Intelligent power distribution equipment cloud edge collaborative multi-source data fusion security management and control system

Abstract

The application discloses a kind of intelligent power distribution equipment cloud edge coordination multi-source data fusion security management and control systems, it is related to intelligent power grid and power distribution automation technical field, for 10kV~35kV power distribution equipment management and control, contain edge data acquisition, pre-processing, cloud storage management, cloud edge coordination scheduling, multi-source data fusion, intelligent risk assessment, security control execution, security protection, man-machine interaction module;Module is interacted by 5G / industrial ethernet;Acquisition module obtains multiple parameters, pre-processing module washes standardization data, cloud hierarchical storage, scheduling module distributes task, fusion module integrates data, assessment module graded risk, protection module safeguards security, and interaction module visual alarm.The application improves power distribution data quality and risk assessment accuracy, optimizes cloud edge coordination efficiency, enhances security protection capability, realizes equipment preventive operation and maintenance, reduces fault and outage time, reduces operation and maintenance cost, and provides support for safe and efficient operation of power distribution network.

Description

Technical Field

[0001] This invention relates to the field of smart grid and distribution automation technology, and in particular to a cloud-edge collaborative multi-source data fusion security management and control system for smart power distribution equipment. Background Technology

[0002] In 10kV~35kV medium-voltage distribution networks, power distribution equipment such as switchgear, transformers, and circuit breakers are the core components ensuring stable power transmission. However, with the expansion of distribution network scale and the growth of electricity demand, the operating environment of equipment is becoming increasingly complex, and traditional management and control methods are no longer sufficient to cope with multi-dimensional safety challenges. Currently, the management and control of power distribution equipment largely relies on local monitoring and manual maintenance. Data collection is limited to single electrical parameters, lacking the coordinated collection of environmental parameters and equipment status parameters. This results in a single data dimension, failing to comprehensively reflect the operating status of the equipment. For example, monitoring only current overload while ignoring temperature rise can easily lead to missed faults such as overheating of equipment contacts. Relying solely on local data collection terminals for storage, without cloud backup and long-term analysis capabilities, makes it difficult to trace historical fault patterns, and maintenance decisions lack data support.

[0003] Some existing management and control systems have attempted to introduce a cloud-edge collaborative architecture, but they suffer from several technical shortcomings: At the data processing level, collected data often contains both obvious and latent anomalies. Traditional cleaning algorithms can only remove obvious anomalies, with a latent anomaly identification rate of less than 80%. Furthermore, the inconsistent formats of multi-source data make standardization and fusion difficult, resulting in data errors exceeding 5% after fusion, directly impacting the accuracy of subsequent risk assessments. At the risk assessment level, fixed thresholds are used to determine risk levels, failing to consider dynamic factors such as peak loads, equipment aging, and seasonal changes, leading to false alarms or missed alarms. At the control execution level, there is a lack of control effectiveness evaluation mechanisms. After implementing measures such as overload load transfer and temperature dissipation, it is impossible to determine whether parameters have returned to normal or whether the risk has decreased. The lack of closed-loop optimization capabilities results in the repeated execution of some ineffective control measures, wasting operational resources.

[0004] Furthermore, insufficient security protection and collaborative scheduling capabilities also restrict the system's practicality: Device access lacks strict identity authentication, allowing unauthorized devices to easily forge identities and tamper with control commands; when cloud-edge transmission bandwidth fluctuates, there is no dynamic adaptation strategy, resulting in non-critical data still being transmitted at high frequency when bandwidth is insufficient, leading to delays in critical data; uneven load distribution at edge nodes, with some nodes experiencing CPU utilization exceeding 80% due to managing too many devices, data processing latency exceeding 200ms, and data loss rate exceeding 3%. With the accelerated intelligent transformation of power distribution networks, the demand for "multi-source data fusion, dynamic risk assessment, secure closed-loop control, and efficient cloud-edge collaboration" is becoming increasingly urgent, and traditional management and control systems can no longer meet the requirements for the safe and efficient operation of medium-voltage power distribution equipment. Summary of the Invention

[0005] The present invention proposes a cloud-edge collaborative multi-source data fusion security management and control system for intelligent power distribution equipment to solve the problems mentioned in the prior art.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A cloud-edge collaborative multi-source data fusion security management and control system for intelligent power distribution equipment includes the following modules:

[0008] Edge-end multi-source data acquisition module: Deployed at the edge nodes of power distribution rooms / substations, it includes multiple types of sensors and data acquisition terminals. Current sensors (model ACS712, measurement range 0~500A, accuracy ±2%, sampling frequency 1kHz) and voltage sensors (model LV28-P, measurement range 0~10kV, accuracy ±1%, sampling frequency 1kHz) acquire electrical parameters of the equipment; temperature sensors (model DS18B20, measurement range -20~85℃, accuracy ±0.5℃, sampling frequency 0.1Hz) and humidity sensors (model SHT30, accuracy ±3%RH, sampling frequency 0.1Hz) acquire environmental parameters; switch status sensors... The device uses a GY-31 sensor (with a response time of ≤10ms) and an insulation monitoring sensor (JCY-10, with a measurement range of 0~1000MΩ and an accuracy of ±5%) to collect device status parameters. The data acquisition terminal (equipped with an ARM Cortex-A9 processor, ≥2GB of memory, and ≥32GB of storage) supports Modbus RTU and IEC61850 protocols. It caches the collected data locally (with a cache validity of 24 hours), and uploads it to the edge preprocessing module after AES-256 encryption. The integrity of the collected data is ≥99.5%.

