Fault prediction method, device and equipment, computer readable storage medium and computer program product

Abstract

The invention provides a fault prediction method, device and equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: acquiring real-time line data of each first power transmission line in the overhead line system; constructing a volt-ampere characteristic curve graph for each first power transmission line; performing prediction based on the volt-ampere characteristic curve graph by using the fault prediction sub-model corresponding to the first power transmission line to obtain a first fault prediction result of the first power transmission line; and obtaining a first adjacent matrix corresponding to the overhead line system, and performing prediction based on the first adjacent matrix, a first fault prediction result corresponding to the first power transmission line and the real-time line data by using the fault prediction total model to obtain a second fault prediction result of the overhead line system. According to the invention, the fault prediction sub-models suitable for different line lengths can be trained, and the fault prediction sub-models are combined with the fault prediction general model to carry out fault prediction and positioning, so that the accuracy and pertinence of fault prediction and positioning are improved.

Description

Technical Field

[0001] The present application relates to fault prediction technology, and in particular to a fault prediction method, apparatus, device, computer-readable storage medium, and computer program product. Background Art

[0002] In the related art, the fault prediction of the transmission line depends on the pre-set fault characteristics. Generally, various types of fault characteristics that may occur in the transmission line are entered into the expert system in advance, and then a threshold value is set for the transmission line. Once the characteristic value of the transmission line reaches the pre-set threshold value and a certain logical relationship is established, the transmission line is judged to have a fault, and the fault diagnosis result is given according to the fault code pre-set by the system. However, due to the large number of overhead lines and their different lengths, the fault types and causes on each line are different, and the pre-set fault characteristics cannot handle all fault conditions. Therefore, it is impossible to perform fault prediction extensively, and the accuracy of fault prediction is low. Summary of the invention

[0003] The embodiments of the present application provide a fault prediction method, device, computer-readable storage medium and computer program product, which can train fault prediction sub-models suitable for different line lengths, and perform fault prediction and location together with the fault prediction overall model, thereby improving the accuracy and pertinence of fault prediction and location.

[0004] The technical solution of the embodiment of the present application is implemented as follows: The present application provides a fault prediction method, the method comprising: Acquiring real-time line data of each first transmission line in the overhead line system; For each first power transmission line, construct a volt-ampere characteristic curve diagram based on the real-time line data of the first power transmission line; Using the fault prediction sub-model corresponding to the first transmission line, and based on the volt-ampere characteristic curve diagram, a prediction is performed to obtain a first fault prediction result of the first transmission line; A first adjacency matrix corresponding to the overhead line system is obtained, and a fault prediction overall model is used to perform prediction based on the first adjacency matrix, the first fault prediction result corresponding to the first transmission line, and real-time line data to obtain a second fault prediction result of the overhead line system.

[0005] The present application provides a fault prediction device, including: A data acquisition module, used to acquire real-time line data of each first transmission line in the overhead line system; A construction module, configured to construct a volt-ampere characteristic curve diagram for each first power transmission line based on real-time line data of the first power transmission line; A first fault prediction module, configured to use a fault prediction sub-model corresponding to the first transmission line to perform prediction based on the volt-ampere characteristic curve diagram to obtain a first fault prediction result of the first transmission line; The second fault prediction module is used to obtain a first adjacency matrix corresponding to the overhead line system, and use a total fault prediction model to perform prediction based on the first adjacency matrix, the first fault prediction result corresponding to the first transmission line and real-time line data to obtain a second fault prediction result of the overhead line system.

[0006] An embodiment of the present application provides an electronic device, including: A memory for storing computer executable instructions or computer programs; The processor is used to implement the fault prediction method provided in the embodiment of the present application when executing the computer executable instructions or computer programs stored in the memory.

[0007] An embodiment of the present application provides a computer-readable storage medium storing a computer program or computer-executable instructions for implementing the fault prediction method provided in the embodiment of the present application when executed by a processor.

[0008] An embodiment of the present application provides a computer program product, including a computer program or computer executable instructions. When the computer program or computer executable instructions are executed by a processor, the fault prediction method provided by the embodiment of the present application is implemented.

[0009] The embodiments of the present application have the following beneficial effects: The embodiment of the present application provides a fault prediction method, which obtains the real-time line data of each first transmission line in the overhead line system; for each first transmission line, based on the real-time line data of the first transmission line, a volt-ampere characteristic curve is constructed; using the fault prediction sub-model corresponding to the first transmission line, prediction is performed based on the volt-ampere characteristic curve to obtain the first fault prediction result of the first transmission line; wherein the fault prediction sub-model corresponding to the first transmission line is determined based on the length of the first transmission line, so that the pertinence of the fault prediction of transmission lines of different lengths and the accuracy of the first fault prediction result can be improved, and then the first adjacency matrix corresponding to the overhead line system is obtained, and the fault prediction total model is used to predict based on the first adjacency matrix, the first fault prediction result corresponding to the first transmission line and the real-time line data to obtain the second fault prediction result of the overhead line system. In this way, after obtaining the first prediction result of each first transmission line in the overhead line system, the fault prediction total model is used to perform fault prediction on the entire overhead line system, and the fault prediction sub-model and the fault prediction total model are combined to predict and locate the fault, thereby improving the accuracy of the fault prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 It is a structural diagram of the fault prediction system architecture provided in an embodiment of the present application; Figure 2 is a schematic diagram of the structure of a fault prediction device provided in an embodiment of the present application; Figure 3 is a first flow chart of the fault prediction method provided in an embodiment of the present application; Figure 4A It is a second flow chart of the first fault prediction sub-model training method provided in an embodiment of the present application; Figure 4B It is a third flow chart of the second fault prediction sub-model training method provided in the embodiment of the present application; Figure 5 is a fourth flow chart of a method for determining a fault prediction sub-model corresponding to a first transmission line provided in an embodiment of the present application; Fig. 6A is a fifth flow chart of the first adjacency matrix construction method provided in an embodiment of the present application; Figure 6B is a sixth flow chart of the second adjacency matrix construction method provided in an embodiment of the present application; Figure 6C 7 is a schematic diagram of a seventh flow chart of a third adjacency matrix construction method provided in an embodiment of the present application; Figure 7 is an eighth flow chart of the second fault prediction result obtaining method provided in an embodiment of the present application; Figure 8 This is a ninth flow chart of the fault prediction overall model training method provided in an embodiment of the present application; Fig. 9 is a deployment diagram of a power grid line system provided in an embodiment of the present application; Fig.10 It is a deployment diagram of a simplified power grid line system provided in an embodiment of the present application.

[0011] It should be pointed out that the above-mentioned "first" and "second" are only used to distinguish different solutions, and do not represent the degree of superiority or inferiority of the solutions or the priority in the implementation process. DETAILED DESCRIPTION

[0012] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0013] In the following description, reference is made to “some embodiments”, which describe a subset of all possible embodiments, but it will be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0014] In the following description, the terms "first\second\third" involved are merely used to distinguish similar objects and do not represent a specific ordering of the objects. It can be understood that "first\second\third" can be interchanged with a specific order or sequence where permitted, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein.

[0015] In the embodiments of the present application, the term "module" or "unit" refers to a computer program or a part of a computer program that has a predetermined function and works together with other related parts to achieve a predetermined goal, and can be implemented in whole or in part by using software, hardware (such as processing circuits or memories), or a combination thereof. Similarly, a processor (or multiple processors or memories) can be used to implement one or more modules or units. In addition, each module or unit can be part of an overall module or unit that includes the function of the module or unit.

[0016] Unless otherwise defined, all technical and scientific terms used in the embodiments of the present application have the same meanings as those commonly understood by those skilled in the art. The terms used in the embodiments of the present application are only for the purpose of describing the embodiments of the present application and are not intended to limit the present application.

[0017] The relevant data collection and processing in the embodiments of this application should be strictly in accordance with the requirements of relevant laws and regulations when applied in examples, obtain the informed consent or separate consent of the personal information subject, and carry out subsequent data use and processing within the scope of authorization of laws and regulations and the personal information subject.

[0018] Before further describing the embodiments of the present application in detail, the nouns and terms involved in the embodiments of the present application are explained. The nouns and terms involved in the embodiments of the present application are subject to the following interpretations.

[0019] 1) Overhead lines: mainly refers to overhead open lines, which are bare-conductor telecommunication lines supported by poles and erected on the ground.

[0020] 2) Insulator: A device installed between conductors of different potentials or between conductors and grounded structures, capable of withstanding voltage and mechanical stress. Insulators are installed on transmission line towers to support and fix conductors while providing electrical insulation.

[0021] 3) Transmission corridor: A strip area under the overhead line with a specified width extending to both sides along the side conductors of the overhead line.

[0022] 4) Secondary three-phase voltage: The three-phase voltage generated on the secondary side, i.e. the output side, of electrical equipment such as power transformers or generators. Three-phase voltage refers to three AC voltages with a phase difference of 120°, which work together in a three-phase system to provide electrical energy for various three-phase loads.

