Centralized FA fault section positioning method and system in multi-energy scene

By adopting a centralized FA fault segment positioning method in multi-energy scenarios, using the phase mode transformation, wavelet packet decomposition, quantitative encoding and deep learning models to identify the fault segments, the complexity of fault characteristics and poor positioning effect in the distribution network are solved. The problem of this problem achieves high accuracy and adaptability of fault location.

CN118937883APending Publication Date: 2024-11-12ELECTRIC POWER RES INST OF GUANGXI POWER GRID CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410854657.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-06-28
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

In multi-energy scenarios, the fault characteristics of distributed power distribution networks are complex, and the fault characteristics of medium and low voltage distribution networks are weak, resulting in the adaptability of traditional fault segment positioning algorithms, and the failure segment cannot be accurately located. Based on threshold adjustment or empirical judgment The method is difficult to cover all fault conditions and the positioning effect is poor.

Method used

The centralized FA fault segment positioning method in multi-energy scenarios is adopted, and the phase mode transformation is obtained by obtaining three-phase voltage and current, wavelet packet decomposition and sorting, time frequency matrix is ​​constructed, quantized encoding and rainbow encoding processing is performed, and fault color images are generated. It is also used to identify the deep learning model (AlexNet-SVM) and output the fault segment label.

Benefits of technology

It improves the accuracy and adaptability of fault segment positioning, enhances the ability to identify weak fault features, improves the anti-interference and generalization capabilities of the distribution network, and achieves the data-driven positioning effect without manual intervention.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118937883A_ABST
    Figure CN118937883A_ABST
Patent Text Reader

Abstract

The invention discloses a centralized FA fault section positioning method in a multi-energy scene, and the method comprises the steps: obtaining fault information which comprises a three-phase voltage and a three-phase current, carrying out the phase-mode conversion of the three-phase current, and obtaining a phase-mode current; performing wavelet packet decomposition on the electrical signals of the three-phase voltage and the phase-mode current to obtain a fault sub-band of each electrical signal, sorting the fault sub-bands of each electrical signal, constructing a time-frequency matrix of each node, and sorting the time-frequency matrix of each node to obtain a time-frequency matrix of the power distribution network; performing quantization coding on the time-frequency matrix of the power distribution network to obtain a pixel matrix, and performing pseudo-color processing on the pixel matrix by using rainbow coding to obtain a fault color image; and identifying the fault color image by using a deep learning model, and outputting a fault color image label as a positioning result, so that an excellent positioning effect can be kept in an actual power distribution network scene.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of power system distribution network automation, and in particular to a centralized FA fault section positioning method and system in a multi-energy scenario. Background Art

[0002] Feeder Automation (FA) is a technical system that uses modern communication, control and protection technologies to automatically control the distribution network. It is mainly used on the power supply side of the distribution network to achieve automatic control and protection of the feeder and improve the reliability, safety and economy of the power grid. According to different control principles, it can be roughly divided into three categories: local FA, centralized FA, and distributed FA. The centralized FA is composed of primary equipment, distribution terminal unit FTU (Feeder Terminal Unit, distribution switch monitoring terminal), communication system, and distribution master station. It is based on the communication network, with centralized control as the core, and integrates multiple control modes such as current protection, distribution terminal remote control, and reclosing to complete the rapid positioning of the fault section.

[0003] At present, distributed multi-energy scenarios have become a trend in distribution networks in new power systems. However, in multi-energy scenarios, the fault characteristics of distributed generation (DG) distribution networks become more complex. At the same time, the fault characteristics in medium and low voltage distribution networks are relatively weak, resulting in a decrease in the adaptability of the fault section location algorithm of the traditional distribution network, and the inability to accurately locate the fault section. In addition, the fault section location method based on threshold setting or empirical judgment is also difficult to cover all fault conditions, and the positioning effect is not good. Therefore, studying the fault section location algorithm that can adapt to all fault conditions with complex and weak fault characteristics in multi-energy scenarios has become a current research hotspot. Summary of the invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is: the present invention solves the problem that the fault characteristics of distributed power distribution networks have become more complicated, and the fault characteristics in medium and low voltage distribution networks are weak, resulting in a decrease in the adaptability of the fault section location algorithm of the traditional distribution network, and the inability to accurately locate the fault section. The fault section location method based on threshold setting or experience judgment is difficult to cover all fault conditions and the positioning effect is poor.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a centralized FA fault section location method in a multi-energy scenario, comprising:

[0008] Acquire fault information, the fault information including three-phase voltage and three-phase current, and perform phase-mode transformation on the three-phase current to obtain phase-mode current;

[0009] Decomposing the electrical signals of the three-phase voltage and the phase mode current by wavelet packets to obtain the fault sub-frequency band of each electrical signal, sorting the fault sub-frequency band of each electrical signal, constructing the time-frequency matrix of each node, sorting the time-frequency matrix of each node, and obtaining the time-frequency matrix of the distribution network;

[0010] Quantizing and encoding the time-frequency matrix of the distribution network to obtain a pixel matrix, and performing pseudo-color processing on the pixel matrix using rainbow coding to obtain a fault color image;

[0011] The fault color image is identified using a deep learning model, and a fault color image label is output as a positioning result.

