An active power distribution network fault section identification method of characteristic current signal energy

Abstract

The application provides a kind of feature current signal energy active power distribution network fault section identification method, it is related to fault identification technical field, comprising: according to preliminary positioning result, using distributed optimization algorithm to analyze preliminary fault area, fault positioning task is divided into multiple sub-regional corresponding tasks according to power distribution network topological structure, and each sub-regional corresponding task is distributed to distributed node, each distributed node carries out optimization calculation of node current signal data in each sub-region according to each sub-regional corresponding task, obtains each sub-regional corresponding target value;Calculate the weight of each sub-region, the weighted average of the target value corresponding to all sub-regions is carried out, and the global final fault positioning result is generated.The method filters out abnormal nodes and locates them through global energy analysis, calculates the energy value of each node and identifies the region composed of multiple nodes, which avoids the error caused by independent calculation of nodes in traditional methods.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fault identification, and particularly relates to a feature current signal energy active power distribution network fault section identification method. BACKGROUND

[0002] Fault diagnosis and positioning of a power distribution network is a key problem in power system operation and maintenance, especially in large-scale power networks, and quickly and accurately locating the fault section is crucial for improving power supply reliability and reducing power outage time. Conventional fault detection methods mainly rely on current and voltage-based protection devices, but these methods have certain limitations in accuracy and response speed in complex network structures, especially in multi-fault scenarios, which include the following limitations and problems:

[0003] Fault positioning accuracy problem, conventional methods are difficult to achieve accurate fault section positioning, especially in complex power distribution network structure and various fault types, which is prone to misjudgment or omission.

[0004] Real-time and response speed problem, in conventional methods, the fault positioning process is relatively slow, which delays the fault recovery time.

[0005] Insufficient data processing and analysis capability, for the massive current signal data in the power distribution network, the conventional method is difficult to efficiently and timely process and obtain accurate fault judgment results. SUMMARY

[0006] The present application provides a feature current signal energy active power distribution network fault section identification method, which aims to solve the problem of fault positioning accuracy in related technologies, avoid real-time and response speed problems, and further avoid the problem of insufficient data processing and analysis capability.

[0007] In order to achieve the above purpose, the present application provides the following technical scheme:

[0008] A feature current signal energy active power distribution network fault section identification method, the method comprising the following steps:

[0009] Obtain the original current signal data of each node in the power distribution network, and pre-process the original current signal data.

[0010] According to the pre-processed current signal data, the energy value of each node is calculated by using a global energy analysis method, the area composed of multiple nodes is identified, the nodes with abnormal energy value change rate in the area are connected, and the preliminary fault area is obtained.

[0011] Local features are extracted from the preprocessed current signal data in the preliminary fault region, and weights are dynamically adjusted according to the correlation between the local features and the fault, to form a dynamically adjusted multi-dimensional feature vector.

[0012] The dynamically adjusted multi-dimensional feature vector is standardized to obtain a standardized multi-dimensional feature vector, and principal component analysis is used to reduce the dimension of the standardized multi-dimensional feature vector to obtain a reduced multi-dimensional feature vector.

[0013] The reduced multi-dimensional feature vector is fused and learned by using a deep neural network to output a fault type and a preliminary positioning result.

[0014] According to the preliminary positioning result, a distributed optimization algorithm is used to analyze the preliminary fault region, and the fault positioning task is divided into multiple sub-regional corresponding tasks according to the distribution network topology, and each sub-regional corresponding task is distributed to a distributed node, and each distributed node independently executes the optimization calculation of the node current signal data in each sub-region according to the sub-regional corresponding task, to obtain a target value corresponding to each sub-region.

[0015] The weights of each sub-region are calculated, and the target values corresponding to all sub-regions are weighted and averaged to generate a global final fault positioning result.

[0016] According to the final fault positioning result, the fault section in the distribution network topology is matched.

[0017] As a preferred scheme of the present application, the original current signal data of each node in the distribution network is obtained, and the original current signal data is preprocessed, and the specific steps are as follows:

[0018] The current sensor pre-installed at the distribution network node is used to obtain the original current signal data of each node.

[0019] The original current signal data of each node is removed from the high-frequency noise caused by external environmental interference and sensor error.

[0020] The denoised current signal data is normalized, and the normalized current signal data is segmented according to the time window to obtain preprocessed current signal data.

