Method, system and equipment for identifying and repairing power grid fault and medium

Abstract

The invention discloses a power grid fault identification and repair method, system and device, and a medium. The method comprises the following steps: constructing a power distribution network digital twin model containing a doubly-fed induction generator; simulating fault conditions of different nodes in a power grid according to the digital twin model of the power distribution network, and generating real-time fault data; the real-time fault data is input into a preset InceptionTime neural network model for fault recognition, and a fault recognition result is obtained; and scheduling the maintenance unit to a fault node for fault repair by adopting a scheduling algorithm based on deep reinforcement learning according to a fault identification result. According to the method, the power grid fault is identified and repaired by introducing the digital twin model, the InceptionTime neural network, the deep reinforcement learning algorithm and other technologies, so that the fault identification precision and robustness are improved, the fault repair path planning is optimized, the fault repair time is shortened, and the fault repair efficiency is improved.

Description

Technical Field

[0001] This application belongs to the technical field of smart grids, and particularly relates to a method, system, device, and medium for identifying and repairing power grid faults. Background Art

[0002] The main impacts of storm surges on power grids are urban waterlogging, as well as accompanying disasters such as storms and heavy rains. The superposition of storm surges and astronomical tides is extremely likely to cause abnormal water level rises, leading to seawater backflow. If accompanied by river floods and waterlogging caused by heavy rains, it is extremely easy to cause floods beyond the urban drainage capacity due to the pushing of high tide levels, resulting in urban waterlogging and the inundation of low-lying areas. Storm surges can also damage other urban infrastructure, causing failures in water conservancy and transportation networks. The floods brought by storm surges are very likely to cause three-phase short-circuit faults in the distribution network. At the same time, strong winds may blow down trees, causing three-phase short circuits in overhead lines. Heavy rains will also greatly increase the risk of short circuits at cable joints with damaged insulation layers. If three-phase short-circuit faults are not properly handled, power outages are extremely likely to occur.

[0003] Currently, in terms of fault identification, there are already methods that use convolutional neural networks (CNNs) to classify and identify the two-dimensional time-frequency energy spectra of fault signals, achieving remarkable results. Another patent combines CNN with BiGRU to achieve accurate identification of power grid fault points. However, these traditional fault analysis methods are often post-mortem analyses and cannot achieve online real-time tracking and prediction. Digital twin (DT) technology can achieve online identification of faults through real-time simulation analysis and even predict potential abnormal events. For example, a patent has achieved accurate simulation by establishing a digital twin model of a doubly-fed induction generator (DFIG) distribution network and used extended causal convolution with skip connections to extract power grid fault features, achieving good fault identification effects. Although digital twin technology can achieve online identification of faults through real-time simulation analysis, the existing patent methods require manually specifying the optimal time window size during the training process, which increases the training cost and may affect the identification accuracy.

[0004] In terms of fault repair, existing patents have proposed some restoration strategies, including fault restoration strategies based on the intelligent software development point (SOP) of new power electronic devices, restoration decisions considering the time-varying characteristics of photovoltaic and loads, and resilient operation strategies based on robust model predictive control. However, these patented technologies have not yet integrated the identification and restoration of faults into a coordinated workflow and have not fully considered the impact of traffic network status on maintenance scheduling. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method, system, device, and medium for identifying and repairing power grid faults to solve the limitations in the identification and repair of power grid faults under extreme weather conditions in related technologies.

[0006] To solve the above technical problems, the present application provides the following technical solutions: In a first aspect, the present application provides a method for identifying and repairing power grid faults, including: Constructing a digital twin model of a distribution network containing doubly-fed induction generators; According to the digital twin model of the distribution network, simulating fault conditions at different nodes in the power grid to generate real-time fault data; Inputting the real-time fault data into a preset InceptionTime neural network model for fault identification to obtain the results of fault identification; According to the results of the fault identification, using a scheduling algorithm based on deep reinforcement learning to schedule the maintenance unit to the fault node for fault repair.