[0012] As a preferred solution of the centralized FA fault section location method in the multi-energy scenario described in the present invention, wherein: performing phase mode transformation on the three-phase current includes:

[0013] Assume the three-phase current is i a 、i b 、i c ;

[0014] The three-phase current vector [i a ,i b ,i c ] T Multiplying with the Karenbauer transformation matrix K, the transformed vector is Among them, i α is the positive sequence component, i0 is the zero sequence component, and is obtained by matrix multiplication:

[0015]

[0016]

[0017]

[0018] The vector transformed by the Karenbauer transformation matrix is

[0019] As a preferred solution of the centralized FA fault section location method in the multi-energy scenario described in the present invention, wherein: quantizing and encoding the time-frequency matrix of the distribution network includes:

[0020] The distribution master station quantizes and encodes the time-frequency matrix of the distribution network, maps each element in the original data sequence to a set mapping range, and obtains a normalized linear mapping value by subtracting the minimum value in the original data sequence from the element in the original data sequence divided by the maximum value in the original data sequence minus the minimum value in the original data sequence. The normalized linear mapping is multiplied by the set maximum value of the mapping range to obtain a standardized linear mapping value.

[0021] The function value obtained by rounding off the value of the linear mapping standardized by the rounding function is the result value of the quantization encoding.

[0022] As a preferred solution of the centralized FA fault section location method in the multi-energy scenario described in the present invention, the pseudo-color processing of the pixel matrix using rainbow coding includes:

[0023] Use rainbow coding to convert each pixel of the pixel matrix, input each pixel into the rainbow coding function, output the color value, and the three primary color conversion function is:

[0024] The red transformation function includes three interval ranges;

[0025] If the pixel value is in [0, T1), the red transformation function value is 0; if the pixel value is in [T1, T2), the red transformation function value is the pixel value minus T1 multiplied by T3 divided by T2-T1; if the pixel value is in [T 2, T3], then the red transformation function value is T3;

[0026] The green transformation function includes five interval ranges;

[0027] If the pixel value is in [0, T4), the green transformation function value is 0; if the pixel value is in [T 4, If the pixel value is within [T5, T6), the green transformation function value is the pixel value minus T4 multiplied by T3, divided by T4; if the pixel value is within [T5, T6), the green transformation function value is T3; if the pixel value is within [T6, T7), the green transformation function value is T7 minus the pixel value multiplied by T3, divided by T7-T6; if the pixel value is within [T7, T3], the green transformation function value is the pixel value minus T7 multiplied by T3, divided by T3-T7.

[0028] As a preferred solution of the centralized FA fault section location method in the multi-energy scenario described in the present invention, it also includes:

[0029] The blue transformation function includes five interval ranges;

[0030] If the pixel value is in [0, T8), the blue transformation function value is the pixel value multiplied by T3 divided by T8; if the pixel value is in [T 8,T9), the blue transformation function value is T3; if the pixel value is between [T9, T 10 ), then the blue transformation function value is T 10 Subtract the pixel value multiplied by T3 divided by T 10 -T9; if the pixel value is between [T 10 ,T 11 ), the blue transformation function value is 0; if the pixel value is between [T 11 ,T3], then the blue transformation function value is the pixel value minus T 11 Multiply by T3, divide by T3-T 11 .

[0031] As a preferred solution of the centralized FA fault section positioning method in the multi-energy scenario described in the present invention, wherein: using the deep learning model to identify the fault color image includes:

[0032] The size of the fault color image is adjusted to a first threshold value to obtain a first fault color image, and the first fault color image is input into the AlexNet model. The first fault color image is forward propagated through a convolution layer, an activation function, and a pooling layer, and a fault image feature vector is output in a second fully connected layer of the AlexNet model;

[0033] The fault image feature vector is input into the SVM model. The SVM model classifies the input feature vector according to the decision boundary learned from the training data and identifies the segment code mark corresponding to the fault image.

[0034] The segment coding mark corresponding to the fault image is identified according to the SVM model to obtain the segment where the fault occurs.

