Power distribution network fault diagnosis method

Abstract

The invention provides a power distribution network fault diagnosis method. The method comprises the following steps: S1, multi-source data access of a power distribution network; s2, checking the data of the power distribution network; s3, performing credible diagnosis on the state of the power distribution network; s4, power distribution network fault diagnosis based on perturbation phase-mode transformation; s5, power distribution network fault positioning, isolation and power supply recovery; according to the method, the distribution network sensing risk prevention and control technology is improved, the reliability of sensing data is improved, the single data checking means is improved, multi-dimensional and high-precision data collaborative checking is achieved, and due to the fact that information and physical risks of the power distribution Internet of Things are coupled and superposed, the active fault recognition capacity is improved, the power distribution network prevention and control means are enhanced, and the reliability of the power distribution network is improved. The situations of risk expansion and overflow are avoided, the problem of single means of fault study and judgment and power supply recovery is solved, and safe and reliable operation of the power distribution network is guaranteed; according to the invention, guidance suggestions are provided for distribution network multi-source data aggregation and equipment state credibility, and accurate processing of power distribution faults is realized synchronously based on the Internet of Things cloud edge cooperation technology.

Description

Technical Field

[0001] The present invention relates to a method for diagnosing faults in a distribution network, and particularly to a method for diagnosing faults in a distribution network, belonging to the technical field of fault detection. Background Art

[0002] As an important public infrastructure connecting substations and users, the distribution network is a key link to improve the power supply quality at the end of the smart grid and the user experience. Deeply exploring the application potential of edge computing in the intelligentization of the distribution network has become the main research hotspot. More and more methods of terminalizing the processing, marginalizing the calculation, and localizing the solution of distribution network data information have become an important way. However, the marginal integration of complex multi-source heterogeneous data brings new challenges to the efficient edge computing of the smart distribution network. Therefore, it is urgent to realize the processing and fusion in the marginalization mode of multi-source data in the smart distribution network, which is an important basis for promoting and ensuring the intelligent development of the distribution network based on edge computing;

[0003] At present, the conventional dispatching system in the distribution network still uses the dispatching automation system to process raw data of fault information. In the case of a large amount of information, it will cause the inundation of alarm and accident information, thus greatly affecting the fault handling speed. The accurate judgment of the fault type in the distribution network is of great significance for formulating a reasonable operation and maintenance strategy and the reliable operation of the distribution system. Reasonably distributing the large-scale data information processing tasks to the edge nodes can greatly relieve the data processing pressure on the upper-layer cloud platform;

[0004] The existing distribution Internet of Things has changed from a traditional physical system to a system with a high degree of coupling between information and physics. The random fluctuations of the source network load and the superposition of information-physics risks have brought new problems to the risk prevention and control of the distribution Internet of Things, and also put forward higher requirements for the perception ability of the distribution Internet of Things;

[0005] To solve the above technical problems, a method for diagnosing faults in a distribution network is proposed. Summary of the Invention

[0006] In view of this, the present invention provides a method for diagnosing faults in a distribution network to solve or alleviate the technical problems existing in the prior art, and at least provide a beneficial option.

[0007] The technical solution of the present invention is realized as follows: A method for diagnosing faults in a distribution network includes the following steps:

[0008] S1. Accessing multi-source data of the distribution network;

[0009] S2. Verifying the distribution network data;

[0010] S3. Credible diagnosis of the distribution network state;

[0011] S4. Fault diagnosis of the distribution network based on the phase-mode transformation by the perturbation method;

[0012] S5, Fault location, isolation and power supply restoration of the distribution network.

[0013] Further preferably, in the said S1, collect multi-source data of the low-voltage distribution network, sort out and collect the data, strengthen the system's collection ability, enhance the system's theoretical operation performance, continuously protect the data from external interference factors by enhancing its own protection performance, use the data matching method to centrally match the data, obtain the required data information, on the premise of obtaining the data information, realize the preprocessing of the data, combine the data management and monitoring methods, increase the data control and analysis efforts, conduct centralized processing and data fusion operations on the data, divide different data sets according to the specific information of the data, record the coexistence states of different sets, and obtain the result parameters that meet the system requirements.

[0014] By adopting the above technical solutions, the research on multi-source data processing and fusion technology of low-voltage distribution network based on edge computing can expand the basic performance of the system to a higher degree, improve the system's operability, and have better development prospects.