[0009] Edge data preprocessing module: includes a data cleaning unit, a data standardization unit, and a data compression unit. The data cleaning unit uses a "rule filtering + isolated forest algorithm". The rule layer removes obvious outliers such as "current < 0A" and "voltage > 12kV" (removal rate ≤ 1%), while the algorithm layer (100 trees, 256 sample subsets) identifies hidden anomalies such as "temperature surge > 5℃ / min" (identification accuracy ≥ 95%). The data standardization unit uses Z-Score standardization for electrical parameters (mapping current and voltage to the [-1,1] range) and one-hot encoding for status parameters (e.g., "switch open = 01, switch closed = 10"). The data compression unit uses the LZ4 compression algorithm (compression ratio 3:1, compression time ≤ 10ms / data) to reduce data transmission volume. The data error after compression is ≤ 0.5%. The preprocessed data is marked in the format of "device ID - acquisition time" and simultaneously uploaded to the cloud and local backup.

[0010] The cloud-based data storage and management module adopts a "distributed HDFS cluster + cloud object storage" architecture. The distributed cluster (≥5 nodes, single-node computing power ≥4 TOPS, storage capacity ≥100TB) stores high-frequency interaction data from the past year (read / write speed ≥1GB / s, IO latency ≤10ms), supporting tiered data storage by "region-device type-time" (e.g., "East China-10kV switchgear-202406"). The cloud object storage (compatible with S3 protocol, storage capacity ≥1PB) stores historical data (retention period ≥3 years, supports cold backup), employing a RAID5 redundancy strategy (fault tolerance of 1 node), with a data loss rate ≤10%. -9 / year; The data management unit implements data indexing (index update frequency 1min / time), data retrieval (retrieval response time ≤1s) and data desensitization (desensitization processing of device IP and location information), and connects to the power distribution management system (PMS) and the power consumption information collection system to synchronize equipment ledgers (update cycle 1h / time) and power load data.

[0011] The cloud-edge collaborative scheduling module includes a task allocation unit and a bandwidth adaptation unit. The task allocation unit, based on the principle of "edge priority, cloud supplementation," allocates tasks with high real-time requirements (such as overload warnings and switch control, with a response time ≤100ms) to the edge, and assigns complex computing tasks (such as monthly risk trend analysis and annual equipment health assessments) to the cloud. The bandwidth adaptation unit monitors the cloud-edge transmission bandwidth (monitoring frequency 1s / time). When the bandwidth is <50Mbps, it automatically reduces the transmission frequency of non-critical data (such as ambient humidity) (from 0.1Hz to 0.05Hz). When the bandwidth is ≥200Mbps, it improves the transmission quality of video surveillance data (such as power distribution room cameras, 1080P resolution, 25fps frame rate). It also supports local autonomy during edge disconnection (the edge independently executes control tasks after disconnection, and incremental data is synchronized to the cloud after connection recovery), maintaining system continuity.

[0012] Furthermore, it also includes:

[0013] Multi-source data fusion module: Includes a homogeneous data fusion unit and a heterogeneous data fusion unit. The homogeneous data fusion unit fuses multi-sensor data from the same device (e.g., the same phase current collected by three current sensors) using a weighted average method. The weights are dynamically allocated based on the sensor accuracy (sensors with ±1% accuracy have a weight of 0.6, and those with ±2% accuracy have a weight of 0.4), with a fusion error ≤1%. The heterogeneous data fusion unit uses a federated learning algorithm (≥10 clients, 50 iterations, learning rate 0.01) to fuse electrical parameters, environmental parameters, and state parameters, mining the correlation features of "current overload + temperature rise + insulation degradation". The feature dimension after fusion is controlled at 50~80 dimensions, providing comprehensive data support for risk assessment, with a fusion time ≤500ms / time.