[0035] As a preferred solution of the centralized FA fault section location method in the multi-energy scenario described in the present invention, the decision boundary includes:

[0036] The objective function is defined as maximizing the interval width between different fault image feature vectors. The objective function is converted into a Lagrangian function using the Lagrangian multiplier method. The Lagrangian function is subjected to a dual transformation to obtain the objective function of the dual problem. The objective function of the dual problem is solved using the KKT condition.

[0037] Derivative the objective function of the dual problem, set the derivative to 0, and solve to obtain the optimal Lagrange multiplier. In the solution process, there are non-zero Lagrange multipliers, and the corresponding sample points are support vectors.

[0038] The decision function is constructed using the support vectors and the corresponding non-zero Lagrange multipliers, and the decision boundary is the weight vector and the bias term.

[0039] In a second aspect, the present invention provides a system for centralized FA fault section location in a multi-energy scenario, comprising:

[0040] A phase-mode conversion module is used to obtain fault information, wherein the fault information includes three-phase voltage and three-phase current, and performs phase-mode conversion on the three-phase current to obtain phase-mode current;

[0041] A decomposition module, used to perform wavelet packet decomposition on the electrical signals of the three-phase voltage and the phase mode current to obtain a fault sub-frequency band of each electrical signal, sort the fault sub-frequency band of each electrical signal, construct a time-frequency matrix of each node, sort the time-frequency matrix of each node, and obtain a time-frequency matrix of the distribution network;

[0042] An encoding module, used for quantizing and encoding the time-frequency matrix of the distribution network to obtain a pixel matrix, and performing pseudo-color processing on the pixel matrix using rainbow coding to obtain a fault color image;

[0043] The recognition module is used to use a deep learning model to recognize the fault color image and output the fault color image label as a positioning result.

[0044] In a third aspect, the present invention provides a computing device, comprising:

[0045] Memory and processor;

[0046] The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions. When the computer executable instructions are executed by the processor, the steps of the centralized FA fault section locating method in the multi-energy scenario are implemented.

[0047] In a fourth aspect, the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the centralized FA fault section locating method in the multi-energy scenario.

[0048] Beneficial effects of the present invention: The centralized FA fault section location method under multi-energy scenarios proposed by the present invention uses Karenbauer transform to decouple the three-phase current during the information processing of the power distribution master station, which can amplify weak fault features; wavelet packet transform is used to extract fault sub-bands, which can enhance the difference of fault features; rainbow coding is used to convert the pixel matrix, which can make the fault color image rich in layers and fully retain the fault features; transfer learning based on model transfer is used to avoid the problems of small training data, difficult training, long training time, etc., and improve the actual application effect; AlexNet-SVM is used to identify fault color images, so that the classification effect is excellent and the accuracy of fault section location can be improved. The present invention locates the fault section in a data-driven form, without manual intervention, and has strong anti-interference and generalization capabilities, and can maintain excellent positioning effects in actual distribution network scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0050] Figure 1 The overall flow chart of a centralized FA fault section location method in a multi-energy scenario provided by the present invention;

[0051] Figure 2 A schematic diagram of the structure of a centralized FA system of a distribution network in a multi-energy scenario of a centralized FA fault section locating method in a multi-energy scenario provided by the present invention;

[0052] Figure 3 A photovoltaic low voltage ride-through control strategy diagram of a centralized FA fault section positioning method in a multi-energy scenario provided by the present invention;

[0053] Figure 4 A schematic diagram of a fault color image generated when a fault occurs in section 3 of a centralized FA fault section location method in a multi-energy scenario provided by the present invention;

[0054] Figure 5 This is a pre-training result diagram of a centralized FA fault section locating method in a multi-energy scenario provided by the present invention. DETAILED DESCRIPTION

[0055] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0056] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0057] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0058] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0059] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0060] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0061] Example 1

[0062] Reference Figure 1-3, is an embodiment of the present invention, and provides a centralized FA fault section location method in a multi-energy scenario, comprising:

[0063] S100: Acquire fault information, which includes three-phase voltage and three-phase current, and perform phase-mode transformation on the three-phase current to obtain phase-mode current;

[0064] In a possible embodiment, let the three-phase current be i a 、i b 、i c ;

[0065] The three-phase current vector [i a ,i b ,i c ] T Multiplying with the Karenbauer transformation matrix K, the transformed vector is Among them, i α is the positive sequence component, i0 is the zero sequence component, and is obtained by matrix multiplication:

[0066]

[0067]

[0068]