[0015] Further preferably, in the said S2, conduct distribution network data verification based on the Disruptor framework, design a high-speed buffer based on the lock-free ring queue technology of the Disruptor framework, conduct in-depth research on the implementation method of the data buffer, analyze the disadvantages of the implementation method in traditional ELT tools, and based on the above methods and tools, propose a high-speed buffer implementation scheme based on the lock-free ring queue Disruptor. The implementation scheme realizes the above content through the following design principles: lock-free design, memory pre-allocation, avoiding false sharing, batch processing effect, and Disruptor consumer waiting strategy.

[0016] By adopting the above technical solutions, while enabling the ELT system to have its own inherent qualities, research and optimization will be carried out on the problems existing in traditional ELT tools, and finally an efficient, reliable and practical ELT system will be realized.

[0017] Further preferably, in the said S3, based on the theory of trusted computing, introduce the concept of "trusted" into the security protection of the distribution network system, propose the concept of "power trusted network", research and construct the theoretical model of the power trusted network, and according to the general research methods of information security engineering, starting from the security object, clarify the security goal, research the security strategy to meet the security goal, and research the implementation of relevant security mechanisms according to the security strategy. The main research ideas include security goal, security strategy, security mechanism, etc.

[0018] Further preferably, in S4, when diagnosing faults in a distribution network based on the phase-mode transformation of the perturbation method, the fault criteria for a three-phase symmetrical low-voltage distribution network with DG, the section fault criteria for an asymmetrical low-voltage distribution network based on the perturbation method, and the fault location process for a low-voltage active distribution network are carried out in sequence.

[0019] By adopting the above technical solutions, based on the differential protection principle, the mode component theory is introduced, the decoupling of the asymmetrical three-phase network is realized by using the perturbation method, and by analyzing the change in the phase angle difference of the phase-mode transformation current fault components before and after the section fault in the low-voltage active distribution network, a method for realizing the section fault location of the low-voltage distribution network with distributed power sources based on the phase-mode transformation of the perturbation method is proposed, which is not affected by the transition resistance and solves the adverse effect of the asymmetry of the distribution network parameters on the location accuracy.

[0020] Further preferably, the main fault criteria for a three-phase symmetrical low-voltage distribution network with DG are divided into the section fault criteria for a two-terminal power supply based on the positive-sequence fault component current, the section fault criteria for a multi-power supply based on the positive-sequence fault component current, the influence of the existence of the transition resistance on the section fault location of the two / multi-terminal power supply, and the section fault criteria for a single power supply based on the intelligent fusion terminal.

[0021] Further preferably, in S5, the fault location, isolation, and power supply restoration of the active distribution network with edge-cloud collaboration are studied, and the fault location of the active distribution network with edge-cloud collaboration is carried out.

[0022] Further preferably, the main research contents include feature extraction based on ST processing on the edge side, fault feature training based on DBN on the cloud side, a fault location method based on ST-DBN, and fault location accuracy analysis.

[0023] Further preferably, in S5, the isolation and power supply restoration with edge-cloud collaboration are carried out. With the SCU distributed at each switch of the intelligent distribution network as the core, each SCU exchanges information such as network topology, switch attributes, and electrical quantities with other SCUs through peer-to-peer communication, and with the support of local information, the topology search of the non-fault power outage area and the simplified solution of the power supply restoration constraint conditions are completed.

[0024] Due to the adoption of the above technical solutions in the embodiments of the present invention, it has the following advantages:

[0025] 1. The present invention improves the risk prevention and control technology of distribution network perception, enhances the credibility of perception data, improves the single data verification method, realizes multi-dimensional and high-precision data collaborative verification. Due to the coupling and superposition of information and physical risks in the distribution Internet of Things, the active recognition ability of faults is improved through the present invention, the prevention and control means of the distribution network are enhanced, the situation of risk expansion and spillover is avoided, the problem of single fault judgment and power supply restoration means is improved, and the safe and reliable operation of the distribution network is guaranteed;

[0026] Second, the present invention provides guiding suggestions for the convergence of multi-source data in the distribution network and the credibility of equipment status. Meanwhile, based on the Internet of Things cloud-edge collaboration technology, it realizes the precise handling of distribution faults, assists inspection personnel in intelligent inspection, and through the research and application of intelligent key technologies for the distribution network based on the Internet of Things, improves the multi-source data aggregation ability of the distribution network and enhances the safety and stability of the distribution network, thus bringing huge direct and indirect economic benefits.

