Flink power distribution network fault report information processing method and device based on real-time stream calculation

Abstract

The invention discloses a power distribution network fault report information processing method and device based on Flink of real-time flow calculation, the method is applied to an Flink flow processing engine, and the method comprises the following steps: obtaining real-time monitoring data and actually measured environment data of a power distribution network; if it is judged that a fault exists in the power distribution network based on the real-time monitoring data, processing the real-time monitoring data and the actually measured environment data to obtain processed data, and obtaining a fault type based on the processed data by adopting a fault prediction model; and determining a fault position in the power distribution network based on the fault type, the real-time monitoring data and a line impedance constraint relationship, and reporting the fault position and the fault type, so that rapid and accurate positioning of the fault can be realized.

Description

Technical Field

[0001] This application relates to the field of smart grid fault handling technology. In particular, it relates to a method and apparatus for processing distribution network fault reporting information based on real-time stream computing using Flink. Background Technology

[0002] With the accelerated construction of new power systems, distribution network fault monitoring faces severe temporal and spatial constraints and technical challenges: substation bays are surrounded by dense cabinets, and the bandwidth of fault information transmission channels is often less than 10Mbps; the renovation of old urban areas is limited by the space of weak current wells, and the deployment height of data acquisition equipment is less than 0.5 meters; underground power distribution rooms need to complete fault signal capture in harsh environments with temperatures >50℃ and humidity >90%, while the instantaneous disturbances of frequent coastal typhoon areas and inland thunderstorm belts cause the signal-to-noise ratio of fault signals to drop sharply.

[0003] Existing methods are insufficient for rapid and accurate fault location in complex environments and with a large number of connected terminals. Summary of the Invention

[0004] This application provides a method and apparatus for processing distribution network fault reporting information based on real-time stream computing in Flink. It takes into account the impact of the environment on the fault and can achieve rapid and accurate fault location.

[0005] In a first aspect, embodiments of this application provide a method for processing distribution network fault reporting information based on real-time stream computing in Flink. The method is applied to the Flink stream processing engine and includes:

[0006] Acquire real-time monitoring data and measured environmental data of the power distribution network;

[0007] If a fault is determined to exist in the power distribution network based on the real-time monitoring data, the real-time monitoring data and the measured environmental data are processed to obtain processed data, and a fault prediction model is used to obtain the fault type based on the processed data.

[0008] The location of the fault in the distribution network is determined based on the fault type, the real-time monitoring data, and the line impedance constraint relationship, and the fault location and the fault type are reported.

[0009] Secondly, embodiments of this application provide a distribution network fault reporting information processing device based on real-time stream computing in Flink. The device is applied to the Flink stream processing engine and includes:

[0010] The acquisition module is used to acquire real-time monitoring data and measured environmental data of the power distribution network.

[0011] The fault type output module is used to process the real-time monitoring data and the measured environmental data if a fault is determined to exist in the distribution network based on the real-time monitoring data, to obtain processed data, and to obtain the fault type based on the processed data using a fault prediction model.

[0012] The fault location determination module is used to determine the fault location in the distribution network based on the fault type, the real-time monitoring data, and the line impedance constraint relationship, and to report the fault location and the fault type.

[0013] Thirdly, embodiments of this application provide an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the method provided in embodiments of this application.

[0014] Fourthly, embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to perform the method provided in embodiments of this application.

[0015] The technical solution provided in this application embodiment is applied to the Flink stream processing engine. It acquires real-time monitoring data and measured environmental data of the distribution network; processes the real-time monitoring data and the measured environmental data, and uses a fault prediction model based on the processed data to obtain the fault type; determines the fault location in the distribution network based on the fault type, real-time monitoring data, and line impedance constraints, and reports the fault location and fault type. That is, when performing fault judgment through the Flink stream processing engine, it can identify faults in the distribution network through data from different sources, which can quickly locate faults, take into account the influence of the external environment, and accurately identify faults. Attached Figure Description

[0016] Figure 1 A flowchart of a distribution network fault reporting information processing method based on real-time stream computing in Flink is provided for embodiments of this application;

[0017] Figure 2 This is a diagram of the real-time stream processing framework.