[0023] 5) Wave recording data: Waveform data of voltage, current and other electrical quantities of the power system or electrical equipment recorded by wave recording equipment within a specific time. Usually includes time domain waveform data of voltage and current, as well as possible spectrum analysis data, harmonic analysis data, etc. These data can reflect the changes in electrical quantities of the power system within a specific time, including waveform amplitude, phase, frequency and other characteristics.

[0024] 6) Zero-sequence voltage: In a three-phase three-wire circuit, when the three-phase voltage is unbalanced, the vector sum of the three-phase voltage is not zero, and this non-zero vector sum is the zero-sequence voltage. It is an abnormal voltage state in the power system, usually related to conditions such as three-phase imbalance and ground fault.

[0025] 7) Three-phase three-wire circuit: A circuit system that uses three AC power sources with a phase difference of 120° to transmit and distribute power through three conductors, namely phase lines. In a three-phase three-wire circuit, the ends or neutral points of the three power sources are usually connected together, but the neutral line or zero line is not drawn out, so only three phase lines are used to transmit electrical energy.

[0026] 8) Zero-sequence current: In a three-phase AC system, when the three-phase current is unbalanced due to some reason such as system imbalance, ground fault or equipment failure, the common current generated by all conductors or equipment relative to the ground.

[0027] 9) Volt-ampere characteristic curve: IU curve or IV curve, with the vertical axis representing current I and the horizontal axis representing voltage U. The IU curve is drawn to show the current response of electrical components under different voltages.

[0028] 10) Fully connected layer: A basic layer in a neural network. In this layer, each neuron in the network is connected to all neurons in the previous layer. This connection allows the layer to capture global features or patterns from the input data and pass these features to subsequent layers for further processing.

[0029] 11) Activation function: It is a key component in neural networks, which determines whether neurons in the network should be activated. It includes linear and nonlinear activation functions. The linear activation function is the simplest activation function, and its output is a linear transformation of the input. For example, the common linear activation function is Nonlinear activation functions can introduce nonlinear characteristics, enabling neural networks to learn and represent complex functions and decision boundaries. Common nonlinear activation functions include Sigmoid, ReLU, Tanh, etc.

[0030] 12) Degree matrix: It is a diagonal matrix whose diagonal elements represent the degree of each vertex in the graph. In an undirected graph, the degree of a vertex is the number of edges connected to the vertex; in a directed graph, the degree of a vertex is divided into out-degree and in-degree. The out-degree is the number of directed edges starting from the vertex, and the in-degree is the number of directed edges entering the vertex.

[0031] 13) Graph Convolutional Network (GCN): A deep learning model specifically designed to process graph structured data. It extracts useful feature information from graph data and can be used for tasks such as node classification, graph classification, edge prediction, and obtaining graph embedding representations.

[0032] 14) Classifier: A mathematical model designed for a specific problem that can classify the features of target samples that need to be identified and classified through some mathematical calculations and assign corresponding labels. In data mining, classifier is a general term for sample classification methods, which includes a variety of algorithms, such as decision trees, logistic regression, naive Bayes, neural networks, etc.

[0033] 15) Break node: refers to a specific point pre-marked in the sample data, which represents the location where a break may occur in the circuit. A break is a state where a part of the circuit is disconnected, causing the circuit to be interrupted and the current cannot flow.

[0034] 16) Transfer learning: An important method in machine learning that allows the model to transfer knowledge learned in one task or field to another related task or field to accelerate and improve learning and problem-solving capabilities in the new field.

[0035] 17) Current trend: The law of change of current in a circuit over time or other variables. This trend can be affected by many factors, such as the properties of circuit components such as power supply voltage, resistance, capacitance, inductance, and changes in external conditions such as temperature and pressure. The positive and negative signs of the current trend are usually used in electricity to indicate the relationship between the direction of the current and the reference direction. When the current trend is positive, it means that the actual direction of the current is the same as the reference direction, that is, it flows from the positive electrode to the negative electrode. When the current trend is negative, it means that the actual direction of the current is opposite to the reference direction, that is, the current actually flows from the negative electrode to the positive electrode.

[0036] 18) Voltage trend: The direction and rate at which the voltage in a circuit changes with time or spatial position. When the voltage trend is positive, it means that the potential at that point or that section of the circuit is higher than the reference point. When the voltage trend is negative, it means that the potential at that point or that section of the circuit is lower than the reference point.

[0037] In the related art, fault prediction of transmission lines relies on pre-set fault characteristics. However, due to the large number of overhead lines and their different lengths, the fault types and causes on each line are different. The pre-set fault characteristics cannot handle all fault conditions. Therefore, fault prediction cannot be carried out extensively, and the accuracy of fault prediction is low.

[0038] The embodiments of the present application provide a fault prediction method, apparatus, device, computer-readable storage medium, and computer program product, which can train fault prediction sub-models suitable for different line lengths, and perform fault prediction and location in combination with the fault prediction overall model, thereby improving the accuracy and pertinence of fault prediction and location. The following describes an exemplary application of the fault prediction device provided by the embodiments of the present application. The device provided by the embodiments of the present application can be implemented as a terminal such as a desktop computer and a laptop computer, or can be implemented as a server. The following describes an exemplary application when the device is implemented as a server.

[0039] See also Figure 1 , Figure 1 is a schematic diagram of the architecture of a fault prediction system 100 provided in an embodiment of the present application. The fault prediction system 100 includes a plurality of sensors (in Figure 1 , a first sensor 400-1 and a second sensor 400-2) and a server 200 are exemplarily shown. To support a fault prediction application, the first sensor 400-1 and the second sensor 400-2 are connected to the server 200 via a network 300. The network 300 may be a wide area network or a local area network, or a combination of the two.

[0040] The first sensor 400-1 and the second sensor 400-2 can be arranged on the transmission line in the overhead line system, and are used to collect the real-time line data of the transmission line, and send the collected real-time line data to the server 200. A plurality of fault prediction sub-models can be deployed in the server 200, and one fault prediction sub-model corresponds to one sensor, which can be understood as one fault prediction sub-model corresponds to one transmission line. After acquiring the real-time line data collected by the sensor, the server 200 constructs a volt-ampere characteristic curve diagram according to the real-time line data, and uses the fault prediction sub-model corresponding to the transmission line where the sensor is located to perform fault prediction based on the volt-ampere characteristic curve diagram, and obtains the fault prediction result corresponding to the transmission line. A total fault prediction model is also deployed in the server 200. The server 200 uses the total fault prediction model to perform fault prediction based on the pre-constructed adjacency matrix, the fault prediction results corresponding to each transmission line, and the real-time line data, and obtains the fault prediction result corresponding to the overhead line system, wherein the fault prediction result may include the fault type and the fault location. The fault type indicates whether a fault occurs and, if a fault occurs, the specific type of the fault, such as phase short circuit, circuit open circuit, line grounding fault, etc. The fault location refers to the location where the fault occurs, such as a certain tower, or a line section between any two towers.

[0041] In some embodiments, in addition to a plurality of sensors and a server, the fault prediction system may also include a line prediction terminal, which is arranged on the transmission line in the overhead line system. The line prediction terminal corresponds to the sensor one by one, that is, the position of the sensor will correspond to a line prediction terminal with a fault prediction sub-model adapted to the line length. It should be noted that the one-to-one corresponding sensors and line prediction terminals are arranged on the same tower. And a fault prediction sub-model is deployed in the line prediction terminal. After collecting the real-time line data of the transmission line, the sensor sends the real-time line data to the line prediction terminal. The line prediction terminal constructs a volt-ampere characteristic curve based on the real-time line data, thereby performing fault prediction based on the fault prediction sub-model to obtain the fault prediction result of the transmission line. The line prediction terminal sends the fault prediction result of the transmission line to the server. The server uses the fault prediction total model to perform fault prediction based on the pre-constructed adjacency matrix, the fault prediction results of each transmission line, and the real-time line data to obtain the final fault type and fault location.

[0042] Here, the pre-constructed adjacency matrix is ​​the sum of the adjacency matrices corresponding to each line segment. The specific construction method of the adjacency matrix is ​​described in the following steps 401 to 403, steps 4021 to 4023, and steps 4031 to 4032, and will not be repeated here.

[0043] In some embodiments, the server 200 may be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms. The terminal and the server may be connected via wired or wireless communication, which is not limited in the embodiments of the present application.

[0044] See also Figure 2 , Figure 2 is a schematic diagram of the structure of the server 200 provided in an embodiment of the present application, Figure 2 The server 200 shown includes: at least one processor 210, a memory 250, at least one network interface 220 and a user interface 230. The various components in the server 200 are coupled together via a bus system 240. It is understood that the bus system 240 is used to achieve connection and communication between these components. In addition to the data bus, the bus system 240 also includes a power bus, a control bus and a status signal bus. However, for the sake of clarity, the bus system 240 is not described in detail. Figure 2 Various buses are labeled as bus system 240 .