[0021] As a preferred scheme of the present application, according to the preprocessed current signal data, a global energy analysis method is used to calculate the energy value of each node, and the region composed of multiple nodes is identified, and the nodes with abnormal energy value change rate in the region are connected to obtain a preliminary fault region, and the specific steps are as follows:

[0022] A global energy analysis method is used to perform time integration on the preprocessed current signal data of each node to calculate the energy value of the current signal data of each node within the time window; based on the energy values ​​of multiple consecutive time windows, the rate of change of energy value of each node is calculated.

[0023] Determine whether the rate of change of energy value of each node in the region meets the anomaly judgment criteria; if the anomaly judgment criteria are met, it means that the node is a node with an abnormal rate of change of energy value.

[0024] Based on the topology of the distribution network, nodes with abnormal energy value change rates within the region are connected to obtain the preliminary fault area.

[0025] As a preferred embodiment of the present invention, local features are extracted from the preprocessed current signal data in the initial fault region to obtain local features. The weights are dynamically adjusted according to the correlation between the local features and the fault to form a dynamically adjusted multidimensional feature vector. The specific steps are as follows:

[0026] Local features are extracted from the preprocessed current signal data in the initial fault region to obtain multiple local features. Correlation analysis is then performed between these local features and the fault, and the dynamic adjustment weights corresponding to each local feature are calculated using the following formula:

[0027]

[0028] In the formula, Represents the i-th local feature F i The mutual information value between fault L and fault L is used to measure the i-th local feature F. i Correlation with fault L; W fi This represents the dynamically adjusted weight of the correlation corresponding to the i-th local feature.

[0029] The weights are dynamically adjusted based on the correlations corresponding to each local feature, and the dynamically adjusted weights w for each local feature are calculated. i The formula is:

[0030]

[0031] In the formula, w i W represents the dynamically adjusted weight of the i-th local feature. fi Z represents the dynamic adjustment weight of the correlation corresponding to the i-th local feature, and Z represents the total number of local features.

[0032] We perform weighted processing on each local feature and its dynamically adjusted weights to obtain multiple weighted features, represented as follows:

[0033]

[0034] wherein F i represents the weighted i-th weighted feature, F i represents the original i-th local feature.

[0035] All the weighted weighted features are combined to form a dynamically adjusted multi-dimensional feature vector, which is represented as:

[0036]

[0037] wherein F i represents the weighted i-th weighted feature, represents the dynamically adjusted multi-dimensional feature vector.

[0038] As a preferred scheme of the present application, a standardized multi-dimensional feature vector is obtained by standardizing the dynamically adjusted multi-dimensional feature vector, and a reduced multi-dimensional feature vector is obtained by reducing the dimension of the standardized multi-dimensional feature vector using principal component analysis, and the specific steps are as follows:

[0039] The dynamically adjusted multi-dimensional feature vector is standardized to eliminate the dimensional and range differences of the feature vector, and all the standardized feature vectors are combined to form a standardized multi-dimensional feature vector.

[0040] The standardized multi-dimensional feature vector is subjected to principal component analysis, the main components are selected through eigenvalue decomposition, and a dimension reduction matrix is generated.

[0041] The standardized multi-dimensional feature vector is subjected to linear transformation using the dimension reduction matrix to generate a reduced multi-dimensional feature vector.

[0042] As a preferred scheme of the present application, the deep neural network includes an input layer, a hidden layer and an output layer; the input layer inputs the reduced multi-dimensional feature vector; the hidden layer realizes the fusion and learning between multi-dimensional features through a nonlinear activation function; and the output layer outputs the fault type classification and preliminary positioning results.

[0043] As a preferred scheme of the present application, each distributed node independently performs optimization calculation of the node current signal data in the sub-region according to the corresponding task of each sub-region to obtain the target value corresponding to each sub-region, and the specific calculation formula is as follows:

[0044]

[0045] wherein T k represents the target value corresponding to the k-th sub-region, n k represents the total number of nodes in the k-th sub-region, I mi and I pirespectively represent the actual current value and the predicted current value of the i th node.

[0046] As a preferred scheme of the present application, the weight of each sub-region is calculated, and the target values corresponding to all sub-regions are weighted and averaged to generate a global final fault location result, and the specific steps are as follows:

[0047] The weight of each sub-region is calculated based on the target value corresponding to each sub-region, and the calculation formula is:

[0048]

[0049] In the formula, W k represents the weight of the k th sub-region, K represents the total number of sub-regions, T j represents the target value of the j th sub-region, T k represents the target value of the k th sub-region.