[0018] Figure 3 Diagram of the Flink stream processing engine runtime framework;

[0019] Figure 4 This is an event-driven process framework diagram;

[0020] Figure 5 A flowchart illustrating the implementation framework provided in this application's embodiments;

[0021] Figure 6A structural block diagram of a distribution network fault reporting information processing device based on real-time stream computing in Flink is provided for embodiments of this application.

[0022] Figure 7 This is a schematic diagram of an electronic device structure provided in an embodiment of this application. Detailed Implementation

[0023] The present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0024] Figure 1 This application provides a flowchart of a distribution network fault reporting information processing method based on real-time stream computing using Flink. The method can be executed by a distribution network fault reporting information processing device based on real-time stream computing using Flink. This device can be implemented in software and / or hardware and can be configured within the Flink stream processing engine, which can be configured in electronic devices such as computers. Flink is a state-managed stream computing framework that offers significant advantages in system fault tolerance and real-time performance.

[0025] like Figure 1 As shown, the method includes the following steps:

[0026] S110: Acquire real-time monitoring data and measured environmental data of the power distribution network.

[0027] In this embodiment, the real-time monitoring data of the distribution network may include data such as current and voltage from the nodes of the distribution network; the measured environmental data may include meteorological data such as ambient temperature, wind speed, whether there is lightning strike, whether it is raining, and whether it is snowing. Among them, the key nodes of the distribution network may be equipment such as transformers and towers.

[0028] S120: If a fault is determined to exist in the power distribution network based on the real-time monitoring data, the real-time monitoring data and the measured environmental data are processed to obtain processed data, and a fault prediction model is used to obtain the fault type based on the processed data.

[0029] In this embodiment, if the real-time monitoring data contains data exceeding a preset threshold, a fault is determined to exist in the distribution network. The real-time monitoring data and measured environmental data can be processed to obtain processed data. A fault prediction model is then used to determine the fault type based on the processed data.

[0030] In existing technologies, fault diagnosis often uses data from a single source without correlating the data from different sources. This application's embodiment determines the fault type by analyzing the same event from different sources, including real-time monitoring data and measured environmental data, while also considering the environmental impact on the fault type. This allows for accurate fault type output. Fault types often differ under different environments; introducing measured environmental data for fault type determination ensures precise fault identification.

[0031] In one embodiment of this example, optionally, the step of processing the real-time monitoring data and the measured environment data to obtain processed data includes: aligning the real-time monitoring data and the measured environment data in time, and constructing a real-time feature matrix based on the real-time monitoring data and the measured environment data within the fault propagation time window;

[0032] The fault prediction model, based on the processed data, determines the fault type, including:

[0033] The real-time feature matrix is ​​input into the fault prediction model, and the corresponding fault type is output.

[0034] In this embodiment, real-time monitoring data originates from the detection equipment at nodes in the power distribution network, while measured environmental data may come from weather forecast systems. Since there may be a time delay between the two types of data, processing can be performed based on data within the fault propagation time window. The real-time feature matrix can be a matrix formed by real-time monitoring data and measured environmental data at various time points within the time propagation window.

[0035] In this embodiment, training of the fault prediction model may also be included. The specific process is as follows: a decision model is constructed based on the power grid topology; historical monitoring data and historical environmental data in the training dataset are processed to obtain a historical feature matrix; the decision model is trained using the historical feature matrix and the corresponding fault types to obtain a fault training model. The decision model can be a decision tree model, constructed using nodes in the distribution network as nodes and the relationships between nodes as the relationships between upper and lower level nodes. By training this model, a fault prediction model is obtained, which is used to determine the fault type.

[0036] In one embodiment of this example, optionally, the real-time monitoring data and the measured environmental data are processed to obtain processed data, and a fault prediction model is used to obtain the fault type based on the processed data. This includes: constructing a real-time feature matrix based on the real-time monitoring data and measured environmental data within the fault propagation time window; extracting spatiotemporal features from the real-time feature matrix using a spatiotemporal network model incorporating an attention mechanism to obtain spatiotemporal features; and inputting the spatiotemporal features into the fault prediction model to output the fault type.