[0069] The vector transformed by the Karenbauer transformation matrix is

[0070] Specifically, a 10kV feeder is derived from the substation busbar, and a circuit breaker CB1 is provided at the outlet as an outgoing line switch. The feeder includes 8 nodes, which are divided into 7 sections. Figure 2 As shown, the short vertical thick black lines represent nodes, not busbars. Each section head is equipped with a section switch FSn (FS1, FS2, FS3, FS4, FS4, FS6, FS7), and the two tail ends of the feeder are equipped with a connecting switch, namely FD5 and FD8. Section switches, circuit breakers, and connecting switches are all referred to as "switches". During normal operation, except for the connecting switch in the open state, the section switches and circuit breakers are in the closed state. The switch is equipped with a distribution terminal unit FTU, and the distribution terminal unit FTU maintains a communication connection with the system's distribution master station.

[0071] Define the feeder between two nodes as a section. The section name is the number of the node at the beginning of the section, such as section 1 is the feeder between node 1 and node 2. To avoid repeated naming, when a branch line appears, the first section of the branch line needs to be named according to the actual situation. For example, section 5 is defined as the feeder between node 2 and node 6.

[0072] Node 4 in this feeder is connected to a distributed renewable energy DG. In actual feeders, multiple DGs can be connected. DG is controlled by using low voltage ride-through control strategy and negative sequence suppression strategy. Figure 3 As shown;

[0073] It should be noted that when a section on the feeder fails, the FTU on the head-end switch of all sections starts the sampling function to collect information on the three-phase voltage and three-phase current flowing through the switch. The length of the collection is one power frequency cycle. After the electrical signal is collected, it is uploaded to the main distribution station. The main distribution station uses Karenbauer transformation on the three-phase current to obtain α, β, and 0 mode currents. The Karenbauer transformation matrix K is expressed as:

[0074]

[0075] i α is the positive sequence component, representing the balanced part of the three-phase current, i0 is the zero sequence component, representing the unbalanced part of the three-phase current, i γ It is the negative sequence component, representing another unbalanced component in the three-phase current.

[0076] Specifically, the phase model transformation can be Clarke transformation, McNamee transformation and Karenbauer transformation, and the present invention preferably uses Karenbauer transformation;

[0077] It should be noted that the Karenbauer transform is suitable for processing asymmetric three-phase systems and provides a simpler calculation. Compared with the Clarke transform, the Karenbauer transform provides higher accuracy when decoupling three-phase systems. The McNamee transform is mainly used to eliminate the impact of unbalanced loads, while the Karenbauer transform is more general and applicable to a wider range of asymmetric conditions and fault analysis.

[0078] S200: performing wavelet packet decomposition on the electrical signals of the three-phase voltage and the phase mode current to obtain the fault sub-frequency band of each electrical signal, sorting the fault sub-frequency band of each electrical signal, constructing the time-frequency matrix of each node, sorting the time-frequency matrix of each node, and obtaining the time-frequency matrix of the distribution network;

[0079] Specifically, the distribution master station uses wavelet packet transform to perform three-layer wavelet packet decomposition on six electrical signals, namely, phase A voltage, phase B voltage, phase C voltage, α, β, and 0-mode current. Each electrical signal is decomposed into eight sub-bands, and the sub-bands are arranged from low to high according to the frequency. The first six sub-bands are selected as fault sub-bands, and the fault sub-bands of each electrical signal are arranged from top to bottom according to the frequency from low to high to construct the time-frequency matrix of the electrical signal; the time-frequency matrix of each electrical signal is arranged from top to bottom in the order of phase A voltage, phase B voltage, phase C voltage, α, β, and 0-mode current to construct the time-frequency matrix of the node; the time-frequency matrix of each node is arranged from top to bottom in the order of sections to construct the time-frequency matrix of the distribution network;

[0080] It should be noted that when the electrical signal is decomposed into n-layer wavelet packets, each electrical signal is decomposed into 2 n Sub-bands, six electrical signals are decomposed by three layers of wavelet packets. In the first layer, each signal is decomposed into two sub-bands. In the second layer, each sub-band of the first layer is decomposed into two sub-bands. At this time, each signal has four sub-bands. In the third layer, each sub-band of the second layer is further decomposed. Finally, each signal has eight sub-bands.

[0081] It should also be noted that the segment order is to sort all segments and customize the segment order, but it is necessary to ensure that the same segment order is used during the analysis process. The segment order list can be [segment 1, segment 2, ..., segment n].

