A power distribution network operation state risk early warning system

Abstract

The application discloses a power distribution network operation state risk early warning system and relates to the technical field of power distribution network operation and maintenance.The system comprises a terminal sensing module, which is used for collecting multi-source operation data of key nodes of a power distribution network; a data fusion processing module, which is used for receiving data transmitted by the terminal sensing module; an intelligent early warning analysis module, which is based on a data set output by the data fusion processing module, adopts a mechanism and data double-driving model, and realizes quantitative evaluation of multi-dimensional risks of the power distribution network; an application cooperation module; and a cloud-side management and control platform module, which is used for realizing data storage, model training and iteration, early warning information display, permission management and global optimization decision-making.The power distribution network operation state risk early warning system realizes comprehensive sensing of key nodes through multiple types of intelligent sensing terminals, realizes multi-source data fusion through an improved D-S evidence theory, and improves data integrity, accuracy and synchronism in combination with a joint cleaning algorithm and a quality evaluation mechanism.

Description

Technical Field

[0001] This invention relates to the field of power distribution network operation and maintenance technology, specifically a power distribution network operation status risk early warning system. Background Technology

[0002] Distribution network operation and maintenance combines cloud computing, the Internet of Things, big data, artificial intelligence and Internet service concepts. Its core objective is to improve the reliability of distribution system operation, reduce operation and maintenance costs, and ensure stable power supply. With the large-scale grid connection of distributed power sources, the popularization of electric vehicle charging piles and the diversified development of user loads, the operating conditions of distribution networks are becoming increasingly complex. The risks of faults such as overload, insulation degradation, short circuit and grounding have increased significantly, which puts forward higher requirements for real-time monitoring and risk warning of distribution network operation status.

[0003] For example, patent CN108805427B discloses a big data-based distribution network operation status risk early warning system. This system includes: a data acquisition module for real-time acquisition of distribution network operation status data; a data processing module for processing the acquired data to obtain status information describing the distribution network operation status; an operation status assessment module for assessing the distribution network operation status based on the data processed by the data processing module; and a risk early warning module for determining the risk level of the distribution network operation status based on the assessment results and issuing corresponding alarm signals according to the obtained risk level. This invention analyzes the real-time acquired distribution network operation status data to perceive threatening behaviors in the distribution network, assesses and predicts the distribution network operation status based on the perceived threatening behaviors, and then issues an alarm, enabling power workers to detect problems early and minimize the risks and losses that threatening behaviors in the distribution network may cause.

[0004] For example, patent CN111740500A discloses an IoT-based distribution network operation status risk early warning system, including a monitoring module, a distribution network operation status risk early warning module, and a user terminal. The monitoring module collects operation status data of the main equipment in the distribution network based on a wireless sensor network. The distribution network operation status risk early warning module receives, stores, and displays the operation status data, and compares the operation status data with the boundary values ​​of a pre-set normal threshold range. If the data exceeds the normal threshold range, a risk early warning message is output. The user terminal is used to access the operation status data in the distribution network operation status risk early warning module in real time. This invention utilizes IoT technology to monitor the main equipment in the distribution network and provides risk early warnings when the operation status data is abnormal, facilitating remote monitoring by relevant personnel.

[0005] Existing distribution network operation status risk early warning systems have significant data deficiencies in actual operation. The sensing data is fragmented and lacks coverage. Many old distribution network key nodes lack the necessary monitoring equipment. The sensor data sampling rate is low and the synchronization is poor, making it difficult to capture transient fault characteristics. Moreover, most systems still rely on simple threshold over-limit alarms, lacking true risk prediction capabilities and failing to identify equipment trend degradation characteristics. The AI ​​model has weak generalization ability and relies on historical fault samples. However, the distribution network fault samples are few and unevenly distributed, leading to the problem that the model is prone to failure in new areas and on new equipment.

[0006] To address the aforementioned issues, there is an urgent need for innovative design based on the existing early warning system. Summary of the Invention

[0007] The purpose of this invention is to provide a distribution network operation status risk early warning system to address the significant data-level deficiencies in existing distribution network operation status risk early warning systems mentioned above. These deficiencies include fragmented and insufficiently covered sensing data, a lack of necessary monitoring equipment at many critical nodes in older distribution networks, low sensor data sampling rates and poor synchronization, making it difficult to capture transient fault characteristics. Furthermore, most systems still rely on simple threshold-over-limit alarms, lacking true risk prediction capabilities and failing to identify equipment trend degradation characteristics. Additionally, the AI ​​models have weak generalization capabilities, relying on historical fault samples, but the limited and unevenly distributed distribution network fault samples lead to model failures in new areas and on new equipment.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a power distribution network operation status risk early warning system, comprising:

[0009] The terminal sensing module is used to collect multi-source operation data of key nodes in the power distribution network, and to perform preliminary preprocessing and synchronous transmission of the collected data;

[0010] The data fusion and processing module is used to receive data transmitted from the terminal sensing module, perform data cleaning, standardization, fusion and quality assessment, break down data silos and build a unified distribution network operation status dataset;

[0011] The intelligent early warning analysis module, based on the dataset output by the data fusion processing module, adopts a dual-driven model of mechanism and data to achieve quantitative assessment of multi-dimensional risks in the distribution network, early identification of minor faults, accurate fault location, and graded early warning.

[0012] The application collaboration module is used to receive early warning information and decision suggestions output by the intelligent early warning analysis module, realize the collaborative linkage between early warning and operation and maintenance, emergency repair and dispatching business, and complete the closed-loop management of the entire fault handling process;

[0013] The cloud-based management platform module is used to realize data storage, model training and iteration, early warning information display, access control and global optimization decision-making. At the same time, it collaborates with edge devices to realize cloud-edge collaborative data processing and early warning response.

[0014] Preferably, the terminal sensing module includes multiple types of intelligent sensing terminals, data acquisition devices and edge communication modules, and the intelligent sensing terminals are deployed at key nodes such as power distribution lines, ring network cabinets, distribution transformers, cable joints, and disconnectors, including electrical quantity sensors, status quantity sensors, environmental quantity sensors and user-side sensors.