[0037] Optionally, the spatiotemporal network model incorporating an attention mechanism includes a temporal convolutional layer, a graph convolutional layer, an attention layer, and a fully connected layer. The step of extracting spatiotemporal features from the real-time feature matrix using the spatiotemporal network model incorporating an attention mechanism includes: constructing an adjacency matrix based on the impedance between nodes in the power grid topology; extracting temporal features from the real-time feature matrix using the temporal convolutional layer; inputting the temporal features and the adjacency matrix into the graph convolutional layer to obtain spatial features; fusing the spatial features using the attention layer and compressing the output through the fully connected layer to obtain the spatiotemporal features.

[0038] Specifically, based on the impedance between nodes (mainly substations / power plants) in the power grid topology, an adjacency matrix is ​​established. The elements in the adjacency matrix can have a certain functional relationship with the impedance between nodes. This allows for the construction of the adjacency matrix. A time convolutional layer can be used to extract time features from the real-time feature matrix using a sliding window. A graph convolutional layer can aggregate neighborhood information according to the power grid topology to obtain spatial features. An attention layer can be used to fuse the spatial features, and a fully connected layer can be used to process them to obtain spatiotemporal features. Thus, by extracting spatiotemporal features through a spatiotemporal network model incorporating an attention mechanism, the correlation between time and space can be captured, making it easier to more accurately identify fault types.

[0039] Accordingly, the method provided in this application embodiment may further include: constructing a historical feature matrix from historical monitoring data and historical environmental data in the training dataset, and extracting spatiotemporal features from the historical feature matrix to obtain historical spatiotemporal features; training an LSTM network model using the historical spatiotemporal features and the corresponding fault types to obtain a fault prediction model. The processing procedure for obtaining historical spatiotemporal features from the training dataset is the same as the method for obtaining spatiotemporal features in the application process.

[0040] In one embodiment of this example, optionally, the fault prediction model includes a first attention layer, a second attention layer, a splicing layer, and an LSTM layer; the first attention layer and the second attention layer are two parallel attention layers;

[0041] Accordingly, the real-time monitoring data and the measured environmental data are processed to obtain processed data. A fault prediction model is then used to determine the fault type based on the processed data. This process includes: constructing a first real-time feature matrix based on the real-time monitoring data within the fault propagation time window, and constructing a second real-time feature matrix based on the measured environmental data within the fault propagation window; assigning corresponding weights through a first attention layer, and processing the first real-time feature matrix based on the assigned weights to obtain a first output feature matrix; assigning corresponding weights through a second attention layer, and processing the second real-time feature matrix based on the assigned weights to obtain a second output feature matrix; concatenating the first and second output features to obtain a concatenated feature, and inputting the concatenated feature into an LSTM layer to obtain the fault type. The fault prediction model can be trained by assigning different weights to two attention layers to process the input data, concatenating the features, and inputting the concatenated feature into an LSTM layer to obtain the fault type. During the training process, the model learns the importance of the influence of real-time monitoring data and the external environment on the fault type, fully considering the weight ratio of external environmental factors, and can accurately determine the fault type.

[0042] In this embodiment, if the fault type output by the fault prediction model is a communication distortion fault, then the abnormal data in the real-time monitoring data is filtered, and the filtered data is stored and reported. During transmission, the real-time monitoring data may be distorted due to communication network issues, which can be considered a communication distortion fault. In this case, during the training process of the fault prediction model, the training samples in the training dataset may include historical datasets under communication distortion fault conditions and historical datasets under distribution network fault conditions.

[0043] S130: Determine the location of the fault in the distribution network based on the fault type, the real-time monitoring data, and the line impedance constraint relationship, and report the fault location and the fault type.

