Power grid fault positioning method and device based on adaptive fuzzy Markov model

Abstract

The embodiment of the invention discloses a power grid fault positioning method and device based on an adaptive fuzzy Markov model, and the method and device are used for power grid fault positioning in extreme weather, and the method comprises the steps: collecting power grid data and environment data related to the extreme weather; generating a multi-modal data set according to a multi-modal attention mechanism by using power grid data and the environment data; defining an uncertain state of the power grid based on fuzzy logic; dynamically adjusting fuzzy membership function parameters through an online learning algorithm; calculating the fuzzy transition probability between power grid states in real time according to the fuzzy membership function parameters based on the multi-modal data set by using an adaptive fuzzy Markov model; predicting a fault propagation path according to the fuzzy transition probability; and generating fault positioning information according to the topological data by using a graph neural network model in combination with the fault propagation path and the multi-modal data set. According to the method, the fault propagation process can be accurately simulated, and the fault positioning accuracy is remarkably improved.

Description

Technical Field

[0001] The present application relates to the fields of smart grid fault location and neural network technology, and in particular to a grid fault location method and device based on an adaptive fuzzy Markov model. Background Art

[0002] Smart grids are the core of modern power systems, enabling efficient and reliable power transmission and management through advanced sensors, communication technologies, and data analytics. However, grid faults caused by extreme weather (such as lightning strikes, storms, and extreme heat) are characterized by dynamic topological changes, uncertain states (such as partial failures and potential overloads), and complex propagation paths, posing significant challenges to fault location. Traditional methods struggle to effectively handle the ambiguous states (such as partial failures) induced by extreme weather, resulting in reduced location accuracy.

[0003] Therefore, how to achieve high-precision, low-latency fault location in extreme weather with high variability has become an urgent problem to be solved. Summary of the Invention

[0004] The present application provides a power grid fault location method and device based on an adaptive fuzzy Markov model, which achieves high-precision fault location in extreme weather conditions.

[0005] This application provides the following solutions:

[0006] According to a first aspect, a power grid fault location method based on an adaptive fuzzy Markov model is provided, which is used for power grid fault location under extreme weather conditions, and comprises: collecting power grid data and environmental data, wherein the power grid data comprises topology data, voltage data, and current data, and the environmental data comprises wind speed, temperature, and lightning activity data; utilizing the power grid data and the environmental data, dynamically allocating fusion weights of the environmental data and the power grid data according to a multimodal attention mechanism, and generating a multimodal data set; defining the uncertain state of the power grid based on fuzzy logic, wherein the uncertain state comprises: normal, partial fault, and potential overload, and dynamically adjusting the fuzzy membership function parameters through an online learning algorithm; utilizing an adaptive fuzzy Markov model, based on the multimodal data set, and according to the fuzzy membership function parameters, calculating in real time the fuzzy transition probabilities between power grid states, and predicting the fault propagation path according to the fuzzy transition probabilities; utilizing a graph neural network model, combining the fault propagation path and the multimodal data set, and generating fault location information according to the topology data.

[0007] According to an achievable method in an embodiment of the present application, the method further includes: performing anomaly detection on the currently collected power grid data and the environmental data based on historical data patterns to identify abnormal fluctuations or missing information; after the anomaly detection, adopting a differentiated processing strategy according to the data source type, prioritizing interpolation and redundant cross-validation on the power grid data, and prioritizing time alignment and credibility weighted correction on the environmental data.

[0008] According to an achievable method in an embodiment of the present application, the dynamic allocation of the fusion weights of the environmental data and the power grid data based on the multimodal attention mechanism includes: when the lightning strike intensity is greater than 10 kiloamperes or the wind speed is greater than 30m / s, the fusion weight of the environmental data is adjusted to 0.6 or above.

[0009] According to an achievable method in an embodiment of the present application, when dynamically adjusting the fuzzy membership function parameters through an online learning algorithm, the method further includes: continuously updating the attribution boundary of the fuzzy state based on the real-time input of the power grid data and the weather data; when a sudden change in the power grid state or a drastic fluctuation in the weather environment is detected, triggering a rapid convergence mechanism to adaptively fine-tune the core parameters of the fuzzy membership function.

[0010] According to an achievable method in an embodiment of the present application, the dynamically adjusting the fuzzy membership function parameters through an online learning algorithm includes: dynamically adjusting the mean and variance of the fuzzy membership function using a fuzzy C-means clustering algorithm.

[0011] According to an implementable method in an embodiment of the present application, the fuzzy transition probability between power grid states is calculated in real time based on the multimodal data set according to the fuzzy membership function parameters, and the fault propagation path is predicted according to the fuzzy transition probability, including: constructing a fuzzy state transition probability matrix based on the multimodal data set, the fuzzy states including: normal, partial fault and potential overload; using a fuzzy Viterbi inference algorithm, the fuzzy state transition probability matrix is ​​updated in real time according to the multimodal data set, and the path with the highest cumulative probability is selected as the predicted fault propagation path according to the transition probability.

[0012] According to an achievable method in an embodiment of the present application, the use of a graph neural network model, in combination with the fault propagation path and the multimodal dataset, to analyze the power grid topology and generate fault location information includes: constructing a dynamic graph structure based on the topology data and the fault propagation path, in which nodes represent power grid equipment, and edges represent connection relationships between devices and fuzzy weather impact weights; utilizing a spatiotemporal graph neural network model and an attention mechanism to process the multimodal dataset and the fault propagation path to generate fault location information.

[0013] According to a second aspect, a power grid fault location device based on an adaptive fuzzy Markov model is provided, the device being used for power grid fault location in extreme weather conditions, the device comprising: a data acquisition unit configured to acquire power grid data and environmental data, the power grid data comprising topology data, voltage data, and current data, and the environmental data comprising wind speed, temperature, and lightning activity data; a multimodal dataset generation unit configured to utilize the power grid data and the environmental data and dynamically assign fusion weights of the environmental data and the power grid data according to a multimodal attention mechanism to generate a multimodal dataset; a fuzzy logic determination unit configured to define the uncertain state of the power grid based on fuzzy logic, the uncertain states comprising normal, partial fault, and potential overload, and dynamically adjust fuzzy membership function parameters through an online learning algorithm; a fault path prediction unit configured to utilize an adaptive fuzzy Markov model, based on the multimodal dataset, and according to the fuzzy membership function parameters, to calculate in real time the fuzzy transition probabilities between power grid states; and to predict the fault propagation path based on the fuzzy transition probabilities; and a fault location information generation unit configured to utilize a graph neural network model, in combination with the fault propagation path and the multimodal dataset, to generate fault location information based on the topology data.

