Power outage area prediction method, power outage sensitive user identification method and system

Abstract

The application belongs to the technical field of smart grid. A power outage area prediction method, a power outage sensitive user identification method and system are provided. For nodes in a device node set, initial node features are obtained by feature extraction according to historical power data, and initial edge weights are determined according to physical or logical relationships. For nodes with changed states, the initial node features and the initial edge weights are updated to obtain updated node features and updated edge weights. The node outage probability is determined according to the updated node features, the edge fault propagation probability is determined according to the updated node features and the updated edge weights, all fault nodes are determined according to the node outage probability and the edge fault propagation probability, and the area where all the fault nodes are located is taken as the power outage area prediction result, and then the power outage sensitive user is identified. The application not only improves the accuracy and reliability of the power outage area prediction, but also ensures the accuracy of the sensitive user identification.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of smart grid, in particular to a power outage area prediction method, a power outage sensitive user identification method and system. BACKGROUND

[0002] The statements in this section merely provide background technology related to the present application and do not necessarily constitute prior art.

[0003] With the rapid development of smart grid technology and the continuous updating of power system equipment, the intelligent level of power grid has been significantly improved. Power outage events, especially sudden power outages caused by faults, are occasional events that are difficult to completely avoid in power system operation. In order to more efficiently identify power outage areas, power companies have installed a large number of power outage monitoring devices at key nodes and important areas of the power grid to monitor the operation state of the power grid in real time, and automatically generate power outage events and quickly report to the power information collection system when monitoring voltage, current and other data anomalies reach a certain threshold; however, although the popularity of power outage monitoring devices has greatly improved the response speed of power outage events, due to technical reasons, these devices often have false positives, false negatives and other problems in actual application, affecting the accuracy and efficiency of power outage area research and judgment.

[0004] In addition, during the power outage, the power demand of users cannot be met, which may cause a series of negative effects, and the severity of these effects will vary depending on the type and demand of the user, therefore, how to quickly and accurately identify the most sensitive user group to power outage after the occurrence of power outage event and prioritize the power supply has become an important challenge for power companies, traditional methods analyze historical complaint data of users, and users with more complaints are considered sensitive customers; or rely on staff to manually enter information of important power users, the above-mentioned solutions are low in efficiency, limited in coverage, and out of date in data updating, and cannot meet the requirements of modern power system for accuracy and timeliness, and the data source of the above-mentioned solutions is single, lacking in-depth analysis of user power consumption behavior. SUMMARY

[0005] In order to solve the problems of the prior art, the present application provides a power outage area prediction method, a power outage sensitive user identification method and system, a tree-shaped device file topology graph is constructed based on the physical structure of the power grid and the node features are initialized, the node feature update is triggered when the node state changes, all fault nodes are determined according to the node power outage probability and the edge fault propagation probability, and the area where all the fault nodes are located is taken as the power outage area prediction result, which has high accuracy and reliability.

[0006] In order to achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0007] In a first aspect, the present application provides a power outage area prediction method.

[0008] A power outage area prediction method comprises the following processes:

[0009] Determine a device node set and a connection relationship set according to a device profile topology map;

[0010] For a node in the device node set, perform feature extraction according to historical power data to obtain initial node features, and determine initial edge weights according to physical or logical relationships;

[0011] For a node whose state changes, update the initial node features and the initial edge weights to obtain updated node features and updated edge weights;

[0012] Determine node outage probabilities according to the updated node features, determine edge fault propagation probabilities according to the updated node features and the updated edge weights, determine all fault nodes according to the node outage probabilities and the edge fault propagation probabilities, and take an area where the all fault nodes are located as a power outage area prediction result.

[0013] In a second aspect, the present application provides a power outage area prediction system.