[0044] In this embodiment, optionally, the impedance values ​​between nodes in the distribution network are calculated based on the line impedance constraint relationship using the real-time monitoring data; the corresponding fault node is queried from a preset fault database based on the fault type and the impedance value between the nodes. The line impedance constraint relationship can be understood as a power grid topology relationship. The preset fault database stores various fault types and the corresponding impedance values ​​between nodes. The corresponding fault node can be found by querying the preset fault database. After finding the fault location, the fault location and fault type are reported for analysis or execution of corresponding strategies. The line impedance constraint relationship can also trace the electrical connection path between nodes in reverse, facilitating the location of the anomaly propagation source and identifying hidden fault chains (such as cascade faults caused by insulation degradation), effectively resisting the resulting topology analysis bias. By using physical constraints for pruning, the computational complexity is significantly reduced, and the processing efficiency is improved.

[0045] Based on the above embodiments, the technical solution provided in this application may further include: if the fault type is a transient fault caused by the environment, performing a reclosing operation based on the fault node; if the fault type is a long-term fault caused by the environment, issuing a warning. The transient fault caused by the environment may be a short-circuit fault caused by strong winds; the long-term fault caused by the environment may be a equipment damage fault caused by lightning strikes. If it is a long-term fault, it cannot be immediately restored and a warning needs to be issued to facilitate maintenance by personnel.

[0046] The technical solution provided in this application embodiment is applied to the Flink stream processing engine. It acquires real-time monitoring data and measured environmental data of the distribution network; processes the real-time monitoring data and the measured environmental data, and uses a fault prediction model based on the processed data to obtain the fault type; determines the fault location in the distribution network based on the fault type, real-time monitoring data, and line impedance constraints, and reports the fault location and fault type. That is, when performing fault judgment through the Flink stream processing engine, it can identify faults in the distribution network through data from different sources, which can quickly locate faults, take into account the influence of the external environment, and accurately identify faults.

[0047] This application embodiment can also introduce a multi-source data fusion model to perform cross-source spatiotemporal alignment and semantic fusion of real-time equipment status and environmental data, eliminating data drift and forming a high-confidence status view with a unified spatiotemporal benchmark. The fused status features trigger preset complex business rules (such as fault diagnosis logic and real-time formula calculation), driving alarm event generation and real-time push, thereby constructing a closed-loop event processing chain from "multi-source data perception → spatiotemporal fusion → topology tracing → intelligent decision-making → alarm output", significantly improving the timeliness and accuracy of power system anomaly response.

[0048] The technical solutions provided in this application embodiment utilize the Flink stream processing engine. For example... Figure 2 As shown, the real-time stream processing framework provided in this application embodiment may include several parts such as a data acquisition layer, a data processing layer, a data storage layer, a data application service, and a data visualization platform. The Flink stream processing engine is applied in the data processing layer. Furthermore, the data acquisition layer is the foundation and starting point for processing distribution network fault reporting information. Data acquisition efficiency needs to ensure data timeliness, providing strong support for subsequent data processing. Streaming data in the power system mainly comes from real-time monitoring data collected by sensors. These data are characterized by high acquisition frequency, large data volume, and complex data types. In the data acquisition module, the real-time transmitted streaming data also needs to be formatted and converted into a JSON format that the system can recognize, laying the foundation for the next step of the data processing flow. Real-time acquisition of streaming data requires equipment with good network conditions, enabling real-time acquisition and real-time monitoring of data changes.

[0049] In this embodiment, the data processing layer can introduce a reverse topology tracing algorithm model based on line impedance constraints to perform refined analysis and status identification of real-time incoming power data, effectively improving the ability to locate node anomalies and dynamically perceive network topology. By deploying a multi-source data fusion model, data can be deeply integrated in the spatiotemporal dimensions to form a unified, high-value fused data view. The fused data serves the upper-layer data application services and visualization platform, providing solid real-time data support for intelligent monitoring, precise scheduling, and decision analysis of the power system. The data processing layer can divide the data into five layers: Data Operation Layer (ODS), Detailed Data Layer (DWD), Summary Data Layer (DWS), Application Data Layer (ADS), and Dimensional Data Layer (DIM). Data flow uses a Kafka message queue, and after each layer completes data computation, it is rewritten to Kafka for the next layer to process. The ODS layer stores raw data, the DWD layer is responsible for data cleaning and transformation, the DWS layer summarizes data based on the DWD layer data, typically forming wide tables across topics based on a certain data dimension, and the ADS layer addresses practical application needs, forming various statistical reports based on the DWD and DWS layer data, which are ultimately stored in a MySQL database. The DIM layer mainly stores dimensional data, using Flink to store it in HBase.