[0014] According to a third aspect, a computer-readable storage medium is provided, on which a computer program is stored. When the program is executed by a processor, the steps of any one of the methods according to the first aspect are implemented.

[0015] According to a fourth aspect, an electronic device is provided, comprising: one or more processors; and a memory associated with the one or more processors, the memory being used to store program instructions, which, when read and executed by the one or more processors, execute the steps of the method described in any one of the above-mentioned first aspects.

[0016] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0017] This application achieves accurate positioning of power grid faults under extreme weather conditions by introducing an adaptive fuzzy Markov model and graph neural network. This application comprehensively collects the voltage, current, topology, wind speed, temperature, lightning and other environmental information of the power grid to construct a multimodal data set, which enhances the model's perception of complex external factors. Through adaptive fuzzy logic and online learning mechanisms, the model can dynamically optimize membership functions and state transition probabilities, thereby more accurately simulating the fault propagation process. Further combined with the in-depth modeling of power grid topological relationships by graph neural networks, the accuracy and real-time performance of fault positioning are significantly improved, with greater robustness and adaptability.

[0018] Of course, any product implementing the present application does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0020] Figure 1 A diagram of the system architecture applicable to the embodiments of the present application;

[0021] Figure 2 A flowchart of a power grid fault location method based on an adaptive fuzzy Markov model provided in an embodiment of the present application;

[0022] Figure 3 A structural block diagram of a power grid fault location device based on an adaptive fuzzy Markov model provided in an embodiment of the present application;

[0023] Figure 4 A schematic block diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0024] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field are within the scope of protection of this application.

[0025] The terms used in the embodiments of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The singular forms "a", "an", "the" and "the" used in the embodiments of the present invention and the appended claims are also intended to include plural forms unless the context clearly indicates otherwise.

[0026] It should be understood that the term "and / or" as used herein is merely a description of the relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, or B exists alone. Furthermore, the character " / " in this document generally indicates that the associated objects are in an "or" relationship.

[0027] The word "if," as used herein, may be interpreted as "at the time of" or "when" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrases "if it is determined" or "if (stated condition or event) is detected" may be interpreted as "when it is determined" or "in response to the determination" or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)," depending on the context.

[0028] Several grid fault location technologies currently exist, but these technologies struggle to effectively handle ambiguous conditions caused by extreme weather, such as partial faults, which can lead to reduced location accuracy. In light of this, this application offers a new approach. To facilitate understanding of this application, we first describe the system architecture upon which it is based. Figure 1 An exemplary system architecture to which the embodiments of the present application can be applied is shown. Figure 1 As shown in , the system architecture may include user equipment and a power grid fault location device based on an adaptive fuzzy Markov model located at a server side.

[0029] A user can input topology data, voltage data, current data, and environmental data related to extreme weather conditions through a user device, which then transmits the data to a server-side power grid fault location device. The power grid fault location device can utilize the methods provided in the embodiments of this application to locate the fault and obtain fault location information. The server-side can then transmit the fault location information to the user terminal.

[0030] User devices may include, but are not limited to, smart mobile terminals, smart home devices, wearable devices, and personal computers (PCs). Smart mobile devices may include mobile phones, tablets, laptops, PDAs (Personal Digital Assistants), and internet-connected cars. Smart home devices may include smart TVs and smart refrigerators. Wearable devices may include smart watches, smart glasses, virtual reality devices, augmented reality devices, and mixed reality devices (i.e., devices that support both virtual reality and augmented reality).

[0031] The power grid fault location device can be set up as an independent server, a server group, or a cloud server. A cloud server, also known as a cloud computing server or cloud host, is a host product in the cloud computing service system. It solves the problems of difficult management and weak service scalability in traditional physical hosts and virtual private servers (VPS). Figure 1In addition to the shown architecture, the power grid fault locating device can also be set on a computer terminal with strong computing capabilities.

[0032] It should be understood that Figure 1 The user equipment and power grid fault location device in the figure are only illustrative. According to the implementation requirements, there may be any number of user equipment and power grid fault location devices.

[0033] Figure 2 The flowchart of the power grid fault location method based on the adaptive fuzzy Markov model provided in the embodiment of the present application can be Figure 1 The power grid fault location device in the system shown in the figure is executed. The method is used for fault location in extreme weather conditions. Figure 2 As shown in , the method may include the following steps:

[0034] Step 201: Collect power grid data and environmental data, where the power grid data includes topology data, voltage data, and current data, and the environmental data includes wind speed, temperature, and lightning activity data.

[0035] Step 202: Utilizing the power grid data and the environmental data, dynamically allocating fusion weights of the environmental data and the power grid data according to a multimodal attention mechanism to generate a multimodal dataset.

[0036] Step 203: defining the uncertain state of the power grid based on fuzzy logic, wherein the uncertain state includes normal, partial fault and potential overload, and dynamically adjusting the fuzzy membership function parameters through an online learning algorithm.

[0037] Step 204: using an adaptive fuzzy Markov model, based on the multimodal data set and according to the fuzzy membership function parameters, calculate the fuzzy transition probabilities between power grid states in real time, and predict the fault propagation path according to the fuzzy transition probabilities.

[0038] Step 205: Utilize a graph neural network model, combine the fault propagation path and the multimodal dataset, and generate fault location information based on the topology data.