[0014] A power outage area prediction system comprises:

[0015] A node and connection relationship determination unit is configured to determine a device node set and a connection relationship set according to a device profile topology map;

[0016] A node feature processing unit is configured to, for a node in the device node set, perform feature extraction according to historical power data to obtain initial node features, and determine initial edge weights according to physical or logical relationships;

[0017] A node feature updating unit is configured to, for a node whose state changes, update the initial node features and the initial edge weights to obtain updated node features and updated edge weights;

[0018] A power outage area prediction result generation unit is configured to determine node outage probabilities according to the updated node features, determine edge fault propagation probabilities according to the updated node features and the updated edge weights, determine all fault nodes according to the node outage probabilities and the edge fault propagation probabilities, and take an area where the all fault nodes are located as a power outage area prediction result.

[0019] In a third aspect, the present application provides a power outage sensitive user identification method.

[0020] A power outage sensitive user identification method comprises the following processes:

[0021] The power outage area prediction method in the first aspect of the present application is used to determine a power outage area.

[0022] The system acquires and preprocesses power-related data from multiple data sources to determine user data for each user within the power outage area. The power-related data includes real-time power consumption data from the power consumption acquisition system, power outage event records from the power outage event acquisition system, and user profile data from the data platform marketing system.

[0023] Based on the user data and the pre-trained user identification model, a user sensitivity probability value is obtained. The user sensitivity level is determined by comparing the user sensitivity probability value with a set threshold.

[0024] Fourthly, the present invention provides a power outage-sensitive user identification system.

[0025] A power outage-sensitive user identification system, comprising:

[0026] The power outage area prediction unit is configured to: determine the power outage area using the power outage area prediction method described in the first aspect of the present invention;

[0027] The user data preprocessing unit is configured to: acquire and preprocess power-related data from multiple data sources, determine user data for each user within the power outage area, wherein the power-related data includes real-time power consumption data from the power consumption acquisition system, power outage event records from the power outage event acquisition system, and user profile data from the data platform marketing system.

[0028] The sensitivity classification unit is configured to: obtain a user sensitivity probability value based on the user data and a pre-trained user identification model, and determine the user sensitivity level by comparing the user sensitivity probability value with a set threshold.

[0029] Fifthly, the present invention provides a computer device, comprising: a processor and a computer-readable storage medium;

[0030] A processor, adapted to execute computer programs;

[0031] A computer-readable storage medium storing a computer program, which, when executed by the processor, implements the power outage area prediction method as described in the first aspect of the present invention; or implements the power outage sensitive user identification method as described in the third aspect of the present invention.

[0032] In a sixth aspect, the present invention provides a computer-readable storage medium storing a computer program adapted to be loaded by a processor and execute the power outage area prediction method as described in the first aspect of the present invention; or execute the power outage sensitive user identification method as described in the third aspect of the present invention.

[0033] Compared with the prior art, the beneficial effects of the present invention are:

[0034] 1. This invention innovatively proposes a power outage area prediction method. Based on the physical structure of the power grid, a tree-shaped equipment file topology map is constructed and node features are initialized. When the node state changes, the node feature is updated. All faulty nodes are determined according to the node outage probability and the edge fault propagation probability. The area where all faulty nodes are located is used as the power outage area prediction result, which improves the accuracy of power outage area prediction.

[0035] 2. This invention innovatively employs hierarchical topology modeling combined with a DGNN (Dynamic Graph Neural Network) incremental update mechanism to construct a tree-like topology graph based on the physical structure of the power grid, and uses a temporal convolutional network to initialize node features. When the monitored value exceeds a threshold, a local feature update is triggered. The node outage probability is output through an MLP model that integrates its own features with the risk peaks of its neighbors. Simultaneously, the edge fault propagation probability is calculated using a dot product attention mechanism based on physical constraints. The derived outage area has high accuracy and reliability.

[0036] 3. This invention adopts a multi-source data fusion analysis framework. By integrating power consumption data collection system, power outage records and user profile data, a three-dimensional feature evaluation system is constructed, which greatly improves the accuracy of sensitive customer identification. The simulation system pushes outage areas and sensitive customer lists in real time, which helps to form a closed-loop link of "fault location - priority dispatch - on-site feedback - model fine-tuning".