[0050] The target values corresponding to all sub-regions and the weights are summarized, and the global final fault location result is calculated by weighted average calculation, and the calculation formula is:

[0051]

[0052] In the formula, T k represents the target value corresponding to the k th sub-region, R represents the final fault location result, K represents the total number of sub-regions, and the final fault location result is calculated according to the calculation, and k represents the sub-region number currently calculated.

[0053] As a preferred scheme of the present application, according to the final fault location result, the fault section in the power distribution network topology structure is matched, and the specific steps are as follows:

[0054] The final fault location result is compared with the target value T i of each region in the power distribution network topology structure, and the comparison calculation formula is:

[0055]

[0056] In the formula, M i represents the matching degree of the final fault location result and the i th region, T i represents the characteristic target value of the i th region, which is obtained by statistical analysis of the node data in the region; σ i represents the standard deviation of the target value of the i th region, which is used to reflect the consistency of the nodes in the region.

[0057] All the calculated matching degrees are sorted, and the region with the smallest matching degree is selected, and the calculation formula is:

[0058]

[0059] wherein i min represents the region with the minimum matching degree, Mi represents the matching degree between the final fault location result and the ith region, min represents the minimum operator, represents that i is an index in the integer set from 1 to N.

[0060] The region i with the minimum matching degree is determined as the fault section, and the section number, node number and position coordinates corresponding to the fault section are recorded. min The region i with the minimum matching degree is determined as the fault section, and the section number, node number and position coordinates corresponding to the fault section are recorded.

[0061] The beneficial effects of the present application are: through the global energy analysis method, the energy values of each node are calculated and the region composed of multiple nodes is identified, avoiding the error caused by independent calculation of nodes in the traditional method. The nodes with abnormal energy value change rate in the region are connected to screen out the initial fault region, improve the accuracy of fault location, and reduce the occurrence of misjudgment and omission; the distributed optimization algorithm is used to decompose the fault location task into multiple subtasks, and the optimization calculation is performed in parallel in the distributed nodes, shortening the fault location time; compared with the slow fault location process in the conventional method, the present application improves the real-time response ability through parallel calculation, and reduces the fault recovery time.

[0062] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the following preferred embodiments are specifically described below, and the accompanying drawings are described in detail as follows. BRIEF DESCRIPTION OF DRAWINGS

[0063] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments of the present application. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained without creative labor on the basis of these drawings.

[0064] Figure 1 A flow chart of a feature current signal energy active power distribution network fault section identification method provided by the embodiments of the present application. DETAILED DESCRIPTION

[0065] The technical solutions in the embodiments of the present application will be described below in combination with the drawings in the embodiments of the present application.

[0066] Please refer to Figure 1 , Figure 1 A flow chart of a feature current signal energy active power distribution network fault section identification method provided by the embodiments of the present application.

[0067] In the embodiment, the active power distribution network fault section identification method of a characteristic current signal energy comprises steps S101, S102, S103, S104, S105, S106 and S107.

[0068] In step S101, original current signal data of each node in the power distribution network is acquired, and the original current signal data is preprocessed, and the specific steps are as follows:

[0069] The original current signal data of each node is acquired based on a current sensor preset at a node in the power distribution network.

[0070] The original current signal data of each node is removed from high-frequency noise caused by external environmental interference and sensor errors.

[0071] The denoised current signal data is normalized, and the normalized current signal data is segmented according to a time window to obtain preprocessed current signal data.

[0072] It should be noted that the acquired original current signal data includes transient and steady-state data of three-phase current, the original current signal data is sampled at a fixed time interval, and the sampling frequency is higher than the upper limit of the characteristic frequency of the fault signal.

[0073] It should be further noted that the denoising processing adopts a wavelet transform method, which filters out high-frequency noise by decomposing the multi-scale components of the original current signal data, and the high-frequency noise is defined as: the current signal data exceeding the range of 1 kHz to 10 kHz is regarded as high-frequency noise.

[0074] It should be further noted that the normalization formula is:

[0075]

[0076] In the formula, I' is the normalized current signal data, I min and I max are the minimum and maximum values of the current signal data of the current node, respectively.

[0077] It should be further noted that the segment length is dynamically set according to the characteristic period of the fault current signal data.