[0015] Preferably, the electrical quantity sensors are used to collect data such as three-phase voltage, current, power, frequency, and harmonics; the status quantity sensors are used to collect data such as contact temperature, cabinet temperature and humidity, partial discharge, SF6 gas concentration, insulation leakage current, switch position, and protection action; the environmental quantity sensors are used to collect data such as wind speed, rainfall, ice thickness, number of lightning strikes, pollution index, and distance from tree obstacles; and the user-side sensors are used to collect data such as charging pile load, industrial and commercial load, and distributed power output.

[0016] Preferably, the data acquisition device includes an FTU, DTU, TTU, smart meter, fault indicator, traveling wave monitoring device, and video surveillance equipment, used to aggregate data collected by the smart sensing terminal, realize synchronous sampling and preliminary filtering of data. The edge communication module adopts redundant deployment of multiple communication methods such as power line carrier, LoRaWAN, 5G, industrial Ethernet and optical fiber, used to transmit the collected data to the data fusion processing layer in real time, and at the same time receive instructions issued by the cloud management platform and the intelligent early warning analysis layer.

[0017] Preferably, the data fusion processing module includes a data access module, a data cleaning module, a data standardization module, a multi-source data fusion module, and a data quality assessment module;

[0018] The data access module is used to access the data transmitted by the terminal perception layer, and at the same time connect to external systems such as SCADA, PMS, GIS, marketing, meteorology, and lightning monitoring to achieve unified access to multi-source heterogeneous data.

[0019] The data cleaning module employs a joint cleaning algorithm based on isolated forest and sliding window to identify and correct noisy data such as packet loss, drift, interference, and abnormal jumps, while removing redundant data and retaining valid data.

[0020] The data standardization module, based on the distribution network data standard, performs unified format conversion, unit unification and coding standardization on data from different sources and in different formats, establishes a unified data dictionary, and solves the problem of inconsistent standards for multi-source data.

[0021] The multi-source data fusion module adopts an improved DS evidence theory to perform hierarchical fusion of electrical quantity data, equipment status data, environmental data, user load data and external system data, and combines distribution network topology information to construct a distribution network operation status dataset containing static and dynamic data.

[0022] The data quality assessment module establishes multi-dimensional data quality assessment indicators, uses the entropy weight method to quantify and assess the data quality level, marks low-quality data and feeds it back to the terminal perception layer, triggering sensor calibration or data re-acquisition commands.

[0023] Preferably, the intelligent early warning analysis module includes a mechanism-data dual-driven model module, a weak fault identification module, a fault location module, and a graded early warning module.

[0024] Preferably, the mechanism-data dual-driven model module combines the physical mechanism of the distribution network with AI algorithms to construct a hybrid early warning model that integrates LSTM, CNN, and the distribution network power flow calculation model. The AI ​​algorithm is used to learn the statistical patterns of historical data, while the distribution network power flow calculation model provides physical mechanism constraints to avoid early warning conclusions that violate electrical laws. Simultaneously, the model introduces an attention mechanism to focus on key influencing factors such as distributed power source fluctuations and load time-varying characteristics, thereby improving the model's generalization ability.

[0025] The weak fault identification module uses wavelet packet transform and an improved isolated forest algorithm to extract and amplify features of weak fault signals such as early partial discharge, minor contact failure, hidden grounding, and high-resistance grounding, so as to achieve early identification of weak faults and issue warnings 1-24 hours in advance.

[0026] The fault location module integrates traveling wave positioning technology and topology matching algorithm, combined with GIS geographic information, to accurately locate the fault point with a positioning error of ≤50 meters. At the same time, for scenarios with many distribution network branches and complex topology, fault transient waveform analysis is introduced to correct the positioning deviation and achieve accurate fault point location.

[0027] The tiered early warning module establishes a refined risk level classification standard based on the results of multi-dimensional risk coupling assessment, dividing the risk into four levels: red, orange, yellow, and blue. It also adds risk index, equipment health index, and remaining life prediction indicators. Among them, the risk assessment adopts a modified CRITIC-entropy weighting method to dynamically assign weights, quantitatively calculate the comprehensive risk value of operational risk, equipment risk, environmental risk, and compound risk, determine the early warning level based on the comprehensive risk value, and generate targeted early warning information and handling suggestions.

[0028] Preferably, the application collaboration module includes an early warning push module, an operation and maintenance scheduling module, a fault handling module, and a closed-loop feedback module.

[0029] Preferably, the early warning push module pushes the early warning information output by the intelligent early warning analysis layer to relevant personnel in operation and maintenance, dispatch, and emergency repair through various means such as monitoring screens, mobile apps, and SMS, to ensure that the early warning information is transmitted quickly.

[0030] The operation and maintenance scheduling module automatically generates operation and maintenance strategies, maintenance priorities and power transfer schemes by combining the real-time operation status and early warning information of the distribution network. It optimizes the power transfer path through a genetic algorithm to achieve rapid power restoration in non-faulty areas. At the same time, it realizes the coordination between the main and distribution networks, receives information such as voltage support and fault propagation transmitted by the main network system, and predicts the impact of main network faults on the distribution network in advance.

[0031] The fault handling module realizes closed-loop management of the entire process, including early warning, dispatching, on-site investigation, defect elimination and review. It connects with the operation and maintenance work order system, automatically generates emergency repair work orders, assigns them to the corresponding emergency repair personnel, and tracks the progress of work order handling in real time.

[0032] The closed-loop feedback module collects fault handling results and operation and maintenance data, and feeds them back to the intelligent early warning analysis layer and cloud management and control platform for model parameter iterative optimization and data quality improvement, forming a self-learning closed loop of data collection, early warning analysis, fault handling and feedback optimization.

[0033] Preferably, the cloud-based management platform includes a data storage module, a model management module, a visualization module, a permission management module, and a cloud-edge collaboration module.

[0034] Compared with the prior art, the beneficial effects of the present invention are:

[0035] 1. Achieve comprehensive perception of key nodes through multiple types of intelligent sensing terminals, realize multi-source data fusion by adopting improved DS evidence theory, and improve data integrity, accuracy and synchronization by combining joint cleaning algorithms and quality assessment mechanisms, breaking down data silos and providing reliable data support for accurate early warning; at the same time, realize sensor calibration and data resampling through data quality feedback mechanism to further improve data quality.