[0027] The above summary is only for the purpose of the specification and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments, and features described above, further aspects, embodiments, and features of the present invention will be readily apparent by reference to the drawings and the following detailed description. Brief Description of the Drawings

[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present application or in the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0029] Figure 1 is the method flowchart of the present invention;

[0030] Figure 2 is the data monitoring flowchart of the present invention;

[0031] Figure 3 is the data matching process diagram of the present invention;

[0032] Figure 4 is the data conversion process diagram of the present invention;

[0033] Figure 5 is the architecture diagram of the distribution Internet of Things of the present invention. Detailed Embodiments

[0034] In the following, only some exemplary embodiments are briefly described. As those skilled in the art can recognize, the described embodiments can be modified in various different ways without departing from the spirit or scope of the present invention. Therefore, the drawings and the description are considered to be exemplary in nature rather than restrictive.

[0035] The embodiments of the present invention will be described in detail below with reference to the drawings.

[0036] As Figures 1-5 shown, the embodiments of the present invention provide a distribution network fault diagnosis method, including the following steps:

[0037] S1. Access of multi-source data in the distribution network:

[0038] (1) Collection of multi-source data in low-voltage distribution network;

[0039] (2) Preprocessing of multi-source data in low-voltage distribution network;

[0040] (3) Processing and fusion of multi-source data in low-voltage distribution network.

[0041] In one embodiment, to enhance the technical performance of processing and fusion of multi-source data in the low-voltage distribution network of the system, first, collect the multi-source data of the distribution network to obtain relatively accurate multi-source data information. It is planned to study the substation address as a text parameter, select the parameters that meet the system research as the initial data input value, perform data storage operations on it, and divide the data into categories according to the corresponding data arrangement method to obtain the final calculation result, and apply the result function to the central management system to ensure the integrity of the data. The data monitoring process is as Figure 2 shown.

[0042] In one embodiment, after collecting the data, preprocess the collected data. Perform a secondary filtering operation on the preliminarily filtered data to remove the attached interference factors in the data, and match the filtered data. The data matching process is as Figure 3 shown.

[0043] In one embodiment, perform the final processing and data fusion operation on the preprocessed multi-source data. Use the data sensor to obtain the data feature function, compare the data features, detect the connection between the data features, and perform system conversion on the feature information. The data conversion process is as Figure 4 shown.

[0044] S2. Data verification of distribution network:

[0045] (1) Lock-free design. The storage core of Disruptor is implemented based on the ring buffer Ring buffer. Different from the traditional blocking queue sharing the same member variable, all producers and consumers in the Ring buffer each maintain a private Cursor (sequence number). When a producer wants to write data into the Ring buffer, it will first declare (or preempt) a position pointed to by the Cursor. After the declaration is successful, it will write the data. After the elements are added, the producer will update the value of the latest readable Cursor on the Ring buffer, and this Cursor is visible to the consumer. Then the consumer can start consuming data according to the latest Cursor value. At the same time, the consumer also maintains a Cursor recording its own consumption status, and the producer will also ensure that it will not overwrite the data that the consumer has not had time to consume according to the consumer's Cursor;

[0046] (2) Memory pre-allocation: All the memory of the Ring buffer is pre-allocated at startup, which means that all elements in the array will exist throughout the life cycle of the Disruptor. Each element object (Entry) in it is not the data itself, but a container. When the producer stores data in the Ring buffer, instead of simply pointing the pointer to another object like a simple array, it updates the value inside the Entry. This enables the element objects in the Ring buffer to survive and be reused throughout the life cycle of the Disruptor, thereby reducing the occurrence of GC.

[0047] (3) Avoiding false sharing: The Disruptor adopts the method of cache line padding to avoid false sharing problems, thereby reducing unnecessary cache misses. It fills the cache line in the form of padding 7 long-type data before and after.

[0048] (4) Batch processing effect: Whether it is the producer or the consumer, when performing read and write operations, it will pre-declare a section of space, and then perform batch read and write on the data within that range. After the read and write are completed, it will perform operations such as the Ring buffer to uniformly publish the data changes within that range. This batch processing method reduces the production and consumption publishing operations and improves the overall processing performance.