[0045] The processor 210 may be an integrated circuit chip having signal processing capabilities, such as a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc., wherein the general-purpose processor may be a microprocessor or any conventional processor, etc.

[0046] The user interface 230 includes one or more output devices 231 that enable presentation of media content, including one or more speakers and / or one or more visual display screens. The user interface 230 also includes one or more input devices 232, including user interface components that facilitate user input, such as a keyboard, mouse, microphone, touch screen display, camera, other input buttons and controls.

[0047] The memory 250 may be removable, non-removable, or a combination thereof. Exemplary hardware devices include solid-state memory, hard disk drives, optical disk drives, etc. The memory 250 may optionally include one or more storage devices that are physically remote from the processor 210.

[0048] The memory 250 includes a volatile memory or a non-volatile memory, and may also include both volatile and non-volatile memories. The non-volatile memory may be a read-only memory (ROM), and the volatile memory may be a random access memory (RAM). The memory 250 described in the embodiment of the present application is intended to include any suitable type of memory.

[0049] In some embodiments, memory 250 can store data to support various operations, examples of which include programs, modules, and data structures, or a subset or superset thereof, as exemplarily described below.

[0050] Operating system 251, including system programs for processing various basic system services and performing hardware-related tasks, such as a framework layer, a core library layer, a driver layer, etc., for implementing various basic services and processing hardware-based tasks; A network communication module 252, used to reach other electronic devices via one or more (wired or wireless) network interfaces 220, exemplary network interfaces 220 include: Bluetooth, Wireless Compatibility Certification (WiFi), and Universal Serial Bus (USB), etc.; a presentation module 253 for enabling presentation of information via one or more output devices 231 (e.g., display screen, speaker, etc.) associated with the user interface 230 (e.g., a user interface for operating peripherals and displaying content and information); The input processing module 254 is used to detect one or more user inputs or interactions from one of the one or more input devices 232 and translate the detected inputs or interactions.

[0051] In some embodiments, the device provided in the embodiments of the present application can be implemented in software. Figure 2 The fault prediction device 255 stored in the memory 250 is shown, which can be software in the form of a program and a plug-in, etc., including the following software modules: a data acquisition module 2551, a construction module 2552, a first fault prediction module 2553 and a second fault prediction module 2554. These modules are logical, so they can be arbitrarily combined or further split according to the functions implemented. The functions of each module will be explained below.

[0052] In other embodiments, the device provided in the embodiments of the present application can be implemented in hardware. As an example, the device provided in the embodiments of the present application can be a processor in the form of a hardware decoding processor, which is programmed to execute the fault prediction method provided in the embodiments of the present application. For example, the processor in the form of a hardware decoding processor can adopt one or more application specific integrated circuits (Application Specific Integrated Circuit, ASIC), digital signal processors (Digital Signal Processor, DSP), programmable logic devices (Programmable Logic Device, PLD), complex programmable logic devices (Complex Programmable Logic Device, CPLD), field programmable gate arrays (Field-Programmable Gate Array, FPGA) or other electronic components.

[0053] In some embodiments, the terminal or server can implement the fault prediction method provided in the embodiments of the present application by running various computer executable instructions or computer programs. For example, computer executable instructions can be microprogram-level commands, machine instructions, or software instructions. The computer program can be a native program or software module in the operating system; it can be a native application (APPlication, APP), that is, a program that needs to be installed in the operating system to run, such as a fault prediction APP. In short, the above-mentioned computer executable instructions can be instructions in any form, and the above-mentioned computer program can be an application, module, or plug-in in any form.

[0054] The fault prediction method provided in the embodiment of the present application will be explained in combination with the exemplary application and implementation of the server side provided in the embodiment of the present application.

[0055] The following describes the fault prediction method provided by the embodiment of the present application, which can be applied to scenarios such as power transmission line fault prediction. The fault prediction method provided by the embodiment of the present application can accurately predict faults on transmission lines of different lengths. As mentioned above, the electronic device that implements the fault prediction method of the embodiment of the present application can be a terminal, a server, or a combination of the two. Therefore, the execution subject of each step will not be repeated below.

[0056] See also Figure 3 , Figure 3 is a flowchart of the image processing method provided in the embodiment of the present application, which will be combined with Figure 3 The steps shown are explained. Figure 3 The body of the step is the server.

[0057] In step 101, real-time line data of each first transmission line in an overhead line system is acquired.

[0058] In some embodiments, real-time data on each first transmission line is collected by sensors, and the collected data includes recorded data of secondary side three-phase voltage and current signals, and recorded data of zero-sequence voltage and zero-sequence current signals at both ends of each first transmission line.

[0059] In step 102, for each first power transmission line, a volt-ampere characteristic curve diagram is constructed based on the real-time line data of the first power transmission line.

[0060] It should be noted that each volt-ampere characteristic curve diagram uniquely corresponds to a first transmission line.

[0061] In some embodiments, a volt-ampere characteristic curve diagram needs to be constructed for the current and voltage of each first transmission line. In the volt-ampere characteristic curve diagram, the horizontal axis represents the voltage U in volts (V), and the vertical axis represents the current I in amperes (A). The volt-ampere characteristic curve diagram needs to be constructed by changing the voltage or current in the first transmission line and recording the corresponding voltage and current values. It is usually necessary to perform multiple measurements to obtain enough data points to draw the volt-ampere characteristic curve diagram.

[0062] In step 103, a fault prediction sub-model corresponding to the first transmission line is used to perform prediction based on the volt-ampere characteristic curve diagram to obtain a first fault prediction result of the first transmission line.

[0063] In some embodiments, each volt-ampere characteristic curve diagram is input into the fault prediction sub-model closest to the sensor in the corresponding first transmission line, and fault prediction is performed based on the fault prediction sub-model to obtain a fault prediction result, which includes the fault type and fault location.

[0064] Here, the fault prediction sub-model corresponds to the sensor one-to-one. It can be understood that the location of the sensor will be configured with a fault prediction sub-model that is adapted to the length of the transmission line. Therefore, there is no situation where there are multiple fault prediction sub-models closest to the sensor. The fault type includes whether a fault occurs, and the specific type of fault after the fault occurs, such as phase short circuit, circuit break, line grounding fault, etc. The fault location refers to the location where the fault occurs, such as a tower, or the transmission line between any two towers.

[0065] As an example of step 103, the volt-ampere characteristic curve corresponding to the first transmission line is input into the fault prediction sub-model corresponding to the first transmission line to perform fault prediction, and the first fault prediction result of the first transmission line obtained is: a fault occurs, specifically a circuit breaker fault, and the location of the fault is tower A.

[0066] Through step 103, faults can be predicted for first transmission lines of different lengths using fault prediction sub-models suitable for first transmission lines of different lengths, thereby improving the pertinence of the fault prediction sub-model prediction. The first fault prediction result obtained by the fault prediction sub-model prediction can be used as a basis for fault prediction by the fault prediction overall model, thereby further improving the accuracy of fault prediction.

[0067] In some embodiments, see Figure 4A ,exist Figure 3 Before step 103 shown, steps 201 to 205 may also be performed to train the fault prediction sub-model, which is described in detail below.

[0068] In step 201, a fault prediction sub-model to be trained and fault waveform data corresponding to a plurality of second transmission lines of different candidate lengths are obtained.

[0069] As an example of step 201, the fault waveform data corresponding to multiple second transmission lines of different candidate lengths may be 500km overhead line fault waveform data, 200km overhead line fault waveform data, 100km overhead line fault waveform data, 50km overhead line fault waveform data and 12km overhead line fault waveform data.

[0070] In step 202, for each of the fault waveform data corresponding to the second transmission line of the candidate length, the fault waveform data is preprocessed to obtain training data corresponding to the second transmission line of the candidate length.

[0071] In some embodiments, before training the fault prediction sub-models corresponding to multiple second transmission lines of different candidate lengths, the acquired fault waveform data corresponding to the multiple second transmission lines of different candidate lengths need to be preprocessed, including denoising, normalization, etc., to obtain the training data corresponding to the second transmission lines of each candidate length.

[0072] In step 203, a first length is determined based on an average length of the plurality of different candidate lengths.

[0073] In some embodiments, in order to train and obtain fault prediction sub-models corresponding to different candidate lengths, it is necessary to pre-train a fault prediction sub-model corresponding to the second transmission line of the first length. In order to reduce the adaptation gap between the pre-trained fault prediction sub-model and the fault prediction sub-models corresponding to other different candidate lengths, it is necessary to select a candidate length close to the average value among multiple different candidate lengths as the first length, and the pre-trained fault prediction sub-model corresponds to the first length. In practical applications, the first length can be the candidate length closest to the average value among multiple different candidate lengths, or it can be any one of multiple different candidate lengths, and there is no limitation here.