[0036] 2. Improved the intelligence and generalization ability of the early warning model: Adopted a mechanism-data dual-driven model, combining the physical mechanism of the distribution network with AI algorithms to avoid early warning conclusions that violate electrical laws; Introduced attention mechanism and dynamic weighting algorithm to consider the coupled effects of distributed power source fluctuations, load time-varying nature, and environmental factors, solving the problems of weak generalization ability and inability to quantify composite risks in existing models; At the same time, through the model self-learning closed loop, the model parameters are continuously optimized, with an early warning accuracy of ≥98% and a false alarm rate of <2%.

[0037] 3. Achieved early identification and precise fault location of minor faults: By adopting wavelet packet transform and improved isolated forest algorithm, it can effectively identify early minor faults and issue early warnings 1-24 hours in advance, truly achieving proactive prevention; by integrating traveling wave positioning and topology matching algorithm, combined with GIS geographic information, the positioning error is ≤50 meters, solving the problems of poor positioning accuracy and low on-site troubleshooting efficiency of existing systems, and significantly shortening the fault handling time.

[0038] 4. Enhanced system collaboration capabilities and engineering feasibility, enabling collaboration between main and distribution networks, early warning and operation and maintenance, emergency repair, and dispatch services. Automatically generated operation and maintenance strategies and supply transfer plans, achieving closed-loop management of the entire process and addressing the shortcomings of existing systems that only report and do not manage. Adopting a cloud-edge collaborative architecture, the edge can independently complete local early warning and fault diagnosis, avoiding system paralysis due to network interruptions and improving system reliability. At the same time, it supports personalized model configuration, improving system standardization and portability, and reducing implementation costs.

[0039] 5. Improved the comprehensiveness and precision of risk warning, added a module for identifying human-caused hazards and a module for predicting long-term trends, enabling real-time perception of human-caused hazards and prediction of risk trends for the next 1-7 days; established a refined risk level classification standard, added risk index, equipment health index and remaining life prediction, and solved the problems of simple risk classification and lack of long-term prediction in the existing system. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the overall early warning method of the present invention.

[0041] Figure 2 This is a schematic diagram of the terminal sensing module of the present invention. Detailed Implementation

[0042] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] This application provides a power distribution network operation status risk early warning system, including:

[0044] The terminal sensing module is used to collect multi-source operation data of key nodes in the power distribution network, and to perform preliminary preprocessing and synchronous transmission of the collected data;

[0045] The data fusion and processing module is used to receive data transmitted from the terminal sensing module, perform data cleaning, standardization, fusion and quality assessment, break down data silos and build a unified distribution network operation status dataset;

[0046] The intelligent early warning analysis module, based on the dataset output by the data fusion processing module, adopts a dual-driven model of mechanism and data to achieve quantitative assessment of multi-dimensional risks in the distribution network, early identification of minor faults, accurate fault location, and graded early warning.

[0047] The application collaboration module is used to receive early warning information and decision suggestions output by the intelligent early warning analysis module, realize the collaborative linkage between early warning and operation and maintenance, emergency repair and dispatching business, and complete the closed-loop management of the entire fault handling process;

[0048] The cloud-based management platform module is used to realize data storage, model training and iteration, early warning information display, access control and global optimization decision-making. At the same time, it collaborates with edge devices to realize cloud-edge collaborative data processing and early warning response.

[0049] In order to better understand the above technical solutions, the following will describe the above technical solutions in detail with reference to the accompanying drawings and specific embodiments. Figures 1-2 As shown in this embodiment of the present application, a power distribution network operation status risk early warning system includes the following steps in its early warning method:

[0050] S1, the terminal sensing module, is used to collect multi-source operation data of key nodes in the distribution network, and to perform preliminary preprocessing and synchronous transmission of the collected data.

[0051] It should be noted that in this application, the terminal sensing module includes multiple types of intelligent sensing terminals, data acquisition devices, and edge communication modules. It also integrates a human-caused hazard identification module. The intelligent sensing terminals are deployed at key nodes such as distribution network lines, ring main units, distribution transformers, cable joints, and disconnectors. They include electrical quantity sensors, status quantity sensors, environmental quantity sensors, and user-side sensors. The electrical quantity sensors are used to collect data such as three-phase voltage, current, power, frequency, and harmonics; the status quantity sensors are used to collect data such as contact temperature, cabinet temperature and humidity, partial discharge, SF6 gas concentration, insulation leakage current, switch position, and protection actions; the environmental quantity sensors are used to collect data such as wind speed, current, power, frequency, and harmonics. Data includes rainfall, ice thickness, lightning strike frequency, pollution index, and distance to tree obstacles; user-side sensors collect data on charging pile load, industrial and commercial load, and distributed power output; data acquisition devices include FTU, DTU, TTU, smart meters, fault indicators, traveling wave monitoring devices, and video surveillance equipment, used to aggregate data collected by smart sensing terminals, achieve synchronous sampling and preliminary filtering of data; the edge communication module adopts redundant deployment of multiple communication methods such as power line carrier, LoRaWAN, 5G, industrial Ethernet, and fiber optics, used to transmit the collected data to the data fusion processing layer in real time, and simultaneously receive instructions from the cloud management platform and the intelligent early warning analysis layer.