[0050] The primary responsibility of the data storage layer is to store data of different types, purposes, and scales. The data acquisition layer needs to transmit data to a Kafka message queue, using Canal or Flink-CDC to pass changed data into Kafka. Processed ADS layer data is stored in ClickHouse, allowing for direct querying or further processing based on specific application needs. Data requiring dimensional correlation is stored in ClickHouse and Redis to meet the requirements of fast querying and high availability. Final calculation results and intermediate state data are stored in a MySQL database for subsequent querying and display. By selecting an appropriate storage engine, efficient storage and query services are provided for different types of data.

[0051] In smart grids, the most common streaming data comes from device runtime sequence data collected by sensors, such as measurement point codes, acquisition timestamps, device voltage, current, and phase. The Flink streaming engine in a real-time streaming framework can specifically mine and analyze the collected data, handling data aggregation, computation, and transformation to generate datasets that applications can query. The data storage layer increases the reusability of data computation. Each time a new requirement arises, it's not necessary to start calculating from the original data again; instead, the intermediate stored dimensional data and status can be used to continue aggregation and analysis, reducing the workload for developers. The ODS layer is the first stop in the data inflow framework, primarily responsible for the collection, storage, and cleaning of raw data. The purpose of this layer is to integrate data from different sources, formats, and types into a standardized format and prepare it for use by subsequent layers. The ODS layer is typically an independent data storage area to ensure decoupling of the data source from the data processing framework and data traceability. Its design needs to consider the differences between different data sources and data quality assurance to ensure data integrity and reliability. It also needs to fully consider data scalability, reliability, and security to meet the requirements of the real-time streaming framework. The primary function of the DWD layer is to clean, transform, integrate, and deduplicate the data within it. It typically employs a topic-oriented modeling approach, grouping related data under the same theme together. The DWD layer also performs dimensional modeling, categorizing data with common attributes to facilitate multi-dimensional analysis and querying. The DWS layer is the core of the entire real-time data processing framework. Its data primarily originates from the DWD layer. ETL tools clean, process, and summarize the data in the DWD layer, creating higher-level data metrics, such as daily electricity consumption calculated from frozen user electricity consumption and electricity consumption of various household appliances. Data storage in the DWS layer generally uses distributed or columnar databases to support faster data querying and analysis. Furthermore, the DWS layer provides data query interfaces for use by upper-layer data application modules. The ADS layer is the highest-level data computation module in the entire data processing flow. It is primarily responsible for real-time computation of the data processed by the DWS layer to generate the final application data results. This data typically exists in a multi-dimensional aggregated form. The data in this layer mainly includes summary results, statistical results, and indicator results. The ADS layer data is characterized by its extremely high real-time performance and accuracy, and can be directly used by the application layer. DIM refers to the dimensional data layer, which mainly provides standardized dimensional data to facilitate correlation and querying in the fact table.

[0052] The data application service layer and data visualization platform primarily analyze and apply data to support business decision-making and real-time monitoring. The data visualization platform enables real-time monitoring and fault alarms. Specifically, it displays various power system indicators such as grid load, voltage, frequency, and current. Many transmission and transformation devices in the smart grid require high safety, necessitating real-time monitoring and real-time alarms and notifications based on pre-set complex rules to help maintenance personnel promptly identify and resolve problems. Real-time reporting and visualization analysis are also provided, displaying the calculation results of data streams in real time. This visualizes and analyzes data, reducing the time spent on manual data collection and analysis, improving operational efficiency, and contributing to grid dispatching and intelligent management.

