A big data driven power distribution network line loss intelligent prediction system, device and medium

Abstract

The application provides a big data driven power distribution network line loss intelligent prediction system, equipment and medium, relates to line loss prediction technical field, and the system includes: data lake construction module, integrates the distribution network from the distribution network topology data, measurement system data, environmental data and equipment account data, constructs unified data lake; knowledge graph construction module, based on unified data lake, through graph calculation, the family relationship of distribution network is analyzed, and the dynamic family genetic knowledge graph is constructed; line loss analysis module, receiving multi-source monitoring data, based on dynamic family genetic knowledge graph, line loss analysis is carried out; fusion positioning module, according to the operation analysis result of theoretical line loss calculation analysis, abnormal detection diagnosis analysis, fusion positioning is carried out, and line loss prediction positioning information is fed back. Through the application, the technical problem that the line loss prediction accuracy is low in the prior art can be solved, and the technical effect that line loss anomaly can be accurately identified is achieved.

Description

Technical Field

[0001] This application relates to the field of line loss prediction technology, and in particular to a big data-driven intelligent prediction system, equipment and medium for distribution network line losses. Background Technology

[0002] With the continuous expansion of the distribution network and the large-scale integration of distributed energy, the power grid structure is characterized by a large number of nodes, complex hierarchical structure, and rapid dynamic changes in operating status, making line loss management an important indicator for measuring the operating efficiency and safety level of the distribution system.

[0003] Currently, existing methods for analyzing line losses in distribution networks largely rely on traditional energy balance calculations, segmented and zoned statistical analysis, and empirical judgment methods based on metering differences. While these methods can generally meet the needs of line loss estimation and anomaly detection in small-scale, simple power grids, their shortcomings are becoming increasingly apparent in modern distribution networks with diverse equipment types, complex topologies, and frequently changing operating conditions. Traditional line loss models generally assume fixed equipment parameters and stable topologies, making it difficult to reflect the impact of equipment aging, load fluctuations, and changes in the external environment on line losses. This leads to a continuous accumulation of theoretical line loss calculation errors over time. For example, when line temperature rises, resistance increases linearly, but static models struggle to reflect this change in real time, causing the deviation between theoretical and actual line losses to widen continuously.

[0004] In summary, existing technologies suffer from the inability to accurately reflect the coupling mechanism between changes in equipment genetic parameters and line loss performance due to the lack of a unified model that integrates physical connection relationships, electrical influence relationships, equipment attribute characteristics, and time-series correlation information. This further affects the technical problems of identifying the causes of abnormal line loss, spatial location, impact range assessment, and subsequent governance work order generation. Summary of the Invention

[0005] The purpose of this application is to provide a big data-driven intelligent prediction system, equipment, and medium for distribution network line loss, in order to solve the technical problems in the prior art where the lack of a unified model integrating physical connection relationships, electrical influence relationships, equipment attribute characteristics, and time-series correlation information leads to the inability to accurately reflect the coupling mechanism between changes in equipment genetic parameters and line loss performance, further affecting the identification of abnormal line loss causes, spatial positioning, impact range assessment, and subsequent management work order generation.

[0006] In view of the above problems, this application provides a big data-driven intelligent prediction system, equipment and medium for distribution network line loss.

[0007] Firstly, this application provides a big data-driven intelligent prediction system for distribution network line losses, comprising: a data lake construction module, used to integrate operation and distribution topology data, measurement system data, environmental data, and equipment ledger data from the distribution network to construct a unified data lake; a knowledge graph construction module, used to perform distribution network family relationship analysis through graph computing based on the unified data lake to construct a dynamic family genetic knowledge graph; a line loss analysis module, used to receive multi-source monitoring data and perform line loss analysis based on the dynamic family genetic knowledge graph, including theoretical line loss calculation analysis and anomaly detection and diagnosis analysis; and a fusion positioning module, used to perform fusion positioning based on the calculation and analysis results of the theoretical line loss calculation analysis and anomaly detection and diagnosis analysis, and to feed back line loss prediction positioning information.

[0008] Preferably, the big data-driven intelligent prediction system for distribution network line losses further includes: a multi-dimensional relationship network construction unit, used to sort out the genetic relationships of equipment based on the operation and distribution topology data, and construct a multi-dimensional relationship network containing an equipment gene library; a response relationship determination unit, used to perform data comparison and fitting based on the measurement system data, environmental data, and equipment ledger data, and determine the equipment gene response relationship; and a knowledge graph construction unit, used to perform topological relationship adaptive evolution based on the equipment gene response relationship, track the signal genetic relationship between equipment, add it to the multi-dimensional relationship network, and construct the dynamic family genetic knowledge graph.

[0009] Preferably, the big data-driven intelligent prediction system for distribution network line losses further includes: a topology graph network construction channel, used to construct a topology graph network with devices as nodes and connection relationships as edges based on the device connection relationships, electrical parameters, and device hierarchical structure in the operation and distribution topology data; and a multidimensional relationship network acquisition channel, used to establish a device gene library in the topology graph network, which stores the inherent electrical characteristic parameters of each device, including line impedance parameters, transformer no-load loss and load loss curves, meter accuracy characteristics, and device health status indicators, to obtain the multidimensional relationship network.