[0037] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0038] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0039] Figure 1 A flowchart illustrating a power outage area prediction method provided as an exemplary embodiment of the present invention;

[0040] Figure 2 A schematic diagram of the principle of a power outage area prediction system provided as an exemplary embodiment of the present invention;

[0041] Figure 3 A schematic diagram of a power outage-sensitive user identification system provided as an exemplary embodiment of the present invention;

[0042] Figure 4A schematic diagram of a computer device provided for an exemplary embodiment of the present invention. Detailed Implementation

[0043] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0044] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0045] This implementation proposes a method for predicting power outage areas. Relying on declassified data provided by the Shandong Provincial Key Laboratory of Artificial Intelligence Applications in Public Services and the Shandong Provincial Laboratory for Future Industries of General Artificial Intelligence in Higher Education Institutions, a hierarchical topology graph is constructed based on the actual power grid structure. Based on historical electricity consumption data, a temporal convolutional network is used for feature extraction, and the weights of graph nodes and edges are initialized. Based on real-time monitoring data, a graph neural network is used for incremental node feature updates and edge weight adjustments. Based on the updated graph data, a multilayer perceptron and a dot product attention mechanism are used to output the node outage probability and edge fault propagation probability, respectively, and the output values ​​are used to accurately predict power outage areas. Simultaneously, based on the predicted power outage areas, this invention proposes a method for identifying power outage-sensitive users. This involves constructing a unified data model by associating and integrating multi-source data. Subsequently, the original data is cleaned and supplemented based on user electricity consumption behavior patterns. Finally, a logistic regression algorithm is used to model the sensitivity level of customers, integrating a customer tag library. The power outage area information determined using the aforementioned method is used as the core input to assess the sensitivity level of customers.

[0046] More specifically, such as Figure 1 As shown, the process includes the following:

[0047] S1: Collect and organize power grid-related data, and preprocess this data.

[0048] S101: Organize the power equipment and their connection relationships in the power system, and construct a hierarchical equipment file topology diagram. The equipment file topology diagram is a tree structure with distinct levels.

[0049] S102: Collect historical power data from the power consumption information collection system and preprocess this data, including missing value marking, non-missing value standardization, data definition and storage, etc.

[0050] S103: Acquire relevant data from multiple data sources, including real-time electricity consumption data from the electricity consumption data collection system, power outage event records from the power outage event data collection system, and user profile data from the data platform marketing system. By linking and integrating these multi-source data, a unified user data model is constructed. Based on users' electricity consumption behavior patterns, the raw data is cleaned using techniques such as data completion, noise filtering, and rule-based judgment to preprocess missing, abnormal, or inconsistent data, thereby significantly improving the quality of the basic data.

[0051] Specifically, data collection was conducted against the backdrop of a city's core power grid (including 120 public transformer substations and 850 electricity meter nodes) and associated data from 100,000 users. The historical electricity data centrally includes electricity consumption data from June to November 2024, with user electricity consumption data collected every hour.

[0052] S2: The topology diagram of this implementation based on the original power system is as follows:

[0053] Device node set ,node The equipment in the power grid is divided into four layers from top to bottom: public transformer substations, dedicated transformer terminals, metering boxes, and electricity meters; the connection relationships are set. ,side Represents device node With nodes Physical or logical connections between them, such as the connection between the meter and the metering box, or the connection between the metering box and the public transformer area.

[0054] S3: This implementation method will define historical electricity data from the electricity information collection system as... , representing nodes exist The original monitoring data at each moment, This indicates the node status, where 0 represents a normal state and 1 represents a fault state.

[0055] S4: Based on historical power data, initialize the weights of the topology graph constructed in step B.

[0056] S401: Based on historical electricity consumption data from the electricity consumption information collection system, a Temporal Convolutional Network (TCN) is used to extract features from the electricity consumption data. TCN, with its dilated convolutional structure, can effectively capture complex long-term dependencies and periodic dynamic patterns in the electricity data, thereby performing highly expressive feature extraction and outputting results. As the initial node embedding for a Dynamic Graph Neural Network (DGNN), this step provides an information-rich and temporally relevant initial node representation foundation for subsequent graph neural networks.

[0057] (1);

[0058] in, for Time Node The original monitoring data, output As the initial node embedding for a dynamic graph neural network.