[0007] Further, the preset InceptionTime neural network model adopts the following loss function: ; Wherein, represents the loss function of the InceptionTime neural network model, represents the probability of identifying the fault category as , represents the fault label corresponding to the time series signal sample, M represents the total number of fault categories, and c is the fault category.

[0008] Further, the scheduling algorithm based on deep reinforcement learning adopts the following loss function: ; Wherein, is the loss function of the scheduling algorithm based on deep reinforcement learning; is the value predicted by the network with parameter Q at the current state, where Q the s j value represents the expected value of the long-term cumulative reward that the agent can obtain after taking the action a j in the state; is the maximum s j+1 value in the next state Q ; is the reward during the state transition, and is the discount rate parameter.

[0009] Further, according to the result of the fault identification, a scheduling algorithm based on deep reinforcement learning is adopted to schedule the maintenance unit to the fault node for fault repair, including: Establish a coupled power and transportation network; According to the coupled network of power and transportation and the constraint conditions of the preset maintenance path, a scheduling algorithm based on deep reinforcement learning is used to generate the maintenance path of the maintenance unit; According to the maintenance path, control the maintenance unit to repair the fault.

[0010] Further, when using the scheduling algorithm based on deep reinforcement learning to generate the maintenance path of the maintenance unit, the reward function is set as follows: ; where, is the reward function, represents the time step the abscissa of the node where the maintenance unit is located in the transportation network at time represents the abscissa of the fault node; represents the time step the ordinate of the node where the maintenance unit is located in the transportation network at time represents the ordinate of the fault node; is the discount factor, is the distance between the maintenance unit and the fault node.

[0011] Further, the constraint conditions of the maintenance path are as follows: ; where, is a binary variable indicating whether the maintenance unit repairs node n at time step ; n is a maintenance node; N represents the set of all maintenance nodes; t is the time step, is the set of time steps.

[0012] Further, when scheduling the maintenance unit to the fault node for fault repair, the following constraint conditions need to be satisfied: ; where, is a binary variable indicating whether the damaged component is repaired at time step ; is a binary variable indicating whether the maintenance unit is repairing the component j at time step w ; is the set of damaged components; represents repairing the component at time step ​w Required duration; Indicates repairing damaged components w Required quantity of resources; Indicates the resource capacity of the maintenance unit.

[0013] In a second aspect, the present application also provides a power grid fault identification and repair system, including: A construction module for constructing a digital twin model of a distribution network containing a doubly-fed induction generator; A data generation module for simulating fault conditions of different nodes in the power grid according to the digital twin model of the distribution network and generating real-time fault data; A fault identification module for inputting the real-time fault data into a preset InceptionTime neural network model for fault identification to obtain the result of fault identification; A fault repair module for scheduling a maintenance unit to a fault node for fault repair according to the result of the fault identification by using a scheduling algorithm based on deep reinforcement learning.

[0014] In a third aspect, the present application also provides a computer electronic device, including a memory and a processor, where the memory stores a computer program, and when the processor executes the computer program, the steps of the power grid fault identification and repair method described in any one of the above are implemented.

[0015] In a fourth aspect, the present application also provides a computer-readable storage medium, where the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the power grid fault identification and repair method described in any one of the above are implemented.

[0016] The power grid fault identification and repair method, system, device and medium provided by the present application have the following beneficial effects: 1. The digital twin model of the doubly-fed distribution network constructed based on the electromagnetic transient program (EMTP) of the present application generates more stable real-time data than Simulink (a visualization multi-domain simulation and model design tool based on MATLAB), providing highly reliable samples for the InceptionTime neural network; through multi-scale parallel convolution and adaptive time window technology, the recognition accuracy of 16 typical faults reaches 98%, overcoming the limitation of traditional methods relying on manual parameter tuning (such as CNN / BiGRU requiring manual specification of the time window).

[0017] 2. Combining the power grid - traffic network coupling model, the deep reinforcement learning (DRL) algorithm can still dynamically optimize the path decision of the maintenance unit (RU) under the interference of a 20% road accident probability. Faults can be repaired within 10 time steps in three groups of test scenarios, shortening the power outage time by more than 50% compared with manual scheduling.