[0074] As an embodiment of step 203, referring to the example of step 201, the lengths of the overhead lines are 500 km, 200 km, 100 km, 50 km and 12 km, and the average length of the overhead lines is 172.4 km, so the first length is determined to be 200 km.

[0075] In step 204, the fault prediction sub-model to be trained is trained using the training data corresponding to the second transmission line of the first length to obtain a trained fault prediction sub-model corresponding to the second transmission line of the first length.

[0076] In some embodiments, a fault prediction sub-model is trained based on training data corresponding to the fault waveform data of the second transmission line of the first length to obtain a trained fault prediction sub-model corresponding to the second transmission line of the first length, wherein the fault prediction sub-model includes multiple convolutional layers and fully connected layers.

[0077] In step 205, the parameters of the trained fault prediction sub-model corresponding to the second transmission line of the first length are adjusted using the training data corresponding to the second transmission line of other lengths to obtain the trained fault prediction sub-model corresponding to the second transmission line of other lengths.

[0078] The other length is any candidate length among the multiple different candidate lengths except the first length.

[0079] In some embodiments, see Figure 4B , Figure 4A Step 205 shown can be implemented by steps 2051 to 2053 as follows. Figure 4B Provide explanation.

[0080] In step 2051, the parameters of the convolutional layer in the trained fault prediction sub-model corresponding to the second power transmission line of the first length are kept unchanged.

[0081] In step 2052, the parameters of the fully connected layer in the trained fault prediction sub-model corresponding to the second power transmission line of the first length are set to a fourth value to obtain an intermediate fault prediction sub-model.

[0082] In some embodiments, the parameters of the convolutional layer in the trained fault prediction sub-model corresponding to the second transmission line of the first length are retained, and the parameters of the fully connected layer are cleared, that is, the fourth value is 0, to obtain an intermediate fault prediction sub-model.

[0083] In step 2053, the training data corresponding to the second transmission line of other lengths are used to adjust the parameters of the fully connected layer in the intermediate fault prediction sub-model to obtain the trained fault prediction sub-model corresponding to the second transmission line of other lengths.

[0084] Here, the parameters of the fully connected layer in the intermediate fault prediction sub-model are adjusted using the training data corresponding to the second transmission line of other lengths. The fully connected layer in the intermediate fault prediction sub-model is retrained using the training data corresponding to the second transmission line of other lengths to obtain corresponding multiple fault prediction sub-models.

[0085] In the above steps 2051 to 2053, the parameters of the convolutional layer in the trained fault prediction sub-model are kept unchanged, and the parameters of the fully connected layer are adjusted using the training data corresponding to the second transmission lines of other lengths. This allows for rapid training of multiple types of models using limited data, which not only improves the training speed, but also facilitates the subsequent targeted layout of different fault prediction sub-models suitable for different transmission line lengths for transmission lines of different lengths, thereby improving the accuracy of fault detection.

[0086] Through steps 201 to 205, it is possible to train fault prediction sub-models suitable for transmission lines of different lengths, thereby improving the pertinence and accuracy of the fault prediction sub-models in predicting faults.

[0087] In some embodiments, the fault prediction sub-models corresponding to the transmission lines of different candidate lengths are obtained through the above steps 201 to 205. In actual application, the length of the transmission line in the overhead line system may be different from the candidate length. In this case, it is necessary to determine the fault prediction sub-model that matches the actual transmission line from multiple fault prediction sub-models. Figure 5 ,exist Figure 4A After step 205 shown, steps 301 to 303 may be performed to determine a fault prediction sub-model corresponding to the first transmission line, which is described in detail below.

[0088] In step 301, the second length of each of the first transmission lines is obtained.

[0089] Here, the second length of the first power transmission line is the actual length of the first power transmission line measured in actual application. For example, the second length may be 150 km.

[0090] In step 302, for each of the first transmission lines, a first length having the smallest difference with the second length is determined from a plurality of candidate lengths.

[0091] In some embodiments, when determining the fault prediction sub-model corresponding to the second length based on the already trained fault prediction sub-models corresponding to different candidate lengths, it is necessary to determine the candidate difference length closest to the second length among multiple candidate lengths, that is, the candidate length with the smallest difference between the second length.

[0092] In step 303, the trained fault prediction sub-model corresponding to the first length is determined as the fault prediction sub-model corresponding to the first transmission line.

[0093] As an example of steps 301 to 303, for example, the multiple candidate lengths are 200km, 180km, 50km, and 100km respectively; the second length of a first transmission line is 185km, and the first length with the smallest difference between the second length and the multiple candidate lengths is 180km. Therefore, the fault prediction sub-model corresponding to the second length 185km is the fault prediction sub-model corresponding to the first length 180km.

[0094] Through steps 301 to 303, a fault prediction sub-model that best matches the first transmission line length can be determined from a plurality of trained fault prediction sub-models, thereby ensuring the accuracy of the first fault prediction result.

[0095] Continue to see Figure 3 In step 104, a first adjacency matrix corresponding to the overhead line system is obtained. Here, the element value in the first adjacency matrix represents the degree to which the line sections between different sensors are affected by other line sections. The larger the element value in the first adjacency matrix, the greater the degree to which the line sections between different sensors corresponding to the element are affected by other line sections.

[0096] In some embodiments, see Fig. 6A ,exist Figure 3 Before step 104 shown, steps 401 to 403 may also be performed to construct a first adjacency matrix, which is described in detail below.

[0097] In step 401, an electrical diagram of the overhead line system is obtained.

[0098] Wherein, the electrical diagram includes: at least one line section included in each of the first transmission lines, the connection relationship between each of the line sections, and the deployment position of each sensor, and the sensor is used to collect real-time line data of the first transmission line.

[0099] In step 402, for each of the route sections, based on the connection relationship between the route sections and the deployment positions of the sensors, a second adjacency matrix corresponding to the route section is constructed.

[0100] In some embodiments, the electrical diagram needs to be abstracted to obtain an abstract electrical diagram. The generation process of the abstract electrical diagram is as follows: the intersection of the transmission lines is taken as a node, and the transmission lines are divided into multiple sections based on the position of each node. Since the sensor and the fault prediction sub-model are deployed at the same location in the electrical diagram, the sensor and the fault prediction sub-model are abstracted into one entity in the abstract electrical diagram, which can be considered as one sensor.

[0101] In some embodiments, after the fault prediction sub-model predicts the fault of a certain section of the transmission line, it is necessary to make another prediction in the fault prediction overall model. Therefore, it is necessary to construct the status information of sensors on all transmission line sections in the abstract electrical diagram, that is, to construct a second adjacency matrix corresponding to the line section. The elements in the matrix are used to indicate whether there is an available line section between two sensors when a line section fails.

[0102] In some embodiments, see Figure 6B , Fig. 6A Step 402 shown in FIG. 402 can be implemented by steps 4021 to 4023 as follows. Figure 6B Provide explanation.

[0103] In step 4021, for each of the line sections, it is determined whether there is an available line section between the ith sensor and the jth sensor when a fault occurs in the line section.

[0104] In some embodiments, it is necessary to determine the status of other line sections when a line section fails based on the current direction. For example, if the current passes through line sections A, B, C, and D in sequence, and it is assumed that line section C fails, since the current flows from line section C to line section D, line section D also fails. At this time, it is necessary to determine whether there is an available line section between every two sensors, that is, a line section that has not failed.

[0105] If there is an available line section between the ith sensor and the jth sensor, the process proceeds to step 4022 ; if there is no available line section between the ith sensor and the jth sensor, the process proceeds to step 4023 .

[0106] In step 4022, the element in the i-th row and the j-th column of the second adjacency matrix is ​​determined as a first value.

[0107] Wherein, i=1,2,…,N, j=1,2,…,N, and N is the total number of sensors in the overhead line system.

[0108] In some embodiments, as long as there is an available line section between sensor i and sensor j, that is, a line section without any faults, sensor i is considered to have no impact on sensor j, and the values ​​corresponding to the i-th row and j-th column of sensor i and sensor j in the second adjacency matrix are set to the first value, and the first value is 0.

[0109] As an example of step 4022, the line sections between the first sensor and the second sensor are A, B, C, and D. At this time, it is assumed that line section C fails. Since the current flows from line section C to line section D, line section D also fails. At this time, there is an available line section A between the first sensor and the second sensor, so the first sensor has no effect on the second sensor, and the value of the first row and second column in the second adjacency matrix is ​​0.

[0110] In step 4023, the element in the i-th row and the j-th column of the second adjacency matrix is ​​determined as a second value.

[0111] Wherein, i=1,2,…,N, j=1,2,…,N, and N is the total number of sensors in the overhead line system.

[0112] In some embodiments, if there are no available line sections between sensor i and sensor j, that is, all line sections are faulty, sensor i is considered to have an impact on sensor j, and the values ​​corresponding to the i-th row and j-th column of sensor i and sensor j in the second adjacency matrix are set to the second value, and the second value is 1.

