Power grid line loss reduction method and system based on fault prediction

Abstract

The invention provides a power grid line loss restoration method and system based on fault prediction, and relates to the technical field of fault prediction, and the method comprises the steps: carrying out the topology modeling according to the real-time operation data flow of a power grid in a power supply region, and restoring the dynamic power grid topology; integrating multi-source node monitoring data, and performing state monitoring to obtain a plurality of real-time health degrees; performing fault conduction influence prediction, and outputting power grid fault probability topology; performing line loss factor mapping to generate a plurality of simulation operation scenes; performing layered line loss increment responsibility distribution calculation, and restoring and outputting a plurality of line loss responsibility distribution maps; fusing and outputting a loss reduction intervention strategy; and after active regulation and control, closed-loop tracking and iterative optimization of the line loss space-time associated data are executed. The technical problems that in the prior art, line loss prediction is generally carried out based on a static topology model of a power grid, dynamic changes of the power grid cannot be captured in real time, the line loss prediction precision is low, and dispatching and optimization of the power grid are affected are solved.

Description

Technical Field

[0001] This invention relates to the field of fault prediction technology, and specifically to a method and system for restoring power grid line losses based on fault prediction. Background Technology

[0002] As a fundamental infrastructure of modern society, the power grid bears the crucial responsibility of transmitting and distributing electricity. With the continuous growth of electricity demand and the expansion of the power grid, its operation has become increasingly complex. Line losses in the power grid refer to the power loss that occurs during transmission due to current flowing through lines, transformers, and other equipment. Line losses not only increase the operating costs of the power grid but also affect its efficiency, stability, and security. Therefore, accurately predicting power grid line losses and taking timely and effective measures to reduce them has become a key issue in power grid management.

[0003] Existing technologies typically predict line losses based on static topology models of the power grid, relying on fixed load data and power grid structure parameters. This method cannot capture dynamic changes in the power grid in real time, such as changes in the power grid topology, load fluctuations, and changes in equipment health status. As a result, it cannot accurately reflect the complexity of the power grid in actual operation, leading to low accuracy in line loss prediction. It is unable to effectively cope with complex situations such as sudden failures and load fluctuations, affecting the scheduling and optimization of the power grid. Summary of the Invention

[0004] This application provides a method and system for power grid line loss restoration based on fault prediction, aiming to solve the technical problem that existing technologies typically predict line losses based on static topology models of the power grid, which cannot capture dynamic changes in the power grid in real time, resulting in low accuracy of line loss prediction and affecting the scheduling and optimization of the power grid.

[0005] The first aspect disclosed in this application provides a power grid line loss restoration method based on fault prediction. The method includes: performing topology modeling based on real-time power grid operation data streams in the power supply area to restore a dynamic power grid topology; integrating multi-source node monitoring data to monitor the status of multiple key power grid components and obtain multiple real-time health values; projecting the multiple real-time health values ​​onto multiple key power grid nodes in the dynamic power grid topology based on the multiple key power grid components to predict the impact of fault propagation and outputting a power grid fault probability topology; performing line loss factor mapping based on the power grid fault probability topology to generate multiple simulated operation scenarios; calculating the hierarchical line loss incremental responsibility distribution for the multiple simulated operation scenarios to restore and output multiple line loss responsibility distribution maps; fusing the multiple line loss responsibility distribution maps to output a loss reduction intervention strategy; and after actively regulating the power supply area using the loss reduction intervention strategy, performing closed-loop tracking and iterative optimization of the spatiotemporal correlation data of line losses.

[0006] The second aspect of this application discloses a power grid line loss restoration system based on fault prediction. The system is used in the aforementioned power grid line loss restoration method based on fault prediction. The system includes: a topology modeling module for performing topology modeling based on real-time power grid operation data streams in the power supply area to restore a dynamic power grid topology; a status monitoring module for integrating multi-source node monitoring data to monitor the status of multiple key power grid components and obtain multiple real-time health values; a fault propagation impact prediction module for projecting the multiple real-time health values ​​onto multiple key power grid nodes in the dynamic power grid topology based on the multiple key power grid components, performing fault propagation impact prediction, and outputting a power grid fault probability topology; a line loss factor mapping module for performing line loss factor mapping based on the power grid fault probability topology to generate multiple simulated operation scenarios; a responsibility distribution calculation module for performing hierarchical incremental line loss responsibility distribution calculations on the multiple simulated operation scenarios, restoring and outputting multiple line loss responsibility distribution maps; an intervention strategy output module for fusing the multiple line loss responsibility distribution maps and outputting a loss reduction intervention strategy; and an iterative optimization module for performing closed-loop tracking and iterative optimization of line loss spatiotemporal correlation data after actively regulating the power supply area using the loss reduction intervention strategy.