[0018] 3. Train the robustness of the RU through probabilistic random blinding processing, enabling it to complete tasks through strategies such as pausing and detouring even in case of communication interruption or road paralysis; the collaborative design of digital twin and DRL can be extended to multi-disaster coupling scenarios such as typhoons and heavy rains. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0020] Figure 1 is a schematic flowchart of a method for identifying and repairing power grid faults in an embodiment of the present application; Figure 2 is a schematic diagram of a digital twin model of a distribution network with a doubly-fed induction generator in an embodiment of the present application; Figure 3 is a schematic diagram of the process of the identification and repair method in an embodiment of the present application; Figure 4 is a schematic diagram of the process of an InceptionTime neural network for identifying power grid faults in an embodiment of the present application; Figure 5 is a simulation schematic diagram of the path planning of a maintenance unit in an embodiment of the present application; Figure 6 is a schematic structural diagram of a power grid fault identification and repair system in an embodiment of the present application; Figure 7 is a schematic structural diagram of a computer electronic device in an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present application.

[0022] It should be noted that when an element is referred to as "fixed to" another element, it can be directly on the other element or there can also be an intermediate element. When an element is considered to be "connected" to another element, it can be directly connected to the other element or there may be an intermediate element at the same time. In contrast, when an element is referred to as being "directly on" another element, there is no intermediate element. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only.

[0023] In this application, unless otherwise clearly defined and limited, terms such as "installed", "connected", "joined", "fixed" and the like shall be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or integrated; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two elements or the interaction relationship between two elements. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0024] In addition, the terms "first" and "second" are used only for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of this application, "a plurality of" means two or more unless otherwise clearly and specifically defined.

[0025] The terms used in one or more embodiments of this application are only for the purpose of describing specific embodiments and are not intended to limit one or more embodiments of this application. The singular forms of "a", "the" and "said" used in one or more embodiments of this application are also intended to include the plural forms unless the context clearly indicates otherwise.

[0026] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which this application belongs. The terms used in the specification of this template are only for the purpose of describing specific embodiments and are not intended to limit this application. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0027] It should be understood that although the terms first, second, etc. may be used in one or more embodiments of the present application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from each other. For example, without departing from the scope of one or more embodiments of the present application, the first may also be referred to as the second, and similarly, the second may also be referred to as the first. Depending on the context, the word "if" as used herein may be interpreted as "when" or "while".

[0028] Currently, the following problems exist in the prior art in dealing with power grid fault identification and repair under extreme weather conditions: 1. Existing algorithms are difficult to achieve a balance among identification accuracy, real-time performance, and adaptability, especially the ability to identify new fault characteristics after the access of new energy is insufficient; 2. There is a lack of efficient parallel computing methods to process massive real-time monitoring data; 3. Existing methods fail to effectively coordinate the relationship between fault identification and maintenance resource scheduling, and lack a robust strategy to cope with the uncertainty of the transportation network; 4. The two links of fault identification and repair are often designed independently, and an overall solution for closed-loop optimization cannot be formed.

[0029] The technical solutions of the present application and how the technical solutions of the present application solve the above technical problems will be described in detail below with specific embodiments. These specific embodiments below can be combined with each other, and the same or similar concepts or processes will not be repeated in some embodiments. The embodiments of the present application will be described below with reference to the drawings.

[0030] Please refer to Figure 1 , a method for identifying and repairing power grid faults provided by an embodiment of the present application includes at least the following steps: S10. Construct a digital twin model of a distribution network containing a doubly-fed induction generator.

[0031] Specifically, in this embodiment, the electromagnetic transient program (EMTP) is used to model the electrical part and mechanical part of the doubly-fed wind turbine, and a full electromagnetic transient model of the doubly-fed wind turbine is constructed. The control part of this model includes pitch angle control, drive train control, rotor side converter (RSC) control, grid side converter (GSC) control, and phase-locked loop (PLL) control. The system network corresponding to the digital twin of the wind turbine connected to the grid is described by the node voltage equation, assuming that the system node admittance matrix is known; the corresponding system wind turbine equipment is described by the state equation, assuming that the state matrix is known. The digital twin model of the distribution network can be referred to Figure 2 .