[0082] S300: quantize and encode the time-frequency matrix of the distribution network to obtain a pixel matrix, perform pseudo-color processing on the pixel matrix using rainbow coding to obtain a fault color image;

[0083] In a possible embodiment, the power distribution master station quantizes and encodes the time-frequency matrix of the distribution network, maps each element in the original data sequence to a set mapping range, and obtains a normalized linear mapping value by subtracting the minimum value in the original data sequence from the element in the original data sequence divided by the maximum value in the original data sequence minus the minimum value in the original data sequence, and multiplies the normalized linear mapping by the set maximum value of the mapping range to obtain a standardized linear mapping value;

[0084] The function value obtained by rounding off the value of the linear mapping standardized by the rounding function is the result value of the quantization encoding.

[0085] Exemplarily, the power distribution master station quantizes and encodes the time-frequency matrix of the distribution network and converts it into a pixel matrix with matrix elements ranging from 0 to 255. The quantization encoding formula can be expressed as:

[0086]

[0087] J i =R(j i )

[0088] Among them, k i is a sequence k = [k1, k2, ..., k n ], k max , k min is the maximum and minimum value in series k, j i is the value obtained after linear mapping, R(·) is the rounding function, J i for j i The function value obtained after rounding is the result value of the quantization encoding.

[0089] In a possible embodiment, each pixel of the pixel matrix is ​​converted using rainbow coding, each pixel is input into a rainbow coding function, and a color value is output. The three primary color conversion function is:

[0090] The red transformation function includes three interval ranges;

[0091] If the pixel value is in [0, T1), the red transformation function value is 0; if the pixel value is in [T1, T2), the red transformation function value is the pixel value minus T1 multiplied by T3 divided by T2-T1; if the pixel value is in [T2, T3], the red transformation function value is T3;

[0092] The green transformation function includes five interval ranges;

[0093] If the pixel value is in [0, T4), the green transformation function value is 0; if the pixel value is in [T4, T5), the green transformation function value is the pixel value minus T4 multiplied by T3, divided by T4; if the pixel value is in [T5, T6), the green transformation function value is T3; if the pixel value is in [T6, T7), the green transformation function value is T7 minus the pixel value multiplied by T3, divided by T7-T6; if the pixel value is in [T7, T3], the green transformation function value is the pixel value minus T7 multiplied by T3, divided by T3-T7;

[0094] The blue transformation function includes five interval ranges;

[0095] If the pixel value is in [0, T8), the blue transformation function value is the pixel value multiplied by T3 divided by T8; if the pixel value is in [T8, T9), the blue transformation function value is T3; if the pixel value is in [T9, T 10 ), then the blue transformation function value is T 10 Subtract the pixel value multiplied by T3 divided by T 10 -T9; if the pixel value is between [T 10 ,T 11), the blue transformation function value is 0; if the pixel value is between [T 11 ,T3], then the blue transformation function value is the pixel value minus T 11 Multiply by T3, divide by T3-T 11 .

[0096] Exemplarily, each element in the original data sequence is mapped to a set mapping range of 0 to 255, and the three primary color conversion function of the rainbow code can be expressed as:

[0097]

[0098]

[0099]

[0100] Among them, R(x, y) is the red transformation function, G(x, y) is the green transformation function, B(x, y) is the blue transformation function, f(x, y) is the pixel value at the pixel point (x, y), and x and y are the position information of the pixel point in the matrix or image.

[0101] S400: Use a deep learning model to identify the fault color image and output the fault color image label as a positioning result.

[0102] In a possible embodiment, the size of the fault color image is adjusted to a first threshold to obtain a first fault color image, the first fault color image is input into the AlexNet model, the first fault color image is forward propagated through a convolution layer, an activation function, and a pooling layer, and a fault image feature vector is output in a second fully connected layer of the AlexNet model;

[0103] The fault image feature vector is input into the SVM model. The SVM model classifies the input feature vector according to the decision boundary learned from the training data and identifies the segment code mark corresponding to the fault image.

[0104] The segment coding mark corresponding to the fault image is identified according to the SVM model to obtain the segment where the fault occurs.

[0105] It should be noted that the size of the fault color image is adjusted to the first threshold, which is able to meet the input requirements of the AlexNet model. The first threshold can be 227*227*3. The AlexNet model performs feature extraction and the SVM model performs fault feature classification. During the fault section location, the distributed power sources on the feeder remain connected to the grid. After determining the fault section, the distribution master station also sends a control command to the head-end switch of the fault section and the head-end switch of the adjacent section downstream of the fault section to control the two head-end switches to trip at the same time, complete the isolation of the fault section, and achieve fault removal.

[0106] Specifically, the deep learning model can be a Transformer model, a Faster R-CNN model, and an AlexNet-SVM model. The present invention preferably uses the AlexNet-SVM model.