[0007] One or more technical solutions provided in this application have at least the following beneficial effects: Dynamic topology modeling enables the timely reconstruction of the actual topology of the power grid based on real-time operational data streams, accurately reflecting changes in grid status and improving the grid's timeliness and dynamic response capabilities. By integrating data from multiple monitoring nodes, the health status of several key components can be monitored in real time, generating multiple real-time health values ​​that reflect the real-time operational status of each component. Mapping the health data of multiple key grid components onto the dynamic grid topology allows for the prediction of fault propagation paths and impact ranges, outputting a fault probability topology map of the grid. This process reveals the fault propagation trend in different parts of the grid and its impact on the entire grid when a specific fault occurs. Mapping line loss factors based on the grid fault probability topology generates multiple simulated operation... The system can comprehensively assess changes in line losses under different scenarios. By calculating line loss responsibility hierarchically and generating multiple line loss responsibility distribution maps, it can accurately formulate power grid optimization plans, specifically reducing line losses in high-responsibility areas and improving the economy and efficiency of power grid operation. Based on multiple line loss responsibility distribution maps, it integrates various data and formulates corresponding loss reduction intervention strategies. Through scientific intervention strategies, it can effectively reduce ineffective losses in power grid operation and improve the efficiency and economy of the power grid. After implementing loss reduction intervention strategies, it uses a closed-loop tracking system to monitor changes in the spatiotemporal correlation data of line losses in real time and perform iterative optimization. This process can continuously provide feedback on the effect of the adjusted strategies, thereby enabling further optimization and ensuring the high efficiency of power grid operation.

[0008] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0009] Figure 1 This is a schematic diagram of the power grid line loss restoration method based on fault prediction provided in an embodiment of this application.

[0010] Figure 2 A schematic diagram of the power grid line loss restoration system based on fault prediction provided in this application embodiment.

[0011] Figure labeling: Topology modeling module 10, status monitoring module 20, fault propagation impact prediction module 30, line loss factor mapping module 40, responsibility distribution calculation module 50, intervention strategy output module 60, iterative optimization module 70. Detailed Implementation

[0012] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0013] Example 1, as Figure 1 As shown in the embodiment of this application, a method for restoring power grid line losses based on fault prediction is provided. The method includes: A100: Performs topology modeling based on the real-time operation data stream of the power grid in the power supply area to reconstruct the dynamic power grid topology.

[0014] A preliminary power grid topology is constructed based on the actual installation of equipment within the power supply area, such as the connection methods of substations, lines, and switches. In this topology, each electrical connection point represents a topology node, and the connections between electrical devices represent topology edges. Based on the topology, and considering the electrical characteristics of the power grid, such as power transmission direction and voltage levels, logical analysis of electrical connections is performed. For example, this determines the direction of current flow and the power system's dispatching methods. The power grid topology is not static; it changes with the activation, deactivation, and faults of power grid equipment. The topology can be dynamically updated by real-time acquisition of switch position change event signals from the SCADA (Supervisory and Data Acquisition) system.

[0015] A200: Integrates multi-source node monitoring data to monitor the status of multiple key power grid components and obtain multiple real-time health statuses.

[0016] To more accurately monitor the status of power grid components, information from different data sources is collected, including sensor data, equipment maintenance records, and external environmental data. This data comes from the node monitoring units of different power grid components, and multiple monitoring points transmit data to a centralized data platform for processing via wireless communication, optical fiber, and other means.

[0017] To improve computational efficiency and real-time performance, multiple edge intelligent units are pre-deployed at key nodes of the power grid. These units can perform preliminary processing on the collected data locally, such as fault feature extraction, data preprocessing, and health inference, thereby reducing communication latency and improving data processing efficiency. The edge intelligent units extract fault characteristics of power grid components through analysis of multi-source data. Using pre-built pruned neural network models, such as miniaturized and optimized neural network models, the edge intelligent units input the extracted fault features into the model for health inference, and the output health value reflects the state of each power grid component.

[0018] A300: Based on the projection of the multiple real-time health values ​​of the multiple key power grid components onto the multiple key power grid nodes of the dynamic power grid topology, perform fault propagation impact prediction, and output the power grid fault probability topology.

[0019] The obtained real-time health values ​​are projected onto multiple critical nodes in the dynamic power grid topology. Each power grid element is considered a critical node in the topology, and the health data reflects the operating status of these nodes. Based on the health of the power grid elements, the probability of failure for each node is estimated. For example, if the health of a transformer is low, the probability of failure for that node is high, and vice versa.

[0020] In real-world power grids, faults not only occur at the faulty node but can also propagate to other nodes through grid connections. Fault propagation models describe this mechanism, predicting the impact on other nodes and potential cascading failures when a fault occurs at one node by analyzing fault propagation paths, such as current flow paths. In fault propagation, a discrete probability propagation mechanism simulates how a fault in the power grid affects other nodes; for example, if a node fails, it may increase the probability of failure at adjacent nodes.

[0021] By iteratively updating and calculating the fault propagation path, a fault probability topology map of the entire power grid is finally generated. This topology map reflects the fault risk of each node in the power grid and can be used to predict the probability of a fault occurring in the power grid in the future.

[0022] A400: Based on the power grid fault probability topology, line loss factor mapping is performed to generate multiple simulated operation scenarios.

[0023] Line loss factors are calculated using current-voltage relationships and fault impact. Once determined, multiple simulated operating scenarios are generated based on these factors. These scenarios are based on different grid operating conditions, fault probabilities, and load distributions, simulating possible grid operating states. These scenarios help assess grid line losses under different conditions.

[0024] A500: Performs hierarchical calculation of incremental line loss responsibility distribution for the multiple simulated operation scenarios, and reconstructs and outputs multiple line loss responsibility distribution maps.