[0010] Preferably, the big data-driven intelligent prediction system for distribution network line losses further includes: a response relationship model establishment channel, used to perform spatiotemporal correlation analysis on equipment operation data and equipment gene parameters in the measurement system data, environmental data, and equipment ledger data through graph neural networks to establish a response relationship model between equipment electrical operation data and gene parameters; and an equipment gene response relationship determination channel, used to verify and correct the topological relationship in the multidimensional relationship network based on the response relationship model to determine the equipment gene response relationship.

[0011] Preferably, the big data-driven intelligent prediction system for distribution network line losses further includes: a multi-dimensional relationship network construction channel, used to construct a multi-dimensional relationship network including static genetic features and dynamic operating features of equipment based on the verified and corrected equipment gene response relationship; a traversal tracking channel, used to establish a signal propagation path model in the multi-dimensional relationship network, and track the propagation path and impact range of power quality anomalies and line loss anomalies in the topology network through a graph traversal algorithm; a dynamic family genetic knowledge graph acquisition channel, used to add signal genetic paths as implicit relationship edges to the multi-dimensional relationship network to obtain a dynamic family genetic knowledge graph containing both physical connections and electrical influences; and also includes: a graph reconstruction process triggering channel, used to establish a graph version management mechanism, which triggers the graph reconstruction process when a change in the power grid structure, an update of equipment parameters, or a change in operating mode is detected; and an adaptive evolution update channel, used to automatically adjust the weights of equipment gene parameters and perform adaptive evolution updates of the knowledge graph based on the graph reconstruction process by continuously comparing real-time measurement data with graph prediction values.

[0012] Preferably, the big data-driven intelligent prediction system for distribution network line losses further includes: a theoretical technical line loss rate calculation unit, used to identify physical connection relationships and electrical attribution relationships based on the dynamic family genetic knowledge graph, calculate the actual line loss rate based on the multi-source monitoring data, and calculate the corresponding theoretical technical line loss rate based on equipment gene parameters; a line loss family abnormal feature extraction unit, used to extract line loss family abnormal features by combining the mapping performance of the multi-source monitoring data in the dynamic family genetic knowledge graph with the theoretical technical line loss rate and the actual line loss rate; and an abnormal detection and diagnosis analysis result acquisition unit, used to input the line loss family abnormal features into a trained and converged AI diagnostic model to perform abnormal detection and diagnosis, and obtain abnormal detection and diagnosis analysis results.

[0013] Preferably, the big data-driven intelligent prediction system for distribution network line losses further includes: a family intrinsic anomaly feature extraction channel, used to extract family intrinsic anomaly features based on the deviation between the theoretical technical line loss rate and the actual line loss rate; a horizontal lineage anomaly feature acquisition channel, used to calculate the dispersion of the horizontal line loss rate distribution of the same-level power grid equipment based on the dynamic family genetic knowledge graph, and obtain horizontal lineage anomaly features; a vertical genetic anomaly feature extraction channel, used to perform temporal correlation analysis of line loss rate changes between upper and lower power grid levels based on the dynamic family genetic knowledge graph, and extract vertical genetic anomaly features; and a line loss family anomaly feature acquisition channel, used to combine the family intrinsic anomaly features, horizontal lineage anomaly features, and vertical genetic anomaly features to obtain the line loss family anomaly features.

[0014] Preferably, the big data-driven intelligent prediction system for distribution network line losses further includes: a probabilistic diagnostic conclusion generation unit, used to perform multi-source evidence fusion on the theoretical line loss distribution characteristics obtained from the theoretical line loss calculation and analysis, and the anomaly root cause probability obtained from the anomaly detection and diagnostic analysis, to calculate the credibility of each potential anomaly cause and generate a probabilistic diagnostic conclusion; a line loss prediction and location report output unit, used to associate and map the anomaly causes whose credibility meets the screening criteria with the equipment nodes and topological paths in the dynamic family genetic knowledge graph based on the probabilistic diagnostic conclusion, and output a line loss prediction and location report containing specific location coordinates, anomaly type confidence, and impact range; and a targeted governance work order generation unit, used to generate targeted governance work orders based on the line loss prediction and location report, wherein the targeted governance work orders include at least one or more of the following: electricity theft inspection work orders, equipment inspection work orders, and file verification work orders.

[0015] Secondly, this application also provides an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the steps of a big data-driven intelligent prediction system for distribution network line losses as described in any one of the first aspects above.

[0016] Thirdly, a computer-readable storage medium storing a computer program that, when executed, implements the steps of the big data-driven intelligent prediction system for distribution network line losses as described in any one of the first aspects.

[0017] The technical solution provided in this application has at least the following technical effects or advantages: by achieving the technical goal of unified modeling and dynamic correlation analysis of the genetic characteristics, topology, electrical influence paths and operating status of equipment in the distribution network, it can accurately identify the causes of abnormal line losses, accurately locate abnormal equipment and its affected areas, and automatically generate targeted treatment work orders to improve the efficiency and intelligence level of line loss treatment.

[0018] The above description is merely an overview of the technical solution of this application. To enable a clearer understanding of the technical means of this application and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, specific embodiments of this application are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent through the following description. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the structure of a big data-driven intelligent prediction system for distribution network line losses according to this application.

