harmonic number, set to 10; C3 - Degree of randomness: The randomness of waveform distortion is used to describe the stable content, and the stable content is defined as the standard deviation of the current waveform distortion: ; in, Indicates the zero sequence current in steady state cycles, N represents the number of cycles in the steady state, set to 10, μ Indicates zero sequence current waveform N Distortion per cycle The average value of C4-information dimension, the calculation formula is as follows: ; Where r represents the box ratio. For different r values, the information entropy is calculated as follows: ; in Indicates The probability of the phase space trajectory in a frame; C5-Time-frequency characteristics based on wavelet transform: According to the wavelet transform, seven features (C5–C11) were extracted as follows: ; in, The zero-sequence current waveform representing a system fault, express Initial time, Indicates the end time of the selected waveform. Indicates the time when the fault occurred. Indicates the end time of the fault transient process, express The middle time, express and The selected time between is used to distinguish the time-frequency characteristics of zero-sequence current before and after the transient process. express The peak time, express After that, the first half peak value of zero sequence current, represents the number of wavelet transform layers, Indicated in The zero sequence current in the cycle The energy of the wavelet transformed signal, Indicated in The zero sequence current in the cycle j The energy of the wavelet transformed signal, U 0 Indicates the line voltage before the fault, M s express The maximum slope of It is a waveform of 1.1 cycles, including 0.1 cycles before the fault and 1 cycle after the fault.

[0008] Preferably, in step S2, the steps of feature statistical analysis based on MANOVA are as follows: The ground fault data were collected from the fault indicators with a sampling rate of 4096 Hz assigned by the power supply company, including 61 cases of arrester breakdown, 61 cases of line-to-crossbar discharge, 61 cases of cable insulation damage, 55 cases of insulator breakdown, 59 cases of tree contact, and 145 cases of other types of ground faults. MANOVA was performed on the data of F1-F5, and the waveform features were ranked according to their F-statistic results. All 11 waveform features were selected as candidate fault features.

[0009] Preferably, in step S3, the steps of constructing the insulation degradation type identification model based on ELM are as follows: S31: Define the ELM model parameters, take 11 waveform features as input, set the number of neurons in the input layer to 11, combine the empirical formula, set the number of neurons in the hidden layer to 13, set the number of neurons in the output layer to 6, the activation function of the hidden layer neurons is the Sigmoid function, and the output vector is processed by the Softmax function, converting the value of each element in the output vector to between 0 and 1; S32: Divide the training set and the test set, then input the feature vector data consisting of 11 waveform features, define the vector label of each insulation degradation type, use the PSO algorithm to optimize the initial weights and deviations, and complete the training and optimization of the model; S33: Use the test set data to evaluate model performance.

[0010] Preferably, in step S4, a recognition model for the insulation degradation type of a single-phase grounding fault is established based on a support vector machine (SVM), which specifically includes the following steps: S41: normalizing the input vector containing nine waveform features; S42: Select radial basis function as kernel function; S43: Select the core parameters c and g of the model, where c represents the penalty factor, indicating the importance of abnormal data in the class, and g is the parameter coefficient of the kernel function; S44: Set the number of cross-validation times for the model to 10; S45: According to the optimal c and g obtained in step S43, the model in step S44 is trained using the svmtrain function in Libsvm, and the classification accuracy is tested using the svmppredict function in Libsvm.

[0011] Preferably, in step S5, the specific steps are as follows: S51: The set of all possible outcomes of the identification frame decision problem is collectively called the identification frame O: ; S52: Select evidence and construct basic probability assignment (BPA); Setting the reliability factor to 0.8, the quality functions for the six insulation degradation types and three pieces of evidence (i=1,2,3) for the uncertainty θ are as follows: ; ; in, It is The reliability coefficient of the evidence; Representation Proposition No. BPA with evidence; Indicates the uncertainty BPA of evidence; Y ij Indicates Fault type The diagnostic output probability of the evidence; S53: Evidence fusion combines the recognition results of ELM and SVM and calculates BPA based on Dempster combination rule: ; Where A, B and C are three subsets of the recognition frame O, and K is a normalization constant; S54: Decision diagnostics uses two rules to determine the type of insulation degradation: Rule 1: ; Rule 2: ; in F type It represents the insulation degradation type finally determined by combining the identification results of ELM and SVM; Rule 1 stipulates that the determined insulation degradation type has the highest BPA; Rule 2 stipulates that the BPA of the determined insulation degradation type is greater than the uncertainty.