[0025] For each simulated operation scenario, responsibility for line losses is assigned. Due to the complexity of the power grid structure, this assignment is layered, specifically divided into component, node, and topology layers. Under each simulation scenario, the incremental responsibility for each component, node, and topology is calculated layer by layer. These incremental responsibilities reflect the contribution of different power grid components to line losses under fault conditions. Finally, this layered line loss responsibility distribution data is mapped onto the power grid topology, generating multiple line loss responsibility distribution maps. These maps demonstrate the distribution of line loss responsibility among various components, nodes, and regions in the power grid under different simulation scenarios.

[0026] A600: Integrates multiple line loss liability distribution maps and outputs loss reduction intervention strategies.

[0027] By fusing multiple line loss responsibility distribution maps, different simulation scenarios can be weighted. The weighted maps are calculated based on factors such as the probability of occurrence and fault risk of different scenarios, resulting in a comprehensive line loss responsibility distribution map. Based on this fused map, loss reduction intervention strategies are formulated. The goal of these strategies is to reduce grid line losses and optimize grid operating efficiency, specifically including equipment optimization, load dispatch optimization, and topology optimization.

[0028] A700: After actively regulating the power supply area using the aforementioned loss reduction intervention strategy, perform closed-loop tracking and iterative optimization of the spatiotemporal correlation data of line loss.

[0029] Active regulation of the power grid is achieved through the established loss reduction intervention strategy. During grid regulation, line losses are monitored in real time, and data analysis is performed by combining temporal and spatial information. By tracking changes in line losses through spatiotemporal data acquired by the monitoring system, the effectiveness of the implemented loss reduction intervention strategy is evaluated. Based on the results of spatiotemporal data analysis, the loss reduction intervention strategy is further adjusted and optimized. Through iterative optimization, the power grid operation mode is gradually improved to maximize the reduction of line losses. This process is a continuous feedback loop. As the power grid operating status changes and the intervention strategy is adjusted, the optimization of line losses will continue to advance, ultimately maximizing the efficiency and economic benefits of power grid operation.

[0030] Furthermore, by integrating multi-source node monitoring data to monitor the status of multiple key power grid components and obtain multiple real-time health indicators, the method includes: A210: Pre-deploy multiple edge intelligent units in the multiple key power grid components; A220: After the multiple edge intelligent units access the original monitoring data of multiple nodes of the multiple key power grid components, they perform fault correlation feature extraction to obtain multiple sets of multimodal fault features; A230: Pre-construct multiple pruned neural network inference models according to multiple component types of the multiple key power grid components; A240: Input the multiple sets of multimodal fault features into the multiple pruned neural network inference models to perform health inference and output the multiple real-time health values; A250: The multiple edge intelligent units transmit the multiple real-time health values ​​to the line loss restoration analysis platform through power fiber optic cables.

[0031] By deploying multiple edge intelligent units on key components of the power grid, such as substations, switches, and transformers, these edge intelligent units can independently process monitoring data, perform preliminary data analysis, and reflect the power grid status in real time. Edge computing avoids reliance on centralized data processing centers, thereby reducing data transmission latency and improving real-time performance and response speed.

[0032] Edge intelligent units connect to monitoring systems of key power grid components to acquire real-time raw monitoring data from various devices, including operating parameters of the power system and equipment such as voltage, current, temperature, pressure, and load. After receiving the raw data, the edge intelligent unit extracts fault-related features. Fault feature extraction identifies fault-related characteristic patterns from the raw monitoring data, such as voltage instability, current fluctuations, excessively high equipment temperatures, and vibration changes, all of which are precursors to faults. Through the fusion of multimodal data, the edge intelligent unit can extract multiple sets of fault features of different types.

[0033] Pruned neural networks are neural network structures that improve computational efficiency and reduce model complexity by optimizing and reducing network parameters (such as weights and nodes). Pruning techniques remove unimportant network connections, allowing the neural network to operate with less computational resources while maintaining accuracy. Each power grid component has different operating characteristics; therefore, corresponding pruned neural network inference models are pre-built for different types of components. Different neural networks can perform health inference for different types of equipment, improving the model's accuracy and efficiency. These pruned neural network models are trained based on historical data from power grid equipment. The model's input is fault features extracted from edge intelligent units, and the output is the corresponding health value. Neural networks optimized through pruning techniques are more computationally efficient and can make accurate health assessments with limited resources.

[0034] Multiple sets of multimodal fault features are input into multiple pruned neural network inference models. These features serve as input data for the neural networks, triggering them to perform inference calculations and output the real-time health status of each key power grid component. The health status reflects the working status of the power grid component, such as whether there is a fault risk, whether the equipment is normal, and whether maintenance is required.

[0035] Edge intelligent units transmit the calculated real-time health data to a centralized analysis platform via power fiber optic cables. Power fiber optic cables are used for high-speed data transmission in large-scale power grid systems, ensuring real-time and stable data transmission to the remote line loss restoration analysis platform. The line loss restoration analysis platform is a centralized management and analysis system used to integrate and analyze data such as power grid health, fault prediction, and line loss.