[0078] In step S102, the energy values of each node are calculated according to the preprocessed current signal data by using a global energy analysis method, the area composed of multiple nodes is identified, the nodes with abnormal energy value change rates in the area are connected, and a preliminary fault area is obtained, and the specific steps are as follows:

[0079] The global energy analysis method is used to time-integrate the preprocessed current signal data of each node to calculate the energy value of the current signal data of each node in a time window; and the energy value variation rate of each node is calculated based on the energy values of multiple continuous time windows.

[0080] It should be noted that the global energy analysis method is used to time-integrate the preprocessed current signal data of each node to calculate the energy value of the current signal data of each node in a time window, and the calculation formula is:

[0081]

[0082] In the formula, E i represents the energy value of the current signal data of the i th node in the time window , I i (t) represents the amplitude of the current signal data of the i th node at time t, t 0, t1 respectively represent the start time and end time of the time window in the global energy analysis, and dt represents the small time increment of integration, i.e. the change amplitude of the time variable t.

[0083] It is determined whether the energy value variation rate of each node in the region meets the abnormality determination condition; if the abnormality determination condition is met, the node is an energy value variation rate abnormal node.

[0084] It should be noted that the energy value variation rate of each node is established based on the time-integrated energy values of multiple continuous time windows, and the node energy value variation rate calculation formula is:

[0085]

[0086] In the formula, ΔE i represents the energy value variation rate of the i th node in the continuous time windows j and j+1, E i,j and E i,j+1 respectively represent the time-integrated energy values of the i th node in the time windows j and j+1.

[0087] According to the node energy value variation rate calculation result, it is determined whether the energy value variation rate ΔE i of each node meets the abnormality determination condition; if the abnormality determination condition is met, the node is an energy value variation rate abnormal node; if the abnormality determination condition is not met, the node is a normal node; and the abnormality determination condition is:

[0088]

[0089] In the formula, ΔE i represents the energy value variation rate of the i th node, σ represents the average energy value change rate of all nodes in the region ΔE σ represents the standard deviation of the energy value change rate of all nodes in the region, k represents a hyperparameter for adjusting the sensitivity of anomaly judgment, and the value range is set to 1.5 to 3.

[0090] It should be noted that when the energy value change rate ΔE i of a node exceeds the average energy value change rate of all nodes in the region k times the standard deviation σ ΔE , the node is determined to be an energy value change rate abnormal node.

[0091] Based on the topology of the power distribution network, the region connected by multiple energy value change rate abnormal nodes is identified as the preliminary fault region.

[0092] Further, by the topology of the power distribution network, i.e. the connection relationship between nodes and lines, each node in the power distribution network is connected into a graph; the connection between nodes reflects the physical connection of power lines and transformers; the power distribution network topology is represented by a network graph, where nodes represent each element in the power distribution network, and edges represent the power transmission lines between them; for each energy value change rate abnormal node, find out which adjacent nodes it is connected to based on the node connection relationship in the power distribution network topology.

[0093] Specifically, the connection between nodes is represented by an adjacency matrix or adjacency list; if node i and node j are connected and they both belong to energy value change rate abnormal nodes, then these two nodes are considered to form part of a candidate fault section; and based on this rule, all abnormal nodes and their adjacent nodes are traversed to identify a group of connected abnormal nodes, forming a preliminary fault region; here, a breadth-first search is used to identify a connected subgraph composed of all connected energy value change rate abnormal nodes, and the connected subgraph is the preliminary fault region; the breadth-first search starts from a node and traverses all directly or indirectly connected abnormal nodes layer by layer.

[0094] Step S103, local feature extraction is performed on the preprocessed current signal data in the preliminary fault region to obtain local features, and the weight is dynamically adjusted according to the correlation between the local features and the fault to form a dynamically adjusted multi-dimensional feature vector, the specific steps are as follows:

[0095] Local feature extraction is performed on the preprocessed current signal data in the preliminary fault region to obtain multiple local features, and correlation analysis is performed on the multiple local features and the fault to calculate the correlation dynamic adjustment weight corresponding to each local feature, and the calculation formula is:

[0096]

[0097] In the formula, represents the ith local feature F i and the failure L, for measuring the correlation between the ith local feature F i and the failure L; W fi represents the relevance dynamic adjustment weight corresponding to the ith local feature.

[0098] According to the relevance dynamic adjustment weight corresponding to each local feature, the weight w i of each local feature after dynamic adjustment is calculated, and the formula is:

[0099]

[0100] wherein, w i represents the weight of the ith local feature after dynamic adjustment, W fi represents the relevance dynamic adjustment weight corresponding to the ith local feature, and Z represents the total number of local features;

[0101] The weight of each local feature and each feature after dynamic adjustment is weighted to obtain a plurality of weighted features, denoted as:

[0102]

[0103] wherein, F i ' represents the ith weighted feature after weighting, F i represents the original ith local feature.