[0032] S20. According to the digital twin model of the distribution network, simulate the fault conditions of different nodes in the power grid to generate real-time fault data.

[0033] Specifically, in this embodiment, a digital twin model (DT) composed of a doubly-fed induction generator (DFIG) and a distribution network will generate the data required for the subsequent fault identification stage.

[0034] It should be noted that the modeling method based on the electromagnetic transient program (EMTP) has a faster response speed than time-domain simulation, can generate more stable and accurate data than Simulink (an important tool in MATLAB, mainly used for modeling, simulation, and analysis of dynamic systems, and has a wide range of applications in multiple fields such as control engineering, signal processing, and communication systems), and helps to achieve subsequent online fault identification. In addition, this DFIG-based power grid digital twin model does not rely on the change of the nodal admittance matrix for fault identification, making it more flexible and reliable in the face of power grid structure changes.

[0035] S30. Input the real-time fault data into a preset InceptionTime neural network model for fault identification to obtain the result of fault identification.

[0036] It can be understood that during a storm surge, the terminal blocks in low-lying areas are prone to three-phase short-circuit faults due to moisture or water immersion, and may even cause permanent faults in severe cases. After a serious fault occurs, the power grid will take measures to isolate the fault point. The circuit breakers and disconnectors will be activated, and the lines around the faulty busbar will be cut off. Subsequently, temporary power supply measures will be implemented in the power outage area. After the flood recedes, the power grid dispatching center will dispatch the RU to the transmission network RU station, and then the RU to the permanent fault location of the nearby power grid, consuming maintenance resources to repair the faulty line or component. The process of the identification and repair stage is as Figure 3 .

[0037] It should be noted that InceptionTime is a deep learning model specifically for time series classification (TSC). It combines the Inception module idea in the convolutional neural network (CNN), optimizes for the temporal characteristics of time series data, aims to efficiently capture features at different time scales, and at the same time maintains the lightweight and generalization ability of the model.

[0038] Specifically, after obtaining the real-time fault data, in this embodiment, first, normalize the multi-dimensional time series signal of the implementation fault data: ; Among them, is the time element of the collected signal, Indicates finding the minimum value of the element, Indicates finding the maximum value of the element.

[0039] Secondly, an InceptionTime model network is composed of five randomly initialized convolutional neural networks (CNNs). Two residual blocks are set up to form the classifier of the initial network. A single component in the initial spatial network consists of three initial spatial modules. The input of each residual block is transmitted to the input of the next component through a linear connection, thus alleviating the problem of vanishing gradients by allowing the gradient to flow directly. Subsequently, a global average pooling layer is used to average the time series output in all dimensions. Finally, it is connected to a fully connected softmax layer for outputting the recognition result.

[0040] It should be noted that the first main component of the initial spatial module is called the "bottleneck" layer. The "bottleneck" layer is a structure designed to reduce computational complexity and model complexity. It reduces the dimensionality of the input features by using convolutional kernels to reduce the computational complexity and extract key features. Set a sliding filter with a length of 1 and a stride of 1 pieces. This will convert the time series collected from the distribution network from - dimensional multivariate time series to - dimensional ( ), and use a filter of to reduce the dimensionality of the input time series, thus alleviating the overfitting problem and the complexity of the power grid dataset.

[0041] The second main component of the initial spatial module is multiple filters of different lengths that slide simultaneously on the same input time series, and the input is the output of the bottleneck layer. The Inception module enables the model to adaptively learn the features of each time scale in the data without manually specifying the optimal time window size. This multi - scale parallel processing method ensures that the performance of the model does not degrade due to an overly large receptive field. The InceptionTime neural network structure and the power grid fault recognition process are as Figure 4 shown.

[0042] In a certain embodiment of this application, during the training process of InceptionTime, we set the cross - entropy loss function as follows: ; Among them, represents the loss function of the InceptionTime neural network model, represents the probability of identifying the fault category as , represents the fault label corresponding to the time series signal sample,M represents the total number of fault categories, and c represents the fault category.