[0113] As an example of step 4023, the line sections between the first sensor and the second sensor are A and B. At this time, it is assumed that line section A fails. Since the current flows from line section A to line section B, line section B also fails. At this time, there is no available line section between the first sensor and the second sensor, so the first sensor has an impact on the second sensor, and the value of the first row and second column in the second adjacency matrix is ​​1.

[0114] Continue to see Fig. 6A In step 403, based on the second adjacency matrix corresponding to each of the line sections, a first adjacency matrix corresponding to the overhead line system is determined.

[0115] In some embodiments, see Figure 6C , Fig. 6AStep 403 shown in FIG. 403 can be implemented by steps 4031 to 4032 as follows. Figure 6C Provide explanation.

[0116] In step 4031, the second adjacency matrix corresponding to each of the route sections is bitwise summed to obtain a third adjacency matrix.

[0117] It should be noted that in the training process of the overall fault prediction model, only one adjacency matrix is ​​needed as the convolution kernel for training, so it is necessary to sum the second adjacency matrices corresponding to each line section bit by bit, and the obtained third adjacency matrix is ​​used as the convolution kernel in the training process of the overall fault prediction model. Here, the bitwise summation can unify the adjacency matrix. In the obtained third adjacency matrix, the larger the value of the element, the more it can be said that the line section between the two sensors is easily affected by other line sections.

[0118] As an example of step 4031, for example, the route segments are A, B, and C, and the second adjacency matrix corresponding to route segment A is , the second adjacency matrix corresponding to route segment B is , the second adjacency matrix corresponding to line segment C is , then the third adjacency matrix is .

[0119] In step 4032, the elements in the third adjacency matrix that are smaller than a preset threshold are set to a third value to obtain the first adjacency matrix.

[0120] Here, the third value may be 0, and the values ​​in the third adjacency matrix corresponding to the line sections between the two sensors that are not easily affected by other line sections may be appropriately cleared to reduce the amount of calculation. The value of the preset threshold depends on actual requirements and is not limited here.

[0121] The second adjacency matrix constructed through steps 4021 to 4023 reflects whether there is an influence between sensors when each line fails. The third adjacency matrix constructed through steps 4031 to 4032 can reflect the influence status between sensors in the entire system, and can clarify whether a certain line section is more easily affected by other line sections. The elements in the third adjacency matrix that are less than the preset threshold are set to 0 to obtain the first adjacency matrix, and the state information of each node in the first adjacency matrix is ​​used as the state matrix. The state matrix is ​​used as the input of the overall fault prediction model, which can improve the accuracy of fault prediction.

[0122] Continue to see Figure 3In step 105, the overall fault prediction model is used to perform prediction based on the first adjacency matrix, the first fault prediction result corresponding to the first transmission line and the real-time line data to obtain a second fault prediction result of the overhead line system.

[0123] In some embodiments, see Figure 7 , Figure 3 Step 105 shown can be implemented by the following steps 1051 to 1054, as shown below. Figure 7 Provide explanation.

[0124] In step 1051, a state matrix is ​​constructed based on the first fault prediction result corresponding to the first transmission line and the real-time line data.

[0125] Here, the information in the state matrix includes sensors / corresponding fault prediction sub-models, current constants, real-time current values, current trends, voltage constants, real-time voltage values, voltage trends, zero-sequence voltage, zero-sequence current, whether there is a fault, and fault type, etc.

[0126] It should be noted that, if there is a fault, 1 indicates a fault and 0 indicates no fault; fault type 1 corresponds to short circuit, 2 corresponds to open circuit, and 3 corresponds to grounding. In the current trend and voltage trend, 1 indicates positive current or positive voltage, and 2 indicates reverse current or reverse voltage. Under normal circumstances, the current trend and voltage trend are in the same direction. The range of real-time voltage is -380 to +380v. The real-time current is determined according to the voltage and power factor of the three-phase electricity. Assuming U=380V, =0.9, the power P is about 950W to 1300W, then the current I can be obtained according to the following formula (1) to be about 1.5-2A. Taking 2A as an example, in the zero-sequence voltage, 1 indicates the existence of zero-sequence voltage, and 0 indicates the absence of zero-sequence voltage. In the zero-sequence current, 1 indicates the existence of zero-sequence current, and 0 indicates the absence of zero-sequence current.

[0127] (1) Where I represents current, P represents power, and U represents voltage. is the power factor, Used to convert line voltage into phase voltage.

[0128] In some embodiments, the state matrix uses the state information of each node in the first adjacency matrix as the state matrix, and the state matrix is ​​input into the fault prediction overall model as the input to predict the fault.

[0129] In step 1052, the first adjacency matrix is ​​determined as the convolution kernel of the overall fault prediction model.

[0130] In step 1053, the overall fault prediction model is used to extract features from the state matrix to obtain a feature matrix.

[0131] In some embodiments, feature extraction is implemented using a graph convolutional neural network (GCN), which includes N layers, where N is greater than or equal to 2, and the i-th layer performs feature extraction according to the following formula (2): (2) in represents a nonlinear activation function, D represents the degree matrix of A, A represents the undirected adjacency matrix, i.e., the first adjacency matrix, Indicates the input The feature matrix of the layer, That is, the state matrix, Represents the parameter matrix of the i-th layer. The parameter matrix W needs to be determined by learning during model training. Indicates The feature matrix output by the layer is also used as the feature matrix of the input layer i+1. After the feature extraction of the Nth layer, the feature matrix extracted by GCN is obtained.

[0132] In step 1054, prediction is performed based on the characteristic matrix to obtain a second fault prediction result of the overhead line system.

[0133] Through steps 1051 to 1054, the overall fault prediction model can be used to further perform fault prediction based on the fault prediction by the fault prediction sub-model, thereby improving the accuracy of fault prediction.

[0134] In some embodiments, see Figure 8 ,exist Figure 3 Before step 105 shown, steps 501 to 505 may also be performed to train the overall fault prediction model, which is described in detail below.

[0135] In step 501, the overall fault prediction model to be trained, the historical fault prediction results and historical line data corresponding to each of the first transmission lines, and the actual fault line are obtained.

[0136] It should be noted that after the fault prediction sub-model is trained, it can be deployed first without deploying the overall fault prediction model. The overall fault prediction model can be trained after sufficient data is collected. The historical status data refers to the status data generated by the fault prediction sub-model in the transmission line in the past.

[0137] In some embodiments, the feature matrix extracted by GCN needs to be input into a classifier, which includes a fully connected layer and an activation function such as softmax, and obtains the probability of whether each node fails. The transmission lines with a failure probability higher than a threshold are regarded as lines that may fail, that is, actual fault lines.

[0138] It should be noted that the GCN training steps are the same as those in step 1053 and will not be repeated here.

[0139] In step 502, a historical state matrix is ​​constructed based on the historical fault prediction results and the historical line data.

[0140] In some embodiments, training the overall fault prediction model requires using the historical state information of each node in the first adjacency matrix as a historical state matrix. The nodes in the first adjacency matrix may include sensors, and the historical state information of the sensors includes the historical prediction results and historical line data of the corresponding fault prediction sub-models. The historical state matrix requirements are consistent with the state matrix requirements constructed in step 1051, and are not described in detail here.

[0141] In step 503, the overall fault prediction model to be trained is used to perform prediction based on the historical state matrix to obtain a third prediction result.

[0142] In step 504, a loss value is determined based on the third prediction result and the actual fault line.

[0143] When training GCN and classifiers, the disconnected nodes in the sample data are marked in advance, and the loss value is calculated according to the disconnected probability of each node output by the classifier and the actual disconnected nodes according to the loss function. This application does not impose any restrictions on the setting of the loss function, and the loss function will be selected according to the needs of the actual application.

[0144] In step 505, the loss value is used to perform back-propagation training on the fault prediction overall model to be trained to obtain a trained fault prediction overall model.

[0145] It should be noted that when using the loss value to perform back-propagation training on the trained fault prediction overall model, it is necessary to update the GCN and classifier based on the loss value.

[0146] The training of the overall fault prediction model is implemented through steps 501 to 505. By updating the GCN and the classifier, the training effect of the overall fault prediction model can be further improved, thereby improving the accuracy of fault prediction.

[0147] The following is an explanation of an exemplary application of the embodiments of the present application in a practical application scenario.

[0148] Overhead lines mainly refer to overhead open lines, which are erected above the ground. Insulators are used to fix the transmission wires on the towers erected on the ground and used to transmit electric energy. This type of installation and maintenance is relatively convenient and has low costs, but it is easily affected by weather and environment such as strong winds, lightning strikes, dirt, ice and snow, etc., causing failures. At the same time, the entire transmission corridor occupies a large area of ​​land, which is easy to cause electromagnetic interference to the surrounding environment.

[0149] When an overhead line fails, timely intervention is required and the fault must be handled promptly, quickly and accurately to enhance power supply reliability and improve user satisfaction.