[0052] In practical implementation, the intelligent sensing terminals are first distributed according to the key nodes of the distribution network, deployed at locations such as 10kV lines, ring main units, distribution transformers, cable joints, and disconnectors. Specifically: For electrical quantity sensors: high-precision Hall effect sensors are used to collect three-phase voltage (measurement range 0-35kV), current (measurement range 0-1000A), power, frequency (50Hz±0.5Hz), and harmonics (2nd-21st harmonics), with a sampling rate set to 10kHz to ensure the capture of transient fault characteristics; For status quantity sensors: temperature sensors (measurement range -40℃-150℃, accuracy ±0.5℃) and partial discharge sensors (detection range 10pC-100℃) are deployed at ring main units and cable joints. 0pC), SF6 gas concentration sensor (measurement range 0-1000ppm), insulation leakage current sensor (measurement range 0-10mA), and integrated switch position sensor and protection action sensor to collect real-time equipment operating status; for environmental sensors: wind speed sensor (measurement range 0-60m / s), rainfall sensor (measurement range 0-500mm / 24h), ice thickness sensor (measurement range 0-100mm), lightning strike counter (records the number and intensity of lightning strikes), pollution index sensor, tree obstacle distance sensor (measurement range 0-5m) deployed on line towers to collect real-time environmental impact factor data; for user-side sensors: deployed on charging piles, industrial and commercial user distribution... The system collects data on charging pile load, industrial and commercial load, and photovoltaic output in the power room and distributed photovoltaic power station, with a sampling rate set to 1kHz. The data acquisition device includes FTUs, DTUs, TTUs, smart meters, fault indicators, traveling wave monitoring devices, and high-definition video surveillance equipment. The FTUs, DTUs, and TTUs are responsible for summarizing line and transformer data; the fault indicators capture transient signals of line faults in real time; the traveling wave monitoring device collects fault traveling wave signals; and the video surveillance equipment is deployed around the line and in construction areas to capture images of potential human-caused hazards. The data acquisition device performs preliminary filtering of the collected data, removing obviously abnormal data to achieve synchronous data sampling. The edge communication module adopts redundant deployment of 5G and fiber optics in remote areas. The domain-assisted system uses LoRaWAN communication, with power line carrier used for data transmission between the distribution transformer and the smart meter to ensure the stability and real-time performance of data transmission. The edge communication module transmits the collected data to the data fusion processing module in real time, while simultaneously receiving instructions from the cloud management platform module and the intelligent early warning analysis module for sensor calibration, data re-sampling, etc. Then, the human-caused hazard identification module uses the YOLOv8 algorithm to perform image recognition on the images collected by the video monitoring equipment to identify human-caused hazards such as construction machinery, unauthorized workers, and tree obstacles. Combined with tree obstacle distance sensor data, when the distance to the tree obstacle is less than the safety threshold (e.g., 1.5m) or unauthorized work is detected, a human-caused hazard warning is immediately generated and transmitted to the intelligent early warning analysis module.

[0053] In this embodiment, multiple types of sensors deployed at key nodes such as lines and ring main units can comprehensively collect multi-source data including electrical quantities, equipment status, environment, and user load. Combined with high-precision sampling and preliminary filtering, this ensures data integrity and synchronization, providing a reliable foundation for subsequent data fusion and early warning analysis. This addresses the problems of insufficient sensing coverage and difficulty in capturing transient fault characteristics in existing systems. By capturing weak fault signals such as partial discharge and insulation degradation through status sensors, and combining this with a human-caused hazard identification module, the system can identify early faults and human-caused hazards such as construction violations and tree obstructions 1-24 hours in advance, preventing fault escalation and truly achieving proactive risk prevention. Redundant deployment using multiple communication methods ensures real-time data transmission. The data fusion processing layer supports receiving instructions from both the cloud and the early warning layer, enabling sensor calibration and data resampling. In extreme weather conditions, it can collaborate with cloud-edge systems to ensure independent local early warning at the edge, preventing system paralysis due to network outages. It can also aggregate and synchronously transmit multi-source data, providing a prerequisite for the cleaning, standardization, and fusion of the data fusion processing layer. Simultaneously, it connects to external system data, facilitating the accurate operation of the dual-drive model of the distribution network operation status risk early warning system and the distribution network operation status risk early warning system. This, combined with the application collaboration layer, achieves a closed loop of early warning, operation and maintenance, and emergency repair, enhancing the overall system's collaborative capabilities. Through comprehensive perception and early warning, it reduces blind inspections, and with the support of fault location-related data, significantly shortens fault investigation and handling time, reduces labor costs, and improves power supply reliability.

[0054] S2, the data fusion processing module, is used to receive data transmitted from the terminal sensing module, perform data cleaning, standardization, fusion and quality assessment, break down data silos and build a unified distribution network operation status dataset;

[0055] It should be noted that in this application, the data fusion processing module includes a data access module, a data cleaning module, a data standardization module, a multi-source data fusion module, and a data quality assessment module. The data access module is used to access data transmitted from the terminal sensing layer and also interfaces with external systems such as SCADA, PMS, GIS, marketing, meteorology, and lightning monitoring to achieve unified access to multi-source heterogeneous data. The data cleaning module uses a joint cleaning algorithm based on isolated forest and sliding window to identify and correct noise data such as packet loss, drift, interference, and abnormal jumps, while removing redundant data and retaining valid data. The data standardization module is based on distribution network data standards. The system performs unified format conversion, unitization, and coding standardization on data from different sources and in different formats, establishes a unified data dictionary, and solves the problem of inconsistent standards for multi-source data. The multi-source data fusion module adopts an improved DS evidence theory to perform hierarchical fusion of electrical quantity data, equipment status data, environmental data, user load data, and external system data. Combined with distribution network topology information, it constructs a distribution network operation status dataset containing both static and dynamic data. The data quality assessment module establishes multi-dimensional data quality assessment indicators, uses the entropy weight method to quantify and assess data quality levels, marks low-quality data, and feeds it back to the terminal sensing layer to trigger sensor calibration or data re-acquisition commands.

[0056] In practical implementation, the data access module first connects to the data transmitted by the terminal sensing module via the edge communication module. Simultaneously, it interfaces with external systems such as SCADA, PMS, GIS, marketing, meteorology, and lightning monitoring, using standardized interface protocols to achieve unified access to multi-source heterogeneous data. The accessed data includes real-time operational data, equipment ledger data, meteorological data, and historical fault data. Next, the data cleaning module employs a joint cleaning algorithm based on isolated forest and sliding window. The isolated forest algorithm identifies outliers (such as packet loss, drift, and interference) in the data, and then a sliding window (with a window size set to 10 seconds) corrects these outliers, replacing them with the average of the valid data within the window. Simultaneously, redundant data (such as duplicated data) is removed, retaining valid data to ensure accuracy. Finally, the data standardization module, based on distribution network data standards, performs unified processing on data from different sources and in different formats. Electrical quantity data undergoes per-unit conversion, environmental data undergoes normalization, and equipment status data undergoes standardization. Standardized coding is adopted to establish a unified data dictionary, solving the problems of inconsistent standards and formats among multi-source data and ensuring data consistency and comparability. Then, the multi-source data fusion module adopts an improved DS evidence theory to perform layered fusion of electrical quantity data, equipment status data, environmental data, user load data, and external system data. The first layer is the fusion of data of the same type, and the second layer is the fusion of data of different types. Combined with GIS topology information, a distribution network operation status dataset containing static and dynamic data is constructed to improve data completeness and relevance. Finally, data quality assessment is carried out. The data quality assessment module establishes multi-dimensional data quality assessment indicators, including data completeness (missing rate ≤5%), accuracy (error ≤1%), synchronization (synchronization error ≤10ms), and consistency. The entropy weight method is used to quantify the data quality level (excellent, good, qualified, unqualified). Unqualified low-quality data is marked and fed back to the terminal sensing module through the edge communication module to trigger sensor calibration or data re-acquisition commands to ensure data quality.