[0104] All weighted weighted features are combined to form a multi-dimensional feature vector after dynamic adjustment, and the multi-dimensional feature vector after dynamic adjustment is denoted as:

[0105]

[0106] wherein, F i ' represents the ith weighted feature after weighting, represents the multi-dimensional feature vector after dynamic adjustment.

[0107] It should be noted that the local features include time domain features, frequency domain features and energy features.

[0108] Time domain feature extraction, statistical analysis is performed on the current signal of the preliminary fault area, and the mean, variance and peak factor of the signal are extracted as time domain features, which are used to represent the amplitude fluctuation characteristics of the signal.

[0109] Frequency domain feature extraction, the current signal is converted from time domain to frequency domain through fast Fourier transform, and the main frequency amplitude, harmonic content and frequency center are extracted as frequency domain features, which are used to analyze the frequency component distribution of the signal.

[0110] Energy feature extraction, the energy of the current signal in the time domain and the frequency domain is calculated, the total energy and the high frequency energy proportion are obtained as the energy feature, and the energy feature is used to reflect the distribution and change of the signal energy.

[0111] In step S104, the standardized multi-dimensional feature vector is obtained by performing standardization processing on the dynamically adjusted multi-dimensional feature vector, and the dimensionality-reduced multi-dimensional feature vector is obtained by performing principal component analysis on the standardized multi-dimensional feature vector. The specific steps are as follows:

[0112] The standardized multi-dimensional feature vector is obtained by performing standardization processing on the dynamically adjusted multi-dimensional feature vector, eliminating the dimensional and range differences of the feature vector, and combining all the standardized feature vectors to form the standardized multi-dimensional feature vector.

[0113] It should be noted that the standardization processing adopts z-score standardization, according to which all features are ensured to be within the same range, and a reference is provided for subsequent dimensionality reduction processing, avoiding that a specific feature has too great an impact on the analysis result due to dimensional differences.

[0114] It should be further noted that the standardized feature vectors are combined to form a new multi-dimensional feature vector, which contains the standardized feature vectors of each node in each dimension and can fully reflect the state characteristics of the node. The standardized multi-dimensional feature vector is represented as:

[0115]

[0116] In the formula, are the standardized feature vectors, respectively, n is the total number of features, represents the standardized feature vector.

[0117] After obtaining the standardized feature vector, the covariance matrix is calculated to provide input for principal component analysis.

[0118] The principal component analysis is performed on the standardized multi-dimensional feature vector, the main components are selected through eigenvalue decomposition, and the dimensionality reduction matrix is generated.

[0119] It should be noted that the principal component analysis method is used to perform eigenvalue decomposition on the covariance matrix to obtain the decomposed eigenvalues and eigenvectors. The eigenvalues reflect the variance contribution size of each principal component, and the eigenvectors represent the direction of the corresponding principal component. The main components with a cumulative contribution rate of 90% to 95% are selected to construct the dimensionality reduction matrix P.

[0120] The dimensionality reduction matrix is used to perform linear transformation on the standardized multi-dimensional feature vector to generate the dimensionality-reduced multi-dimensional feature vector.

[0121] It should be noted that the reduced dimension matrix is used to linearly transform the normalized multi-dimensional feature vector to generate a reduced multi-dimensional feature vector, specifically as follows:

[0122]

[0123] In the formula, P is a reduced dimension matrix, which is constructed from the selected principal components in the eigenvectors of the covariance matrix; represents the normalized multi-dimensional feature vector, represents the reduced multi-dimensional feature vector.

[0124] Step S105, the deep neural network includes an input layer, a hidden layer and an output layer; the input layer inputs the reduced multi-dimensional feature vector; the hidden layer realizes the fusion and learning between multi-dimensional features through a nonlinear activation function; the output layer simultaneously completes the output of the fault type classification and the preliminary positioning result, and the specific steps are as follows:

[0125] The reduced multi-dimensional feature vector obtained in step S104 is taken as the input data of the deep neural network, and the form of the input data is as follows:

[0126]

[0127] In the formula, represents the input reduced multi-dimensional feature vector, represents the k-dimensional feature vector obtained after reduction, and k is the dimension of the feature vector after reduction.