[0036] Furthermore, based on the projection of the multiple real-time health values ​​of the multiple key power grid components onto multiple key power grid nodes in the dynamic power grid topology, fault propagation impact prediction is performed, and a power grid fault probability topology is output. The method includes: A310: Map the multiple real-time health values ​​to multiple node-level fault probability values; A320: Based on the real-time connection relationship of the dynamic power grid topology, establish a fault cascading propagation model and a cascaded diffusion path model; A330: Using the multiple node-level fault probability values ​​as the initial probability distribution, perform discrete probability propagation driven by protection failure in the fault cascading propagation model, and perform continuous probability diffusion driven by power flow transfer in the cascaded diffusion path model, iteratively update the fault probability distribution along the fault propagation path, and generate a probability propagation map; A340: Superimpose the probability propagation map onto the dynamic power grid topology to output the power grid fault probability topology.

[0037] Using linear or nonlinear functions for mapping, real-time health is transformed into a probability value of failure occurrence, resulting in multiple node-level failure probability values, which represent the likelihood of a node failing.

[0038] The interconnected relationships in a dynamic power grid topology define how current, power, and other quantities flow within the grid, thus directly influencing fault propagation paths. The cascading fault propagation model describes the transmission of a fault from one node to others, particularly when equipment protection systems fail to respond promptly. In such cases, the fault may bypass some protection nodes and continue propagating. If the grid's protection system fails to disconnect the faulty node in time, the fault may continue to affect other equipment, whether near or far from the fault. In this model, the failure of the grid's protection system to operate is the driving factor for fault propagation. The cascading diffusion path model describes how a fault in one node affects other nodes and ultimately triggers a cascading effect. This model considers power flow transfer between nodes in the grid and the cascading effects caused by the fault. When a node fails, the power flow (such as current and voltage) in the grid is redistributed. This power flow transfer may increase the burden on some nodes, leading to an increased probability of failure in those nodes, thus creating a cascading fault effect.

[0039] The obtained multiple node-level fault probability values ​​are used as the initial probability distribution for fault propagation. In the fault cascading propagation model, discrete probability propagation is performed along the connection relationship of the power grid topology. That is, starting from the node where the fault occurs, the fault probability of that node is propagated to adjacent nodes according to the fault propagation path. If the protection of a node fails to operate, causing the fault to propagate to other nodes, then the fault probability values ​​of these nodes will increase. In the cascading diffusion path model, the impact of the fault leads to the transfer of power flow in the power grid. These transfers affect the health status of other nodes. Therefore, continuous probability diffusion plays a role in this model. Through this process, the fault probability value of a node changes gradually with the power flow transfer in the power grid. The probability diffusion is iteratively updated along the fault path in the power grid. Each iteration updates the fault probability of the node until an equilibrium state is reached or a certain termination condition is met. Through iterative updates, the generated probability propagation map shows the process of fault propagation from the initial node to other nodes. Each node in the map corresponds to a fault probability value, which reflects the fault risk of each node in the power grid.

[0040] The generated probability propagation map is superimposed on the dynamic power grid topology. The superimposed power grid fault probability topology is a comprehensive map that shows the fault probability of each node in the power grid and their mutual influence.

[0041] Furthermore, based on the power grid fault probability topology, line loss factor mapping is performed to generate multiple simulated operation scenarios. The method includes: A410: Perform node-level line loss nonlinear mapping on the power grid fault probability topology to obtain the real-time power grid topology; A420: Predefine multiple load level modes and perform orthogonal experimental design processing on the real-time power grid topology to generate multiple initial simulation scenarios; A430: Derive multiple line loss increment parameters based on the multiple node-level fault probability values; A440: Inject the multiple line loss increment parameters into the multiple initial simulation scenarios to output the multiple simulated operation scenarios.

[0042] The failure probability of each power grid node directly affects its operating efficiency and power flow. Nodes with higher failure probabilities may experience greater line losses. Therefore, the failure probability value is converted into a node-level line loss factor using a nonlinear mapping method. Line losses in a power grid are nonlinear, meaning that changes in current and voltage are not simply linearly related to power loss, especially when a grid fault occurs. Nonlinear functions such as exponential, logarithmic, and piecewise functions can be used to convert the failure probability value to obtain the line loss factor for each node. The calculated node-level line loss factors are then mapped back to the power grid topology to obtain the real-time power grid topology.

[0043] The load level of a power grid determines the distribution of power flow and potential line losses. Load level patterns reflect the operating state of the power grid under different load conditions, such as high load, low load, and normal load. Orthogonal experimental design is a statistical method that optimizes the experimental process and results by rationally selecting experimental conditions. In this step, orthogonal design is used to systematically generate power grid simulation scenarios under various load level patterns. Based on different combinations of load levels and the real-time state of the power grid topology, multiple load level patterns are designed for experiments to generate corresponding simulation scenarios. These simulation scenarios can cover different load states, fault conditions, and changes in power grid topology.

[0044] Based on the fault probability value of each node, the line loss increment parameters of the power grid are derived. These increment parameters represent the additional line loss added to each node or area of ​​the power grid when a fault occurs. When a node fails, the power flow of the power grid will change, which may lead to an increase in the load of other nodes or a failure of power flow, thereby increasing the line loss. These line loss increments are closely related to factors such as the health of the faulty node, topology, and load level.