[0012] Preferably, in step S6, based on the collected 442 SPGF data sets, the recognition performance of the ELM and SVM models is tested, the recognition effect of the fusion model based on the DS evidence theory is tested, and the recognition performance of different models is compared.

[0013] Therefore, the present invention adopts a medium voltage power distribution single-phase grounding fault fusion identification model of the above structure, which has the following beneficial effects: (1) The comprehensive model based on DS evidence theory proposed in this paper is superior to each single model.

[0014] (2) The method proposed in the present invention can improve the efficiency of fault detection and maintenance and prevent the development of SPGF caused by insulation degradation.

[0015] (3) The method proposed in the present invention can be embedded in the existing distribution automation system for engineering applications.

[0016] The method scheme of the present invention is further described in detail below through the drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 2 is a diagram illustrating feature selection in the time-frequency domain of the present disclosure. DETAILED DESCRIPTION

[0018] The method scheme of the present invention is further described below through drawings and embodiments.

[0019] Unless otherwise defined, method terms or scientific terms used in the present invention shall have the common meanings understood by one of ordinary skill in the art to which the present invention belongs.

[0020] The words "include" or "comprises" and the like used in the present invention mean that the elements before the word include the elements listed after the word, and do not exclude the possibility of also including other elements. The orientation or position relationship indicated by the terms "inside", "outside", "upper", "lower", etc. is based on the orientation or position relationship shown in the drawings, which is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation of the present invention. When the absolute position of the described object changes, the relative position relationship may also change accordingly. In the present invention, unless otherwise clearly specified and limited, the terms "attachment" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral body; it can be directly connected, or indirectly connected through an intermediate medium, and it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0021] Example The present invention provides a medium voltage power distribution single-phase grounding fault fusion identification model, comprising the following steps: Step S1: Establish identification rules and models to extract effective multi-domain features reflecting different types of insulation degradation; a total of 11 electrical features were selected, of which four separate multi-dimensional features were selected, and seven features were extracted based on wavelet transform, such as Figure 1 Description of feature selection in the time-frequency domain; Step S2: Multivariate analysis of variance (MANOVA) is used to perform feature statistical analysis. MANOVA is performed on the data of F1-F5 in Table 1, and the waveform features are sorted according to the F statistical results, as shown in Table 2: Table 1 ;

[0022] Table 2 ;

[0023] The statistical results show that all 11 waveform features can be used as candidate fault features; In step S2, the multi-waveform feature quantity selected is: C1-After Fault and Ratio: ; ; C2-Distortion degree: The distortion of a waveform compared to a sine wave is expressed by total harmonic distortion THD: ; in represents the number of harmonics, Indicates The RMS value of the subharmonic current, M Represents the highest harmonic number, set to 10; C3 - Degree of randomness: The randomness of waveform distortion is used to describe the stable content, and the stable content is defined as the standard deviation of the current waveform distortion: ; in, Indicates the zero sequence current in steady state cycles, N represents the number of cycles in the steady state, set to 10, μ Indicates zero sequence current waveform N Distortion per cycle The average value of C4-information dimension, the calculation formula is as follows: ; Where r represents the box ratio. For different r values, the information entropy is calculated as follows: ; in Indicates The probability of the phase space trajectory in a frame; C5-Time-frequency characteristics based on wavelet transform: According to the wavelet transform, seven features (C5–C11) were extracted as follows: ; in, The zero-sequence current waveform representing a system fault, express Initial time, Indicates the end time of the selected waveform. Indicates the time when the fault occurred. Indicates the end time of the fault transient process, express The middle time, express and The selected time between is used to distinguish the time-frequency characteristics of zero-sequence current before and after the transient process. express The peak time, express After that, the first half peak value of zero sequence current, represents the number of wavelet transform layers, Indicated in The zero sequence current in the cycle The energy of the wavelet transformed signal, Indicated in The zero sequence current in the cycle j The energy of the wavelet transformed signal, U 0 Indicates the line voltage before the fault, M s express The maximum slope of It is a waveform of 1.1 cycles, including 0.1 cycles before the fault and 1 cycle after the fault.