[0128] It should be noted that the input layer receives the reduced multi-dimensional feature vector, and these input data are transmitted to the hidden layer through the neural network; in the input layer, each node represents a feature value, and these feature values are transmitted to the hidden layer of the neural network; the hidden layer fuses the input multi-dimensional features through a nonlinear activation function, the activation function adopts ReLU, and through the processing of multiple hidden layers, the neural network can learn the complex nonlinear relationship between the input features; and through the activation function, the neural network can capture the relevance between different features, optimize its internal representation, and make the subsequent classification and regression tasks more accurate; the output layer is responsible for two tasks, one of which is to use a soft max function classification method for multi-class classification and output the category of the fault; the other is to output the position of the fault, i.e. the identification of the fault section, which is a regression task and outputs a continuous value representing the specific position or relative position of the fault.

[0129] It should be further noted that the deep neural network is trained using labeled data, wherein the labels include classification labels of fault types and regression labels of fault positions; the training data set contains multiple samples, and each sample is composed of an input feature vector and corresponding fault type and fault position labels.

[0130] After the training is completed, the neural network outputs the prediction result of the fault type according to the input reduced dimension feature vector, and obtains the specific category of the fault through the output of the classification layer; meanwhile, the neural network outputs the preliminary positioning result of the fault, i.e., the fault section, through the regression task, and this result is expressed as a specific section identifier, which is used for subsequent fault processing.

[0131] In step S106, according to the preliminary positioning result, a distributed optimization algorithm is used to analyze the preliminary fault region, the fault positioning task is divided into tasks corresponding to multiple sub-regions according to the topology of the distribution network, and each task corresponding to a sub-region is distributed to a distributed node, each distributed node independently performs optimization calculation of the node current signal data in the sub-region according to the task corresponding to the sub-region, and obtains the target value corresponding to each sub-region, and the specific calculation formula is:

[0132]

[0133] In the formula, T k represents the target value corresponding to the kth sub-region, n k represents the total number of nodes in the kth sub-region, I mi and I pi represent the actual current value and the predicted current value of the ith node, respectively.

[0134] It should be noted that according to the preliminary positioning result, when the distributed optimization algorithm is used, the basis for dividing the fault region into sub-regions is the topology of the distribution network, the physical location of the nodes and the current signal characteristics, which specifically includes:

[0135] According to the electrical connection relationship of the distribution network, the nodes in the region are preferentially divided into sub-regions to reduce the calculation burden across regions; according to the distribution characteristics of the nodes in the fault region, the nodes close to the center of the fault are preferentially divided into separate sub-regions to improve the efficiency of distributed optimization; if the current signal of some nodes fluctuates sharply, they are classified into independent sub-regions to more accurately analyze the influence of the independent sub-regions on fault positioning.

[0136] It should be noted that in step S105, the prediction result of the fault type and the preliminary positioning result are obtained through the deep neural network; according to the preliminary positioning result, the analysis task of the fault section is further decomposed into tasks corresponding to multiple sub-regions; the initial fault region of the distribution network is divided into multiple sub-regions; each sub-region corresponds to a distributed node for further optimization calculation; the basis for dividing each sub-region is the spatial distribution of the fault region and the relative position of the fault point; if the preliminary positioning result shows that the fault is located in the center region of a specific section, the region is divided into multiple small regions according to its characteristics to make the distributed calculation more efficient.

[0137] Step S107, the weight of each sub-region is calculated, and the target values corresponding to all sub-regions are weighted and averaged to generate a global final fault location result, the specific steps are as follows:

[0138] The weight of each sub-region is calculated based on the target value corresponding to each sub-region, and the calculation formula is:

[0139]

[0140] In the formula, W k represents the weight of the kth sub-region, K represents the total number of sub-regions, T j represents the target value of the jth sub-region, T k represents the target value of the kth sub-region.

[0141] The target values corresponding to all sub-regions and the weights are summarized, and the global final fault location result is generated by weighted average calculation, and the calculation formula is:

[0142]

[0143] In the formula, T k represents the target value corresponding to the kth sub-region, R represents the final fault location result, K represents the total number of sub-regions, and the final fault location result is calculated according to the calculation, and k represents the sub-region number currently calculated.