[0021] Figure 2 This is a schematic diagram of the knowledge graph construction module in a big data-driven intelligent prediction system for distribution network line losses according to this application.

[0022] Figure 3 This is a schematic diagram of the structure of an exemplary electronic device of this application.

[0023] Figure labeling: Data lake construction module 1, knowledge graph construction module 2, line loss analysis module 3, fusion positioning module 4, bus 300, receiver 301, processor 302, transmitter 303, memory 304, bus interface 305. Detailed Implementation

[0024] This application provides a big data-driven intelligent prediction system, equipment, and medium for distribution network line losses. It addresses the technical problem in existing technologies where the lack of a unified model integrating physical connections, electrical influence relationships, equipment attribute characteristics, and temporal correlation information leads to an inability to accurately reflect the coupling mechanism between changes in equipment genetic parameters and line loss performance. This further affects the identification of line loss anomaly causes, spatial location, impact range assessment, and subsequent management work order generation. The system achieves the technical goal of unified modeling and dynamic correlation analysis of equipment genetic characteristics, topology, electrical influence paths, and operating status in the distribution network. This results in the ability to accurately identify the causes of line loss anomalies, accurately locate abnormal equipment and their affected areas, and automatically generate targeted management work orders, thereby improving the efficiency and intelligence level of line loss management.

[0025] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. It should be understood that this application is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. It should also be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all of them.

[0026] Example 1, please refer to Figure 1 and Figure 2 This application provides a big data-driven intelligent prediction system for distribution network line losses, specifically including:

[0027] Data Lake Construction Module 1 is used to integrate operation and distribution topology data, measurement system data, environmental data, and equipment ledger data from the distribution network to build a unified data lake.

[0028] Specifically, integrating distribution network topology data, measurement system data, environmental data, and equipment ledger data refers to aggregating topological information reflecting the connection relationships, power supply paths, and network hierarchy among electrical equipment at all levels within the distribution network. This allows for a comprehensive identification of the physical connection structure from the main lines to end users. Measurement system data refers to operating status quantities such as current, voltage, active power, reactive power, and power factor collected by distribution automation terminals, smart meters, and online monitoring devices. Integrating this type of data reflects the real-time electrical characteristics of the distribution network at different operating stages. Environmental data refers to external environmental factors affecting the operating performance of the distribution network, including information on temperature, humidity, wind speed, rainfall, and seasonal load changes. This data is used to explain the regularity of line losses changing with external conditions. Equipment ledger data refers to basic attribute data of various equipment recorded in the distribution operation and maintenance system, including equipment model, rated capacity, commissioning time, maintenance records, and health status assessments. This data is used to understand the inherent characteristics and historical changes in the equipment's condition. By unifying and integrating operational topology data, measurement system data, environmental data, and equipment ledger data, a unified data lake that can store both structured and unstructured data can be constructed to meet the needs of subsequent multi-source data fusion analysis.

[0029] Knowledge graph construction module 2 is used to analyze the family relationships of the power distribution network through graph computing based on the unified data lake, and to construct a dynamic family genetic knowledge graph.

[0030] Furthermore, this application also includes: a multidimensional relationship network construction unit, used to sort out the genetic relationships of equipment based on the breeding topology data, and construct a multidimensional relationship network containing an equipment gene library; a response relationship determination unit, used to perform data comparison and fitting based on the measurement system data, environmental data, and equipment ledger data, and determine the genetic response relationship of equipment genes; and a knowledge graph construction unit, used to perform adaptive evolution of topological relationships based on the genetic response relationship of equipment, track the signal genetic relationship between equipment, add it to the multidimensional relationship network, and construct the dynamic family genetic knowledge graph.

[0031] Furthermore, this application also includes: a topology graph network construction channel, used to construct a topology graph network with devices as nodes and connection relationships as edges based on device connection relationships, electrical parameters, and device hierarchical structure in the operation and maintenance topology data; and a multidimensional relationship network acquisition channel, used to establish a device gene library in the topology graph network, wherein the inherent electrical characteristic parameters of each device are stored, including line impedance parameters, transformer no-load loss and load loss curves, meter accuracy characteristics, and device health status indicators, to obtain the multidimensional relationship network.

[0032] Furthermore, this application also includes: a response relationship model establishment channel, used to perform spatiotemporal correlation analysis on equipment operation data and equipment gene parameters in the measurement system data, environmental data and equipment ledger data through graph neural networks, and establish a response relationship model between equipment electrical operation data and gene parameters; and an equipment gene response relationship determination channel, used to verify and correct the topological relationship in the multidimensional relationship network based on the response relationship model, and determine the equipment gene response relationship.

[0033] Furthermore, this application also includes: a multi-dimensional relationship network construction channel, used to construct a multi-dimensional relationship network including static genetic features and dynamic operating features of the equipment based on the verified and corrected equipment gene response relationship; a traversal tracking channel, used to establish a signal propagation path model in the multi-dimensional relationship network, and track the propagation path and impact range of power quality anomalies and line loss anomalies in the topology network through a graph traversal algorithm; a dynamic family genetic knowledge graph acquisition channel, used to add the signal genetic path as a recessive relationship edge to the multi-dimensional relationship network to obtain a dynamic family genetic knowledge graph containing both physical connection and electrical influence relationships; and also includes: a graph reconstruction process triggering channel, used to establish a graph version management mechanism, which triggers the graph reconstruction process when a change in power grid structure, equipment parameter update, or change in operating mode is detected; and an adaptive evolution update channel, used to automatically adjust the weight of equipment gene parameters and perform adaptive evolution update of the knowledge graph based on the graph reconstruction process by continuously comparing real-time measurement data and graph prediction values.