[0150] In order to improve the accuracy of overhead line fault prediction, this application deploys sensors and fault prediction sub-models at multiple key nodes of the power grid line, and uses the fault prediction sub-models that match the power grid line to perform regional fault prediction, which can effectively improve the accuracy and efficiency of fault prediction. At the same time, the overall fault prediction model is used to perform overall fault prediction based on multiple deployed sensors and fault prediction sub-models to further improve the accuracy of the prediction.

[0151] The entire fault prediction system includes multiple sensors, fault prediction sub-models, and a fault prediction overall model. Fig. 9 ,The deployment of the fault prediction system in the power grid line system is as follows: The main line 600 with a total length of 500 km includes four branch lines, namely the first branch line 601, the second branch line 602, the third branch line 603 and the fourth branch line 604. The length of the first branch line 601 is 100 km, the length of the second branch line 602 is 200 km, the length of the third branch line 603 is 50 km, and the length of the fourth branch line 604 is 12 km. Towers 605 are set up at certain intervals on each line.

[0152] Since the location of the fault is different from the location of the sensor, the fault waveform data collected is also different. If a single fault prediction model is used for fault prediction, the accuracy of the fault prediction is poor. Therefore, in this application, in order to improve the accuracy of fault prediction, fault prediction sub-models suitable for different lengths are deployed for lines of different lengths.

[0153] See also Fig. 9In the first branch line 601, a fault prediction sub-model 606 for 100km fault prediction is deployed, in the second branch line 602, a fault prediction sub-model 607 for 200km fault prediction is deployed, in the third branch line 603, a fault prediction sub-model 608 for 50km fault prediction is deployed, and in the fourth branch line 604, a fault prediction sub-model 609 for 12km fault prediction is deployed. At the same time, a fault prediction master model is deployed in the monitoring center 610, and the monitoring center 610 is respectively connected to each fault prediction sub-model. Sensors are also deployed at the towers where each fault prediction sub-model is located to obtain line information in real time.

[0154] See also Fig. 9 , Fig. 9 ABC is three-phase alternating current, which is a power system composed of three AC circuits with the same frequency, equal potential amplitude and 120 degrees phase difference. The line voltage of the industrial three-phase four-wire power supply line, that is, the line voltage between the live wire and the live wire, is 380V, and the phase voltage, that is, the line voltage between the live wire and the neutral wire, is 220V. The three phases here can all be understood as live wires.

[0155] In the model inference stage, first, for each line, each sensor obtains real-time line data, which includes the recording data of the secondary side three-phase voltage and current signals at both ends of the transmission line, and the recording data of the zero-sequence voltage and zero-sequence current signals.

[0156] Secondly, a volt-ampere characteristic curve is constructed based on the line data, and the volt-ampere characteristic curve is input into the fault prediction sub-model closest to the sensor. Fault prediction is performed based on the fault prediction sub-model to obtain a fault prediction result, which includes the fault type and the distance of the fault location. Here, a volt-ampere characteristic curve is constructed for each line data collected by each sensor. The fault prediction sub-model corresponds to the sensor, and a fault prediction sub-model adapted to the line length will be configured at the location where the sensor is located, so there is no problem of multiple nearest sensors. The fault type includes whether a fault occurs, and after the fault occurs, the specific type of fault, such as phase-to-phase short circuit, circuit break, line grounding fault, etc. The fault location refers to the location where the fault occurs, such as a certain tower, or the line between any two towers.

[0157] Finally, the monitoring center performs fault prediction based on the fault prediction overall model, the pre-built adjacency matrix, the results of each fault prediction sub-model, and the real-time line data to obtain the final fault type and fault location. The pre-built adjacency matrix is ​​the final adjacency matrix mentioned below.

[0158] The monitoring center obtains the electrical diagram of the overhead line, which includes the connection relationship of the transmission line and the deployment location of the fault prediction sub-model, and constructs an adjacency matrix based on the connection relationship of the transmission line, sensors and the fault prediction sub-model.

[0159] A state matrix is ​​constructed based on the state information of each node in the adjacency matrix. The nodes in the adjacency matrix may include sensors, and the state information of the sensors includes the prediction results of the corresponding fault prediction sub-model and real-time line data. The state matrix is ​​used as the input of the overall fault prediction model, and the adjacency matrix is ​​used as the convolution kernel of the graph neural network in the overall fault prediction model. The overall fault prediction model is used to perform fault prediction and obtain the final fault prediction result. The final fault prediction result represents whether a fault occurs and the fault location when a fault occurs.

[0160] For details about the status information, see Table 1.

[0161] Table 1, Status Information Table

[0162] In Table 1, 1 indicates fault and 0 indicates no fault; fault type 1 corresponds to short circuit, 2 corresponds to open circuit, and 3 corresponds to grounding. In the current trend and voltage trend, 1 indicates positive current or positive voltage, and 2 indicates reverse current or reverse voltage. Under normal circumstances, the current trend and voltage trend are in the same direction. The range of real-time voltage is -380V to +380V. The real-time current is determined according to the voltage and power factor of the three-phase electricity. Assuming U=380V, =0.9, the power P is about 950W to 1300W, then the current I can be obtained according to the following formula (3) to be about 1.5-2A. Taking 2A as an example, in the zero-sequence voltage, 1 indicates the existence of zero-sequence voltage, and 0 indicates the absence of zero-sequence voltage. In the zero-sequence current, 1 indicates the existence of zero-sequence current, and 0 indicates the absence of zero-sequence current.

[0163] (3) Where I represents current, P represents power, and U represents voltage. is the power factor, Used to convert line voltage into phase voltage.

[0164] Will Fig. 9 Simplified, the intersection of transmission lines is represented by a triangle, and the triangle is used as the boundary. Fig. 9 The five transmission lines shown are divided into nine sections, and nine line sections are obtained (corresponding to the line sections in other embodiments). The sensors, i.e., the fault prediction sub-models, are represented by circles, and five circles A, B, C, D, and E are obtained respectively. The simplified Fig. 9 See also Fig.10 .

[0165] For the first line segment 701, the adjacency matrix between sensors is constructed based on the criterion that the failure of the line segment will affect the subsequent line segments. The adjacency matrix is ​​shown in Table 2: Table 2, Adjacency matrix constructed when the first line segment 701 fails

[0166] When the first line segment 701 fails, the subsequent second line segment 702, third line segment 703, fourth line segment 704, fifth line segment 705, sixth line segment 706, seventh line segment 707, eighth line segment 708, and ninth line segment 709 will all be affected, that is, the second line segment 702, sixth line segment 706, third line segment 703, seventh line segment 707, and fourth line segment 704 between sensors A and B are all affected, that is, the number in column B of row A is 1, and at the same time, the sixth line segment 706 in row A will also be affected, that is, the number in column A of row A is also 1, and the same applies to the others.

[0167] For the second line segment 702, an adjacency matrix between sensors is constructed based on whether a failure in the line segment will affect subsequent line segments. The adjacency matrix is ​​shown in Table 3: Table 3, Adjacency matrix constructed when the second line segment 702 fails

[0168] When the second line segment 702 fails, the subsequent third line segment 703, fourth line segment 704, fifth line segment 705, seventh line segment 707, eighth line segment 708, and ninth line segment 709 will be affected. The sixth line segment 706 where sensor A is located and the first line segment 701 are not affected. Therefore, row A and column A are 0. The line segments from A to B are the sixth line segment 706, the second line segment 702, the third line segment 703, and the fourth line segment 704. Because there is an unaffected sixth line segment 706, the value of row A and column B is also 0, that is, as long as there is an unaffected line segment, the corresponding value is 0, and the same applies to others.

[0169] For the third line segment 703, the adjacency matrix between sensors is constructed based on the criterion that the failure of the line segment will affect the subsequent line segments. The adjacency matrix is ​​shown in Table 4: Table 4: Adjacency matrix constructed when the third line segment 703 fails

[0170] When the third line segment 703 fails, the subsequent fourth line segment 704 and fifth line segment 705 will be affected, and the same applies to the others.

[0171] For the fourth line segment 704, the adjacency matrix between sensors is constructed based on the criterion that the failure of the line segment will affect the subsequent line segments. The adjacency matrix is ​​shown in Table 5: Table 5, Adjacency matrix constructed when the fourth line segment 704 fails

[0172] Taking the above adjacency matrix as an example, after each line segment from the first line segment 701 to the ninth line segment 709 fails, an adjacency matrix is ​​constructed to obtain the adjacency matrices corresponding to the 9 line segments. The 9 adjacency matrices are added together to obtain the final adjacency matrix. Optionally, the relationship below 3 can be deleted to obtain the final adjacency matrix. Here, the addition is because the final adjacency matrix only uses 1 as the convolution kernel. There are 9 adjacency matrices here. On the one hand, they need to be unified. On the other hand, the larger the value, the more it means that the line segment between the two sensor nodes is easily affected by other line segments. The deletion below 3 is to reduce the amount of calculation. The values ​​with less impact on the line segment are appropriately deleted and not calculated.