[0039] As can be seen from the above process, this application has achieved accurate positioning of power grid faults under extreme weather conditions by introducing adaptive fuzzy Markov models and graph neural networks. This application comprehensively collects the voltage, current, topology structure and wind speed, temperature, lightning and other environmental information of the power grid to construct a multimodal data set, which enhances the model's perception of complex external factors. Through adaptive fuzzy logic and online learning mechanisms, the model can dynamically optimize membership functions and state transition probabilities, thereby more accurately simulating the fault propagation process. Further combined with the in-depth modeling of power grid topology relationships by graph neural networks, the accuracy and real-time performance of fault location have been significantly improved, with greater robustness and adaptability.

[0040] The following describes in detail the steps in the above process and the effects that can be further produced in conjunction with the embodiments.

[0041] First, the above step 201, namely "collecting power grid data and environmental data, wherein the power grid data includes topology data, voltage data, and current data, and the environmental data includes wind speed, temperature, and lightning activity data" is described in detail in conjunction with an embodiment.

[0042] Grid data collection is used to describe core information about the operating status and structure of the power system. Topological data reflects the physical and logical connections within the grid, with substations, transformers, and other equipment serving as nodes and transmission lines serving as connecting edges, clearly depicting the grid's network structure. Voltage and current data record the electrical operating parameters of grid equipment in real time and can reveal abnormal conditions, such as voltage spikes caused by lightning strikes or current overloads caused by storms. This data is typically collected through IoT sensors and smart meters deployed throughout the grid. The sampling frequency is high, and the accuracy is sufficient to capture transient changes, providing high-resolution foundational information for subsequent analysis.

[0043] Environmental data collection focuses on the impact of extreme weather, including wind speed, temperature, and lightning activity. Wind speed data reflects the physical impact of storms or strong winds on transmission lines, potentially causing them to shake or break. Temperature data captures the impact of high or low temperatures on equipment efficiency, such as transformer overheating. Lightning activity data records the intensity and location of lightning strikes, directly linking them to short circuits or equipment damage caused by lightning. This data, collected through meteorological sensors, radar, or satellite systems, enables real-time monitoring of weather changes, providing an external environmental perspective for locating power grid faults.

[0044] As an implementable method, after acquiring the power grid data and environmental data, the present application also includes a data preprocessing process, specifically: performing anomaly detection on the currently collected power grid data and the environmental data based on historical data patterns to identify abnormal fluctuations or missing information; after the anomaly detection, adopting a differentiated processing strategy according to the data source type, wherein interpolation and redundant cross-validation are preferentially performed on the power grid data, and time alignment and credibility weighted correction are preferentially performed on the environmental data.

[0045] Anomaly detection is the first step in this feature, which aims to identify abnormal fluctuations or missing information in the collected data. Grid data may be abnormal due to equipment failure or communication interruption, such as unreasonable spikes in voltage data or missing records of current data; environmental data may be abnormal due to sensor errors or sudden changes in weather, such as abnormally high wind speed data or incomplete records of lightning activity data. This application uses historical data patterns to detect these anomalies by comparing current data with historical patterns under normal operation or weather conditions. For example, historical data may show a normal voltage fluctuation range. If the current voltage data exceeds this range, it will be marked as an abnormal fluctuation.

[0046] After anomaly detection, the system adopts differentiated processing strategies based on the characteristics of the data source type to maximize data quality. Grid data processing prioritizes interpolation and redundant cross-validation. Interpolation uses data from nearby time points to infer reasonable values ​​for missing voltage or current data, ensuring data continuity. Redundant cross-validation utilizes redundant data collected by multiple sensors (such as voltages recorded by multiple smart meters) to eliminate outliers or confirm data reliability through cross-comparison. This strategy is well-suited to the high-precision and high-reliability requirements of grid data, effectively repairing data defects caused by equipment or communication issues while preserving the true characteristics of the grid's operating status.

[0047] Environmental data processing prioritizes time alignment and credibility weighting. Time alignment addresses inconsistent acquisition frequencies or timestamps across meteorological sensors (such as anemometers and lightning detectors) by synchronizing data to a unified timeline, ensuring temporal consistency between environmental and power grid data. Credibility weighting assesses data credibility based on sensor quality or weather conditions. For example, in heavy rainfall, wind speed sensors may be subject to interference, so the system reduces their data weight and prioritizes highly reliable lightning activity data. This strategy adapts to the dynamic nature and uncertainty of environmental data and can generate reliable feature representations in extreme weather conditions.

[0048] The above step 202, i.e., "using the power grid data and the environmental data, dynamically allocating fusion weights of the environmental data and the power grid data according to a multimodal attention mechanism to generate a multimodal data set" is described in detail below in conjunction with an embodiment.

[0049] This application uses a multimodal attention mechanism to intelligently assign weights based on real-time context, ensuring that the fusion process can highlight the most important information for fault location, such as prioritizing lightning activity data in lightning strike scenarios.

[0050] The multimodal attention mechanism is an advanced technology for processing and fusing multiple heterogeneous data sources (such as text, images, audio or sensor data). It is widely used in the fields of artificial intelligence and machine learning, especially in scenarios where complex information needs to be integrated, such as the fault location of power grids under extreme weather conditions in this application. It dynamically assigns weights to different modal data, highlights the most important information for the task, and generates a unified feature representation, thereby improving the performance and adaptability of the model. In practical applications, the multimodal attention mechanism first extracts features from each modal data. For example, the voltage data of the power grid may extract the frequency of fluctuations, and the wind speed data of the environment may extract the time series features. These features are input into the attention module, which evaluates the importance of each feature based on the real-time context.

[0051] In this application, a multimodal attention mechanism adaptively assesses the relative importance of each type of data based on weather parameters and grid status. For example, when a significant increase in lightning strike intensity is detected, the system automatically increases the weight of environmental data to more accurately reflect the potential impact of lightning on the grid. In contrast, in the case of large voltage fluctuations, the weight of grid data may dominate. This dynamic adjustment is achieved through a neural network that learns a weight allocation strategy based on historical fault patterns and real-time data, thereby generating a multimodal dataset that comprehensively captures the operating status of the grid.