[0053] In Flink's distributed runtime architecture, such as Figure 3 As shown, its core process consists of two key components: the Job Manager and the Task Manager. The Job Manager plays a central role as the system's coordination hub, acting as the master control process for Flink job execution. Each application instance submitted to the cluster is globally managed by a unique JobManager instance, responsible for managing the job's lifecycle, including but not limited to job graph parsing, execution plan generation, task scheduling strategy implementation, checkpoint coordination, and triggering fault recovery mechanisms. It is the ultimate responsible entity for system resource coordination and job execution status monitoring. In contrast, the Task Manager serves as the work execution unit in the Flink cluster, deployed on various physical or virtual nodes. Its core responsibility is to actually carry out and execute the specific computational tasks (Tasks) assigned by the JobManager, responsible for the real-time transfer of data streams between operators, state maintenance, and efficient execution of user-defined computational logic (such as transformations, aggregations, and window operations). The TaskManager manages its own computing resources (such as memory, network buffers, and CPU slots) to ensure low-latency and high-throughput operation of the data stream processing pipeline, serving as the physical carrier of distributed stream processing computing capabilities. The JobManager and TaskManager work together to form the cornerstone of Flink's highly reliable and high-performance real-time stream processing capabilities.

[0054] The data pipeline serves as a core component of the real-time stream processing framework, undertaking the crucial functions of real-time cleaning, transformation, and preliminary analysis during transmission. Compared to the traditional batch-based offline ETL model, to meet the increasingly stringent real-time requirements of big data scenarios, a real-time database architecture with processing latency compressed to minutes or seconds needs to be built. This architecture utilizes real-time Stream ETL technology to achieve dynamic transformation and semantic expansion of upstream data, significantly reducing downstream query latency and improving decision-making timeliness.

[0055] Among them, such as Figure 4 As shown, by treating real-time monitoring data and actual environmental data as the same fault event, and locating the fault through this event, fault alarms can be triggered and event logs can be recorded. The implementation framework of the technical solution provided in this application embodiment can also be referenced. Figure 5 .

[0056] Figure 6 This application provides a structural block diagram of a distribution network fault reporting information processing device based on real-time stream computing in Flink. The device is applied to the Flink stream processing engine and includes:

[0057] The acquisition module 610 is used to acquire real-time monitoring data and measured environmental data of the power distribution network;

[0058] The fault type output module 620 is used to process the real-time monitoring data and the measured environmental data if it is determined that there is a fault in the distribution network based on the real-time monitoring data, obtain the processed data, and use a fault prediction model to obtain the fault type based on the processed data.

[0059] The fault location determination module 630 is used to determine the fault location in the distribution network based on the fault type, the real-time monitoring data and the line impedance constraint relationship, and to report the fault location and the fault type.

[0060] like Figure 7 As shown in the figure, this application provides an electronic device, including a processor 111, a communication interface 112, a memory 113, and a communication bus 114, wherein the processor 111, the communication interface 112, and the memory 113 communicate with each other through the communication bus 114.

[0061] Memory 113 is used to store computer programs;

[0062] In one embodiment of this application, when the processor 111 executes a program stored in the memory 113, it implements the method provided in any of the foregoing method embodiments, including:

[0063] Acquire real-time monitoring data and measured environmental data of the power distribution network;

[0064] If a fault is determined to exist in the power distribution network based on the real-time monitoring data, the real-time monitoring data and the measured environmental data are processed to obtain processed data, and a fault prediction model is used to obtain the fault type based on the processed data.

[0065] The location of the fault in the distribution network is determined based on the fault type, the real-time monitoring data, and the line impedance constraint relationship, and the fault location and the fault type are reported.

[0066] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the method provided in any of the foregoing method embodiments.

[0067] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0068] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented using software plus a general-purpose hardware platform, or of course, using hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0069] The above embodiments are merely illustrative examples and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations. However, obvious variations or modifications derived therefrom are still within the scope of protection of this application.