[0024] In step S2, the steps of feature statistical analysis based on MANOVA are as follows: The ground fault data were collected from the fault indicators with a sampling rate of 4096 Hz assigned by the power supply company, including 61 cases of arrester breakdown, 61 cases of line-to-crossbar discharge, 61 cases of cable insulation damage, 55 cases of insulator breakdown, 59 cases of tree contact, and 145 cases of other types of ground faults. MANOVA was performed on the data of F1-F5, and the waveform features were ranked according to their F-statistic results. All 11 waveform features were selected as candidate fault features.

[0025] Step S3: constructing an insulation degradation type identification model based on ELM; In step S3, the steps of constructing the insulation degradation type identification model based on ELM are as follows: S31: Define the ELM model parameters, take 11 waveform features as input, set the number of neurons in the input layer to 11, combine the empirical formula, set the number of neurons in the hidden layer to 13, set the number of neurons in the output layer to 6, the activation function of the hidden layer neurons is the Sigmoid function, and the output vector is processed by the Softmax function, converting the value of each element in the output vector to between 0 and 1; S32: Divide the training set and the test set, then input the feature vector data consisting of 11 waveform features, define the vector label of each insulation degradation type, use the PSO algorithm to optimize the initial weights and deviations, and complete the training and optimization of the model; S33: Evaluate model performance using test set data.

[0026] Step S4: by combining multiple support vector machines (SVMs), a recognition model for insulation degradation type of single-phase grounding fault is established based on the Libsvm software package; In step S4, a recognition model of insulation degradation type of single-phase grounding fault is established based on support vector machine SVM, which specifically includes the following steps: S41: normalizing the input vector containing nine waveform features; S42: Select radial basis function as kernel function; S43: Select the core parameters c and g of the model, where c represents the penalty factor, indicating the importance of abnormal data in the class, and g is the parameter coefficient of the kernel function; S44: Set the number of cross-validation times for the model to 10; S45: According to the optimal c and g obtained in step S43, the model in step S44 is trained using the svmtrain function in Libsvm, and the classification accuracy is tested using the svmppredict function in Libsvm.

[0027] Step S5: Using the Dempster-Shafer evidence theory to fuse the recognition results of the two recognition models in step S3 and step S4, and establishing a recognition fusion model; In step S5, the specific steps are as follows: S51: The set of all possible outcomes of the identification frame decision problem is collectively called the identification frame O: ; S52: Select evidence and construct basic probability assignment (BPA); Setting the reliability factor to 0.8, the quality functions for the six insulation degradation types and three pieces of evidence (i=1,2,3) for the uncertainty θ are as follows: ; ; in, It is The reliability coefficient of the evidence; Representation Proposition No. BPA with evidence; Indicates the uncertainty BPA of evidence; Y ij Indicates Fault type The diagnostic output probability of the evidence; S53: Evidence fusion combines the recognition results of ELM and SVM and calculates BPA based on Dempster combination rule: ; Where A, B and C are three subsets of the recognition frame O, and K is a normalization constant; S54: Decision diagnostics uses two rules to determine the type of insulation degradation: Rule 1: ; Rule 2: ; in F type It represents the insulation degradation type finally determined by combining the identification results of ELM and SVM; Rule 1 stipulates that the determined insulation degradation type has the highest BPA; Rule 2 stipulates that the BPA of the determined insulation degradation type is greater than the uncertainty.