[0057] In this embodiment, the data access module connects to terminal sensing data of various types and external system data such as the SCADA, PMS, and GIS distribution network operation status risk early warning systems. Combined with a data standardization module, it unifies the format, units, and coding, establishing a unified data dictionary. This addresses the problems of data incompatibility, inconsistent standards, and discrepancies between records and actual conditions in existing systems, providing a unified data foundation for subsequent early warning analysis. Furthermore, a joint cleaning algorithm based on isolated forests and sliding windows effectively identifies and corrects data loss, drift, interference, and other noise data, eliminating redundant data. Combined with the multi-dimensional quantitative evaluation and feedback mechanism of the data quality assessment module, sensor calibration or data resampling can be triggered, significantly improving data integrity, accuracy, and synchronization, avoiding misjudging data errors as equipment failures, and resolving the shortcomings of poor data quality and high false alarm rates in existing systems. An improved distribution network operation status risk early warning system is adopted. The TongDS distribution network operation status risk early warning system achieves multi-source data hierarchical fusion through evidence theory. Combining distribution network topology information, it organically integrates data such as electrical quantities, equipment status, environment, and user load to construct a complete dataset containing both static and dynamic data. This highlights the correlation between data from different dimensions, providing high-quality and highly correlated data support for the intelligent early warning analysis layer's mechanism, the distribution network operation status risk early warning system's dual-drive data model, weak fault identification, and risk coupling assessment, thereby improving the accuracy of the early warning model. The high-quality data, after cleaning, standardization, and fusion, can be efficiently connected to the application collaboration layer and cloud management platform, providing reliable data support for operation and maintenance strategy formulation, work order dispatch, and closed-loop management. At the same time, the data quality assessment results and processed data can be fed back to the terminal perception layer and cloud model management module, assisting in sensor optimization and early warning model parameter iteration, forming a data-driven self-optimization closed loop.

[0058] S3, the intelligent early warning analysis module, is based on the dataset output by the data fusion processing module. It adopts a dual-driven model of mechanism and data to realize the quantitative assessment of multi-dimensional risks in the distribution network, early identification of minor faults, accurate fault location, and hierarchical early warning.

[0059] It should be noted that in this application, the intelligent early warning analysis module includes a mechanism-data dual-driven model module, a weak fault identification module, a fault location module, and a hierarchical early warning module. The mechanism-data dual-driven model module combines the physical mechanisms of the distribution network with AI algorithms to construct a hybrid early warning model integrating LSTM, CNN, and the distribution network power flow calculation model. The AI ​​algorithm is used to learn the statistical patterns of historical data, while the distribution network power flow calculation model provides physical mechanism constraints to avoid early warning conclusions that violate electrical laws. Simultaneously, the model introduces an attention mechanism to focus on key influencing factors such as distributed power source fluctuations and load time-varying characteristics, improving the model's generalization ability. The weak fault identification module uses wavelet packet transform and an improved isolated forest algorithm to extract and amplify features of weak fault signals such as early partial discharge, minor contact defects, hidden grounding, and high-resistance grounding. The system enables early identification of minor faults, issuing warnings 1-24 hours in advance. The fault location module integrates traveling wave positioning technology and topology matching algorithms with GIS geographic information to accurately locate fault points with a positioning error ≤50 meters. Furthermore, for scenarios with numerous distribution network branches and complex topologies, transient fault recording analysis is introduced to correct positioning deviations and achieve precise fault location. The tiered early warning module, based on multi-dimensional risk coupling assessment results, establishes a refined risk level classification standard, dividing warnings into four levels: red, orange, yellow, and blue. It also adds risk indices, equipment health indices, and remaining life prediction indicators. The risk assessment employs an improved CRITIC-entropy weighting method for dynamic weighting, quantifying the comprehensive risk value of operational risk, equipment risk, environmental risk, and composite risk. Based on the comprehensive risk value, the warning level is determined, and targeted warning information and handling suggestions are generated.

[0060] In practical implementation, firstly, a hybrid early warning model is constructed by combining the physical mechanism of the distribution network with AI algorithms, integrating LSTM, CNN, and the distribution network power flow calculation model. CNN is used to extract spatial features from the data, LSTM is used to extract temporal features, and the distribution network power flow calculation model uses the Newton-Raphson method to calculate the power flow distribution of the distribution network, providing physical mechanism constraints and avoiding model outputs that violate electrical laws. The model introduces an attention mechanism, focusing on the coupled effects of distributed power source output fluctuations, load time-varying characteristics, and environmental factors. The model is trained using historical fault data and operation and maintenance data to improve its generalization ability. Then, a small... Wave packet transform decomposes the acquired transient signal, extracts and amplifies weak fault features, and then uses an improved isolated forest algorithm to identify these features. This allows for the identification of weak faults such as early partial discharge, minor contact defects, and hidden grounding, providing early warnings 1-24 hours in advance to prevent fault escalation. Furthermore, by fusing traveling wave positioning technology with a topology matching algorithm, the traveling wave monitoring device acquires the fault traveling wave signal, calculates the traveling wave propagation time, and, combined with the traveling wave velocity (1.5×10⁻⁸ m / s), preliminarily determines the fault location. Finally, by incorporating GIS topology information, a topology matching algorithm corrects the positioning error, particularly useful in scenarios with numerous distribution network branches and complex topologies. Transient fault recording analysis is introduced to capture transient fault characteristics, further improving positioning accuracy to an error of ≤50 meters. Simultaneously, the outage range is predicted, providing support for power transfer scheme development. Then, an improved CRITIC-entropy weighting method is used to dynamically assign weights to operational risks (overvoltage / undervoltage, three-phase imbalance, heavy load), equipment risks (aging, insulation degradation, joint overheating), environmental risks (typhoons, rainstorms, thunderstorms), and composite risks (such as high temperature, heavy load, equipment aging), quantifying the comprehensive risk value (0-100). A refined risk level classification standard is established: Red Alert (comprehensive risk value ≥80, emergency), Orange Alert (60 ≤ comprehensive risk value ≥80, emergency), and others. The system generates three risk levels: a yellow alert (comprehensive risk value < 80, severe), a blue alert (comprehensive risk value < 60, caution), and a yellow alert (comprehensive risk value < 20, warning). It also calculates the equipment health index (0-100) and remaining lifespan prediction, generating targeted warning information and handling suggestions (e.g., immediate power outage and isolation for red alerts, on-site handling within 2 hours for orange alerts). Finally, a spatiotemporal sequence prediction model is used, combined with historical operating data, meteorological forecast data (for the next 7 days), and load forecast data, to predict the trend of the comprehensive risk value of the distribution network over the next 1-7 days, generating a trend analysis report that identifies risk escalation and descent points, providing support for the development of operation and maintenance plans.