[0045] Multiple incremental line loss parameters are injected into different initial simulation scenarios. These parameters provide additional load effects to these scenarios. After injection, the grid operation status of the simulation scenarios will reflect the incremental changes in line losses in the grid when a fault occurs. These scenarios can simulate the specific impact of factors such as fault propagation, load fluctuations, and power flow on line losses. Ultimately, these simulation scenarios form multiple simulation operation scenarios, each reflecting the operating status of the grid under different conditions and providing corresponding line loss data.

[0046] Furthermore, based on the real-time operation data stream of the power grid in the power supply area, topology modeling is performed to reconstruct the dynamic power grid topology. The method includes: A110: Construct an initial topology diagram based on the physical connection relationship and electrical connection logic of the equipment in the power supply area, wherein the electrical connection points in the initial topology diagram are topology nodes and the topology connection elements are topology edges; A120: After collecting the switch change event signal of the SCADA system, dynamically update the node connectivity status of the initial topology diagram based on the event-driven mechanism, and restore the dynamic power grid topology with timestamps.

[0047] Physical connections determine how electricity is transmitted within the system, requiring an understanding of the physical layout of these devices. Electrical connection logic describes how electrical signals such as current and voltage are transmitted between grid devices. It ensures that the electrical connections between devices meet the requirements of power transmission and enable stable grid operation. An initial topology diagram is constructed, presenting the physical connection structure of the grid. Based on the electrical connections and device layout, each device is considered a node in the topology, representing an electrical connection point, such as a substation, circuit breaker, or transformer. Connecting elements in the grid are considered topology edges, representing the connection relationships between devices and defining the power flow paths between grid devices.

[0048] SCADA systems monitor and control various devices in the power grid in real time, including switches, circuit breakers, and transformers. When the state of these devices changes, such as switch changes or circuit breaker trips, the SCADA system records these events and generates corresponding switch change event signals. Based on these event signals, an event-driven mechanism is used to update the power grid topology in real time. This mechanism dynamically updates the node connection status in the topology based on actual events, such as switch changes and circuit breaker operations. For example, when a switch trips or equipment is under maintenance, the connectivity of the corresponding node changes. This change is reflected in the topology in real time through event signals, thus forming a dynamic power grid topology. This timestamped dynamic power grid topology map can accurately record the evolution of the power grid state, supporting subsequent backtracking and fault analysis of the power grid state.

[0049] Furthermore, the method also includes: A200-1: Traverse the initial topology graph and count the effective in-degree and out-degree of multiple electrical connection points in the active power transmission direction; A200-2: Perform fault simulation on the initial topology graph and output the multiple load loss ratios caused by the failure of the multiple electrical connection points; A200-3: Perform line loss sensitivity quantification on the multiple electrical connection points and output multiple node line loss sensitivity coefficients; A200-4: Retrieve multiple asset value weights based on the multiple associated device IDs of the multiple electrical connection points; A200-5: Combine the multiple effective in-degree and out-degree counts, multiple load loss ratios, multiple node line loss sensitivity coefficients, and multiple asset value weights to output multiple key comprehensive scores; A200-6: Traverse the multiple key comprehensive scores using a preset score threshold to screen and determine the multiple key power grid components.

[0050] In a power grid topology, in-degree and out-degree refer to the number of times electricity flows into and out of each node. Specifically, the out-degree of a node indicates how many lines output electricity from that node, while the in-degree indicates how many lines input electricity into that node. The direction of active power transmission indicates the direction of electricity flow, that is, the process of electricity being transferred from the generator or power source to the load. For power grid nodes, the in-degree and out-degree counts reflect the direction and intensity of electricity flow.

[0051] In the initial topology diagram, fault simulation means intentionally disconnecting certain electrical connection points and simulating the behavior of the power grid when these components fail. The purpose of fault simulation is to predict the impact of faults on the power grid, particularly how they affect load distribution and power transmission. Simulations include various types of faults such as line disconnection, equipment damage, and overload. Load loss ratio refers to the degree of power load loss caused by the failure of a particular electrical connection point or device. For example, when a substation fails, it affects the power supply to multiple connected areas, causing load losses in those areas. The load loss ratio is the ratio of the load in the affected area to the normal load. By calculating the load loss ratio when multiple electrical connection points fail, the stability and disturbance resilience of the power grid during faults can be assessed.

[0052] Line loss sensitivity refers to the degree to which each node in a power grid is sensitive to line losses; that is, how the line loss of a node changes when the power flow in the grid changes. Quantifying the line loss sensitivity of each electrical connection point in the power grid means analyzing the load response of each node during grid operation and how power loss is affected when a fault or load change occurs at that node. Nodes with higher sensitivity will experience a significant increase in line loss when power flow changes; conversely, nodes with lower sensitivity will experience smaller changes in line loss. Line loss changes at each node under different fault scenarios can be simulated by establishing a power flow model of the power grid. For example, based on changes in current and voltage during a fault, the relationship between the increase in line loss at each node and load changes can be calculated. The line loss sensitivity coefficient of each node represents the impact of the node's line loss change on the overall operation of the power grid. The higher the sensitivity coefficient, the greater the impact of the node on the grid's line losses; typically, these nodes are more sensitive to fault responses.