[0059] S402: Edge Weight Initialization is based on the physical distance or logical connection strength between devices, providing a physical and logical basis for the subsequent deduction of fault propagation paths. If there is a physical connection between nodes (such as a cable), then... ,in, Minimal values ​​are used to avoid division by zero; if it is a logical connection (such as a management relationship), then... .

[0060] S5: Continuously monitors node real-time data, triggering updates when equipment status changes abruptly (e.g., current overload) or indicators exceed thresholds (e.g., voltage deviation > 10%). Considering the large number of power grid nodes and their extremely complex relationships, a highly efficient incremental update strategy is adopted to ensure system processing efficiency and real-time performance: updates are only applied to nodes whose status has changed. It updates its direct parent / child nodes, and after the update, it aggregates the parent node features upwards and broadcasts the child node status downwards.

[0061] S501: Continuously monitor the real-time monitoring data stream of the node. When a sudden change in equipment status is detected (e.g., instantaneous current value ≥ 150% of rated current) or a key indicator exceeds the threshold for three consecutive monitoring cycles (sampling cycle of 5 seconds) (e.g., absolute value of voltage deviation > 10%), the node is marked as a potentially faulty node.

[0062] S502: A continuous-time graph convolutional network architecture is used to update the node features of potentially faulty nodes. The graph neural network effectively enhances the feature representation ability of nodes by automatically aggregating information from neighboring nodes and can capture potential signs of the transmission of related fault risks.

[0063] The continuous-time GNN update operation based on neighbor aggregation is as follows:

[0064] (2);

[0065] in, This represents the activation function; in this embodiment, the ReLU function is used. Represents a node The set of direct neighbors; and The weight matrix is ​​a learnable matrix; For nodes Initial node characteristics; represent time, For nodes The initial node characteristics, This represents the updated node characteristics.

[0066] S503: The updated node features are used to adjust the weights of edges related to potentially faulty nodes. Dynamically adjusting the edge weights allows the model to adaptively capture real-time changes in the fault propagation intensity in the power grid topology. The edge weight adjustment operation is as follows:

[0067] (3);

[0068] in, Represents a node With nodes The edge between; For nodes exist Features at any given moment (i.e., updated features); This is a forgetting factor used to control the decay of historical weights. The global maximum feature distance is used to normalize the propagation intensity.

[0069] S504: After updating the graph weights, a two-layer fully connected network is used to calculate the node outage probability. This model integrates the node's own features with the maximum eigenvalue of its neighbors (representing local risk peaks), effectively enhancing the model's ability to perceive potential local fault hotspots. The node outage probability is calculated as follows:

[0070] (4);

[0071] in, This represents the activation function; in this embodiment, the sigmoid function is used. Represents a node The set of direct neighbor nodes; Indicates the first Layer node characteristics, (here) Representative node (updated node features) ; This indicates a splicing operation.

[0072] Simultaneously, a physically constrained dot product attention mechanism is used to quantify the tendency of fault propagation along edges. This mechanism intelligently focuses on the connection path most likely to propagate faults based on node feature relevance and edge weights. The edge fault propagation probability is calculated as follows:

[0073] (5);

[0074] S505: Probability of node power outage Then the node Mark as a faulty node; if the edge fault propagation probability Then Mark the node as a potential failure point and repeat step S502, where and All of these are risk thresholds; setting risk thresholds , This judgment mechanism simulates the logical process of actual fault propagation in the power grid, ensuring the rationality and accuracy of the inference results.

[0075] S6: Based on S5, obtain the set of faulty nodes. The area where the faulty nodes are located is the deduced power outage area. Use the binary cross-entropy loss function to train the model to predict the power outage status of the nodes (fault = 1, normal = 0); let... It is a node The true label (obtained based on actual power outage records from subsequent feedback). If the model predicts the probability of a power outage at this node, then the final loss function is:

[0076] (6);

[0077] in, Represents a set The number of nodes in the system.

[0078] By iteratively solving S5, the minimum value is continuously minimized. This allows us to obtain the learning parameters of the model proposed in this embodiment.