[0052] As an implementable approach, the multimodal attention mechanism dynamically adjusts the fusion weight according to the weather parameters of the environmental data, wherein when the lightning strike intensity is greater than 10kA or the wind speed is greater than 30m / s, the fusion weight of the environmental data is adjusted to 0.6 or above. This example specifies a weather parameter threshold to trigger the weight adjustment. When the lightning strike intensity is greater than 10 kiloamperes or the wind speed is greater than 30 meters per second, the system automatically increases the fusion weight of the environmental data to 0.6 or higher. This means that the contribution of environmental data in the multimodal dataset is at least 60%, thereby highlighting the direct impact of lightning activity or wind speed on power grid failures.

[0053] The following describes in detail step 203 of "defining the uncertain state of the power grid based on fuzzy logic, wherein the uncertain state includes normal, partial fault and potential overload, and dynamically adjusting the fuzzy membership function parameters through an online learning algorithm" in conjunction with an embodiment.

[0054] This application uses fuzzy logic to quantify the complex operating state of the power grid under extreme weather (such as lightning strikes, storms or high temperatures), overcoming the limitations of traditional normal or fault binary state modeling and providing an accurate state description for subsequent fault propagation path prediction.

[0055] Fuzzy logic is a mathematical method for dealing with uncertainty and fuzzy phenomena, and is particularly suitable for describing intermediate states in power grid operation that are neither black nor white. In power grid fault location, the equipment state is often not a simple normal or complete fault, but rather a fuzzy transition state. For example, a transmission line may be slightly damaged by a lightning strike and has not yet been completely disconnected, but its performance has degraded; or the transformer is close to overload due to high temperature operation, but the protection mechanism has not yet been triggered. This application defines three typical uncertain states: normal, partial fault, and potential overload. The normal state indicates that the power grid equipment operates stably without abnormal signals, such as voltage and current fluctuations within the expected range; partial fault refers to slight damage or performance degradation of the equipment, such as a slight short circuit in the line caused by a lightning strike but not completely disconnected; potential overload describes that the equipment is close to the operating limit, such as the transformer's current continues to rise due to high temperature operation but the protection mechanism has not been triggered.

[0056] These states are quantified using fuzzy membership functions, allowing the system to represent the likelihood of the current state of the power grid in a probabilistic manner. Fuzzy membership functions are tools used by fuzzy logic to represent the degree to which an element belongs to a specific set. In power grid fault location, the power grid state is not simply normal or faulty, but rather has fuzzy intermediate states. For example, a partial fault may manifest as a slight voltage fluctuation, and a potential overload may be a current approaching but not exceeding the limit. The membership function describes the likelihood of these states using a value between 0 and 1 (called the membership degree). For example, a line may be assessed as 60% normal, 30% partially faulted, and 10% potentially overloaded. The shape of a membership function is usually a bell, triangle, or trapezoid, and its specific shape and position are controlled by a set of parameters, which are the fuzzy membership function parameters, mainly including the mean and variance.

[0057] The mean parameter determines the center position of the membership function and represents the most typical value of a certain state. For example, for a normal state, the mean may correspond to the standard value of the voltage data, such as 220 volts, indicating that the voltage is most likely to be in a normal state when it is around this value. The variance parameter controls the width of the membership function and reflects the fuzzy range of the state. A wider variance means that the boundary of the state is more blurred, allowing a larger fluctuation range to be considered normal; a narrower variance means that the state is more strictly defined and anomalies may be detected more sensitively. In this application, these parameters are derived from multimodal data sets (such as voltage, current, and lightning intensity) and combined with fuzzy rules (such as "if the voltage fluctuates slightly and the lightning intensity is medium, then there is a partial fault") to quantify the power grid state.

[0058] An online learning algorithm is a machine learning method that aims to dynamically update model parameters by processing data streams in real time to adapt to changing environments or data patterns. This application dynamically adjusts fuzzy membership function parameters through an online learning algorithm, and continuously updates membership function parameters in a real-time multimodal data set. The online learning algorithm can dynamically adjust the parameters of the fuzzy membership function to ensure that the state definition can adapt to the dynamic changes of the power grid and weather in real time. For example, in a lightning strike scenario, if an increase in lightning intensity is detected, the algorithm may adjust the membership function so that the membership of some faults is significantly improved, thereby more accurately reflecting the current risk. The online learning algorithm can be implemented through algorithms such as fuzzy C-means clustering or Bayesian updating, which can quickly respond to data changes and improve the adaptability and robustness of the model.

[0059] As an implementable approach, the present application can utilize the fuzzy C-means clustering algorithm to dynamically adjust the mean and variance of the fuzzy membership function. The fuzzy C-means clustering algorithm is an online learning method that dynamically calculates the center (mean) and dispersion (variance) of each cluster by assigning samples in a multimodal data set to different state clusters (such as normal, partial fault, potential overload). For example, in a lightning strike scenario, the algorithm may detect abnormally high values ​​in the voltage data. After re-clustering, the mean of the partial fault state is adjusted to a higher voltage range, while reducing the variance to more accurately capture the short circuit risk caused by lightning strikes. This clustering process is based on real-time data streams, and the mean and variance of the cluster are incrementally updated each time new data is received, without the need to reprocess the entire data set, significantly improving computational efficiency and response speed. The algorithm also incorporates historical failure modes to ensure that the adjusted parameters maintain long-term stability while reflecting short-term changes.

[0060] As another feasible method, when dynamically adjusting the fuzzy membership function parameters through an online learning algorithm, the present application can continuously update the attribution boundary of the fuzzy state based on the real-time input of the power grid data and the weather data; when a sudden change in the power grid state or a drastic fluctuation in the weather environment is detected, a rapid convergence mechanism is triggered to adaptively fine-tune the core parameters of the fuzzy membership function.

[0061] The attribution boundary of a fuzzy state refers to the range of the fuzzy membership function and is used to quantify the degree of membership of the grid state, for example, determining the extent to which a device is normal, partially faulty, or potentially overloaded. Traditional methods typically use fixed attribution boundaries, which are difficult to cope with rapid state changes in extreme weather, such as instantaneous voltage surges caused by lightning strikes or line overloads caused by storms. This feature uses real-time grid and weather data input to continuously analyze the characteristics of this data and dynamically update the attribution boundary. For example, if real-time data indicates that voltage fluctuations are gradually increasing, the system may expand the attribution boundary of the partial fault state, allowing more abnormal data to be identified as partial fault states.