[0107] It should be noted that compared with other deep learning models using transfer learning, the AlexNet-SVM model is more suitable for specific fault detection tasks, especially when the fault features can be effectively captured by the convolutional layer of AlexNet and SVM can classify them well. SVM, as a simple linear classifier, performs well on small data sets, while models like Transformer and Faster R-CNN require more data to avoid overfitting. AlexNet can be used as a feature extractor, and its output can be easily integrated with other classifiers (such as SVM) to form a powerful classification system. The AlexNet-SVM model used in the present invention is significantly better than other models and has the best recognition effect.

[0108] In a possible embodiment, maximizing the interval width between different fault image feature vectors is defined as an objective function, the objective function is converted into a Lagrangian function using a Lagrangian multiplier method, the Lagrangian function is subjected to a dual transformation to obtain an objective function of the dual problem, and the objective function of the dual problem is solved using a KKT condition;

[0109] Derivative the objective function of the dual problem, set the derivative to 0, and solve to obtain the optimal Lagrange multiplier. In the solution process, there are non-zero Lagrange multipliers, and the corresponding sample points are support vectors.

[0110] The decision function is constructed using the support vectors and the corresponding non-zero Lagrange multipliers, and the decision boundary is the weight vector and the bias term.

[0111] It should be noted that for each training sample x i and its corresponding label y i , introduce the Lagrange multiplier β i , the Lagrangian function L is expressed as:

[0112]

[0113] Among them, w is the weight vector and b is the bias vector;

[0114] The Lagrangian function is dualized to obtain the objective function of the dual problem. The dual problem is usually easier to solve and shares the same optimal solution. The KKT conditions are used, which are a set of necessary conditions in the Lagrange multiplier method, to solve the dual problem.

[0115] Using the support vectors and the corresponding non-zero Lagrange multipliers, the decision function is constructed as:

[0116]

[0117] The bias term b can be calculated through any support vector to ensure that the value of the decision function at this point is the category label. The equation of the decision boundary is determined by the weight vector w and the bias term b, expressed as: w·x+b=0.

[0118] The above is a schematic scheme of a centralized FA fault section location method in a multi-energy scenario of this embodiment. It should be noted that the technical scheme of the device for centralized FA fault section location in a multi-energy scenario and the technical scheme of the centralized FA fault section location method in a multi-energy scenario described above belong to the same concept, and the details not described in detail in the technical scheme of the device for centralized FA fault section location in a multi-energy scenario in this embodiment can all be referred to the description of the technical scheme of the centralized FA fault section location method in a multi-energy scenario described above.

[0119] The device for locating a centralized FA fault section in a multi-energy scenario in this embodiment includes:

[0120] A phase-mode conversion module is used to obtain fault information, wherein the fault information includes three-phase voltage and three-phase current, and performs phase-mode conversion on the three-phase current to obtain phase-mode current;

[0121] A decomposition module, used to perform wavelet packet decomposition on the electrical signals of the three-phase voltage and the phase mode current to obtain a fault sub-frequency band of each electrical signal, sort the fault sub-frequency band of each electrical signal, construct a time-frequency matrix of each node, sort the time-frequency matrix of each node, and obtain a time-frequency matrix of the distribution network;

[0122] An encoding module, used for quantizing and encoding the time-frequency matrix of the distribution network to obtain a pixel matrix, and performing pseudo-color processing on the pixel matrix using rainbow coding to obtain a fault color image;

[0123] The recognition module is used to use a deep learning model to recognize the fault color image and output the fault color image label as a positioning result.

[0124] This embodiment also provides a computing device, which is applicable to the centralized FA fault section location in a multi-energy scenario, including:

[0125] Memory and processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the centralized FA fault section locating method in a multi-energy scenario as proposed in the above embodiment.

[0126] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for locating a centralized FA fault section in a multi-energy scenario as proposed in the above embodiment is implemented.

[0127] The storage medium proposed in this embodiment and the method for realizing centralized FA fault section location in multi-energy scenarios proposed in the above embodiment belong to the same inventive concept. The technical details not described in detail in this embodiment can be referred to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0128] Through the above description of the implementation methods, the technicians in the relevant field can clearly understand that the present invention can be implemented by means of software and necessary general hardware, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ReadOnly, Memory, ROM), random access memory (RandomAccess Memory, RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform the methods of various embodiments of the present invention.

[0129] Example 2

[0130] Referring to Table 1, Figure 4-5 As shown, this embodiment is different from the first embodiment and provides a verification test of a centralized FA fault section locating method in a multi-energy scenario to verify and illustrate the technical effects adopted in this method.