Claims

1. A power distribution network fault reporting information processing method based on real-time stream computing Flink, characterized in that, The method is applied to a Flink stream processing engine, and the method comprises: obtaining real-time monitoring data and measured environmental data of a power distribution network; if it is determined based on the real-time monitoring data that there is a fault in the power distribution network, processing the real-time monitoring data and the measured environmental data to obtain processed data, and using a fault prediction model to obtain a fault type based on the processed data; determining a fault location in the power distribution network based on the fault type, the real-time monitoring data and a line impedance constraint relationship, and reporting the fault location and the fault type.

2. The method of claim 1, wherein, The method further comprises: based on the power grid topology structure, constructing a decision model; processing historical monitoring data and historical environmental data in a training data set to obtain a historical feature matrix, and using the historical feature matrix and the corresponding fault type to train the decision model to obtain a fault training model. The method further comprises: based on the power grid topology structure, constructing a decision model; processing historical monitoring data and historical environmental data in a training data set to obtain a historical feature matrix, and using the historical feature matrix and the corresponding fault type to train the decision model to obtain a fault training model.

3. The method of claim 1, wherein, The method further comprises: based on the power grid topology structure, constructing a decision model; processing historical monitoring data and historical environmental data in a training data set to obtain a historical feature matrix, and using the historical feature matrix and the corresponding fault type to train the decision model to obtain a fault training model. The method further comprises: based on the power grid topology structure, constructing a decision model; processing historical monitoring data and historical environmental data in a training data set to obtain a historical feature matrix, and using the historical feature matrix and the corresponding fault type to train the decision model to obtain a fault training model.

4. The method of claim 3, wherein, The time and space network model incorporating the attention mechanism comprises a time convolution layer, a graph convolution layer, an attention layer and a fully connected layer; The method further comprises: based on the power grid topology structure, constructing a decision model; processing historical monitoring data and historical environmental data in a training data set to obtain a historical feature matrix, and using the historical feature matrix and the corresponding fault type to train the decision model to obtain a fault training model. The fault prediction model comprises a first attention layer, a second attention layer, a splicing layer and an LSTM layer; the first attention layer and the second attention layer are two parallel attention layers. ​ 5. The method of claim 1, wherein, ​ The real-time monitoring data and the measured environment data are processed to obtain processed data, and a fault prediction model is used to obtain a fault type based on the processed data, including: A first real-time feature matrix is constructed based on real-time monitoring data within a fault propagation time window, and a second real-time feature matrix is constructed based on measured environment data within the fault propagation window; A corresponding weight is assigned through a first attention layer, and the first real-time feature matrix is processed based on the assigned weight to obtain a first output feature matrix; A corresponding weight is assigned through the second attention layer, and the second real-time feature matrix is processed based on the assigned weight to obtain a second output feature matrix; The first output feature and the second output feature are spliced to obtain a spliced feature, and the spliced feature is input into an LSTM layer to obtain a fault type.

6. The method of claim 1, wherein, The fault type, the real-time monitoring data, and a line impedance constraint relationship are used to determine a fault location in the power distribution network, including: The real-time monitoring data is used to calculate impedance values between nodes in the power distribution network based on the line impedance constraint relationship; The fault type and the impedance values between the nodes are used to query a corresponding fault node in a pre-set fault database.

7. The method of claim 1, wherein, Further including: If the fault type belongs to an environment-caused transient fault, a reclosing operation is performed based on the fault node; If the fault type belongs to an environment-caused long-time fault, an alarm is given.

8. A power distribution network fault reporting information processing device based on real-time stream computing Flink, characterized in that, The device is applied to a Flink stream processing engine, and the device includes: An acquisition module for acquiring real-time monitoring data and measured environment data of a power distribution network; A fault type output module for processing the real-time monitoring data and the measured environment data to obtain processed data if it is determined based on the real-time monitoring data that there is a fault in the power distribution network, and using a fault prediction model to obtain a fault type based on the processed data; A fault location determination module for determining a fault location in the power distribution network based on the fault type, the real-time monitoring data, and a line impedance constraint relationship, and reporting the fault location and the fault type.

9. An electronic device, comprising: A memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method of any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, A computer program is stored thereon, and when the computer program is executed in a computer, the computer program causes the computer to execute the method of any one of claims 1-7.

Images