[0079] The system assesses the probability of power outages in the test samples, pushes the assessment results, compares them with the actual power outage situation, and updates the underlying data information to continuously optimize the data weight values ​​in the model, thereby significantly improving the accuracy of power outage area projection.

[0080] S7: The preprocessed user data is obtained from S103. Based on user electricity consumption data, customer tag information, customer service work order complaints and other data, a proportional sample of sensitive customer data and a sample of non-sensitive customer electricity consumption data are selected as the model sample set. The corresponding proportion of data is randomly extracted from the model sample set to form the training set data and the test set data.

[0081] S8: For the processed data model, use the logistic regression algorithm to model the sensitivity level of customers. The logistic regression model has a clear structure, is efficient in training, and can directly output probability values, making it suitable for probabilistic classification and determination of the sensitivity level of such customers.

[0082] S801: Define user data vectors from the training set data. The feature weight vector is ,in Represents user data vector eigenvectors The dimension. Then the objective function. T represents the vector transpose, and the set of categories. (Representing sensitive customers and non-sensitive customers respectively), initial iteration count Allowable error Based on a certain range Randomly assign values.

[0083] S802: Entering the iterative solution process. ;

[0084] Find the optimal solution for the objective function. The gradient of the objective function is as follows:

[0085] (7);

[0086] in, express The probability, express The probability, Indicates that the data in the training set belongs to The number of samples.

[0087] The feature weight vector is updated after a series of calculations. , .in This represents the learning rate, which controls the magnitude of the update of the feature weight vector in each iteration. Indicates the first Metrics in the next iteration (such as gradient or error information) are used for feature weight updates.

[0088] S803: Transfer the feature weight vector Substitute into the objective function Determine whether the following expression is true:

[0089] (8);

[0090] If the condition is met, return to S802 to continue the iterative solution process; if the condition is not met, obtain the optimal objective function obtained in this calculation and proceed to the next step.

[0091] S804: A sensitive customer rating model based on an optimized objective function:

[0092] (9);

[0093] Among them, probability value This directly represents the probability that the user is a sensitive customer under the current (or projected) power outage conditions; among which, This is test sample data, belonging to the test set data obtained from S7.

[0094] S805: Conduct sample testing on the sensitive customer rating model, compare probabilities, compare the probabilities with the category proportions to obtain the corresponding class labels, and classify the test samples.

[0095] Customer sensitivity levels are classified as follows:

[0096] like If so, it is classified as a Level 1 sensitive customer (power outage has an extremely high impact and is guaranteed with the highest priority).

[0097] like These are classified as Level 2 sensitive customers (high impact from power outages, second highest priority for protection).

[0098] like Then it is classified as a Level 3 sensitive customer (the impact of power outage is moderate, and it is generally given priority protection).

[0099] like Those that are not sensitive customers are classified as such.

[0100] S806: Calculate the test parameters of the test set data and determine whether they meet the requirements of the initial model for sensitive customer level evaluation. If the test set accuracy is lower than 0.9, return to S801 and try to adjust the initial parameter range and retrain. If it meets the requirements, proceed to the next step.

[0101] S807: Complete the sensitive customer level evaluation model and output the updated model.

[0102] S9: Obtain the accurate power outage area through S6, and extract a list of all affected users within that area. Input user information (including user profiles, electricity usage data, complaint details, etc.) into the sensitive customer level evaluation model obtained in step H, and obtain the output results. According to the preset sensitivity level threshold rules (as described in S805), classify users into Level 1, Level 2, Level 3 sensitive customers or non-sensitive customers; push the final power outage area and the sensitivity level list of users within the area (sorted by priority) to the emergency repair command system in real time to assist in resource scheduling and differentiated service decisions.

[0103] Figure 2 A power outage area prediction system is shown, comprising:

[0104] The node and connection relationship determination unit 201 is configured to: determine the set of device nodes and the set of connection relationships based on the device file topology diagram;

[0105] The node feature processing unit 202 is configured to: extract features from the nodes in the device node set based on historical power data to obtain initial node features, and determine initial edge weights based on physical or logical relationships.

[0106] The node feature update unit 203 is configured to update the initial node features and initial edge weights for nodes whose states have changed, so as to obtain updated node features and updated edge weights.