[0061] In this embodiment, a mechanism-data dual-driven model is used to integrate the physical mechanism of the distribution network with the AI-based distribution network operation status risk warning system algorithm. This model learns statistical patterns from data through LSTM and CNN algorithms, while providing physical constraints through power flow calculation models. This effectively avoids warning conclusions from existing systems that violate electrical principles. Furthermore, an attention mechanism is introduced to adapt to complex operating conditions such as distributed power source fluctuations and load time-varying characteristics, significantly improving the model's generalization ability and addressing the problem of existing models failing in new areas and on new equipment. This achieves accurate early identification of minor faults, strengthening proactive prevention. It also improves fault location accuracy and shortens response time. For scenarios with many distribution network branches and complex topologies, fault transient waveform analysis is introduced to correct location deviations, controlling the fault location error to ≤50 meters. This addresses the shortcomings of existing systems, such as large location drift and low efficiency of on-site segment-by-segment inspection, providing support for repair personnel to quickly locate fault points and significantly shortening fault response time.

[0062] S4, the application collaboration module, is used to receive early warning information and decision suggestions output by the intelligent early warning analysis module, realize the collaborative linkage between early warning and operation and maintenance, emergency repair and dispatching business, and complete the closed-loop management of the entire fault handling process;

[0063] It should be noted that in this application, the application collaboration module includes an early warning push module, an operation and maintenance scheduling module, a fault handling module, and a closed-loop feedback module. The early warning push module pushes early warning information output by the intelligent early warning analysis layer to relevant personnel in operation and maintenance, scheduling, and emergency repair through various means such as monitoring screens, mobile apps, and SMS, ensuring rapid dissemination of early warning information. The operation and maintenance scheduling module, combining the real-time operating status of the distribution network with early warning information, automatically generates operation and maintenance strategies, repair priorities, and power transfer plans. It optimizes the power transfer path through a genetic algorithm to achieve rapid power restoration in non-faulty areas. Simultaneously, it enables the main and distribution networks to... The system integrates various modules: a collaborative module that receives voltage support and fault propagation information from the main network system to predict the impact of main network faults on the distribution network; a fault handling module that implements closed-loop management of the entire process, including early warning, dispatching, on-site investigation, troubleshooting, and verification, and connects to the maintenance work order system to automatically generate emergency repair work orders, assign them to the corresponding repair personnel, and track the progress of work order handling in real time; and a closed-loop feedback module that collects fault handling results and maintenance data and feeds them back to the intelligent early warning analysis layer and cloud management platform for model parameter iteration optimization and data quality improvement, forming a self-learning closed loop of data collection, early warning analysis, fault handling, and feedback optimization.

[0064] In practice, the system first pushes the early warning information (early warning level, fault location, risk description, and handling suggestions) output by the intelligent early warning analysis module to relevant personnel in operation, maintenance, dispatch, and emergency repair through various means such as the monitoring screen of the cloud management platform module, mobile APP, and SMS. Red and orange warnings are pushed via both SMS and APP pop-up notifications to ensure rapid dissemination of warning information, with a push delay of ≤30 seconds. Next, combining the real-time operating status of the distribution network, early warning information, and GIS topology information, the system automatically generates operation and maintenance strategies (such as enhanced monitoring and regular inspections) and maintenance priorities. A genetic algorithm optimizes the power transfer path to achieve rapid power restoration in non-faulty areas. Simultaneously, it connects to the main grid system to receive main grid voltage support and fault propagation data. Information is used to predict the impact of main network failures on distribution networks in advance and formulate contingency plans. Then, it connects with the operation and maintenance work order system to automatically generate emergency repair work orders based on the early warning information, specifying the repair personnel, repair materials, and handling time limits, and assigning them to the corresponding repair personnel. Repair personnel receive work orders through a mobile APP, conduct on-site investigation and handling, and report the handling results. The system automatically records the handling process, realizing a closed-loop management of the entire process of early warning → work order dispatch → on-site investigation → defect elimination → review. Afterwards, the fault handling results and operation and maintenance data (such as inspection records and defect elimination records) are collected and fed back to the intelligent early warning analysis module and the cloud management and control platform module for model parameter iteration optimization (such as adjusting model weights) and data quality improvement (such as optimizing data cleaning parameters), forming a self-learning closed loop.

[0065] In this embodiment, the early warning push module uses multiple methods, including monitoring screens, mobile distribution network operation status risk early warning system APP, and SMS, to push early warning information. Red and orange warnings use a dual push mode to ensure that early warning information is quickly delivered to relevant personnel in operation and maintenance, dispatch, and emergency repair, with a push delay of ≤30s. This solves the problems of untimely delivery of early warning information and inability of relevant personnel to respond quickly in the existing system. The operation and maintenance dispatch module can automatically generate operation and maintenance strategies, maintenance priorities, and power transfer plans by combining early warning information with the real-time status of the distribution network. It optimizes the power transfer path through a genetic algorithm to achieve rapid power restoration in non-faulty areas. At the same time, it realizes the coordination of the main and distribution networks, predicts the impact of main network faults on the distribution network in advance, and breaks the limitations of the existing system and business separation. Through the fault handling module, it completes the closed-loop management of the entire process of early warning → dispatching → on-site investigation → defect elimination → review. It automatically generates emergency repair work orders, allocates emergency repair resources, and tracks the handling progress without the need for manual dispatching and tracking. This solves the problems of cumbersome and inefficient fault handling processes in the existing system and significantly shortens the fault handling cycle.