[0043] In the fault identification stage, the metrics used to measure the identification accuracy include precision ( Pre ), recall ( Rec ), and F1 score: ; where is the number of samples in which the fault type is successfully identified; is the number of samples misidentified; represents the number of samples in which each fault fails to be successfully identified.

[0044] S40. According to the result of the fault identification, adopt a scheduling algorithm based on deep reinforcement learning to schedule the maintenance unit to the fault node for fault repair.

[0045] In a certain embodiment of the present application, step S40 includes: S41. Establish a power and transportation coupling network.

[0046] S42. According to the power and transportation coupling network and the constraint conditions of the preset maintenance path, adopt a scheduling algorithm based on deep reinforcement learning to generate the maintenance path of the maintenance unit.

[0047] S43. According to the maintenance path, control the maintenance unit to repair the fault.

[0048] It should be noted that the maintenance unit can be a power grid maintenance robot or a patented power grid maintenance unit.

[0049] It can be understood that there is usually a strong coupling relationship between the power grid and the transportation network, and both the mobile power source and the maintenance unit have their own transportation routes. The state of the transportation network usually has a restrictive effect on the routes and transportation of the mobile power source and the RU, thus affecting the load restoration and fault repair of the power grid.

[0050] Specifically in this embodiment, in this embodiment, first, model the urban transportation network to form a weighted graph G, and the formula is as follows: ; where each node contains an agency station for a maintenance unit (RU) and is coupled with the surrounding area of a specific busbar in the power grid. The RU can drive from this station to the newly opened line around this busbar. The time required for the RU to travel between different nodes in the transportation network is calculated as follows: ; where represents the RU on the road Actual driving time on it; Indicates the free driving time of RU on the road; and Indicates the road Delay coefficient; Indicates the road Capacity; Indicates the road Traffic volume. To simplify the simulation, we simplify and rewrite it as a driving time model based on accident probability, as shown in the formula: ; Where, Is a binary variable indicating whether an unexpected road condition occurs at time step , with a probability of .

[0051] In a certain embodiment of the present application, the path decision of RU in the traffic network satisfies the following constraints: ; Where, Is a binary variable indicating whether the maintenance unit repairs node n at time step ; n is a maintenance node; N represents the set of all maintenance nodes; t is the time step, Is the set of time steps.

[0052] The above equation indicates that the RU agent can stay at most at one parking station and maintain the lines around a bus at each time step .

[0053] The maintenance operation of RU satisfies the following constraints: ; Where, Is a binary variable indicating whether a damaged component is repaired at time step ; Is a binary variable indicating whether the maintenance unit is repairing component j at time step w ; Is the set of damaged components; Indicates the time required to repair component at time step w ; Indicates the amount of resources required to repair the damaged component w ; Indicates the resource capacity of the maintenance unit.

[0054] The sequential decision-making strategy of RU is trained through model-free deep reinforcement learning. The observation space of RU is: ; In this formula, represents the time step and the abscissa of the traffic network node where the RU is located at time while represents the abscissa of the fault node; represents the time step and the ordinate of the traffic network node where the RU is located at time

[0055] The action space of the RU is: ; In this formula represents the routing decision made by the RU at time step . .

[0056] We set the reward function for the RU : ; where is the reward function, represents the abscissa of the traffic network node where the repair unit is located at time step , while represents the abscissa of the fault node; represents the ordinate of the traffic network node where the repair unit is located at time step , is the discount factor, is the distance between the repair unit and the fault node, and .

[0057] In a certain embodiment of the present application, in order to better handle the discrete observation space and action space, we adopt the RU routing and scheduling method based on Deep Q-Network (DQN). The training loss function of the DQN network is set as: ; where is the loss function of the scheduling algorithm for deep reinforcement learning; is the value predicted by the network with parameter Q at the current state, where Q the s j value represents the expected value of the long-term cumulative reward that the agent can obtain after taking the action a j in the state is the next state s j+1 the largest next Q value; is the reward during state transition, is the discount rate parameter.