[0107] The power outage area prediction result generation unit 204 is configured to: determine the node outage probability based on the updated node features; determine the edge fault propagation probability based on the updated node features and updated edge weights; determine all faulty nodes based on the node outage probability and edge fault propagation probability; and use the area where all faulty nodes are located as the power outage area prediction result.

[0108] Figure 3 A power outage-sensitive user identification system is shown, comprising:

[0109] User data preprocessing unit 301 is configured to: acquire power-related data from multiple data sources and preprocess it to determine user data for each user in the power outage area. The power-related data includes real-time power consumption data from the power consumption acquisition system, power outage event records from the power outage event acquisition system, and user profile data from the data platform marketing system.

[0110] Sensitivity classification unit 302 is configured to: obtain a user sensitivity probability value based on the user data and a pre-trained user identification model, and determine the user sensitivity level based on a comparison between the user sensitivity probability value and a set threshold.

[0111] It is understood that the aforementioned units can be individually or entirely merged into one or more other units, or some of the units can be further divided into multiple functionally smaller units. This achieves the same operation without affecting the technical effects of the embodiments of this application. The aforementioned units are based on logical functional division. In practical applications, the function of one unit can be implemented by multiple units, or the function of multiple units can be implemented by one unit. In other embodiments of this application, the system may also include other units. In practical applications, these functions can also be implemented with the assistance of other units, and can be implemented collaboratively by multiple units.

[0112] According to another embodiment of this application, the system described in this embodiment can be constructed by running a computer program (including program code) capable of performing the steps involved in the corresponding method of the present invention on a general-purpose computing device, such as a computer, which includes processing elements and storage elements such as a central processing unit (CPU), random access memory (RAM), and read-only memory (ROM). The computer program can be recorded on, for example, a computer-readable recording medium, loaded into the aforementioned computing device through the computer-readable recording medium, and run therein.

[0113] Figure 4 A computer device is shown, which includes a processor 401, a communication interface 402, and a computer-readable storage medium 403. The processor 401, communication interface 402, and computer-readable storage medium 403 can be connected via a bus or other means.

[0114] The communication interface 402 is used to receive and send data. The computer-readable storage medium 403 can be stored in the memory of the electronic device. The computer-readable storage medium 403 is used to store computer programs, which include program instructions. The processor 401 is used to execute the program instructions stored in the computer-readable storage medium 403.

[0115] The processor 401 is the computing and control core of the electronic device. It is suitable for implementing one or more instructions, specifically for loading and executing one or more instructions to achieve the corresponding method flow or corresponding function.

[0116] The processor 401 is configured to perform the following process:

[0117] Determine the set of device nodes and the set of connection relationships based on the device file topology diagram;

[0118] For nodes in the device node set, feature extraction is performed based on historical power data to obtain initial node features, and initial edge weights are determined based on physical or logical relationships.

[0119] For nodes whose states have changed, the initial node features and initial edge weights are updated to obtain the updated node features and updated edge weights.

[0120] The power outage probability of a node is determined based on the updated node characteristics. The edge fault propagation probability is determined based on the updated node characteristics and the updated edge weights. All faulty nodes are determined based on the node power outage probability and the edge fault propagation probability. The area where all faulty nodes are located is used as the power outage area prediction result.

[0121] Alternatively, perform the following procedure:

[0122] The system acquires and preprocesses power-related data from multiple data sources to determine user data for each user within the power outage area. The power-related data includes real-time power consumption data from the power consumption acquisition system, power outage event records from the power outage event acquisition system, and user profile data from the data platform marketing system.

[0123] Based on the user data and the pre-trained user identification model, a user sensitivity probability value is obtained. The user sensitivity level is determined by comparing the user sensitivity probability value with a set threshold.

[0124] This invention also provides a computer-readable storage medium, which is a memory device in an electronic device for storing programs and data. It is understood that the computer-readable storage medium here may include both built-in storage media in the electronic device and extended storage media supported by the electronic device. The computer-readable storage medium provides storage space for storing the processing system of the electronic device.