[0028] Step S6: The fusion model is tested and analyzed by the selected data set, as shown in Table 3. The change of fitness in PSO based on ELM relative to the number of iterations and the recognition accuracy of each case are shown in Table 4. The recognition accuracy of the fusion model is 94.1%, which is higher than the recognition models based on ELM and SVM. The data size and recognition accuracy corresponding to each fault type in the test set are shown in Table 5. Finally, several groups of examples are given in which a single model leads to an incorrect recognition result and a correct recognition result is obtained, as shown in Table 6.

[0029] Table 3 ;

[0030] Table 4 ;

[0031] Table 5 ;

[0032] Table 6 ;

[0033] In step S6, based on the 442 SPGF data sets collected by the power supply company, the recognition performance of the ELM and SVM models was tested, the recognition effect of the fusion model based on the DS evidence theory was tested, and the recognition performance of different models was compared. According to the comparison results, the comprehensive model based on the DS evidence theory is superior to each single model. The proposed method can improve the efficiency of fault detection and maintenance and prevent the development of SPGF caused by insulation degradation. The method proposed in the present invention can be embedded in the existing distribution automation system for engineering applications.

[0034] Therefore, the present invention adopts the above-mentioned medium-voltage power distribution single-phase grounding fault fusion identification model. The comprehensive model based on DS evidence theory is superior to each single model. The proposed method can improve the efficiency of fault detection and maintenance and prevent the development of SPGF caused by insulation degradation.

[0035] Finally, it should be noted that the above embodiments are only used to illustrate the method scheme of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, ordinary method personnel in the field should understand that they can still modify or replace the method scheme of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified method scheme to deviate from the spirit and scope of the method scheme of the present invention.

Claims

1. A fusion identification model for single-phase grounding fault in medium voltage power distribution, characterized by: The following steps are involved: Step S1: Establish identification rules and models to extract effective multi-domain features reflecting different insulation degradation types; Step S2: Multivariate analysis of variance (MANOVA) was used for processing, and the waveform features were ranked according to the F statistical results; Step S3: constructing an insulation degradation type identification model based on ELM; Step S4: by combining multiple support vector machines (SVMs), a recognition model for insulation degradation type of single-phase grounding fault is established based on the Libsvm software package; Step S5: Using the Dempster-Shafer evidence theory to fuse the recognition results of the two recognition models in step S3 and step S4, and establishing a recognition fusion model; Step S6: Test and analyze the fusion model using the selected data set to obtain the correct recognition result.

2. A medium voltage power distribution single-phase grounding fault fusion identification model according to claim 1, characterized in that: In step S2, the multi-waveform feature quantity selected is: C1-After Fault and Ratio: ； ； C2-Distortion degree: The distortion of a waveform compared to a sine wave is expressed by total harmonic distortion THD: ； in represents the number of harmonics, Indicates The RMS value of the subharmonic current, M Represents the highest harmonic number, set to 10; C3 - Degree of randomness: The randomness of waveform distortion is used to describe the stable content, and the stable content is defined as the standard deviation of the current waveform distortion: ； in, Indicates the zero sequence current in steady state cycles, N represents the number of cycles in the steady state, set to 10, μ Indicates zero-sequence current waveform N Distortion per cycle The average value of C4-information dimension, the calculation formula is as follows: ； Where r represents the box ratio. For different r values, the information entropy is calculated as follows: ； in Indicates The probability of the phase space trajectory in a frame; C5-Time-frequency characteristics based on wavelet transform: According to the wavelet transform, seven features (C5–C11) were extracted as follows: ； in, The zero-sequence current waveform representing a system fault, t 0 express Initial time, Indicates the end time of the selected waveform. Indicates the time when the fault occurred. t w Indicates the end time of the fault transient process, express The middle time, express and The selected time between is used to distinguish the time-frequency characteristics of zero-sequence current before and after the transient process. express The peak time, express After that, the first half peak value of zero sequence current, represents the number of wavelet transform layers, Indicated in The zero sequence current in the cycle The energy of the wavelet transformed signal, Indicated in The zero sequence current in the cycle The energy of the wavelet transformed signal, U 0 Indicates the line voltage before the fault, M s express The maximum slope of It is a waveform of 1.1 cycles, including 0.1 cycles before the fault and 1 cycle after the fault.