[0066] S5, the cloud-based management platform module, is used to realize data storage, model training and iteration, early warning information display, access control and global optimization decision-making. At the same time, it collaborates with edge devices to realize cloud-edge collaborative data processing and early warning response.

[0067] It should be noted that, in this application, the cloud-based management platform includes a data storage module, a model management module, a visualization module, a permission management module, and a cloud-edge collaboration module. The data storage module uses a combination of time-series databases (InfluxDB, TimescaleDB) and relational databases to store real-time operational data, historical fault data, early warning records, and maintenance data, supporting efficient reading, writing, and querying of massive amounts of data. The model management module is used for model training, validation, deployment, and iteration. Combined with data transmitted from the closed-loop feedback module, it periodically updates model parameters to improve the accuracy and generalization ability of model early warnings. It also supports personalized model configuration to adapt to the needs of power distribution networks in different regions and under different operating conditions. The visualization module uses GIS topology... The system combines map and heatmap to display the real-time operating status, risk distribution, early warning information, fault location, and work order progress of the power distribution network. It supports multi-dimensional data query and statistical analysis, providing intuitive support for dispatching decisions. The access control module assigns different operation permissions based on roles, enabling hierarchical management of system data, early warning information, and work order processing, ensuring system security. The cloud-edge collaboration module enables collaborative work between the cloud and the edge. The edge deploys a lightweight early warning model and data processing module, responsible for real-time processing of local data and emergency early warnings, while the cloud is responsible for global data aggregation, model training, and optimization decisions. In the event of network interruption due to extreme weather, the edge can independently complete local early warning and fault diagnosis. After the network is restored, the data is synchronized to the cloud to ensure continuous system operation.

[0068] In practical implementation, the system first uses the time-series database InfluxDB to store real-time operational data and early warning records (storage period ≥ 3 years), and the relational database MySQL to store device ledgers, historical fault data, and maintenance data (storage period ≥ 5 years), supporting efficient reading, writing, and querying of massive amounts of data with a query response time ≤ 1 second. Then, combined with data transmitted from the closed-loop feedback module 44, model parameters are updated monthly and model verification is performed quarterly to ensure the accuracy of model early warnings. Simultaneously, personalized model configuration is supported, allowing adjustment of model parameters for different regions and operating conditions to improve system applicability. Finally, a combination of GIS topology maps and heat maps is used to display the real-time operating status of the distribution network (voltage, current, load rate), risk distribution (red, yellow, green, and blue heat maps), early warning information, fault locations, and work order progress; supporting multi-dimensional data... Based on query and statistical analysis, intuitive support is provided for scheduling decisions. Different operating permissions are assigned based on roles, divided into administrators, dispatchers, maintenance personnel, and emergency repair personnel. Administrators have full system operation permissions, dispatchers can only view early warning information and operational data and issue scheduling instructions, maintenance personnel can only view early warning information and receive maintenance work orders, and emergency repair personnel can only receive emergency repair work orders and provide feedback on handling results, ensuring system security. Finally, collaborative work between the cloud and the edge is achieved. The edge deploys a lightweight early warning model and data processing module, responsible for real-time processing of local data and emergency early warnings (such as red alerts), while the cloud is responsible for global data aggregation, model training, and optimization decisions. In the event of network interruption under extreme weather conditions, the edge can independently complete local early warning and fault diagnosis. After the network is restored, local data and early warning records are synchronized to the cloud to ensure continuous system operation.

[0069] In this embodiment, the data storage module employs a combination of time-series and relational databases to securely store various types of data, including real-time operational data, historical fault data, and early warning records. It supports efficient reading and writing of massive amounts of data and rapid querying (response time ≤ 1s), while also meeting the long-term storage needs of different data types (early warning records ≥ 3 years of distribution network operation status risk early warning system data, maintenance data ≥ 5 years of distribution network operation status risk early warning system data). This addresses the problems of chaotic data storage, inefficient querying, and insufficient storage periods in existing systems, providing complete data support for system analysis and model iteration. The model management module enables the training, verification, deployment, and iteration of the early warning model. Combined with fault handling and maintenance data fed back from the application collaboration layer, it regularly updates model parameters and conducts model verification. It also supports personalized model configuration, adapting to the needs of distribution networks in different regions and under different operating conditions. This addresses the shortcomings of existing systems, such as model rigidity, weak generalization ability, and inability to adapt to complex operating conditions, continuously improving the accuracy of early warnings.

[0070] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A power distribution network operation status risk early warning system, characterized in that, The power distribution network operation status risk early warning system includes: The terminal sensing module is used to collect multi-source operation data of key nodes in the power distribution network, and to perform preliminary preprocessing and synchronous transmission of the collected data; The data fusion and processing module is used to receive data transmitted from the terminal sensing module, perform data cleaning, standardization, fusion and quality assessment, break down data silos and build a unified distribution network operation status dataset; The intelligent early warning analysis module, based on the dataset output by the data fusion processing module, adopts a dual-driven model of mechanism and data to achieve quantitative assessment of multi-dimensional risks in the distribution network, early identification of minor faults, accurate fault location, and graded early warning. The application collaboration module is used to receive early warning information and decision suggestions output by the intelligent early warning analysis module, realize the collaborative linkage between early warning and operation and maintenance, emergency repair and dispatching business, and complete the closed-loop management of the entire fault handling process; The cloud-based management platform module is used to realize data storage, model training and iteration, early warning information display, access control and global optimization decision-making. At the same time, it collaborates with edge devices to realize cloud-edge collaborative data processing and early warning response.

2. The distribution network operation status risk early warning system according to claim 1, characterized in that: The terminal sensing module includes multiple types of intelligent sensing terminals, data acquisition devices, and edge communication modules. The intelligent sensing terminals are deployed at key nodes such as power distribution lines, ring main units, distribution transformers, cable joints, and disconnectors, and include electrical quantity sensors, status quantity sensors, environmental quantity sensors, and user-side sensors.