[0058] The beneficial effects of a power grid fault identification and repair method provided by this application are as follows: 1. The digital twin model of the doubly-fed distribution network constructed based on the electromagnetic transient program (EMTP) in this application generates more stable real-time data than Simulink (a visualization multi-domain simulation and model design tool based on MATLAB), providing highly reliable samples for the InceptionTime neural network; through multi-scale parallel convolution and adaptive time window technology, the recognition accuracy of 16 typical faults reaches 98%, overcoming the limitations of traditional methods that rely on manual parameter tuning (such as CNN / BiGRU requiring manual specification of the time window).

[0059] 2. Combining the power grid - transportation network coupling model, the deep reinforcement learning (DRL) algorithm can still dynamically optimize the path decision of the repair unit (RU) under the interference of a 20% road accident probability. Faults can be repaired within 10 time steps in three groups of test scenarios, shortening the power outage time by more than 50% compared to manual scheduling.

[0060] 3. The robustness of the RU is trained through probabilistic random blinding, enabling it to complete tasks through strategies such as pausing and detouring even when communication is interrupted or the road is paralyzed; the collaborative design of digital twin and DRL can be extended to multi-disaster coupling scenarios such as typhoons and heavy rains.

[0061] To verify the effectiveness of the identification and repair method in this application, the following two simulation experiments were conducted in this application: Simulation Experiment 1: This application constructed a simulation platform using the actual distribution network parameters of a certain park. The platform contains a 16-node transportation network, which is closely coupled with the distribution network. In the simulation, we set a three-phase short circuit fault to occur at node 21 10 seconds after the load is connected, and a total of 16 different fault conditions were designed. Each fault contains 500 data samples, a single time window contains 200 data samples, and 300 samples are generated for each fault, forming a three-dimensional time series data set. 20% of the samples were randomly selected from each fault as the test set.

[0062] To better demonstrate the superiority of the fault identification method based on InceptionTime, we selected a large number of representative advanced and traditional deep learning methods as baselines, such as: mWDN, ResCNN, gMLP, XCM, OmniScale, TCN, XceptionTime, TST, FCN, GRU, LSTM, as well as the combined models LSTM-FCN and GRU-FCN. The fault identification results, training and testing durations of all the above methods were recorded and the results are shown in Table 1.

[0063] Table 1 Identification Results of Advanced Fault Identification Algorithms

[0064] As can be seen from Table 1, the InceptionTime model has the highest fault identification accuracy, recall rate and F1-score among all the baseline methods, verifying its superior performance in fault identification.

[0065] Simulation Experiment 2: The road conditions in the traffic network have strong uncertainty, and the floods caused by storm surges may also lead to the paralysis of roads at uncertain times. There may also be traffic jams or other accidents on the roads where RU can pass smoothly, and the probability of such accidents occurring is different in different situations, which greatly increases the complexity of the process. To simplify the simulation process, we set μ to the constant 0.2. Considering that communication failures may occur during the process, we applied the probabilistic random blindness method to the observation space of RU and reset it to the initial value to train the robustness of RU. Once encountering the above unexpected situations, the RU agent will get lost and have to stop, turn back or detour. This is reflected in our simulation process based on discrete time steps, where RU will decide to stay at a certain stop for one time step, or make a turn-back or detour action. The above decisions are manifested as the routing and position transfer of RU in the experiment. The free driving time of RU on different sections of the traffic network is uncertain, and with the occurrence of accidents, the time required to complete different sections is also different. After reaching the target station, RU will go to the nearby power grid fault location for repair, consuming the repair parts it carries at the same time, and this process takes a total of Tp,trd. We set the total number of time steps T to 20 and trained RU in different scenarios. After RU interacted with the environment for 200 - 2000 episodes, the routing strategy was obtained and its strategy is presented in Table 2. The P in the table refers to the action that RP did not make a path selection at this time step but chose to stay in place for two reasons: 1. RU encountered an accident on the road; 2. RU has reached the target location and needs to stay for some time to repair the fault. After choosing a path, RU will move autonomously. To demonstrate the process of its position change, we will show the position change of RU step by step in time, and the position change of RU is as Figure 5as shown