[0062] When a sudden change in grid status or a dramatic fluctuation in weather conditions is detected, this feature triggers a rapid convergence mechanism to further optimize the core parameters of the fuzzy membership function, such as the mean and variance. A sudden change in grid status may manifest as an abnormal spike in voltage or current, such as a short circuit caused by a lightning strike; a dramatic fluctuation in weather conditions may include wind speeds suddenly exceeding the storm threshold or a sudden increase in lightning intensity. The rapid convergence mechanism uses an efficient online learning algorithm to rapidly analyze the sudden change data and adjust the membership function parameters to adapt to the new state characteristics. For example, in a lightning strike scenario, the system may shift the mean of some fault states toward a higher voltage fluctuation range while reducing the variance to more accurately capture short circuit risks. This adaptive fine-tuning ensures that the model responds quickly at critical moments and avoids misjudgments caused by parameter lag.

[0063] The following describes in detail step 204, namely, "using an adaptive fuzzy Markov model, based on a multimodal data set, to calculate the fuzzy transition probability between power grid states in real time according to the fuzzy membership function parameters, and predicting the fault propagation path according to the fuzzy transition probability" in conjunction with an embodiment.

[0064] The Markov model is a probabilistic modeling approach used to describe the random transitions of a system between different states. Its core concept is "memorylessness," meaning that the system's future state depends solely on the current state and is unrelated to earlier historical states. However, the limitations of the traditional Markov model are apparent: it assumes discrete and deterministic states and struggles to handle ambiguity or dynamic changes. The fuzzy Markov model is an extension of the Markov model, incorporating fuzzy logic to handle uncertainty and ambiguity.

[0065] In this application, an adaptive fuzzy Markov model is used to predict fault propagation paths. The adaptive fuzzy Markov model combines the advantages of fuzzy logic and Markov chains to model the dynamic evolution of power grid states. This application uses multimodal data sets to provide real-time state information, accurately defines the current state through fuzzy membership function parameters, and then calculates fuzzy transition probabilities between states. These probabilities represent the likelihood that one state will transition to another at the next moment. For example, under the influence of a lightning strike, a normal state may be more likely to transition to a partial fault state.

[0066] Fuzzy membership function parameters play a fundamental role in the real-time calculation of fuzzy transition probabilities. The mean parameter determines the central characteristics of the state; for example, a normal state may be centered around a stable voltage value. The variance parameter controls the fuzzy range of the state and determines the model's sensitivity to anomalies.

[0067] The real-time calculation of fuzzy transition probabilities relies on the dynamic nature of multimodal datasets and the accuracy of fuzzy membership function parameters. Grid data, such as voltage anomalies or current fluctuations, reveal changes in equipment status, while environmental data, such as lightning intensity or wind speed, reflect the impact of weather on faults. For example, in a lightning strike scenario, a multimodal dataset might show a sudden change in voltage and a simultaneous increase in lightning intensity. The model uses algorithms such as fuzzy Bayesian inference, combined with optimized membership function parameters, to calculate the transition probability from normal to partial fault. The system continuously analyzes the real-time data stream and updates the probability values, ensuring that the model can quickly respond to transient changes caused by extreme weather, such as short circuits caused by lightning strikes or line breaks caused by storms. This real-time performance is crucial for predicting fault propagation in complex power grid environments.

[0068] Transition probabilities form a dynamic state transition network that describes how a fault propagates from one device or area to other parts. For example, if a line enters a partial fault state due to a lightning strike, the model may predict a high probability of causing potential overloads on adjacent lines, leading to a more widespread fault. The prediction process analyzes the cumulative effect of transition probabilities to identify the most likely path for the fault to propagate, such as from the struck substation to downstream transmission lines.

[0069] The core of predicting fault propagation paths is to construct a dynamic state transition network by analyzing fuzzy transition probabilities, identifying the most likely path for a fault to propagate from an initial state to other states. As an implementable approach, a fuzzy state transition probability matrix is ​​constructed based on the multimodal dataset. The fuzzy states include normal, partial fault, and potential overload. Using the fuzzy Viterbi inference algorithm, the fuzzy state transition probability matrix is ​​updated in real time based on the multimodal dataset. Based on the transition probabilities, the path with the highest cumulative probability is selected as the predicted fault propagation path. The specific steps are as follows:

[0070] First, an adaptive fuzzy Markov model is used to generate a fuzzy transition probability matrix in real time based on a multimodal dataset and fuzzy membership function parameters. This matrix describes the likelihood of transitions between grid states (e.g., normal, partial fault, and potential overload). For example, in a lightning strike scenario, the normal state may remain unchanged with a high probability, but the probability of transitioning from normal to partial fault increases significantly due to increased lightning intensity. Each probability value in the matrix is ​​calculated using a fuzzy Bayesian inference algorithm, incorporating data such as voltage mutations and lightning intensity to ensure that it reflects the latest dynamics of the power grid and weather. For example, if a substation experiences voltage anomalies due to a lightning strike, the model may predict a 50% probability of entering a partial fault state and a 20% probability of entering a potential overload state.

[0071] Next, the fuzzy Viterbi algorithm is used, starting from the initial state, to iteratively analyze the transition probabilities along the time steps to find the most likely fault propagation path. The fuzzy Viterbi algorithm is a dynamic programming method specifically used for the optimization prediction of fuzzy state sequences. Assuming that the initial state is that a line has entered a partial fault due to a lightning strike, the algorithm checks all possible state transitions (such as partial fault to potential overload or complete failure) at each time step and selects the path with the highest cumulative probability based on the transition probabilities. For example, if a lightning strike causes a short circuit, the algorithm may predict that the fault will propagate from the damaged line to the adjacent transformer, causing a potential overload. The algorithm uses a backtracking mechanism to generate a state sequence from the initial fault point to the final affected area, such as "partial fault → potential overload → complete failure."