[0034] Specifically, based on the equipment connection relationships, electrical parameters, and equipment hierarchy in the distribution network topology data, that is, by utilizing the physical connection methods between different devices in the distribution network, including the electrical connection links from the upstream substation to the distribution transformer, from the distribution transformer to the branch line, and from the branch line to the user-side metering device, the wire connections, contact relationships, and power supply paths between each device are identified to reflect the transmission direction and structural relationship of electrical energy in the network. Using a set of parameters characterizing the physical characteristics of equipment operation, including the resistance, reactance, and capacitance parameters of the line, the rated capacity, impedance voltage, and load characteristic curves of the transformer, the electrical coupling strength between devices is described. Through the hierarchical relationship of the equipment within the distribution system from the backbone network to the end user, such as the main substation being at the highest level, the distribution transformer being at the middle level, and the branch line and user meter being at the lower level, the status and lineage of each device in the network can be clearly identified by identifying the hierarchical structure. By integrating device connection relationships, electrical parameters, and device hierarchical structure, a topological graph network can be constructed with each device as a graph node and the physical connection relationships between devices as graph edges, enabling the network structure to be directly expressed and calculated in the graph model.

[0035] Establishing a device gene library in a topological network refers to attaching inherent electrical characteristics to each device node to describe its static essential attributes. Line impedance parameters refer to the resistance and reactance of a line, used to characterize the inertia and energy loss characteristics of electrical energy during transmission. Transformer no-load and load loss curves represent the transformer's loss levels under no-load and different load conditions, respectively, and are important bases for calculating distribution network line losses. Meter accuracy characteristics refer to the error range of metering equipment under different voltage and current conditions, reflecting the true reliability of energy metering. Equipment health status indicators are quantitative indicators reflecting the degree of equipment aging, operational risks, and failure probabilities, such as equipment operating years, insulation status, or temperature rise levels. By storing multiple types of parameters in the device gene library, a multi-dimensional relationship network containing electrical characteristics, metering features, and health dimensions can be formed, providing a structured data foundation for subsequent calculations of equipment behavior relationships and analysis of line loss formation mechanisms.

[0036] Furthermore, spatiotemporal correlation analysis of equipment operation data and equipment genetic parameters in measurement system data, environmental data, and equipment ledger data using graph neural networks refers to using a neural network model capable of processing graph-structured data. Under the constraint of equipment topology, measurement system data such as voltage, current, power, and power factor are used as time series inputs, combined with environmental data such as temperature, humidity, and seasonal load, as well as ledger data such as equipment rated capacity and historical maintenance records, to jointly model the changes in electrical characteristics of equipment under different times, operating environments, and health states. Spatiotemporal correlation analysis refers to using continuous measurement data to reflect the dynamic changes in equipment operating status over time in the time dimension, and using a topology graph to represent the mutual influence relationships between equipment in the spatial dimension. Through the multi-layer aggregation mechanism of graph neural networks, the electrical response coupling between adjacent nodes can be fully learned, thereby establishing a mapping relationship between equipment electrical operation data and its genetic parameters.

[0037] Verifying and correcting topological relationships in a multidimensional network based on a response relationship model involves using the device characteristic coupling relationships obtained from a graph neural network to compare the consistency of device connections and characteristics within the multidimensional network. By analyzing the correlation, trend consistency, and abnormal deviation of device operating data, areas with potential data anomalies, device file errors, or inconsistent topological records are corrected. For example, if the topology map records that two branches are physically connected, but their operating current curves show no significant correlation over multiple cycles, it can be determined that the actual power supply path may have changed, requiring correction of the topology path. Similarly, if the actual loss curve of a transformer changes with load deviates by more than 20% from the load loss curve recorded in its ledger, it can be determined that its genetic parameters are aging or recorded incorrectly, requiring adjustment of the corresponding attributes. Through this correction process, the device genetic response relationships can be ultimately determined, enabling the multidimensional network to reflect the real, dynamic, and up-to-date response characteristics between devices that change with operating conditions.

[0038] Furthermore, based on the verified and corrected device gene response relationships, a multidimensional relationship network is constructed, including both static gene features and dynamic operational features of the devices. This means that after graph neural network comparison and topology correction, the inherent parameters of the devices are used as static gene features, including device impedance, transformer load loss curves, metering accuracy, and device health status. Simultaneously, the changes in operating voltage, current, and power of the devices at different time periods are used as dynamic operational features. By mapping both static and dynamic information onto a graph structure, the multidimensional relationship network can simultaneously characterize the inherent attributes and real-time operational behavior of the devices. Static gene features are used to depict the fundamental differences in the physical structure and electrical properties of the devices, while dynamic operational features are used to reflect the response patterns of the devices under different operating conditions, enabling the multidimensional relationship network to possess a more comprehensive descriptive capability.