[0066] Table 2 Scheduling Process of RU

[0067] From Table 2 and Figure 5 it can be seen that in three scenarios, the RU made the decision to stop due to an accident when approaching the fault point, and in Scenario 3, it made the decision to turn back along the original path. In Scenario 3, it took a total of 9 time steps from the start of routing to the end of repair, which is the most time-consuming among the three scenarios. The reason is that the RU encountered an accident on the way and turned back on the 12' path. The faults in all three scenarios were repaired within 10 time steps.

[0068] Please refer to Figure 6 , this application also provides a power grid fault identification and repair system 200, including: A construction module 201 for constructing a digital twin model of a distribution network with a doubly-fed induction generator; A data generation module 202 for simulating fault conditions at different nodes in the power grid according to the digital twin model of the distribution network and generating real-time fault data; A fault identification module 203 for inputting the real-time fault data into a preset InceptionTime neural network model for fault identification to obtain the result of fault identification; A fault repair module 204 for scheduling a maintenance unit to the fault node for fault repair according to the result of the fault identification by using a scheduling algorithm based on deep reinforcement learning.

[0069] Please refer to Figure 7 , this embodiment of the application also provides a computer electronic device 300, including a memory 303 and a processor 302. The memory 303 stores a computer program, and when the processor executes the computer program, it implements the steps of the power grid fault identification and repair method described in any one of the above.

[0070] Specifically, the electronic device 300 includes: a transceiver 301, a bus interface, and a processor 302. The processor 302 is used to construct a digital twin model of a distribution network with a doubly-fed induction generator; simulate fault conditions at different nodes in the power grid according to the digital twin model of the distribution network and generate real-time fault data; input the real-time fault data into a preset InceptionTime neural network model for fault identification to obtain the result of fault identification; and schedule a maintenance unit to the fault node for fault repair according to the result of the fault identification by using a scheduling algorithm based on deep reinforcement learning.

[0071] In an embodiment of the present application, the electronic device 300 further includes: a memory 303. In Figure 7 it, the bus architecture may include any number of interconnected buses and bridges. Specifically, various circuits of one or more processors represented by the processor 302 and the memory represented by the memory 303 are linked together. The bus architecture may also link together various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art. Therefore, they will not be further described herein. The bus interface provides an interface. The transceiver 301 may be multiple elements, that is, including a transmitter and a receiver, and provides a unit for communicating with various other devices on the transmission medium. The processor 302 is responsible for managing the bus architecture and general processing, and the memory 303 may store data used by the processor 302 when executing operations.

[0072] The embodiment of the present application further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method for identifying and repairing a power grid fault described in any one of the above are implemented.

[0073] In this embodiment, the computer-readable storage medium may be a non-volatile storage medium or a volatile storage medium. For example, the computer storage medium may include but is not limited to: various media such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disc that can store program codes.

[0074] In all the examples shown and described here, any specific value should be construed as merely exemplary, rather than as a limitation. Therefore, other examples of the exemplary embodiments may have different values.

[0075] It should be noted that: similar reference numerals and letters denote similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0076] In several embodiments provided by this application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely illustrative. For example, the flowcharts and structure diagrams in the accompanying drawings show the possible architectures, functions, and operations of devices, methods, and computer program products according to multiple embodiments of this application. In this regard, each block in the flowchart or block diagram can represent a module, a program segment, or a part of code, and the module, program segment, or part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in an alternative implementation, the functions marked in the blocks can occur in a different order than that marked in the accompanying drawings. For example, two consecutive blocks can actually be executed substantially in parallel, and they can sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the structure diagram and / or flowchart, as well as the combination of blocks in the structure diagram and / or flowchart, can be implemented by a dedicated hardware-based system that performs the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0077] In addition, each functional module or unit in various embodiments of this application can be integrated together to form an independent part, or each module can exist alone, or two or more modules can be integrated to form an independent part.