[0049] S3. Trustworthy diagnosis of the distribution network status:

[0050] (1) Taking the distribution system network as the research object, analyze its characteristics and security protection requirements;

[0051] (2) Propose the concept of a power trustworthy network and study the construction of a power trustworthy network model;

[0052] (3) Study the security policy model of the power trustworthy network to provide a theoretical basis for the construction of the power trustworthy network system;

[0053] (4) Study the key technologies of the power trustworthy network to provide technical support for the construction of the power trustworthy network system.

[0054] In one embodiment, by comparing the differences between the electrical system network and the information system network, analyze the characteristics and special security requirements of the distribution control system network, and clarify that the goal of the distribution control system network security protection is: on the basis of ensuring a clear and controllable network boundary and clear and controllable hardware and systems, achieve the trustworthy guarantee of system operation, the normative guarantee of behavior, and the strong immunity of the system.

[0055] In one embodiment, a trusted concept is introduced into the security protection of the distribution control system network. Based on the trusted network theory, in view of the relatively closed and limited characteristics of the distribution control system network environment, the concept of "power trusted network" is proposed to clarify the significance and connotation of the power trusted network. Starting from two levels of the static environment trust and dynamic environment trust of the network system, the security objectives of the power trusted network are defined, the hierarchical model of the power trusted network is studied, and three trustworthiness assurance levels of the power trusted network are proposed, namely the reliability of hardware and systems, the trustworthiness of system operation, and the normativity of behaviors. To provide the trustworthiness assurance of these three levels, the system is also required to have a clearly controllable network boundary and high survivability.

[0056] In one embodiment, aiming at the security objectives of the power trusted network, its security strategies are studied. The controllability of hardware devices and systems mainly relies on management means and strategies, and a clearly controllable network boundary mainly relies on relevant network isolation technologies and management strategies. The security strategies are mainly studied for three aspects: the trustworthiness of system operation, the normativity of behaviors, and the survivability of the system. For the trustworthiness of system operation, the theory of mandatory operation control based on trusted computing is studied; for the normativity of behaviors, from three aspects of personnel trustworthiness, behavior trustworthiness, and business operation compliance, the role-based access control theory, the personnel trustworthiness theory based on mandatory hardware confirmation control, the behavior trustworthiness auditing theory, and the business process compliance inspection theory based on graph theory are respectively studied; for the survivability of the system, the system survivability theory based on the self-healing theory is studied, and the formal descriptions of these theories are given respectively to support and enrich the power trusted network theory system.

[0057] In one embodiment, to provide technical support for the construction of the power trusted network system, based on the research of the security strategies and theoretical models of the power trusted network, for each security strategy, the implementation of its key technologies is studied. The mandatory operation control technology based on trusted processes, network behavior auditing technology, mandatory hardware confirmation control technology, fault tolerance and intrusion tolerance key technologies, and network boundary isolation technology are respectively studied to enrich the security mechanism of the power trusted network system and provide technical support for the construction of the power trusted network system.

[0058] S4. Distribution network fault diagnosis based on perturbation method phase-mode transformation:

[0059] (1) Fault criterion for three-phase symmetrical low-voltage distribution network with DG;

[0060] (2) Section fault criterion for asymmetrical low-voltage distribution network based on perturbation method;

[0061] (3) Fault location process for low-voltage active distribution network.

[0062] In one embodiment, the connection of DG transforms the low-voltage distribution network from a radial network powered by a single-sided power source into a two-terminal or multi-terminal power network including DG power supply. However, the section where the end of the distribution line is connected to the load remains powered by a single-sided power source without DG connection. Therefore, the distribution network containing DG can be divided into three forms: single / double / multi-power supply sections. The main criteria are as follows: the fault criterion for the two-terminal power section based on the positive-sequence fault component current, the fault criterion for the multi-power section based on the positive-sequence fault component current, the influence of the existence of the transition resistance on the fault location of the two / multi-terminal power section, and the fault criterion for the single-power section based on the intelligent fusion terminal.

[0063] In one embodiment, based on the perturbation method, by constructing the perturbation amount of the line impedance matrix of the low-voltage active distribution network, a phase-mode transformation method that can decouple three-phase coupled asymmetric lines is proposed. The low-voltage active distribution network with asymmetric three-phase parameters is decoupled into three independent networks (moduli). After decoupling, the voltage of each modulus is only related to the current of that mode and has nothing to do with the voltages and currents of other modes. The relationship between these three networks is similar to the relationship between the positive-sequence, negative-sequence, and zero-sequence networks in the symmetrical component method. By generalizing the fault location criterion of the positive-sequence current phase angle difference in the symmetrical network, a fault location criterion applicable to the asymmetric low-voltage active distribution network is obtained, thus realizing the accurate fault location of the actual asymmetric low-voltage active distribution network.