[0039] Establishing a signal propagation path model in a multidimensional network refers to simulating the spread of power quality anomalies or line loss anomalies in the network by constructing a model describing the propagation law of electrical quantity changes in the network based on the topological relationships between nodes and edges in a graph structure. Graph traversal algorithms are algorithms that traverse nodes in a graph structure, including depth-first search and breadth-first search, used to trace the path of abnormal signals spreading from the source to other nodes. For example, when a voltage drop occurs in a branch, the graph traversal algorithm will identify the range of the signal propagation along the upper bus and adjacent branches, and determine the set of nodes affected by the anomaly by analyzing the propagation path and propagation amplitude.

[0040] Adding signal genetic paths as recessive edges to a multidimensional knowledge graph refers to adding logical relationships formed by the propagation and diffusion of anomalies during actual operation as recessive edges, in addition to the existing physical connection edges. This creates a two-layered structure that includes both physical connections and electrical influence relationships. Recessive edges are used to characterize relationships between devices that are not directly connected by wires but have significant correlations when their electrical characteristics change. For example, when two unconnected line branches exhibit synchronous voltage fluctuations due to shared load fluctuations from an upstream transformer, an electrical influence edge can be established between them, ultimately forming a dynamic family genetic knowledge graph.

[0041] Establishing a knowledge graph version management mechanism refers to setting up a module for versioned management of the knowledge graph. This module records the graph structure after each topology change, parameter correction, and implicit relationship update, and automatically triggers a graph reconstruction process when a change in power grid structure, equipment parameter update, or change in operating mode is detected. The graph reconstruction process involves regenerating the graph structure based on the latest topology data, measurement data, and equipment characteristic data after a version update is triggered, thereby ensuring that the graph remains consistent with the actual operating state.

[0042] Based on the knowledge graph reconstruction process, the device gene parameter weights are automatically adjusted by continuously comparing real-time measurement data with the knowledge graph prediction values. This means using real-time data to correct the parameters in the knowledge graph, and dynamically updating the device gene weights by analyzing the deviation between the predicted and actual values, so that the knowledge graph can adaptively evolve and update.

[0043] Line loss analysis module 3 is used to receive multi-source monitoring data and perform line loss analysis based on the dynamic family genetic knowledge graph, including theoretical line loss calculation analysis and abnormal detection and diagnosis analysis.

[0044] Furthermore, this application also includes: a theoretical technical line loss rate calculation unit, used to identify physical connection relationships and electrical attribution relationships based on the dynamic family genetic knowledge graph, calculate the actual line loss rate based on the multi-source monitoring data, and calculate the corresponding theoretical technical line loss rate based on the equipment gene parameters; a line loss family abnormal feature extraction unit, used to extract line loss family abnormal features by combining the mapping representation of the multi-source monitoring data in the dynamic family genetic knowledge graph with the theoretical technical line loss rate and the actual line loss rate; and an abnormal detection and diagnosis analysis result acquisition unit, used to input the line loss family abnormal features into a trained and converged AI diagnostic model to perform abnormal detection and diagnosis, and obtain abnormal detection and diagnosis analysis results.

[0045] Furthermore, this application also includes: a family intrinsic abnormality feature extraction channel, used to extract family intrinsic abnormality features based on the deviation between the theoretical technical line loss rate and the actual line loss rate; a horizontal kinship abnormality feature acquisition channel, used to calculate the dispersion of the horizontal line loss rate distribution of the same-level power grid equipment based on the dynamic family genetic knowledge graph, and obtain horizontal kinship abnormality features; a vertical genetic abnormality feature extraction channel, used to perform temporal correlation analysis of line loss rate changes between upper and lower power grid levels based on the dynamic family genetic knowledge graph, and extract vertical genetic abnormality features; and a line loss family abnormality feature acquisition channel, used to combine the family intrinsic abnormality features, horizontal kinship abnormality features, and vertical genetic abnormality features to obtain the line loss family abnormality features.

[0046] Specifically, based on a dynamic family genetic knowledge graph, identifying physical connections and electrical attribution relationships involves using a two-layer graph structure containing physical connection edges and electrical influence edges to identify the actual conductor connections between devices and the electrical associations formed by load transfer, voltage coupling, and power quality changes, thereby clarifying the physical location and electrical scope of each node in the network. Calculating the actual line loss rate based on multi-source monitoring data involves using energy metering data, current and voltage measurement data, and power data to calculate the actual energy difference between the start and end points of the line, and dividing it by the input energy to obtain the actual line loss ratio, reflecting the true loss level under line operating conditions. Furthermore, calculating the corresponding theoretical technical line loss rate based on device genetic parameters involves using inherent device parameters such as line impedance, conductor length, transformer load loss curves, and no-load loss parameters, calculating the theoretically expected technical loss value through electrical formulas or simulation models, and then comparing this with the device input energy to obtain the theoretical line loss rate, thus obtaining an ideal line loss reference based on static device characteristics.