[0078] If the described functions are implemented in the form of software functional modules and sold or used as an independent product, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to enable a terminal device (which can be a smart phone, a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of this application.

[0079] As described above, the above is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed by this application can easily think of changes or substitutions, which should all be covered by the protection scope of this application.

Claims

1. A method for identifying and repairing power grid faults, characterized in that, Including: Constructing a digital twin model of a distribution network containing a doubly-fed induction generator; According to the digital twin model of the distribution network, simulating the fault conditions of different nodes in the power grid to generate real-time fault data; Inputting the real-time fault data into a preset InceptionTime neural network model for fault identification to obtain the results of fault identification; According to the results of the fault identification, adopting a scheduling algorithm based on deep reinforcement learning to schedule the maintenance unit to the fault node for fault repair.

2. The recognition and repair method according to claim 1, wherein The preset InceptionTime neural network model adopts the following loss function: ； Among them, represents the loss function of the InceptionTime neural network model, represents the probability of identifying the fault category as the probability, represents the fault label corresponding to the time series signal sample, M represents the total number of fault categories, and c is the fault category.

3. The recognition and repair method according to claim 1, wherein The scheduling algorithm based on deep reinforcement learning adopts the following loss function: ； Among them, is the loss function of the scheduling algorithm for deep reinforcement learning; is the value predicted by the network with parameter Q at the current state, where Q the s j value represents the expected value of the long-term cumulative reward that the agent can obtain after taking the action a j in the state is the next state s j+1 the maximum Q value at is the reward during the state transition, is the discount rate parameter.

4. The recognition and repair method according to claim 1, wherein The step of, according to the results of the fault identification, adopting a scheduling algorithm based on deep reinforcement learning to schedule the maintenance unit to the fault node for fault repair includes: Establishing a coupled network of power and transportation; According to the coupled network of power and transportation and the constraint conditions of the preset maintenance path, adopting a scheduling algorithm based on deep reinforcement learning to generate the maintenance path of the maintenance unit; According to the maintenance path, controlling the maintenance unit to repair the fault.

5. The recognition and repair method according to claim 4, characterized in that, When adopting the scheduling algorithm based on deep reinforcement learning to generate the maintenance path of the maintenance unit, the reward function is set as follows: ； Among them, is the reward function, represents the time step the abscissa of the traffic network node where the maintenance unit is located at time step represents the abscissa of the fault node; represents the time step the ordinate of the traffic network node where the maintenance unit is located at time step represents the ordinate of the fault node; is the discount factor, is the distance between the maintenance unit and the fault node.

6. The recognition and repair method according to claim 4, characterized in that, The constraint conditions of the maintenance path are as follows: ； wherein, is a binary variable indicating whether to repair node n by the maintenance unit at time step ; n is a maintenance node; N represents the set of all maintenance nodes; t is a time step, and is the set of time steps.

7. The recognition and repair method according to claim 1, wherein When scheduling the maintenance unit to the fault node for fault repair, the following constraint conditions need to be satisfied: ； Among them, is a binary variable indicating whether a damaged component is repaired at time step ; is a binary variable indicating whether the repair unit is repairing the component j at time step w ; is the set of damaged components; indicates the duration required to repair the component at time step w ; indicates the quantity of resources required to repair the damaged component w ; indicates the resource capacity of the repair unit.

8. A power grid fault identification and repair system, characterized in that, Including: A construction module for constructing a digital twin model of a distribution network containing a doubly-fed induction generator; A data generation module for simulating the fault conditions of different nodes in the power grid according to the digital twin model of the distribution network to generate real-time fault data; A fault identification module for inputting the real-time fault data into a preset InceptionTime neural network model for fault identification to obtain the results of fault identification; A fault repair module for, according to the results of the fault identification, adopting a scheduling algorithm based on deep reinforcement learning to schedule the maintenance unit to the fault node for fault repair.

9. A computer electronic device, characterized in that, Including a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, the steps of the method for identifying and repairing power grid faults described in any one of claims 1-7 are implemented.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method for identifying and repairing power grid faults described in any one of claims 1-7 are implemented.

Images