[0131] When a single-phase grounding fault occurs in section 2 on the feeder, the FTU on the head-end switch of all sections starts the sampling function to collect information on the three-phase voltage and three-phase current flowing through the switch. The sampling frequency is 10kHz, and the collection length is one power frequency cycle. The electrical signal of one power frequency cycle contains 200 sampling points. After the electrical signal is collected, it is uploaded to the main distribution station. The main distribution station uses Karenbauer transformation on the three-phase current to obtain α, β, and 0 mode currents.

[0132] The distribution master station uses wavelet packet transform to perform three-layer wavelet packet decomposition on six electrical signals, namely, phase A voltage, phase B voltage, phase C voltage, α, β, and 0-mode current. Each electrical signal is decomposed into 8 sub-bands, and 0-3000 Hz is selected as the main frequency band of the fault feature. According to the frequency from low to high, the first five sub-bands of each electrical signal, namely, 0-3125 Hz, are selected as the fault sub-bands. The fault sub-bands of each electrical signal are arranged from top to bottom according to the frequency from low to high, and the time-frequency matrix of the electrical signal is constructed. At this time, the size of the time-frequency matrix of the electrical signal is 5*200.

[0133] Arrange the time-frequency matrix of each electrical signal from top to bottom in the order of phase A voltage, phase B voltage, phase C voltage, α, β, and 0-mode current to construct the time-frequency matrix of the node. At this time, the size of the time-frequency matrix of the node is 30*200; arrange the time-frequency matrix of each node from top to bottom in the order of sections to construct the time-frequency matrix of the distribution network. At this time, the size of the time-frequency matrix of the distribution network is 210*200.

[0134] The distribution master station quantizes and encodes the time-frequency matrix of the distribution network and converts it into a pixel matrix with matrix elements ranging from 0 to 255;

[0135] After obtaining the pixel matrix, rainbow coding is used to transform the pixel matrix to obtain the fault color image, such as Figure 4 As shown;

[0136] After obtaining the fault color image, the size of the fault color image is adjusted to 227*227*3, and then input into the AlexNet model. The fault image feature vector is output in the second fully connected layer of the AlexNet model. The fault image feature vector is input into the SVM model. After classification and recognition by the SVM model, the fault color image label is output. The section corresponding to the coding mark corresponding to the fault color image label is judged as the fault section. Both the AlexNet model and the SVM model have been pre-trained. The pre-training results are shown in Figure 2. Figure 5 As shown, it can be seen that the fault section positioning accuracy of the present invention is as high as 97.8836%

[0137] One-hot encoding is used to encode the fault segment, as shown in Table 1:

[0138] Table 1 Fault section coding table

[0139]

[0140]

[0141] The output code of the AlexNet-SVM model is [0,1,0,0,0,0,0], so section 2 is judged as the faulty section, the head-end switch of section 2 and the head-end switch of section 3 trip, and the fault is eliminated.

[0142] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A centralized FA fault section location method in a multi-energy scenario, characterized in that: include: Acquire fault information, the fault information including three-phase voltage and three-phase current, and perform phase-mode transformation on the three-phase current to obtain phase-mode current; Decomposing the electrical signals of the three-phase voltage and the phase mode current by wavelet packets to obtain the fault sub-frequency band of each electrical signal, sorting the fault sub-frequency band of each electrical signal, constructing the time-frequency matrix of each node, sorting the time-frequency matrix of each node, and obtaining the time-frequency matrix of the distribution network; Quantizing and encoding the time-frequency matrix of the distribution network to obtain a pixel matrix, and performing pseudo-color processing on the pixel matrix using rainbow coding to obtain a fault color image; The fault color image is identified using a deep learning model, and a fault color image label is output as a positioning result.

2. The centralized FA fault section location method in a multi-energy scenario according to claim 1 is characterized in that: The three-phase current is subjected to phase mode transformation, comprising: Assume the three-phase current is i a 、i b 、i c ; The three-phase current vector [i a ,i b ,i c ] T Multiplying with the Karenbauer transformation matrix K, the transformed vector is Among them, i α is the positive sequence component, i0 is the zero sequence component, and is obtained by matrix multiplication: The vector transformed by the Karenbauer transformation matrix is 3. The centralized FA fault section location method in a multi-energy scenario as claimed in claim 2 is characterized in that: Quantizing and encoding the time-frequency matrix of the distribution network includes: The distribution master station quantizes and encodes the time-frequency matrix of the distribution network, maps each element in the original data sequence to a set mapping range, and obtains a normalized linear mapping value by subtracting the minimum value in the original data sequence from the element in the original data sequence divided by the maximum value in the original data sequence minus the minimum value in the original data sequence. The normalized linear mapping is multiplied by the set maximum value of the mapping range to obtain a standardized linear mapping value. The function value obtained by rounding off the value of the linear mapping standardized by the rounding function is the result value of the quantization encoding.