3. The distribution network operation status risk early warning system according to claim 2, characterized in that: The electrical quantity sensors are used to collect data such as three-phase voltage, current, power, frequency, and harmonics; the status quantity sensors are used to collect data such as contact temperature, cabinet temperature and humidity, partial discharge, SF6 gas concentration, insulation leakage current, switch position, and protection action; the environmental quantity sensors are used to collect data such as wind speed, rainfall, ice thickness, number of lightning strikes, pollution index, and distance from tree obstacles; and the user-side sensors are used to collect data such as charging pile load, industrial and commercial load, and distributed power output.

4. The distribution network operation status risk early warning system according to claim 2, characterized in that: The data acquisition device includes an FTU, DTU, TTU, smart meter, fault indicator, traveling wave monitoring device, and video surveillance equipment. It is used to aggregate the data collected by the smart sensing terminal and realize synchronous sampling and preliminary filtering of the data. The edge communication module adopts redundant deployment of multiple communication methods such as power line carrier, LoRaWAN, 5G, industrial Ethernet, and optical fiber. It is used to transmit the collected data to the data fusion processing layer in real time, and at the same time receive instructions issued by the cloud management platform and the intelligent early warning analysis layer.

5. The distribution network operation status risk early warning system according to claim 1, characterized in that: The data fusion processing module includes a data access module, a data cleaning module, a data standardization module, a multi-source data fusion module, and a data quality assessment module. The data access module is used to access the data transmitted by the terminal perception layer, and at the same time connect to external systems such as SCADA, PMS, GIS, marketing, meteorology, and lightning monitoring to achieve unified access to multi-source heterogeneous data. The data cleaning module employs a joint cleaning algorithm based on isolated forest and sliding window to identify and correct noisy data such as packet loss, drift, interference, and abnormal jumps, while removing redundant data and retaining valid data. The data standardization module, based on the distribution network data standard, performs unified format conversion, unit unification and coding standardization on data from different sources and in different formats, establishes a unified data dictionary, and solves the problem of inconsistent standards for multi-source data. The multi-source data fusion module adopts an improved DS evidence theory to perform hierarchical fusion of electrical quantity data, equipment status data, environmental data, user load data and external system data, and combines distribution network topology information to construct a distribution network operation status dataset containing static and dynamic data. The data quality assessment module establishes multi-dimensional data quality assessment indicators, uses the entropy weight method to quantify and assess the data quality level, marks low-quality data and feeds it back to the terminal perception layer, triggering sensor calibration or data re-acquisition commands.

6. The distribution network operation status risk early warning system according to claim 1, characterized in that: The intelligent early warning analysis module includes a mechanism-data dual-driven model module, a weak fault identification module, a fault location module, and a graded early warning module.

7. The distribution network operation status risk early warning system according to claim 6, characterized in that: The mechanism-data dual-driven model module combines the physical mechanisms of the distribution network with AI algorithms to construct a hybrid early warning model that integrates LSTM, CNN, and the distribution network power flow calculation model. The AI ​​algorithm is used to learn the statistical patterns of historical data, while the distribution network power flow calculation model provides physical mechanism constraints to avoid early warning conclusions that violate electrical principles. Simultaneously, the model introduces an attention mechanism to focus on key influencing factors such as distributed power source fluctuations and load time-varying characteristics, thereby improving the model's generalization ability. The weak fault identification module uses wavelet packet transform and an improved isolated forest algorithm to extract and amplify features of weak fault signals such as early partial discharge, minor contact failure, hidden grounding, and high-resistance grounding, so as to achieve early identification of weak faults and issue warnings 1-24 hours in advance. The fault location module integrates traveling wave positioning technology and topology matching algorithm, combined with GIS geographic information, to accurately locate the fault point with a positioning error of ≤50 meters. At the same time, for scenarios with many distribution network branches and complex topology, fault transient waveform analysis is introduced to correct the positioning deviation and achieve accurate fault point location. The tiered early warning module establishes a refined risk level classification standard based on the results of multi-dimensional risk coupling assessment, dividing the risk into four levels: red, orange, yellow, and blue. It also adds risk index, equipment health index, and remaining life prediction indicators. Among them, the risk assessment adopts a modified CRITIC-entropy weighting method to dynamically assign weights, quantitatively calculate the comprehensive risk value of operational risk, equipment risk, environmental risk, and compound risk, determine the early warning level based on the comprehensive risk value, and generate targeted early warning information and handling suggestions.

8. The distribution network operation status risk early warning system according to claim 1, characterized in that: The application collaboration module includes an early warning push module, an operation and maintenance scheduling module, a fault handling module, and a closed-loop feedback module.

9. A power distribution network operation status risk early warning system according to claim 7, characterized in that: The aforementioned early warning push module pushes the early warning information output by the intelligent early warning analysis layer to relevant personnel in operation and maintenance, dispatch, and emergency repair through various means such as monitoring screens, mobile apps, and SMS messages, ensuring that the early warning information is delivered quickly. The operation and maintenance scheduling module automatically generates operation and maintenance strategies, maintenance priorities and power transfer schemes by combining the real-time operation status and early warning information of the distribution network. It optimizes the power transfer path through a genetic algorithm to achieve rapid power restoration in non-faulty areas. At the same time, it realizes the coordination between the main and distribution networks, receives information such as voltage support and fault propagation transmitted by the main network system, and predicts the impact of main network faults on the distribution network in advance. The fault handling module realizes closed-loop management of the entire process, including early warning, dispatching, on-site investigation, defect elimination and review. It connects with the operation and maintenance work order system, automatically generates emergency repair work orders, assigns them to the corresponding emergency repair personnel, and tracks the progress of work order handling in real time. The closed-loop feedback module collects fault handling results and operation and maintenance data, and feeds them back to the intelligent early warning analysis layer and cloud management and control platform for model parameter iterative optimization and data quality improvement, forming a self-learning closed loop of data collection, early warning analysis, fault handling and feedback optimization.

10. A power distribution network operation status risk early warning system according to claim 1, characterized in that: The cloud-based management platform includes a data storage module, a model management module, a visualization module, a permission management module, and a cloud-edge collaboration module.

Images