[0064] In one embodiment, a fault location process for the low-voltage active distribution network is proposed. According to the topological structure information of the active distribution network, the network is divided into single / double / multi-terminal power sections. The measuring devices at both ends of the two-terminal and multi-terminal power sections synchronously collect three-phase current information, and after filtering, the phase angle differences of the modal fault component currents on both sides / upstream and downstream are obtained. For the single-terminal power section, the over-current information is collected through the intelligent fusion terminal to form an over-current coding matrix. The phase angle differences of the modal fault component currents and the over-current coding matrix are transmitted to the control center. The control center performs fault location according to the fault location criteria for the single-terminal / double-terminal / multi-terminal power sections. When a fault is detected in a certain section, an action signal is sent to the protection device in that section to isolate the fault in that section.

[0065] S5. Fault Location, Isolation, and Power Supply Restoration of the Distribution Network:

[0066] (1) Fault Location of the Active Distribution Network with Edge-Cloud Collaboration;

[0067] (2) Isolation and Power Supply Restoration with Edge-Cloud Collaboration.

[0068] In one embodiment, based on the new power distribution Internet of Things (IoT) architecture system of "cloud - pipe - edge - end", applying the powerful edge computing capabilities on the edge side, the ST processing is used to capture the fault current signal from the intelligent terminal on the end side, and the total harmonic distortion rate is calculated through different fault feeders. The ST coefficient of the fault signal is calculated to extract the fault characteristics. On the "cloud" side, the deep belief network is used to learn the characteristic attributes of different faults and classify them. The root mean square value of the weight of the ST coefficient is adjusted according to the attribute classification factor to adapt to the processing and location of different types of fault characteristics. The power distribution IoT architecture based on "cloud - pipe - edge - end" is as Figure 5 shown. The "cloud" end is the power distribution IoT cloud platform, which calculates, stores, and uniformly schedules the data collected on the edge side in a software - defined manner. The "pipe" is the communication network, which uses optical fiber or 4 / 5G network to achieve cloud - edge data interaction. The "edge" is the IoT proxy terminal close to the data source, which pre - processes the local data and provides intelligent decision - making locally. The "end" is the intelligent terminal of the sub - station area at the end, which realizes the collection of the data of the distribution network line. Based on this architecture, the ST processing is embedded in the IoT proxy terminal, and the edge - side computing capabilities are used to detect and pre - process the fault information locally; then through the cloud master station platform, the DBN is used for different fault characteristic training, learning classification, and fault location. The main research contents include feature extraction based on ST processing on the edge side, fault characteristic training based on DBN on the cloud side, fault location method based on ST - DBN, and fault location accuracy analysis.

[0069] In one embodiment, with the SCU installed dispersedly at each switch of the intelligent distribution network as the core, each SCU interacts information such as network topology, switch attributes, and electrical quantities with other SCUs through peer - to - peer communication. With the support of local information, the topology search of the non - fault power outage area and the simplified solution of the power supply restoration constraint conditions are completed.

[0070] In one embodiment, the basic principle of distributed power supply restoration is proposed. The structure and functions of a distributed power supply restoration system with the SCU as the core are introduced. The management areas of each SCU are divided. Each SCU can communicate peer-to-peer with adjacent terminals to exchange information such as network topology and electrical quantities, laying a foundation for solving the power supply restoration constraint conditions and completing the topological search of the entire power grid. A method for distributed search of network topology based on relay search of adjacent terminals is proposed, and the main process of distributed power supply restoration is presented. The topological search is divided into the following three processes: search for non-fault power outage areas, search for the actual recoverable power supply range, and DG island search. The search process is introduced from three aspects: selecting the starting point, searching for stop nodes, and relay search of intermediate nodes, aiming to solve the coordination control problem in the power supply restoration process. A distributed simplified solution method for constraint conditions is proposed. This method is coordinated with the distributed power supply restoration topology search process and is realized based on network step-by-step Thevenin equivalent. The constraint conditions are expressed by simplified equations and can be solved locally at the SCU in a fast and simple way. The constraint conditions in the presence of branches are corrected. A distributed power supply restoration method for distribution networks considering soft-switching characteristics is proposed. The basic structure and control mode of the SOP are analyzed. Based on the powerful power transmission control ability and power supply restoration ability of the SOP, a power supply restoration model of the SOP is proposed, and the proposed distributed network topology search method and power supply restoration process control are corrected and improved.