[0072] To enhance the robustness of path prediction, the adaptive fuzzy Markov model can also introduce multi-step probability accumulation analysis to consider the transition probabilities of multiple time steps in the future, rather than focusing only on single-step transitions. This approach can capture the cascading effects of faults. For example, a short circuit caused by a lightning strike may spread to downstream equipment within seconds. Based on multimodal data sets, the system analyzes the grid topology (such as device connection relationships) and environmental data (such as persistently high wind speeds) to predict the long-term trend of fault propagation. For example, in a storm scenario, if the wind speed data indicates an increased risk of line rupture, the algorithm may predict that the partial fault state will continue to transition to a complete fault, affecting multiple adjacent nodes. Multi-step analysis generates a more comprehensive propagation path by accumulating transition probabilities, such as "partial failure of a certain line → potential overload of adjacent lines → regional complete failure."

[0073] Finally, the predicted fault propagation path is output as a state sequence or propagation graph, describing the propagation of the fault from its initial point to the affected area. For example, in a lightning strike scenario, the path might be represented as "partial fault at substation A → potential overload at transmission line B → complete failure at substation C." This path provides input for topological analysis in the graph neural network, which further integrates grid topology data to precisely locate the fault point, such as confirming the location of a short circuit at substation C. The output path can also be visualized using augmented reality devices, assisting operations and maintenance personnel in quickly formulating maintenance strategies, such as prioritizing inspection of damaged lines.

[0074] The above step 205, i.e., "generating fault location information based on topology data by utilizing a graph neural network model, combining the fault propagation path and the multimodal data set," is described in detail below with reference to an embodiment.

[0075] Graph neural network is a deep learning model that specializes in processing graph-structured data and is particularly suitable for analyzing networked systems such as power grids. Grid topology data describes the connection relationship between devices as nodes and transmission lines as edges, forming a complex graph structure. Graph neural networks capture the dependencies between devices and the impact of fault propagation by analyzing the characteristics of nodes and edges. This application uses graph neural networks to analyze topological structures, and combines the propagation paths predicted by fuzzy Markov models with the rich features of multimodal data sets to accurately identify the source and scope of the fault. For example, in a storm scenario, the model may locate a transmission line that has been broken due to strong winds, and predict its potential impact on downstream substations. This comprehensive analysis significantly improves positioning accuracy and is particularly suitable for complex scenarios such as microgrids or vehicles to the grid.

[0076] As an implementable method, the fault location information of the present application can be generated in the following manner: constructing a dynamic graph structure based on the topology data and the fault propagation path, in which the nodes represent the power grid equipment and the edges represent the connection relationship between the equipment and the fuzzy weather impact weight; using the spatiotemporal graph neural network model and attention mechanism, processing the multimodal data set and the fault propagation path to generate fault location information.

[0077] First, a dynamic graph structure is constructed based on the topological data. For example, in a high-voltage power grid, substation A is connected to substation C via line B, forming a graph structure. The system maps the features of the multimodal dataset (such as voltage, current, and lightning intensity) to the nodes and edges of the graph. For example, the node features of substation A include its voltage fluctuation value, and the edge features of line B include the lightning impact weight (based on lightning intensity data). The fault propagation path (predicted by the fuzzy Markov model) is input as a state sequence, such as "partial fault at substation A → potential overload at line B → complete fault at substation C", providing dynamic guidance for fault propagation in the graph neural network.

[0078] Next, a spatiotemporal graph neural network is used to process dynamic graph structures. The spatiotemporal graph neural network combines spatial analysis (based on topological node neighbor relationships) and temporal analysis (based on time series changes in multimodal datasets) to extract fault-related features. In specific implementation, the network first aggregates each node's neighbor information through spatial convolution. For example, the features of substation A are integrated with the voltage and current data of line B and substation C to generate a more comprehensive node representation. The network then analyzes the temporal changes of the multimodal dataset through temporal convolution, such as detecting sudden increases in lightning intensity over the past few seconds. Combined with the trend of voltage mutations, this enhances the perception of short-circuit risks. The fault propagation path serves as a guide, prompting the network to focus on affected nodes and edges, for example, prioritizing the analysis of propagation characteristics from substation A to line B.

[0079] In order to improve positioning accuracy, this application can introduce an attention mechanism to dynamically weight multimodal features and propagation path information. The attention mechanism evaluates the importance of each feature based on real-time data. For example, in scenarios with high lightning intensity, the weight of lightning data is increased, and the weight of voltage mutations is also increased accordingly to highlight short-circuit features. The weight of the propagation path is adjusted according to its predicted probability. For example, if the path predicts that the probability of fault propagation from substation A to line B is 70%, the network will prioritize allocating computing resources to analyze this area. The attention mechanism is implemented through a neural network, which learns a weight allocation strategy based on historical fault patterns and real-time data to ensure that the model can focus on the most important information for fault location, such as the short circuit point of substation A caused by lightning.

[0080] Subsequently, fault location information is generated through the classification module of the graph neural network. The network inputs the fused node features into the fully connected layer to predict whether each node is a fault point and its fault type (such as short circuit, break). For example, substation A may be predicted as a short circuit fault point, and line B as a potential overload risk point. The prediction results are further optimized in combination with the topological data to eliminate unreasonable positioning. For example, if there is no abnormal data in substation C, the system eliminates the possibility of it as a fault point. Finally, the system generates fault location information, such as "substation A short circuit, line B needs to be checked for potential overload", and visualizes it through augmented reality equipment to display the location of the fault point on the power grid map, assisting operation and maintenance personnel in formulating maintenance plans.

[0081] The above method provided in the embodiment of the present application has a wide range of application scenarios in extreme weather conditions, including but not limited to: in the scenario of power grid failure caused by lightning strikes, the method collects voltage, current and lightning intensity data in real time, uses a multimodal attention mechanism for dynamic fusion, combines a fuzzy Markov model to predict the short-circuit propagation path, and uses a graph neural network to accurately locate the damaged substation or line to ensure rapid repair; in microgrid maintenance under storm scenarios, the method analyzes wind speed and humidity data, predicts the risk of line rupture, generates interactive repair guidance, and assists operation and maintenance personnel through augmented reality equipment to improve maintenance efficiency; in the scenario of transformer overload caused by high temperature, the method dynamically adjusts the fuzzy membership function, predicts potential overload propagation, and locates overheating equipment, which is suitable for vehicle-to-grid systems.