[0173] The overall fault prediction model is a pre-trained graph neural network, specifically a graph convolutional neural network GCN.

[0174] In the model training phase, the first step is to train the fault prediction model and obtain the fault waveform data of various transmission lines, including 500km overhead line fault waveform data, 200km overhead line fault waveform data, 100km overhead line fault waveform data, 50km overhead line fault waveform data, and 12km overhead line fault waveform data. The fault waveform data is the same as the data in the inference phase, and also includes the recording data of the three-phase voltage and current signals. The fault waveform data of various transmission lines are preprocessed, including denoising and normalization, to obtain training data.

[0175] The convolutional neural network CNN is trained based on the training data corresponding to the 200km overhead line fault waveform data to obtain the 200km fault prediction sub-model. The convolutional neural network includes multiple convolutional layers and fully connected layers. The parameters of the convolutional layer in the 200km fault prediction sub-model are retained, and the parameters of the fully connected layer are cleared to obtain the intermediate fault prediction sub-model. This operation is to quickly train multiple types of models using limited data. On the one hand, it can improve the training speed, and on the other hand, it can facilitate the subsequent targeted layout of different fault models for different lines to improve the accuracy of fault detection. 200km is chosen because it is necessary to select an intermediate value so that the gap with other data is more moderate.

[0176] The fully connected layer in the intermediate fault prediction sub-model is retrained using the training data corresponding to the 500km overhead line fault waveform data, the training data corresponding to the 100km overhead line fault waveform data, the training data corresponding to the 50km overhead line fault waveform data, and the training data corresponding to the 12km overhead line fault waveform data to obtain the corresponding multiple fault prediction sub-models.

[0177] The second step is to train the overall fault prediction model. The electrical diagram of the overhead line is obtained. The electrical diagram includes the connection relationship of the transmission line and the deployment location of the fault prediction sub-model. The adjacency matrix is ​​constructed based on the connection relationship of the transmission line, the sensor and the fault prediction sub-model. The historical state information of each node in the adjacency matrix is ​​used as the historical state matrix. The nodes in the adjacency matrix may include sensors. The historical state information of the sensor includes the historical prediction results and historical line data of its corresponding fault prediction sub-model. After the fault prediction sub-model is trained, it can be deployed first without deploying the overall fault prediction model. After collecting enough data, the overall fault prediction model can be trained. The historical state data refers to the state data generated by the fault prediction sub-model in the past.

[0178] The graph convolutional neural network GCN is used for feature extraction. GCN consists of N layers, where N is greater than or equal to 2. The feature extraction is performed according to the following formula (4).

[0179] (4) in, represents a nonlinear activation function, D represents the degree matrix of A, A represents an undirected adjacency matrix, Represents the feature matrix of the input layer i, W i Represents the parameter matrix of the i-th layer. The parameter matrix W needs to be determined by learning during model training. It represents the feature matrix output by the i-th layer, and also serves as the feature matrix input to the i+1-th layer. After the feature extraction of the N-th layer, the feature matrix extracted by GCN is obtained.

[0180] Finally, the extracted feature matrix is ​​input into the classifier, which includes a fully connected layer and an activation function such as softmax, to obtain the probability of whether each node has a fault, and the transmission lines with a fault probability higher than the threshold are regarded as lines that may have faults.

[0181] When training GCN and classifiers, the disconnected nodes in the sample data are marked in advance, and the loss value is calculated based on the disconnection probability of each node output by the classifier and the actual disconnected nodes, so as to update the GCN and classifiers.

[0182] Based on the electrical diagram of the overhead line, an adjacency matrix is ​​constructed, and the node information of the adjacency matrix is ​​used as the state matrix. Fault prediction and location are performed based on the adjacency matrix, state matrix and fault prediction overall model, thereby improving the accuracy of fault prediction. The transfer learning method is used to save the convolutional layer parameters in the fault prediction sub-model trained with 200km training data, and the training data of other different lines are used to train the fault prediction sub-models suitable for different line lengths, thereby improving the pertinence and accuracy of fault prediction.

[0183] The following is a description of an exemplary structure of the fault prediction device 255 provided in the embodiment of the present application implemented as a software module. In some embodiments, see Figure 2 , the software modules stored in the fault prediction device 255 of the memory 250 may include: The data acquisition module 2551 is used to acquire the real-time line data of each first transmission line in the overhead line system.

[0184] The construction module 2552 is used to construct a volt-ampere characteristic curve diagram for each first power transmission line based on the real-time line data of the first power transmission line.

[0185] The first fault prediction module 2553 is used to use the fault prediction sub-model corresponding to the first transmission line to perform prediction based on the volt-ampere characteristic curve diagram to obtain a first fault prediction result of the first transmission line.

[0186] The second fault prediction module 2554 is used to obtain the first adjacency matrix corresponding to the overhead line system, and use the overall fault prediction model to perform prediction based on the first adjacency matrix, the first fault prediction result corresponding to the first transmission line and the real-time line data to obtain a second fault prediction result of the overhead line system.

[0187] In some embodiments, the second fault prediction module 2554 is also used to construct a state matrix based on the first fault prediction result corresponding to the first transmission line and the real-time line data; determine the first adjacency matrix as the convolution kernel of the fault prediction overall model, and use the fault prediction overall model to extract features from the state matrix to obtain a feature matrix; and perform prediction based on the feature matrix to obtain a second fault prediction result of the overhead line system.

[0188] The fault prediction device 255 also includes a first adjacency matrix construction module, which is used to obtain an electrical diagram of the overhead line system, wherein the electrical diagram includes: at least one line section included in each of the first transmission lines, the connection relationship between each of the line sections, and the deployment position of each sensor, and the sensor is used to collect real-time line data of the first transmission line; for each of the line sections, based on the connection relationship between each of the line sections and the deployment position of each sensor, a second adjacency matrix corresponding to the line section is constructed; based on the second adjacency matrix corresponding to each of the line sections, a first adjacency matrix corresponding to the overhead line system is determined.

[0189] The fault prediction device 255 also includes a second adjacency matrix construction module, which is used for, for each of the line sections, when a fault occurs in the line section, if there is an available line section between the ith sensor and the jth sensor, determining the element in the ith row and jth column of the second adjacency matrix as a first value, wherein i=1,2,…,N, j=1,2,…,N, and N is the total number of sensors in the overhead line system; if there is no available line section between the ith sensor and the jth sensor, determining the element in the ith row and jth column of the second adjacency matrix as a second value.

[0190] The fault prediction device 255 also includes a third adjacency matrix construction module, which is used to perform bitwise summation of the second adjacency matrices corresponding to each of the line sections to obtain a third adjacency matrix; and set the elements in the third adjacency matrix that are less than a preset threshold to a third value to obtain the first adjacency matrix.

[0191] The fault prediction device 255 also includes a first fault prediction sub-model training module, which is used to obtain the fault prediction sub-model to be trained and the fault waveform data corresponding to multiple second transmission lines of different candidate lengths; for the fault waveform data corresponding to each of the second transmission lines of the candidate length, the fault waveform data is pre-processed to obtain the training data corresponding to the second transmission line of the candidate length; based on the average length of the multiple different candidate lengths, the first length is determined; using the training data corresponding to the second transmission line of the first length, the fault prediction sub-model to be trained is trained to obtain the trained fault prediction sub-model corresponding to the second transmission line of the first length; using the training data corresponding to the second transmission lines of other lengths, the parameters of the trained fault prediction sub-model corresponding to the second transmission line of the first length are adjusted to obtain the trained fault prediction sub-model corresponding to the second transmission lines of the other lengths, and the other lengths are any candidate lengths among the multiple different candidate lengths except the first length.

[0192] The fault prediction device 255 also includes a second fault prediction sub-model training module, which is used to keep the parameters of the convolution layer in the trained fault prediction sub-model corresponding to the second transmission line of the first length unchanged, set the parameters of the fully connected layer in the trained fault prediction sub-model corresponding to the second transmission line of the first length to a fourth value, and obtain an intermediate fault prediction sub-model; use the training data corresponding to the second transmission lines of other lengths to adjust the parameters of the fully connected layer in the intermediate fault prediction sub-model to obtain the trained fault prediction sub-models corresponding to the second transmission lines of other lengths.

[0193] The fault prediction device 255 also includes an actual fault prediction sub-model determination module, which is used to obtain the second length of each of the first transmission lines; for each of the first transmission lines, determine the first length with the smallest difference with the second length from the multiple candidate lengths; and determine the trained fault prediction sub-model corresponding to the first length as the fault prediction sub-model corresponding to the first transmission line.