[0053] Each electrical connection point has multiple associated device IDs, such as transformers, switches, and distribution lines. These device IDs help identify the type of equipment associated with each node and its importance in the power grid. Asset value refers to the relative importance of each critical component in the power grid, and is comprehensively assessed based on factors such as the load the component bears in the grid, equipment maintenance costs, and replacement costs. Each electrical connection point can retrieve the asset value weight of related equipment through its associated device IDs. Nodes with high asset values ​​play a crucial role in power grid operation and have a more severe impact on the grid during failures.

[0054] The obtained effective in-degree counts, load loss ratios, node line loss sensitivity coefficients, and asset value weights are integrated to obtain a key comprehensive score for each node. For example, the above four indicators are weighted to obtain a comprehensive score for each electrical connection point. The weighting coefficients can be set according to actual needs, and indicators with larger weights have a greater impact on the final score.

[0055] Based on the power grid's operational requirements and safety objectives, a score threshold is set to screen critical components within the power grid. The comprehensive score of each electrical connection point is iterated and compared with the preset score threshold; nodes with scores exceeding the threshold are considered critical power grid components.

[0056] Furthermore, the method involves calculating the hierarchical incremental line loss responsibility distribution for the multiple simulated operating scenarios, and reconstructing and outputting multiple line loss responsibility distribution maps. A510: Based on the hierarchical distribution of line loss responsibility, pre-construct line loss responsibility allocation models at the component level, node level, and topology level; A520: Using the multiple simulated operating scenarios as input sources, perform incremental responsibility distribution layer-by-layer tracking in the line loss responsibility allocation models at the component level, node level, and topology level, and output multiple levels of responsibility distribution data; A530: Map the multiple levels of responsibility distribution data to the dynamic power grid topology to generate the multiple line loss responsibility distribution maps.

[0057] Line loss responsibility stratification refers to the hierarchical management of responsibilities at different levels in a power grid to more accurately assess and allocate line losses. In a power grid, line losses are not only closely related to the load but also involve components, nodes, and topology at different levels. Component-level line loss responsibility allocation models analyze the contribution of each component to overall line loss and assess the line loss responsibility caused by component failures. For example, when a transformer fails, the load in other parts of the grid may need to be shifted, thus affecting line losses. The electrical states of each node, such as current and voltage, directly affect the overall line loss of the power grid. Node-level line loss responsibility allocation models assess the impact of node-level faults or load changes on other nodes and their operation. The overall topology, connection methods, and changes in the power grid affect the power transmission efficiency. Topology-level line loss responsibility allocation models comprehensively analyze the relationships between nodes and components to analyze the overall line loss responsibility under different topology conditions.

[0058] Incremental responsibility distribution refers to the additional line loss responsibility borne by nodes or components during power grid operation or fault simulation. For example, after a component fails, the burden on other parts of the power grid increases, leading to incremental line losses. In each simulation scenario, responsibility is allocated layer by layer according to the component layer, node layer, and topology layer models to track changes in responsibility at different levels. The output multi-level responsibility distribution data describes the line loss responsibility borne by each component, node, and the entire topology under different conditions.

[0059] Multiple levels of responsibility distribution data are mapped onto a dynamic power grid topology. This dynamic topology, generated based on real-time changes in the power grid state, reflects the operational status of various components and nodes within the grid. Through data mapping, the line loss responsibility of each component, node, and topology layer can be visually displayed in a graphical interface. The generated multiple line loss responsibility distribution maps illustrate the distribution of line losses in the power grid, including which nodes, components, or regions bear more line loss responsibility, which regions have lower power transmission efficiency, or which regions are prone to significant line losses under fault or high-load conditions.

[0060] Example 2 is based on the same inventive concept as the power grid line loss restoration method based on fault prediction in the previous examples, such as... Figure 2 As shown in the figure, this application provides a power grid line loss restoration system based on fault prediction, the system comprising: The topology modeling module 10 is used to perform topology modeling based on the real-time operation data stream of the power grid in the power supply area, and to restore the dynamic power grid topology. The status monitoring module 20 is used to integrate multi-source node monitoring data, monitor the status of multiple key power grid components, and obtain multiple real-time health values. The fault propagation impact prediction module 30 is used to project the multiple real-time health values ​​onto multiple key power grid nodes in the dynamic power grid topology based on the multiple key power grid components, perform fault propagation impact prediction, and output the power grid fault probability topology. The line loss factor mapping module 40 is used to perform line loss factor mapping based on the power grid fault probability topology to generate multiple simulated operation scenarios. The responsibility distribution calculation module 50 is used to perform hierarchical line loss incremental responsibility distribution calculation on the multiple simulated operation scenarios, and restore and output multiple line loss responsibility distribution maps. The intervention strategy output module 60 is used to fuse the multiple line loss responsibility distribution maps and output a loss reduction intervention strategy. The iterative optimization module 70 is used to perform closed-loop tracking and iterative optimization of line loss spatiotemporal correlation data after the active regulation of the power supply area is carried out using the loss reduction intervention strategy.

[0061] Furthermore, the status monitoring module 20 is used to perform the following operation steps: Multiple edge intelligent units are pre-deployed in multiple key power grid components. After the multiple edge intelligent units access the original monitoring data of multiple nodes of the multiple key power grid components, they perform fault correlation feature extraction to obtain multiple sets of multimodal fault features. Multiple pruned neural network inference models are pre-constructed according to multiple component types of the multiple key power grid components. The multiple sets of multimodal fault features are input into the multiple pruned neural network inference models to perform health inference and output the multiple real-time health values. The multiple edge intelligent units transmit the multiple real-time health values ​​to the line loss restoration analysis platform through power fiber optic cables.