[0082] The foregoing description of this specification describes specific embodiments. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different from that described in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order shown or the sequential order to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0083] According to another embodiment, a power grid fault location device based on an adaptive fuzzy Markov model is provided for power grid fault location under extreme weather conditions. Figure 3 FIG. 1 is a schematic block diagram of a power grid fault location device based on an adaptive fuzzy Markov model according to an embodiment. Figure 3 As shown, the device 300 includes:

[0084] The data acquisition unit 301 is configured to collect power grid data and environmental data, wherein the power grid data includes topology data, voltage data, and current data, and the environmental data includes wind speed, temperature, and lightning activity data;

[0085] A multimodal dataset generating unit 302 is configured to use the power grid data and the environmental data and dynamically assign fusion weights of the environmental data and the power grid data using a multimodal attention mechanism to generate a multimodal dataset;

[0086] A fuzzy logic determination unit 303 is configured to define an uncertain state of the power grid based on fuzzy logic, wherein the uncertain state includes normal, partial fault, and potential overload, and dynamically adjust fuzzy membership function parameters through an online learning algorithm;

[0087] a fault path prediction unit 304 configured to calculate, in real time, fuzzy transition probabilities between power grid states according to the fuzzy membership function parameters using an adaptive fuzzy Markov model based on the multimodal data set, and predict a fault propagation path according to the fuzzy transition probabilities;

[0088] The fault localization information generating unit 305 is configured to utilize a graph neural network model, in combination with the fault propagation path and the multimodal data set, to generate fault localization information according to the topology data.

[0089] As an implementable method, the data acquisition unit 301 can be configured to: perform anomaly detection on the currently collected power grid data and the environmental data based on historical data patterns, and identify abnormal fluctuations or missing information; after the anomaly detection, adopt a differentiated processing strategy according to the data source type, give priority to interpolation and redundant cross-validation on the power grid data, and give priority to time alignment and credibility weighted correction on the environmental data.

[0090] As an implementable manner, the multimodal dataset generation unit 302 can be configured to dynamically allocate the fusion weights of the environmental data and the power grid data according to the multimodal attention mechanism: when the lightning strike intensity is greater than 10 kA or the wind speed is greater than 30 m / s, the fusion weight of the environmental data is adjusted to 0.6 or above.

[0091] As an implementable method, the fuzzy logic determination unit 303 can be configured to: continuously update the attribution boundary of the fuzzy state based on the real-time input of the power grid data and the weather data when dynamically adjusting the fuzzy membership function parameters through an online learning algorithm; trigger a rapid convergence mechanism when a sudden change in the power grid state or a drastic fluctuation in the weather environment is detected, and adaptively fine-tune the core parameters of the fuzzy membership function.

[0092] As an implementable manner, when dynamically adjusting the fuzzy membership function parameters through an online learning algorithm, the fuzzy logic determination unit 303 may be configured to dynamically adjust the mean and variance of the fuzzy membership function using a fuzzy C-means clustering algorithm.

[0093] As an implementable manner, the fault path prediction unit 304 calculates the fuzzy transition probability between power grid states in real time based on the multimodal data set and the fuzzy membership function parameters, and predicts the fault propagation path based on the fuzzy transition probability. It can be configured as follows: based on the multimodal data set, a fuzzy state transition probability matrix is ​​constructed, and the fuzzy states include: normal, partial fault and potential overload; using the fuzzy Viterbi inference algorithm, the fuzzy state transition probability matrix is ​​updated in real time according to the multimodal data set, and the path with the highest cumulative probability is selected as the predicted fault propagation path according to the transition probability.

[0094] As an implementable method, the fault location information generation unit 305 can be configured to: construct a dynamic graph structure based on the topology data and the fault propagation path, in which nodes represent grid equipment and edges represent connection relationships between equipment and fuzzy weather impact weights; and utilize a spatiotemporal graph neural network model and an attention mechanism to process the multimodal dataset and the fault propagation path to generate fault location information.

[0095] Each embodiment in this specification is described in a progressive manner. The same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple. For the relevant parts, refer to the partial description of the method embodiment. The device embodiment described above is merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without expending creative work.

[0096] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0097] In addition, an embodiment of the present application further provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps of any one of the methods in the aforementioned method embodiments are implemented.

[0098] And an electronic device comprising: one or more processors; and a memory associated with the one or more processors, the memory being used to store program instructions, which, when read and executed by the one or more processors, execute the steps of the method described in any one of the aforementioned method embodiments.

[0099] The present application also provides a computer program product, comprising a computer program, which implements the steps of any one of the methods described in the aforementioned method embodiments when executed by a processor.

[0100] in, Figure 4 The electronic device architecture is shown as an example, and may include a processor 410, a video display adapter 411, a disk drive 412, an input / output interface 413, a network interface 414, and a memory 420. The processor 410, the video display adapter 411, the disk drive 412, the input / output interface 413, the network interface 414, and the memory 420 may be communicatively connected via a communication bus 430.

[0101] The processor 410 may be implemented as a general-purpose CPU, a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and may be used to execute relevant programs to implement the technical solutions provided in this application.

[0102] The memory 420 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 420 can store an operating system 421 for controlling the operation of the electronic device 400, and a basic input and output system (BIOS) 422 for controlling the low-level operations of the electronic device 400. In addition, a web browser 423, a data storage management system 424, and a power grid fault location device 425 based on an adaptive fuzzy Markov model can also be stored. The above-mentioned power grid fault location device 425 based on an adaptive fuzzy Markov model can be an application program that specifically implements the operations of the aforementioned steps in the embodiment of the present application. In short, when the technical solution provided by the present application is implemented by software or firmware, the relevant program code is stored in the memory 420 and is called and executed by the processor 410.

[0103] The input / output interface 413 is used to connect input / output modules to implement information input and output. The input / output modules can be configured as components in the device (not shown in the figure) or can be externally connected to the device to provide corresponding functions. Input devices may include a keyboard, mouse, touch screen, microphone, various sensors, etc., and output devices may include a display, speaker, vibrator, indicator light, etc.