[0144] Step S108, according to the final fault location result, match the fault section in the power distribution network topology structure, the specific steps are as follows:

[0145] The final fault location result is compared with the target value T i of each region in the power distribution network topology structure, and the comparison calculation formula is:

[0146]

[0147] In the formula, M i represents the matching degree of the final fault location result and the ith region, T i represents the characteristic target value of the ith region, which is obtained by statistical analysis of the node data in the region; σ i represents the target value standard deviation of the ith region, which is used to reflect the consistency of the nodes in the region.

[0148] All the calculated matching degrees are sorted, and the region with the smallest matching degree is selected, and the calculation formula is:

[0149]

[0150] In the formula, i min represents the region with the smallest matching degree, M irepresents the matching degree of the final fault location result and the ith region, and min represents the minimum operator, represents that i is an index in the integer set from 1 to N.

[0151] The region i with the minimum matching degree min is determined as the fault section, and the fault section corresponding section number, node number and position coordinates are recorded.

[0152] So far, a kind of active power distribution network fault section identification method of characteristic current signal energy is completed.

[0153] The above is only a specific embodiment of the present application, but the protection scope of the embodiments of the present application is not limited to this, any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the embodiments of the present application, which should be covered within the protection scope of the embodiments of the present application. Therefore, the protection scope of the embodiments of the present application should be subject to the protection scope of the claims.

Claims

1. A method for active power distribution network fault section identification featuring current signal energy, characterized in that, The method comprises the following steps: Obtain the original current signal data of each node in the power distribution network, and pre-process the original current signal data; According to the pre-processed current signal data, calculate the energy value of each node using a global energy analysis method, identify a region composed of multiple nodes connected, connect the nodes with abnormal energy value change rate in the region, and obtain a preliminary fault region; Perform local feature extraction on the pre-processed current signal data in the preliminary fault region to obtain local features, dynamically adjust the weights according to the correlation between the local features and the faults, and form a dynamically adjusted multi-dimensional feature vector; According to the dynamically adjusted multi-dimensional feature vector, perform standardization processing to obtain a standardized multi-dimensional feature vector, and perform dimension reduction on the standardized multi-dimensional feature vector using principal component analysis to obtain a reduced multi-dimensional feature vector; Fuse and learn the reduced multi-dimensional feature vector using a deep neural network, and output the fault type and preliminary positioning result; According to the preliminary positioning result, analyze the preliminary fault region using a distributed optimization algorithm, divide the fault positioning task into multiple sub-region corresponding tasks according to the power distribution network topology, and distribute each sub-region corresponding task to a distributed node. Each distributed node independently performs optimization calculation on the node current signal data in the sub-region according to the sub-region corresponding task to obtain the target value of each sub-region. The specific calculation formula is: where T k denotes the target value corresponding to the kth sub-region, n k denotes the total number of nodes in the kth sub-region, I mi and I pi denote the actual current value and the predicted current value of the ith node, respectively; Calculate the weight of each sub-region, and perform weighted average on the target values of all sub-regions to generate the global final fault positioning result, which specifically includes: Calculate the weight of each sub-region based on the target value of each sub-region. The calculation formula is: wherein W k denotes the weight of the kth sub-region, K denotes the total number of sub-regions, T j denotes the target value of the jth sub-region, T k denotes the target value of the kth sub-region; Summarize the target values and weights of all sub-regions, and generate the global final fault positioning result by weighted average calculation. The calculation formula is: In the formula, T k The target value corresponding to the kth sub-region is represented by R, and K represents the total number of sub-regions. The final fault location result is calculated according to the calculation, and k represents the sub-region number currently calculated. According to the final fault positioning result, match the fault section in the power distribution network topology.

2. A method of active power distribution network fault section identification featuring current signal energy as claimed in claim 1 characterized in that, Obtain the original current signal data of each node in the power distribution network, and pre-process the original current signal data, specifically as follows: Obtain the original current signal data of each node based on the pre-set current sensor at the node of the power distribution network; Remove the high-frequency noise caused by external environmental interference and sensor errors from the original current signal data of each node; Perform normalization processing on the denoised current signal data, and segment the normalized current signal data according to the time window to obtain the pre-processed current signal data.

3. A method of active power distribution network fault section identification featuring current signal energy as claimed in claim 2, characterized in that, According to the pre-processed current signal data, calculate the energy value of each node using a global energy analysis method, identify a region composed of multiple nodes connected, connect the nodes with abnormal energy value change rate in the region, and obtain a preliminary fault region, specifically as follows: Using a global energy analysis method, perform time integration on the pre-processed current signal data of each node to calculate the energy value of each node current signal data in the time window; Based on the energy values of multiple consecutive time windows, calculate the energy value change rate of each node; Determine whether the energy value change rate of each node in the region meets the abnormal judgment condition; If the abnormality determination condition is met, it indicates that the node is an energy value rate of change abnormal node; Based on the topology structure of the power distribution network, an area connected by a plurality of energy value rate of change abnormal nodes is identified as a preliminary fault area.