[0125] Furthermore, this storage space also contains one or more instructions suitable for loading and execution by the processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be high-speed RAM memory or unstable memory, such as at least one disk storage device; optionally, it can also be at least one computer-readable storage medium located remotely from the aforementioned processor.

[0126] In one embodiment, the computer-readable storage medium stores one or more instructions; the processor loads and executes the one or more instructions stored in the computer-readable storage medium to perform the following process:

[0127] Determine the set of device nodes and the set of connection relationships based on the device file topology diagram;

[0128] For nodes in the device node set, feature extraction is performed based on historical power data to obtain initial node features, and initial edge weights are determined based on physical or logical relationships.

[0129] For nodes whose states have changed, the initial node features and initial edge weights are updated to obtain the updated node features and updated edge weights.

[0130] The power outage probability of a node is determined based on the updated node characteristics. The edge fault propagation probability is determined based on the updated node characteristics and the updated edge weights. All faulty nodes are determined based on the node power outage probability and the edge fault propagation probability. The area where all faulty nodes are located is used as the power outage area prediction result.

[0131] Alternatively, the following process can be implemented:

[0132] The system acquires and preprocesses power-related data from multiple data sources to determine user data for each user within the power outage area. The power-related data includes real-time power consumption data from the power consumption acquisition system, power outage event records from the power outage event acquisition system, and user profile data from the data platform marketing system.

[0133] Based on the user data and the pre-trained user identification model, a user sensitivity probability value is obtained. The user sensitivity level is determined by comparing the user sensitivity probability value with a set threshold.

[0134] The present invention also provides a computer program product or computer program comprising computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform the following process:

[0135] Determine the set of device nodes and the set of connection relationships based on the device file topology diagram;

[0136] For nodes in the device node set, feature extraction is performed based on historical power data to obtain initial node features, and initial edge weights are determined based on physical or logical relationships.

[0137] For nodes whose states have changed, the initial node features and initial edge weights are updated to obtain the updated node features and updated edge weights.

[0138] The power outage probability of a node is determined based on the updated node characteristics. The edge fault propagation probability is determined based on the updated node characteristics and the updated edge weights. All faulty nodes are determined based on the node power outage probability and the edge fault propagation probability. The area where all faulty nodes are located is used as the power outage area prediction result.

[0139] Alternatively, perform the following procedure:

[0140] The system acquires and preprocesses power-related data from multiple data sources to determine user data for each user within the power outage area. The power-related data includes real-time power consumption data from the power consumption acquisition system, power outage event records from the power outage event acquisition system, and user profile data from the data platform marketing system.

[0141] Based on the user data and the pre-trained user identification model, a user sensitivity probability value is obtained. The user sensitivity level is determined by comparing the user sensitivity probability value with a set threshold.

[0142] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0143] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in or transmitted through a computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic cable, digital cable) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data processing device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive), etc.

[0144] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for predicting power outage areas, characterized in that, Includes the following processes: Determine the set of device nodes and the set of connection relationships based on the device file topology diagram; For nodes in the device node set, feature extraction is performed based on historical power data to obtain initial node features, and initial edge weights are determined based on physical or logical relationships. For nodes whose states have changed, the initial node features and initial edge weights are updated to obtain updated node features and updated edge weights. The power outage probability of a node is determined based on the updated node characteristics. The edge fault propagation probability is determined based on the updated node characteristics and the updated edge weights. All faulty nodes are determined based on the node power outage probability and the edge fault propagation probability. The area where all faulty nodes are located is used as the power outage area prediction result. The initial node features are updated to obtain the updated node features. ,include: ;in, and The weight matrix is ​​a learnable matrix. External environmental bias Represents a node The set of direct neighbors, For nodes The initial node characteristics, Represents the activation function. represent time, For nodes The initial node features are determined; the initial edge weights are updated to obtain the updated edge weights. ,include: ;in, represent Time Node and nodes Edge weights between them Representative node exist The updated node features at any given time. Representative node exist The updated node features at any given time. Forgetting factor, The global maximum feature distance; node The probability of a power outage is: ;node and nodes The probability of fault propagation between them is: ;in, This represents the activation function. Represents a node The set of direct neighbor nodes, Indicates the first Layer node characteristics, , , This indicates a splicing operation. Representative node Updated node features Representing the The updated node characteristics of each node. Representative node and nodes Edge weights between nodes; probability of node power outage Then the node Mark as a faulty node; if the edge fault propagation probability Then Nodes marked as potentially faulty, among which, and All are risk thresholds. The value is greater than The value of .