[0062] Furthermore, the fault propagation impact prediction module 30 is used to perform the following operation steps: The multiple real-time health values ​​are mapped to multiple node-level fault probability values; based on the real-time connection relationship of the dynamic power grid topology, a fault cascading propagation model and a cascaded diffusion path model are established; using the multiple node-level fault probability values ​​as the initial probability distribution, discrete probability propagation driven by protection failure is performed in the fault cascading propagation model, and continuous probability diffusion driven by power flow transfer is performed in the cascaded diffusion path model. The fault probability distribution is iteratively updated along the fault propagation path to generate a probability propagation map; the probability propagation map is superimposed on the dynamic power grid topology to output the power grid fault probability topology.

[0063] Furthermore, the line loss factor mapping module 40 is used to perform the following operation steps: The power grid fault probability topology is nonlinearly mapped at the node level to obtain the real-time power grid topology; multiple load level modes are predefined to perform orthogonal experimental design processing on the real-time power grid topology to generate multiple initial simulation scenarios; multiple line loss increment parameters are derived based on the multiple node-level fault probability values; the multiple line loss increment parameters are injected into the multiple initial simulation scenarios to output the multiple simulated operation scenarios.

[0064] Furthermore, the topology modeling module 10 is used to perform the following operation steps: Based on the physical connection relationship and electrical connection logic of the equipment in the power supply area, an initial topology diagram is constructed, wherein the electrical connection points in the initial topology diagram are topology nodes and the topology connection elements are topology edges; after collecting the switch change event signals of the SCADA system, the node connectivity status of the initial topology diagram is dynamically updated based on the event-driven mechanism to restore the dynamic power grid topology with timestamps.

[0065] Furthermore, the status monitoring module 20 is used to perform the following operation steps: The initial topology is traversed, and multiple effective in-degree and out-degree counts are performed for multiple electrical connection points in the active power transmission direction. Fault simulation is performed on the initial topology, and multiple load loss ratios caused by the failure of the multiple electrical connection points are output. Line loss sensitivity quantification is performed on the multiple electrical connection points, and multiple node line loss sensitivity coefficients are output. Multiple asset value weights are retrieved based on multiple associated device IDs of the multiple electrical connection points. The multiple effective in-degree and out-degree counts, multiple load loss ratios, multiple node line loss sensitivity coefficients, and multiple asset value weights are fused to output multiple key comprehensive scores. The multiple key comprehensive scores are traversed using a preset score threshold to screen and determine the multiple key power grid components.

[0066] Furthermore, the responsibility distribution calculation module 50 is used to perform the following operation steps: Based on the hierarchical distribution of line loss responsibility, a line loss responsibility allocation model at the component level, a line loss responsibility allocation model at the node level, and a line loss responsibility allocation model at the topology level are pre-constructed. Using the multiple simulated operating scenarios as input sources, incremental responsibility distribution is tracked layer by layer in the line loss responsibility allocation models at the component level, the node level, and the topology level, outputting multiple levels of responsibility distribution data. The multiple levels of responsibility distribution data are mapped to the dynamic power grid topology to generate multiple line loss responsibility distribution maps.

[0067] Through the foregoing detailed description of the power grid line loss restoration method based on fault prediction, those skilled in the art can clearly understand the power grid line loss restoration system based on fault prediction in this embodiment. Since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and relevant parts can be referred to the method section.

[0068] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for restoring power grid line losses based on fault prediction, characterized in that, The method includes: Topology modeling is performed based on the real-time operation data stream of the power grid in the power supply area to reconstruct the dynamic power grid topology; By integrating multi-source node monitoring data, the status of multiple key power grid components is monitored, and multiple real-time health values ​​are obtained. Based on the projection of the multiple real-time health values ​​of the multiple key power grid components onto the multiple key power grid nodes of the dynamic power grid topology, the fault propagation impact is predicted, and the power grid fault probability topology is output. Based on the power grid fault probability topology, line loss factor mapping is performed to generate multiple simulated operation scenarios; The hierarchical line loss incremental responsibility distribution is calculated for the multiple simulated operation scenarios, and multiple line loss responsibility distribution maps are reconstructed and output. By integrating the multiple line loss liability distribution maps, a loss reduction intervention strategy is output. After actively regulating the power supply area using the aforementioned loss reduction intervention strategy, closed-loop tracking and iterative optimization of the spatiotemporal correlation data of line loss are performed.

2. The power grid line loss restoration method based on fault prediction as described in claim 1, characterized in that, The method integrates multi-source node monitoring data to monitor the status of multiple key power grid components and obtain multiple real-time health indicators. Multiple edge intelligent units are pre-deployed in the aforementioned key power grid components; After the multiple edge intelligent units access the original monitoring data of multiple nodes of the multiple key power grid components, they perform fault correlation feature extraction to obtain multiple sets of multimodal fault features. Multiple pruned neural network inference models are pre-constructed based on the multiple component types of the multiple key power grid components; The multiple sets of multimodal fault features are input into the multiple pruned neural network inference models to perform health inference, and the multiple real-time health values ​​are output. The multiple edge intelligent units transmit the multiple real-time health values ​​to the line loss restoration analysis platform via power fiber optic cables.