[0194] The fault prediction device 255 also includes a fault prediction total model training module, which is used to obtain the fault prediction total model to be trained, the historical fault prediction results and historical line data corresponding to each of the first transmission lines, and the actual fault line; based on the historical fault prediction results and the historical line data, a historical state matrix is ​​constructed; using the fault prediction total model to be trained, prediction is performed based on the historical state matrix to obtain a third prediction result; based on the third prediction result and the actual fault line, a loss value is determined; and the loss value is used to perform back-propagation training on the fault prediction total model to be trained to obtain a trained fault prediction total model.

[0195] The embodiment of the present application provides a computer program product, which includes a computer program or a computer executable instruction, and the computer program or the computer executable instruction is stored in a computer-readable storage medium. The processor of the electronic device reads the computer executable instruction from the computer-readable storage medium, and the processor executes the computer executable instruction, so that the electronic device executes the fault prediction method described in the embodiment of the present application.

[0196] An embodiment of the present application provides a computer-readable storage medium, which stores computer-executable instructions or a computer program. When the computer-executable instructions or the computer program are executed by a processor, the processor will execute the fault prediction method provided by the embodiment of the present application.

[0197] In some embodiments, the computer-readable storage medium may be a memory such as RAM, ROM, flash memory, magnetic surface memory, optical disk, or CD-ROM; or may be various devices including one or any combination of the above memories.

[0198] In some embodiments, computer executable instructions may be in the form of a program, software, software module, script or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as a stand-alone program or as a module, component, subroutine or other unit suitable for use in a computing environment.

[0199] As an example, computer-executable instructions may, but need not, correspond to a file in a file system, may be stored as part of a file storing other programs or data, such as in one or more scripts in a HyperText Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple coordinated files (e.g., files storing one or more modules, subroutines, or code portions).

[0200] As an example, computer executable instructions may be deployed to be executed on one electronic device, or on multiple electronic devices located at one site, or on multiple electronic devices distributed at multiple sites and interconnected by a communication network.

[0201] In summary, through the embodiments of the present application, it is possible to train fault prediction sub-models suitable for different line lengths, and combine the fault prediction overall model to perform fault prediction and location, thereby improving the accuracy and pertinence of fault prediction and location.

[0202] The above is only an embodiment of the present application and is not intended to limit the protection scope of the present application. Any modifications, equivalent substitutions and improvements made within the spirit and scope of the present application are included in the protection scope of the present application.

Claims

1. A fault prediction method, characterized in that: The method comprises: Acquiring real-time line data of each first transmission line in the overhead line system; For each first power transmission line, construct a volt-ampere characteristic curve diagram based on the real-time line data of the first power transmission line; Using the fault prediction sub-model corresponding to the first transmission line, and based on the volt-ampere characteristic curve diagram, a prediction is performed to obtain a first fault prediction result of the first transmission line; A first adjacency matrix corresponding to the overhead line system is obtained, and a fault prediction overall model is used to perform prediction based on the first adjacency matrix, the first fault prediction result corresponding to the first transmission line, and real-time line data to obtain a second fault prediction result of the overhead line system.

2. The method according to claim 1, characterized in that The method of using the overall fault prediction model to perform prediction based on the first adjacency matrix, the first fault prediction result corresponding to the first transmission line, and the real-time line data to obtain a second fault prediction result of the overhead line system includes: constructing a state matrix based on a first fault prediction result corresponding to the first transmission line and real-time line data; Determine the first adjacency matrix as the convolution kernel of the overall fault prediction model, and use the overall fault prediction model to perform feature extraction on the state matrix to obtain a feature matrix; Prediction is performed based on the characteristic matrix to obtain a second fault prediction result of the overhead line system.

3. The method according to claim 1, characterized in that The method further comprises: Acquire an electrical diagram of the overhead line system, the electrical diagram comprising: at least one line section included in each of the first transmission lines, a connection relationship between each of the line sections, and a deployment position of each sensor, the sensor being used to collect real-time line data of the first transmission line; For each of the route sections, based on the connection relationship between the route sections and the deployment position of each sensor, construct a second adjacency matrix corresponding to the route section; Based on the second adjacency matrix corresponding to each of the line sections, a first adjacency matrix corresponding to the overhead line system is determined.

4. The method according to claim 3, characterized in that The step of constructing, for each of the route sections, a second adjacency matrix corresponding to the route section based on the connection relationship between the route sections and the deployment positions of the sensors, includes: For each of the line sections, when a fault occurs in the line section, if there is an available line section between the i-th sensor and the j-th sensor, determine the element of the i-th row and j-th column of the second adjacency matrix as a first value, where i=1,2,…,N, j=1,2,…,N, and N is the total number of sensors in the overhead line system; If there is no available route section between the i-th sensor and the j-th sensor, the element in the i-th row and j-th column of the second adjacency matrix is ​​determined as a second value.

5. The method according to claim 3, characterized in that: The determining, based on the second adjacency matrix corresponding to each of the line sections, a first adjacency matrix corresponding to the overhead line system comprises: Performing bitwise summation on the second adjacency matrices corresponding to the route sections to obtain a third adjacency matrix; The elements in the third adjacency matrix that are smaller than a preset threshold are set to a third value to obtain the first adjacency matrix.

6. The method according to any one of claims 1 to 5, characterized in that: The method further comprises: Acquire a fault prediction sub-model to be trained and fault waveform data corresponding to a plurality of second transmission lines of different candidate lengths; For each of the fault waveform data corresponding to the second transmission line of the candidate length, preprocessing the fault waveform data to obtain training data corresponding to the second transmission line of the candidate length; Determining a first length based on an average length among the plurality of different candidate lengths; Using the training data corresponding to the second transmission line of the first length, training the fault prediction sub-model to be trained to obtain a trained fault prediction sub-model corresponding to the second transmission line of the first length; By using the training data corresponding to the second transmission lines of other lengths, the parameters of the trained fault prediction sub-model corresponding to the second transmission lines of the first length are adjusted to obtain the trained fault prediction sub-model corresponding to the second transmission lines of the other lengths, wherein the other lengths are any candidate lengths among the multiple different candidate lengths except the first length.

7. The method according to claim 6, characterized in that The method of using the training data corresponding to the second transmission line of other lengths to adjust the parameters of the trained fault prediction sub-model corresponding to the second transmission line of the first length to obtain the trained fault prediction sub-model corresponding to the second transmission line of other lengths includes: Keeping unchanged the parameters of the convolutional layer in the trained fault prediction sub-model corresponding to the second transmission line of the first length, setting the parameters of the fully connected layer in the trained fault prediction sub-model corresponding to the second transmission line of the first length to a fourth value, to obtain an intermediate fault prediction sub-model; The training data corresponding to the second transmission lines of other lengths are used to adjust the parameters of the fully connected layer in the intermediate fault prediction sub-model to obtain the trained fault prediction sub-model corresponding to the second transmission lines of other lengths.

8. The method according to claim 7, characterized in that The method further comprises: Obtaining a second length of each of the first transmission lines; For each of the first transmission lines, determine a first length having the smallest difference with the second length from a plurality of the candidate lengths; The trained fault prediction sub-model corresponding to the first length is determined as the fault prediction sub-model corresponding to the first transmission line.

9. The method according to any one of claims 1 to 5, characterized in that: The method further comprises: Obtaining the overall fault prediction model to be trained, historical fault prediction results and historical line data corresponding to each of the first transmission lines, and actual fault lines; Based on the historical fault prediction results and the historical line data, construct a historical state matrix; Using the overall fault prediction model to be trained, prediction is performed based on the historical state matrix to obtain a third prediction result; Determining a loss value based on the third prediction result and the actual fault line; The loss value is used to perform back-propagation training on the fault prediction overall model to be trained to obtain a trained fault prediction overall model.

10. A fault prediction device, characterized in that: The device comprises: A data acquisition module, used to acquire real-time line data of each first transmission line in the overhead line system; A construction module, configured to construct a volt-ampere characteristic curve diagram for each first power transmission line based on real-time line data of the first power transmission line; A first fault prediction module, configured to use a fault prediction sub-model corresponding to the first transmission line to perform prediction based on the volt-ampere characteristic curve diagram to obtain a first fault prediction result of the first transmission line; The second fault prediction module is used to obtain a first adjacency matrix corresponding to the overhead line system, and use a total fault prediction model to perform prediction based on the first adjacency matrix, the first fault prediction result corresponding to the first transmission line and real-time line data to obtain a second fault prediction result of the overhead line system.

11. An electronic device, characterized in that: The electronic device comprises: A memory for storing computer executable instructions or computer programs; A processor, configured to implement the method according to any one of claims 1 to 9 when executing computer executable instructions or computer programs stored in the memory.

12. A computer-readable storage medium storing computer-executable instructions or a computer program, characterized in that: When the computer executable instructions or computer programs are executed by a processor, the method according to any one of claims 1 to 9 is implemented.

13. A computer program product comprising computer executable instructions or a computer program, characterized in that: When the computer executable instructions or computer programs are executed by a processor, the method according to any one of claims 1 to 9 is implemented.

Images