[0047] Furthermore, based on the deviation between the theoretical technical line loss rate and the actual line loss rate, intrinsic anomalous features of the family are extracted. The difference between the theoretical technical line loss rate and the actual line loss rate is used to determine the inherent anomalous features within the equipment family. The theoretical technical line loss rate refers to the theoretical energy loss ratio calculated by a physical model based on the equipment's genetic parameters, while the actual line loss rate refers to the real-time loss ratio formed by multi-source monitoring data under actual operating conditions. The deviation is used to characterize the magnitude of the difference between the two and serves as a metric for measuring whether the equipment exhibits abnormal behavior that is mismatched with its inherent performance. This forms the intrinsic anomalous features of the family, reflecting the inherent health deviations of the equipment within the same structure.

[0048] Furthermore, based on a dynamic family genetic knowledge graph, the dispersion of the lateral line loss rate distribution of grid equipment at the same level is calculated to obtain lateral lineage abnormality characteristics. This demonstrates that by utilizing the clearly defined lateral structural relationships of grid equipment in the dynamic family genetic knowledge graph, the dispersion of the line loss rate distribution of multiple grid equipment belonging to the same level can be calculated. The dispersion is used to reflect the performance consistency under the same origin relationship. Grid equipment at the same level refers to multiple parallel sub-levels under the same upper-level control, such as different branches of the same transformer area or multiple end users under the same branch. When the dispersion exceeds a preset reasonable range, lateral lineage abnormality characteristics can be identified, indicating that the equipment has deviated significantly from its lateral homologous objects.

[0049] Furthermore, based on a dynamic family genetic knowledge graph, a temporal correlation analysis of line loss rate changes between upper and lower power grid levels is conducted to extract longitudinal genetic anomalies. By analyzing the hierarchical relationships of the power grid structure, the association paths between different levels are retrieved in the dynamic family genetic knowledge graph. By temporally matching the line loss rate change sequences of upper and lower power grid levels, the correlation and transmission delay between the two are calculated. The upper and lower power grid levels refer to the hierarchical structure from upper-level equipment to lower-level equipment in the power transmission path, such as the hierarchical inheritance relationship from distribution transformers to branch lines and then to end users. When the line loss change pattern of the lower level cannot form a normal delayed transmission relationship with the change trend of the upper level, longitudinal genetic anomalies can be identified, reflecting that there may be abnormal energy dissipation or topological state in the longitudinal lineage chain.

[0050] Subsequently, the familial intrinsic abnormality features, horizontal bloodline abnormality features, and vertical genetic abnormality features are combined to obtain the familial abnormality features of line loss, thereby constructing a comprehensive feature set for line loss diagnosis. At the same time, it can characterize the deviation of the inherent performance of the equipment, the relative abnormality of the same level structure, and the temporal abnormality of the upper and lower level transmission chains, which can reveal the overall health status of power grid equipment under the structural bloodline system in a multidimensional way.

[0051] Inputting the abnormal features of the line loss family into a trained and converged AI diagnostic model refers to feeding the extracted abnormal features as input vectors into an AI model that has completed training and reached a convergent state. The AI ​​diagnostic model can be a deep neural network, graph neural network, or ensemble learning model, used to classify and distinguish abnormal types based on feature distributions. Performing anomaly detection and diagnosis, and obtaining anomaly detection and diagnostic analysis results, involves processing the input features through the model's internal parameters and outputting the probability distribution or label of the abnormal type, such as high-resistance line contact, metering device misalignment, electricity theft, or equipment aging, thereby forming anomaly detection and diagnostic results.

[0052] The fusion positioning module 4 is used to perform fusion positioning based on the calculation and analysis results of the theoretical line loss calculation and analysis and the anomaly detection and diagnosis analysis, and to feed back the line loss prediction and positioning information.

[0053] Furthermore, this application also includes: a probabilistic diagnostic conclusion generation unit, used to perform multi-source evidence fusion on the theoretical line loss distribution characteristics obtained from the theoretical line loss calculation and analysis, and the abnormal root cause probability obtained from the abnormal detection and diagnostic analysis, to calculate the credibility of each potential abnormal cause and generate a probabilistic diagnostic conclusion; a line loss prediction and location report output unit, used to associate and map the abnormal causes whose credibility meets the screening criteria with the device nodes and topological paths in the dynamic family genetic knowledge graph based on the probabilistic diagnostic conclusion, and output a line loss prediction and location report containing specific location coordinates, anomaly type confidence, and impact range; and a targeted governance work order generation unit, used to generate a targeted governance work order based on the line loss prediction and location report, wherein the targeted governance work order includes at least one or more of the following: electricity theft inspection work order, equipment inspection work order, and file verification work order.

[0054] Specifically, theoretical line loss distribution characteristics are obtained through theoretical line loss calculation and analysis. These characteristics are spatial distributions and numerical features calculated using a line loss model based on the inherent parameters of power grid equipment, physical topology, and load characteristics. They describe the line loss performance that each piece of equipment should exhibit under normal operating conditions. The theoretical line loss distribution characteristics are then combined with anomaly detection and diagnostic analysis to obtain the probability of root causes of anomalies. This probability is the probability distribution of potential anomaly causes obtained by analyzing the anomaly characteristics of line loss families using an AI diagnostic model, characterizing the likelihood of each anomaly mechanism occurring. Multi-source evidence fusion is then performed on the probability of root causes of anomalies to calculate the credibility of each potential anomaly cause, generating probabilistic diagnostic conclusions. This involves combining theoretical evidence based on physical mechanisms with data-driven statistical evidence, calculating the credibility of each potential anomaly cause using Bayesian fusion or evidence theory methods, thereby forming diagnostic conclusions that include probability values.