4. A method of active power distribution network fault section identification featuring current signal energy as claimed in claim 1 characterized in that, The local feature is extracted from the preprocessed current signal data in the preliminary fault area, and the weight is dynamically adjusted according to the correlation between the local feature and the fault, and a dynamically adjusted multi-dimensional feature vector is formed, and the specific steps are as follows: The local feature is extracted from the preprocessed current signal data in the preliminary fault area, and the local feature is associated with the fault, and the correlation dynamic adjustment weight corresponding to each local feature is calculated, and the formula is: wherein, denotes the ith local feature F i and the mutual information value of the ith local feature F i with the fault L; W fi denotes the relevance dynamic adjustment weight corresponding to the ith local feature. According to the relevance of each local feature, the weight is dynamically adjusted to calculate the dynamic adjusted weight W of each local feature i , the formula is: wherein W i denotes the dynamic adjusted weight of the ith local feature, W fi denotes the dynamic adjusted weight of the ith local feature, Z denotes the total number of local features; The local feature is extracted from the preprocessed current signal data in the preliminary fault area, and the local feature is associated with the fault, and the correlation dynamic adjustment weight corresponding to each local feature is calculated, and the formula is: where F i denotes the i-th weighted feature after weighting, F i denotes the i-th original local feature; The local feature is extracted from the preprocessed current signal data in the preliminary fault area, and the local feature is associated with the fault, and the correlation dynamic adjustment weight corresponding to each local feature is calculated, and the formula is: where F i denotes the weighted i-th weighted feature, denotes the dynamically adjusted multi-dimensional feature vector.

5. A method of identifying a faulted section of an active power distribution network characterized by a signature current signal energy as recited in claim 1, wherein, All weighted features are combined to form a dynamically adjusted multi-dimensional feature vector, and the dynamically adjusted multi-dimensional feature vector is represented as: According to the dynamically adjusted multi-dimensional feature vector, the standardization processing is carried out, and the standardized multi-dimensional feature vector is obtained, and the principal component analysis is used to reduce the dimension of the standardized multi-dimensional feature vector, and the reduced multi-dimensional feature vector is obtained, and the specific steps are as follows: The dynamically adjusted multi-dimensional feature vector is standardized to eliminate the dimension and range difference of the feature vector, and all standardized feature vectors are combined to form a standardized multi-dimensional feature vector; The principal component analysis is performed on the standardized multi-dimensional feature vector, the main components are selected through eigenvalue decomposition, and a dimension reduction matrix is generated; 6. A method of identifying a faulted section of an active power distribution network characterized by a signature current signal energy as defined in claim 1, characterized in that, The dimension reduction matrix is used to linearly transform the standardized multi-dimensional feature vector to generate a reduced multi-dimensional feature vector. The deep neural network includes an input layer, a hidden layer and an output layer; the input layer inputs the reduced multi-dimensional feature vector; the hidden layer realizes the fusion and learning between multi-dimensional features through a nonlinear activation function; 7. A method of identifying a faulted section of an active power distribution network characterized by a signature current signal energy as defined in claim 1, wherein, The output layer completes the output of the fault type classification and the preliminary positioning result. The final fault location result is compared with the target value T corresponding to each region in the power distribution network topology i A comparison calculation is performed, and the comparison calculation formula is: wherein M i represents the matching degree of the final fault location result and the ith region, T i represents the characteristic target value of the ith region, which is obtained by statistical analysis of the node data within the region; σ i σi represents the standard deviation of the target value of the i-th region, and is used to reflect the consistency of the nodes in the region. According to the final fault positioning result, the fault section in the power distribution network topology structure is matched, and the specific steps are as follows: All the calculated matching degrees are sorted, and the area with the smallest matching degree is selected, and the formula is: where i min denotes the region with the minimum matching degree, M i denotes the matching degree of the final fault location result with the i-th region, min denotes the operator of minimum value, denotes i is an index in the set of integers from 1 to N; The region i with the minimum matching degree min Determine the fault section, and record the section number, node number and position coordinates corresponding to the fault section.

Images