2. The power outage area prediction method as described in claim 1, characterized in that, Includes the following processes: Determine the initial edge weights based on physical or logical relationships, including: If node and nodes If there is a physical connection between the nodes, then and nodes The initial edge weights are ,in, The length of the physical connection. A positive real number less than a set threshold; If node and nodes If the relationship between them is logical, then the nodes... and nodes The initial edge weight is 1.

3. A power outage area prediction system, characterized in that, include: The node and connection relationship determination unit is configured to: determine the set of device nodes and the set of connection relationships based on the device file topology diagram; The node feature processing unit is configured to: extract initial node features from nodes in the device node set based on historical power data, and determine initial edge weights based on physical or logical relationships. The node feature update unit is configured to update the initial node features and initial edge weights for nodes whose states have changed, so as to obtain the updated node features and updated edge weights. The power outage area prediction result generation unit is configured to: determine the node outage probability based on the updated node features, determine the edge fault propagation probability based on the updated node features and the updated edge weights, determine all fault nodes based on the node outage probability and the edge fault propagation probability, and take the area where all fault nodes are located as the power outage area prediction result. The initial node features are updated to obtain the updated node features. ,include: ;in, and The weight matrix is ​​a learnable matrix. External environmental bias Represents a node The set of direct neighbors, For nodes The initial node characteristics, Represents the activation function. represent time, For nodes The initial node features are determined; the initial edge weights are updated to obtain the updated edge weights. ,include: ;in, represent Time Node and nodes Edge weights between them Representative node exist The updated node features at any given time. Representative node exist The updated node features at any given time. Forgetting factor, The global maximum feature distance; node The probability of a power outage is: ;node and nodes The probability of fault propagation between them is: ;in, This represents the activation function. Represents a node The set of direct neighbor nodes, Indicates the first Layer node characteristics, , , This indicates a splicing operation. Representative node Updated node features Representing the The updated node characteristics of each node. Representative node and nodes Edge weights between nodes; probability of node power outage Then the node Mark as a faulty node; if the edge fault propagation probability Then Nodes marked as potentially faulty, among which, and All are risk thresholds. The value is greater than The value of .

4. A method for identifying power outage-sensitive users, characterized in that, Includes the following processes: The power outage area is determined using the power outage area prediction method according to any one of claims 1-2; The system acquires and preprocesses power-related data from multiple data sources to determine user data for each user within the power outage area. The power-related data includes real-time power consumption data from the power consumption acquisition system, power outage event records from the power outage event acquisition system, and user profile data from the data platform marketing system. Based on the user data and the pre-trained user identification model, a user sensitivity probability value is obtained. The user sensitivity level is determined by comparing the user sensitivity probability value with a set threshold.

5. A power outage-sensitive user identification system, characterized in that, include: The power outage area prediction unit is configured to: determine the power outage area using the power outage area prediction method according to any one of claims 1-2; The user data preprocessing unit is configured to: acquire and preprocess power-related data from multiple data sources, determine user data for each user within the power outage area, wherein the power-related data includes real-time power consumption data from the power consumption acquisition system, power outage event records from the power outage event acquisition system, and user profile data from the data platform marketing system. The sensitivity classification unit is configured to: obtain a user sensitivity probability value based on the user data and a pre-trained user identification model, and determine the user sensitivity level by comparing the user sensitivity probability value with a set threshold.

6. A computer device, characterized in that, include: Processor and computer-readable storage media; A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, implements the power outage area prediction method as described in any one of claims 1-2; or implements the power outage sensitive user identification method as described in claim 4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program adapted to be loaded by a processor and execute the power outage area prediction method as described in any one of claims 1-2; or execute the power outage sensitive user identification method as described in claim 4.

Images