[0055] Furthermore, based on probabilistic diagnostic conclusions, anomalies meeting the screening criteria are associated and mapped with device nodes and topological paths in the dynamic family genetic knowledge graph. This outputs a line loss prediction and location report containing specific location coordinates, anomaly type confidence level, and impact range. The probabilistic diagnostic conclusions provide confidence values ​​for each anomaly cause, and a screening threshold is set to select anomalies with sufficiently high confidence. The dynamic family genetic knowledge graph is a comprehensive knowledge network composed of physical connections, electrical influence relationships, and multi-level kinship structures. Device nodes represent individual power grid equipment entities, and topological paths represent structural paths for energy flow and impact propagation. The association mapping process determines the physical and electrical location of the anomaly by aligning the anomaly cause with the corresponding device node and its line loss propagation path. The final line loss prediction and location report includes specific coordinates, the confidence level corresponding to the anomaly type, and the potentially affected spatial range.

[0056] Subsequently, targeted remediation work orders are generated based on the line loss prediction and location report. These work orders include at least one or more of the following: electricity theft investigation work orders, equipment inspection work orders, and file verification work orders. The line loss prediction and location report provides accurate information on the location and type of anomalies. Targeted remediation work orders match different types of remediation tasks based on the nature of the anomaly to achieve precise remediation. For example, electricity theft investigation work orders are used to investigate illegal electricity use such as bypassing metering devices and unauthorized wiring; equipment inspection work orders are used to conduct on-site inspections of suspected aging or faulty equipment; and file verification work orders are used to verify line ledgers and update outdated equipment files. When an anomaly involves both equipment aging and metering anomalies, two or three different types of work orders can be generated simultaneously to improve remediation efficiency.

[0057] In summary, the big data-driven intelligent prediction system for distribution network line losses provided in this application has the following technical effects: by achieving the technical goal of unified modeling and dynamic correlation analysis of the genetic characteristics, topology, electrical influence paths and operating status of equipment in the distribution network, it can accurately identify the causes of abnormal line losses, accurately locate abnormal equipment and its affected areas, and automatically generate targeted treatment work orders to improve the efficiency and intelligence level of line loss management.

[0058] Example 2: Based on the inventive concept of a big data-driven intelligent prediction system for distribution network line losses in the foregoing embodiments, this application also provides an electronic device, including: at least one processor; a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the steps of the big data-driven intelligent prediction system for distribution network line losses described in any one of the above embodiments.

[0059] Appendix Figure 3 This is a schematic diagram of the structure of an exemplary electronic device of this application. Figure 3 In this document, the bus architecture is represented by bus 300. Bus 300 may include any number of interconnected buses and bridges, and bus 300 connects various circuits including one or more processors represented by processor 302 and memory represented by memory 304. Bus 300 may also connect various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. Bus interface 305 provides an interface between bus 300 and receiver 301 and transmitter 303. Receiver 301 and transmitter 303 may be the same element, i.e., a transceiver, providing a unit for communicating with various other devices over a transmission medium. Processor 302 is responsible for managing bus 300 and general processing, while memory 304 can be used to store data used by processor 302 during operation.

[0060] In Embodiment 3, based on the same inventive concept as the big data-driven intelligent prediction system for distribution network line losses in the foregoing embodiments, this application also provides a computer-readable storage medium storing a computer program, which, when executed, implements the steps of the big data-driven intelligent prediction system for distribution network line losses described in any one of Embodiment 1.

[0061] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0062] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application also intends to include such modifications and variations.