[0104] The network interface 414 is used to connect to a communication module (not shown) to enable communication between the device and other devices. The communication module can communicate via a wired method (such as USB, network cable, etc.) or a wireless method (such as mobile network, WiFi, Bluetooth, etc.).

[0105] The bus 430 comprises a pathway for transmitting information between the various components of the device (eg, the processor 410 , the video display adapter 411 , the disk drive 412 , the input / output interface 413 , the network interface 414 , and the memory 420 ).

[0106] It should be noted that although the above device only shows a processor 410, a video display adapter 411, a disk drive 412, an input / output interface 413, a network interface 414, a memory 420, a bus 430, etc., in a specific implementation, the device may also include other components necessary for normal operation. In addition, it will be understood by those skilled in the art that the above device may also include only the components necessary to implement the solution of the present application, and does not necessarily include all the components shown in the figure.

[0107] Through the description of the above embodiments, it can be seen that those skilled in the art can clearly understand that the present application can be implemented by means of software plus a necessary general hardware platform. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a computer program product, which can be stored in a storage medium such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in various embodiments of the present application or certain parts of the embodiments.

[0108] The above is a detailed introduction to the technical solutions provided by this application. Specific examples are used herein to illustrate the principles and implementation methods of this application. The description of the above embodiments is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the contents of this specification should not be understood as limiting this application.

Claims

1. A power grid fault location method based on an adaptive fuzzy Markov model, which is used for power grid fault location under extreme weather conditions, characterized in that: The method comprises: Collecting power grid data and environmental data, the power grid data including topology data, voltage data, and current data, and the environmental data including wind speed, temperature, and lightning activity data; Utilizing the power grid data and the environmental data, dynamically allocating fusion weights of the environmental data and the power grid data according to a multimodal attention mechanism to generate a multimodal dataset; Based on fuzzy logic, the uncertain states of the power grid are defined, including normal, partial fault and potential overload, and the fuzzy membership function parameters are dynamically adjusted through an online learning algorithm. Using an adaptive fuzzy Markov model, based on the multimodal data set, and in real time calculating fuzzy transition probabilities between power grid states according to the fuzzy membership function parameters, and predicting a fault propagation path according to the fuzzy transition probabilities; Utilizing a graph neural network model, combined with the fault propagation path and the multimodal dataset, fault location information is generated based on the topological data.

2. The method according to claim 1, characterized in that The method further comprises: Performing anomaly detection on the currently collected power grid data and the environmental data based on historical data patterns to identify abnormal fluctuations or missing information; After the anomaly detection, a differentiated processing strategy is adopted according to the data source type, interpolation and redundant cross-validation are preferentially performed on the power grid data, and time alignment and credibility weighted correction are preferentially performed on the environmental data.

3. The method according to claim 1, characterized in that The dynamically allocating the fusion weights of the environmental data and the power grid data according to the multimodal attention mechanism includes: When the lightning strike intensity is greater than 10 kA or the wind speed is greater than 30 m / s, the fusion weight of the environmental data is adjusted to 0.6 or above.

4. The method according to claim 1, wherein When dynamically adjusting the fuzzy membership function parameters by an online learning algorithm, the method further includes: Continuously updating the attribution boundary of the fuzzy state based on the real-time input of the power grid data and the weather data; When a sudden change in the state of the power grid or a sharp fluctuation in the weather environment is detected, a fast convergence mechanism is triggered to adaptively fine-tune the core parameters of the fuzzy membership function.

5. The method according to claim 1, wherein The dynamic adjustment of fuzzy membership function parameters by online learning algorithm includes: The fuzzy C-means clustering algorithm is used to dynamically adjust the mean and variance of the fuzzy membership function.

6. The method according to claim 4, characterized in that The step of calculating the fuzzy transition probabilities between power grid states in real time based on the multimodal data set and the fuzzy membership function parameters, and predicting the fault propagation path based on the fuzzy transition probabilities includes: Based on the multimodal data set, a fuzzy state transition probability matrix is ​​constructed, wherein the fuzzy state includes: normal, partial fault and potential overload; The fuzzy state transition probability matrix is ​​updated in real time according to the multimodal data set using a fuzzy Viterbi inference algorithm, and a path with the highest cumulative probability is selected as the predicted fault propagation path according to the transition probability.

7. The method according to claim 1, characterized in that The using of a graph neural network model, combining the fault propagation path and the multimodal dataset, analyzing the power grid topology, and generating fault location information includes: Constructing a dynamic graph structure based on the topology data and the fault propagation path, wherein nodes represent power grid devices and edges represent connection relationships between devices and fuzzy weather impact weights; The multimodal dataset and fault propagation path are processed using a spatiotemporal graph neural network model and an attention mechanism to generate fault location information.

8. A power grid fault location device based on an adaptive fuzzy Markov model, which is used for power grid fault location in extreme weather conditions, characterized in that: The device comprises: a data acquisition unit configured to acquire power grid data and environmental data, wherein the power grid data includes topology data, voltage data, and current data, and the environmental data includes wind speed, temperature, and lightning activity data; a multimodal data set generating unit configured to utilize the power grid data and the environmental data, dynamically assign fusion weights of the environmental data and the power grid data according to a multimodal attention mechanism, and generate a multimodal data set; a fuzzy logic determination unit configured to define an uncertain state of the power grid based on fuzzy logic, the uncertain state including: normal, partial fault, and potential overload, and dynamically adjust fuzzy membership function parameters through an online learning algorithm; a fault path prediction unit configured to calculate, in real time, fuzzy transition probabilities between power grid states according to the fuzzy membership function parameters using an adaptive fuzzy Markov model based on the multimodal data set, and predict a fault propagation path according to the fuzzy transition probabilities; The fault location information generating unit is configured to utilize a graph neural network model, combine the fault propagation path and the multimodal data set, and generate fault location information according to the topology data.

9. An electronic device, characterized in that: include: one or more processors; and a memory associated with the one or more processors, the memory being used to store program instructions, wherein the program instructions, when read and executed by the one or more processors, execute the steps of the method according to any one of claims 1 to 7.

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

Images