3. A medium voltage power distribution single-phase grounding fault fusion identification model according to claim 2, characterized in that: In step S2, the steps of feature statistical analysis based on MANOVA are as follows: The ground fault data were collected from the fault indicators with a sampling rate of 4096 Hz assigned by the power supply company, including 61 cases of arrester breakdown, 61 cases of line-to-crossbar discharge, 61 cases of cable insulation damage, 55 cases of insulator breakdown, 59 cases of tree contact, and 145 cases of other types of ground faults. MANOVA was performed on the data of F1-F5, and the waveform features were ranked according to their F-statistic results. All 11 waveform features were selected as candidate fault features.

4. A medium voltage power distribution single-phase grounding fault fusion identification model according to claim 3, characterized in that: In step S3, the steps of constructing the insulation degradation type identification model based on ELM are as follows: S31: Define the ELM model parameters, take 11 waveform features as input, set the number of neurons in the input layer to 11, combine the empirical formula, set the number of neurons in the hidden layer to 13, set the number of neurons in the output layer to 6, the activation function of the hidden layer neurons is the Sigmoid function, and the output vector is processed by the Softmax function, converting the value of each element in the output vector to between 0 and 1; S32: Divide the training set and the test set, then input the feature vector data consisting of 11 waveform features, define the vector label of each insulation degradation type, use the PSO algorithm to optimize the initial weights and deviations, and complete the training and optimization of the model; S33: Evaluate model performance using test set data.

5. A medium voltage power distribution single-phase grounding fault fusion identification model according to claim 4, characterized in that: In step S4, a recognition model of insulation degradation type of single-phase grounding fault is established based on support vector machine SVM, which specifically includes the following steps: S41: normalizing the input vector containing nine waveform features; S42: Select radial basis function as kernel function; S43: Select the core parameters c and g of the model, where c represents the penalty factor, indicating the importance of abnormal data in the class, and g is the parameter coefficient of the kernel function; S44: Set the number of cross-validation times for the model to 10; S45: According to the optimal c and g obtained in step S43, the model in step S44 is trained using the svmtrain function in Libsvm, and the classification accuracy is tested using the svmppredict function in Libsvm.

6. A medium voltage power distribution single-phase grounding fault fusion identification model according to claim 5, characterized in that: In step S5, the specific steps are as follows: S51: The set of all possible outcomes of the identification frame decision problem is collectively called the identification frame O: ； S52: Select evidence and construct basic probability assignment (BPA); Setting the reliability factor to 0.8, the quality functions for the six insulation degradation types and three pieces of evidence (i=1,2,3) for the uncertainty θ are as follows: ； ； in, It is The reliability coefficient of the evidence; Representation Proposition No. BPA with evidence; Indicates the uncertainty BPA of evidence; Y ij Indicates Fault type The diagnostic output probability of the evidence; S53: Evidence fusion combines the recognition results of ELM and SVM and calculates BPA based on Dempster combination rule: ； Where A, B and C are three subsets of the recognition frame O, and K is a normalization constant; S54: Decision diagnostics uses two rules to determine the type of insulation degradation: Rule 1: ; Rule 2: ; in F type It represents the insulation degradation type finally determined by combining the identification results of ELM and SVM; Rule 1 stipulates that the determined insulation degradation type has the highest BPA; Rule 2 stipulates that the BPA of the determined insulation degradation type is greater than the uncertainty.

7. A medium voltage power distribution single-phase grounding fault fusion identification model according to claim 6, characterized in that: In step S6, based on the collected 442 SPGF data sets, the recognition performance of the ELM and SVM models is tested, the recognition effect of the fusion model based on the DS evidence theory is tested, and the recognition performance of different models is compared.

Images