Claims

1. A big data driven power distribution network line loss intelligent prediction system, characterized in that, The method comprises the following steps: a data lake construction module is used to integrate the operation and distribution topology data, the measurement system data, the environmental data and the equipment account data from the power distribution network, and construct a unified data lake; a knowledge graph construction module is used to analyze the family relationship of the power distribution network based on the unified data lake through graph calculation, and construct a dynamic family genetic knowledge graph; a line loss analysis module is used to receive multi-source monitoring data, and perform line loss analysis based on the dynamic family genetic knowledge graph, including theoretical line loss calculation analysis and abnormal detection and diagnosis analysis; a fusion positioning module is used to perform fusion positioning according to the operation analysis results of the theoretical line loss calculation analysis and the abnormal detection and diagnosis analysis, and feed back line loss prediction positioning information; the knowledge graph construction module comprises: a multi-dimensional relationship network construction unit is used to sort out the equipment gene relationship based on the operation and distribution topology data, and construct a multi-dimensional relationship network containing an equipment gene library; a response relationship determination unit is used to compare and fit the data according to the measurement system data, the environmental data and the equipment account data, and determine the equipment gene response relationship; a knowledge graph construction unit is used to adaptively evolve the topological relationship according to the equipment gene response relationship, track the signal genetic relationship between the equipment, add to the multi-dimensional relationship network, and construct the dynamic family genetic knowledge graph; the line loss analysis module comprises: a theoretical technical line loss rate calculation unit is used to identify the physical connection relationship and the electrical belonging relationship based on the dynamic family genetic knowledge graph, calculate the actual line loss rate according to the multi-source monitoring data, and calculate the corresponding theoretical technical line loss rate based on the equipment gene parameters; a line loss family abnormal feature extraction unit is used to extract the line loss family abnormal feature by mapping the multi-source monitoring data on the dynamic family genetic knowledge graph, combining the theoretical technical line loss rate and the actual line loss rate; an abnormal detection and diagnosis analysis result obtaining unit is used to input the line loss family abnormal feature into a trained and converged AI diagnosis model, perform abnormal detection and diagnosis, and obtain an abnormal detection and diagnosis analysis result; the line loss family abnormal feature extraction unit comprises: an intrinsic abnormal feature extraction channel is used to extract the intrinsic abnormal feature according to the deviation degree of the theoretical technical line loss rate and the actual line loss rate; a transverse blood relationship abnormal feature obtaining channel is used to calculate the dispersion degree of the transverse line loss rate distribution of the same level power grid equipment based on the dynamic family genetic knowledge graph, and obtain the transverse blood relationship abnormal feature; a longitudinal genetic abnormal feature extraction channel is used to perform line loss rate change time sequence correlation analysis between the upper and lower power grid levels based on the dynamic family genetic knowledge graph, and extract the longitudinal genetic abnormal feature; a line loss family abnormal feature obtaining channel is used to combine the intrinsic abnormal feature, the transverse blood relationship abnormal feature and the longitudinal genetic abnormal feature, and obtain the line loss family abnormal feature.

2. The big data driven power distribution network line loss intelligent prediction system of claim 1, wherein, the multi-dimensional relationship network construction unit comprises: a topological graph network construction channel is used to construct a topological graph network taking the equipment as nodes and the connection relationship as edges based on the equipment connection relationship, the electrical parameters and the equipment hierarchical structure in the operation and distribution topology data; The multi-dimensional relationship network obtains a channel for establishing a device gene library in the topological graph network, wherein inherent electrical characteristic parameters of each device are stored, including line impedance parameters, transformer no-load loss and load loss curves, electric meter accuracy characteristic quantities, and device health state indicators, and the multi-dimensional relationship network is obtained.

3. The big data driven power distribution network line loss intelligent prediction system of claim 1, wherein, The response relationship determination unit comprises: A response relationship model establishment channel is configured to perform spatio-temporal correlation analysis on the measurement system data, environmental data, and device operation data and device gene parameters in the device account data by a graph neural network, and establish a response relationship model of device electrical operation data and gene parameters. A device gene response relationship determination channel is configured to verify and correct the topological relationship in the multi-dimensional relationship network based on the response relationship model, and determine the device gene response relationship.

4. The big data driven power distribution network line loss intelligent prediction system of claim 3, wherein, The knowledge graph construction unit comprises: A multi-dimensional relationship network construction channel is configured to construct a multi-dimensional relationship network including device static gene characteristics and dynamic operation characteristics based on the verified and corrected device gene response relationship. A traversal tracking channel is configured to establish a signal propagation path model in the multi-dimensional relationship network, and track the propagation path and influence range of power quality abnormalities and line loss abnormalities in the topological network by a graph traversal algorithm. A dynamic family genetic knowledge graph obtaining channel is configured to add a signal genetic path as a recessive relationship edge to the multi-dimensional relationship network, and obtain a dynamic family genetic knowledge graph containing both physical connection and electrical influence relationships. Further comprising: A graph reconstruction process triggering channel is configured to establish a graph version management mechanism, and trigger a graph reconstruction process when detecting changes in power grid structure, device parameter updates, or operation modes. An adaptive evolution update channel is configured to automatically adjust device gene parameter weights and perform adaptive evolution update of the knowledge graph by continuously comparing real-time measurement data and graph prediction values based on the graph reconstruction process.

5. The big data driven power distribution network line loss intelligent prediction system of claim 1, wherein, The fusion positioning module comprises: A probabilistic diagnosis conclusion generation unit is configured to perform multi-source evidence fusion on the theoretical line loss distribution characteristics obtained by the theoretical line loss calculation and analysis and the abnormal root cause probability obtained by the abnormality detection and diagnosis analysis, calculate the credibility of each potential abnormal reason, and generate a probabilistic diagnosis conclusion. A line loss prediction positioning report output unit is configured to associate and map abnormal reasons with credibility reaching a screening condition with device nodes and topological paths in the dynamic family genetic knowledge graph based on the probabilistic diagnosis conclusion, and output a line loss prediction positioning report containing specific positioning coordinates, abnormal type confidence, and influence range. A targeted governance work order generation unit is configured to generate a targeted governance work order according to the line loss prediction positioning report, wherein the targeted governance work order at least includes one or more of electricity stealing inspection work orders, device inspection work orders, and file verification work orders.

6. An electronic device, comprising: The system comprises: at least one processor; a memory connected in communication with the at least one processor; The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the functions of the power distribution network line loss intelligent prediction system driven by big data according to any one of claims 1 to 5.

7. A computer readable storage medium characterized by The computer program stored on the computer readable storage medium, when executed, implements the functions of the power distribution network line loss intelligent prediction system driven by big data according to any one of claims 1 to 5.

Images