4. The centralized FA fault section location method in a multi-energy scenario as claimed in claim 3 is characterized in that: Performing pseudo-color processing on the pixel matrix using rainbow coding includes: Use rainbow coding to convert each pixel of the pixel matrix, input each pixel into the rainbow coding function, output the color value, and the three primary color conversion function is: The red transformation function includes three interval ranges; If the pixel value is in [0, T1), the red transformation function value is 0; if the pixel value is in [T 1, T2), the red transformation function value is the pixel value minus T1 multiplied by T3 divided by T2-T1; if the pixel value is within [T 2, T3], then the red transformation function value is T3; The green transformation function includes five interval ranges; If the pixel value is in [0, T4), the green transformation function value is 0; if the pixel value is in [T 4, T5), the green transformation function value is the pixel value minus T4 multiplied by T3, divided by T4; if the pixel value is within [T 5, T6), the green transformation function value is T3; if the pixel value is within [T 6, T7), the green transformation function value is T7 minus the pixel value multiplied by T3, divided by T7-T6; if the pixel value is within [T 7, T3], then the green transformation function value is the pixel value minus T7 multiplied by T3, divided by T3-T7.

5. The centralized FA fault section location method in a multi-energy scenario as claimed in claim 4 is characterized in that: Also includes: The blue transformation function includes five interval ranges; If the pixel value is in [0, T8), the blue transformation function value is the pixel value multiplied by T3 divided by T8; if the pixel value is in [T 8, T9), the blue transformation function value is T3; if the pixel value is between [T 9, T 10 ), then the blue transformation function value is T 10 Subtract the pixel value multiplied by T3 divided by T 10 -T9; if the pixel value is between [T 10, T 11 ), the blue transformation function value is 0; if the pixel value is between [T 11, T3], then the blue transformation function value is the pixel value minus T 11 Multiply by T3, divide by T3-T 11 .

6. The centralized FA fault section location method in a multi-energy scenario according to claim 3 or 5, characterized in that: Using a deep learning model to identify the fault color image includes: The size of the fault color image is adjusted to a first threshold value to obtain a first fault color image, and the first fault color image is input into the AlexNet model. The first fault color image is forward propagated through a convolution layer, an activation function, and a pooling layer, and a fault image feature vector is output in a second fully connected layer of the AlexNet model; The fault image feature vector is input into the SVM model. The SVM model classifies the input feature vector according to the decision boundary learned from the training data and identifies the segment code mark corresponding to the fault image. The segment coding mark corresponding to the fault image is identified according to the SVM model to obtain the segment where the fault occurs.

7. The centralized FA fault section location method in a multi-energy scenario according to claim 6, characterized in that: The decision boundary consists of: The objective function is defined as maximizing the interval width between different fault image feature vectors. The objective function is converted into a Lagrangian function using the Lagrangian multiplier method. The Lagrangian function is subjected to a dual transformation to obtain the objective function of the dual problem. The objective function of the dual problem is solved using the KKT condition. Derivative the objective function of the dual problem, set the derivative to 0, and solve to obtain the optimal Lagrange multiplier. In the process of solving, there are non-zero Lagrange multipliers, and the corresponding sample points are support vectors. The decision function is constructed using the support vectors and the corresponding non-zero Lagrange multipliers, and the decision boundary is the weight vector and the bias term.

8. A system for centralized FA fault section location method in multi-energy scenarios, characterized in that: include: A phase-mode conversion module is used to obtain fault information, wherein the fault information includes three-phase voltage and three-phase current, and performs phase-mode conversion on the three-phase current to obtain phase-mode current; A decomposition module, used to perform wavelet packet decomposition on the electrical signals of the three-phase voltage and the phase mode current to obtain a fault sub-frequency band of each electrical signal, sort the fault sub-frequency band of each electrical signal, construct a time-frequency matrix of each node, sort the time-frequency matrix of each node, and obtain a time-frequency matrix of the distribution network; An encoding module, used for quantizing and encoding the time-frequency matrix of the distribution network to obtain a pixel matrix, and performing pseudo-color processing on the pixel matrix using rainbow coding to obtain a fault color image; The recognition module is used to use a deep learning model to recognize the fault color image and output the fault color image label as a positioning result.

9. An electronic device, characterized in that: The device comprises: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 7 is implemented.