3. The power grid line loss restoration method based on fault prediction as described in claim 1, characterized in that, Based on the projection of the multiple real-time health values ​​of the multiple key power grid components onto the multiple key power grid nodes of the dynamic power grid topology, the method performs fault propagation impact prediction and outputs the power grid fault probability topology. The multiple real-time health values ​​are mapped to multiple node-level failure probability values; Based on the real-time connection relationship of dynamic power grid topology, a fault cascading propagation model and a cascading diffusion path model are established. Using the multiple node-level fault probability values ​​as the initial probability distribution, discrete probability propagation driven by protection failure is performed in the fault cascading propagation model, and continuous probability diffusion driven by power flow transfer is performed in the cascaded diffusion path model. The fault probability distribution is iteratively updated along the fault propagation path to generate a probability propagation map. The probability propagation map is superimposed on the dynamic power grid topology to output the power grid fault probability topology.

4. The power grid line loss restoration method based on fault prediction as described in claim 3, characterized in that, Based on the power grid fault probability topology, line loss factor mapping is performed to generate multiple simulated operation scenarios. The method includes: Perform node-level line loss nonlinear mapping on the power grid fault probability topology to obtain the real-time power grid topology; Multiple load level modes are predefined for orthogonal experimental design processing of the real-time power grid topology to generate multiple initial simulation scenarios; Multiple line loss increment parameters are derived based on the multiple node-level fault probability values; The multiple line loss increment parameters are injected into the multiple initial simulation scenarios, and the multiple simulation operation scenarios are output.

5. The power grid line loss restoration method based on fault prediction as described in claim 1, characterized in that, The method involves topology modeling based on real-time operational data streams of the power grid in the power supply area to reconstruct the dynamic power grid topology. Based on the physical connection relationship and electrical connection logic of the equipment in the power supply area, an initial topology graph is constructed, wherein the electrical connection points in the initial topology graph are topology nodes, and the topology connection elements are topology edges; After acquiring the switch change event signals from the SCADA system, the node connectivity status of the initial topology is dynamically updated based on the event-driven mechanism, restoring the timestamped dynamic power grid topology.

6. The power grid line loss restoration method based on fault prediction as described in claim 5, characterized in that, The method further includes: Traverse the initial topology graph and count the effective in-degree and out-degree of multiple electrical connection points in the active power transmission direction; A fault simulation is performed on the initial topology diagram, and the proportions of multiple load losses caused by the failure of the multiple electrical connection points are output. Perform line loss sensitivity quantization on the multiple electrical connection points and output multiple node line loss sensitivity coefficients; Retrieve multiple asset value weights based on multiple associated device IDs of the multiple electrical connection points; By integrating multiple effective in-degree and out-degree counts, multiple load loss ratios, multiple node line loss sensitivity coefficients, and multiple asset value weights, multiple key comprehensive scores are output. The multiple key comprehensive scores are traversed using a preset score threshold to screen and determine the multiple key power grid components.

7. The power grid line loss restoration method based on fault prediction as described in claim 1, characterized in that, The method involves calculating the hierarchical incremental line loss responsibility distribution for the multiple simulated operation scenarios and reconstructing multiple line loss responsibility distribution maps. Based on the hierarchical line loss responsibility system, pre-construct line loss responsibility allocation models for the component layer, node layer, and topology layer; Using the multiple simulated operating scenarios as input sources, incremental responsibility distribution is tracked layer by layer in the component layer line loss responsibility allocation model, node layer line loss responsibility allocation model, and topology layer line loss responsibility allocation model, and multiple levels of responsibility distribution data are output. The multiple hierarchical responsibility distribution data are mapped to the dynamic power grid topology to generate the multiple line loss responsibility distribution maps.

8. A power grid line loss restoration system based on fault prediction, characterized in that, The system is used to implement the power grid line loss restoration method based on fault prediction as described in any one of claims 1-7, the system comprising: The topology modeling module is used to perform topology modeling based on the real-time operation data stream of the power grid in the power supply area, and to reconstruct the dynamic power grid topology. The status monitoring module is used to integrate multi-source node monitoring data, monitor the status of multiple key power grid components, and obtain multiple real-time health statuses. The fault propagation impact prediction module is used to predict the fault propagation impact based on the projection of the multiple real-time health values ​​of the multiple key power grid components onto the multiple key power grid nodes of the dynamic power grid topology, and output the power grid fault probability topology. The line loss factor mapping module is used to perform line loss factor mapping based on the power grid fault probability topology and generate multiple simulated operation scenarios. The responsibility distribution calculation module is used to perform hierarchical line loss incremental responsibility distribution calculation on the multiple simulated operation scenarios and reconstruct and output multiple line loss responsibility distribution maps. The intervention strategy output module is used to integrate the multiple line loss responsibility distribution maps and output a loss reduction intervention strategy; The iterative optimization module is used to perform closed-loop tracking and iterative optimization of line loss spatiotemporal correlation data after the active regulation of the power supply area is carried out using the loss reduction intervention strategy.

Images