[0071] As described above, the above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of various changes or substitutions, and these should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A distribution network fault diagnosis method, characterized in that: The following steps are involved: S1, multi-source data access of distribution network; S2, distribution network data verification; S3, reliable diagnosis of distribution network status; S4. Distribution network fault diagnosis based on phase mode transformation of perturbation method; S5. Distribution network fault location, isolation and power supply restoration.

2. A distribution network fault diagnosis method according to claim 1, characterized in that: In the S1, multi-source data of the low-voltage distribution network is collected, the data is sorted and collected, the system collection capability is strengthened, the theoretical operational performance of the system is strengthened, the data is continuously protected from external interference factors by enhancing its own protection performance, and the data is centrally matched using a data matching method to obtain the required data information. On the premise of obtaining the data information, the data is pre-processed, combined with data management and monitoring methods, the data control and analysis efforts are increased, the data is centrally processed and the data is fused, different data sets are divided according to the specific information of the data, the coexistence status of different sets is recorded, and the result parameters that meet the system requirements are obtained.

3. A distribution network fault diagnosis method according to claim 1, characterized in that: In the S2, the distribution network data is verified based on the Disruptor framework, a high-speed buffer based on the lock-free circular queue technology of the Disruptor framework is designed, the implementation method of the data buffer is deeply studied, and the drawbacks of the implementation method in the traditional ELT tool are analyzed. Based on the above methods and tools, a high-speed buffer implementation scheme based on the lock-free circular queue Disruptor is proposed. The implementation scheme realizes the above contents through the following design principles: lock-free design, memory pre-allocation, avoidance of false sharing, batching effect and Disruptor consumer waiting strategy.

4. A distribution network fault diagnosis method according to claim 1, characterized in that: In the S3, based on the trusted computing theory, the concept of "trusted" is introduced into the security protection of the distribution network system, the concept of "power trusted network" is proposed, and the theoretical model of power trusted network is studied and constructed. According to the general research method of information security engineering, starting from the security object, the security goal is clarified, the security strategy that meets the security goal is studied, and the implementation of related security mechanisms is studied based on the security strategy. The main research ideas include security goals, security strategies, security mechanisms, etc.

5. A distribution network fault diagnosis method according to claim 1, characterized in that: In S4, when diagnosing distribution network faults based on phase mode transformation using the perturbation method, the fault judgment criteria for a three-phase symmetrical low-voltage distribution network containing DG, the section fault judgment criteria for an asymmetrical low-voltage distribution network based on the perturbation method, and the low-voltage active distribution network fault location process are performed in sequence.

6. A distribution network fault diagnosis method according to claim 5, characterized in that: The main fault judgment criteria of the three-phase symmetrical low-voltage distribution network containing DG are divided into two-terminal power supply section fault judgment based on positive-sequence fault component current, multi-power supply section fault judgment based on positive-sequence fault component current, the influence of transition resistance on the fault location of two-terminal / multi-terminal power supply section, and single power supply section fault judgment based on intelligent fusion terminal.

7. A distribution network fault diagnosis method according to claim 1, characterized in that: In S5, the fault location, isolation and power supply restoration of the active distribution network in edge-cloud collaboration are studied, and the fault location of the active distribution network in edge-cloud collaboration is performed.

8. A distribution network fault diagnosis method according to claim 7, characterized in that: The main research contents include feature extraction based on ST processing on the edge side, fault feature training based on DBN on the cloud side, fault location method based on ST-DBN, and fault location accuracy analysis.

9. A distribution network fault diagnosis method according to claim 1, characterized in that: In the S5, edge-cloud collaborative isolation and power supply restoration are carried out, with the SCU installed dispersedly at each switch of the smart distribution network as the core. Each SCU interacts with other SCUs through peer-to-peer communication to exchange network topology, switch properties, electrical quantities and other information. With the support of local information, the topology search of non-fault power outage areas and the simplified solution of power supply restoration constraints are completed.

Images