[0014] The intelligent risk assessment module includes a risk indicator construction unit and a risk level determination unit. The risk indicator construction unit establishes a risk indicator system for power distribution equipment, including electrical risks (overload rate = actual current / rated current, voltage deviation rate = |actual voltage - rated voltage| / rated voltage), equipment condition risks (temperature exceedance rate = |actual temperature - rated temperature| / rated temperature, insulation degradation rate = (initial insulation value - current insulation value) / initial insulation value), and environmental risks (humidity exceedance rate = (actual humidity - rated humidity) / rated humidity, dust concentration exceedance rate). The risk level determination unit classifies risks into four levels: Level I (no risk, risk value < 0.3), Level II (low risk, 0.3 ≤ risk value < 0.5), Level III (medium risk, 0.5 ≤ risk value < 0.8), and Level IV (high risk, risk value ≥ 0.8). Risk value calculation incorporates equipment importance weights (e.g., main power transformer weight 1.2, branch circuit breaker weight 0.8), with an assessment cycle ≤ 1 minute and an assessment accuracy ≥ 92%.

[0015] Safety control execution module: Includes automatic control unit and manual intervention unit. The automatic control unit triggers preset measures for Level II and Level III risks: In case of electrical overload, it activates the load transfer device (response time ≤ 500ms) to transfer the overload circuit load to the backup circuit; in case of excessive temperature, it activates the cooling fan / air conditioner (control accuracy ±1℃); in case of insulation degradation, it disconnects the corresponding circuit breaker (opening time ≤ 200ms). The manual intervention unit generates intervention suggestions for Level IV risks (e.g., "Recommend maintenance of the 10kV main transformer, insulation value has dropped to 500MΩ"), providing a remote control entry point (supports Web / APP operation, operation log retention ≥ 1 year) and a local control interface (anti-misoperation lock, requires dual-user authorization); after control execution, it provides real-time feedback on the execution status (e.g., "Load transfer completed, current current 200A < rated 300A"), forming a control closed loop.

[0016] Security Protection Module: Includes a data security unit, a device security unit, and an access security unit. The data security unit employs "transmission encryption (AES-256) + storage encryption (SM4)" to prevent data leakage; the device security unit implements power distribution equipment identity authentication (based on the national cryptographic SM2 algorithm, with an authentication success rate ≥99.8%), rejecting unauthorized device access; the access security unit adopts the RBAC permission model, classifying users into administrators (full permissions), maintenance personnel (control + monitoring permissions), and viewers (monitoring permissions only). Operations require secondary verification (password + verification code), and abnormal access (such as login from a different location or multiple incorrect passwords) triggers account locking (lock duration 30 minutes); simultaneously, an intrusion detection system (IDS, detection rate ≥95%, false alarm rate ≤1%) is deployed to prevent network attacks.

[0017] Human-Computer Interaction Module: Includes a visualization unit and an alarm push unit. The visualization unit supports access via Web / APP and provides a power distribution network topology map (real-time updates of equipment status, red indicates faults, green indicates normal), data trend charts (such as the current change curve of the past 24 hours, the temperature change curve of the past 7 days), and a risk heat map (colored by region / equipment risk value, the darker the color, the higher the risk). The alarm push unit supports SMS (delivery rate ≥98%, response time ≤10s), WeChat / DingTalk (text and image alarms, including risk details and handling suggestions), and audible and visual alarms (deployed in the operation and maintenance center, sound pressure level ≥100dB, flashing frequency 1Hz). Level I risks are only logged, Level II risks are pushed to the operation and maintenance team, Level III risks are pushed to the operation and maintenance manager, and Level IV risks are pushed to senior management and the power regulatory department.

[0018] The data quality assessment unit calculates data credibility using multi-dimensional indicators, selecting high-quality data for fusion and evaluation to avoid risky misjudgments caused by low-quality data. Data credibility comprehensively considers data completeness, timeliness, and consistency; the calculation formula is as follows:

[0019]

[0020] In the formula, The overall credibility of the data (dimensionless, range [0,1], R≥0.8 is used to determine high-quality data). The integrity weight is 0.4 (dimensionless, as integrity is critical to electrical data), and C is the number of fields actually included in the data (e.g., current data should include 3 fields: "sample value, timestamp, and sensor ID"). The total number of standard data fields (3 for current data). The timeliness weight is dimensionless and has a value of 0.35. The time decay coefficient (h) -1 The value is 0.2, and the higher the coefficient, the faster the data's timeliness decays over time. The data upload time (h, e.g., 15:00 on June 10, 2024). The data collection time (h, e.g., 14:30 on June 10, 2024). This is the consistency weight (dimensionless, value 0.25). The number of consistent data from multiple sensors on the same device (e.g., 2 out of 3 current sensors have consistent data). This represents the total number of sensors for the same device (3 in this case).

[0021] Furthermore, it includes a dynamic risk threshold adjustment unit. This unit adjusts the risk assessment indicator thresholds based on changes in power distribution load, equipment operating years, and seasonal factors to avoid false alarms or missed alarms caused by fixed thresholds. The adjustment criteria include: during peak power distribution load periods (such as summer peak electricity consumption, load rate ≥90%), the current overload threshold is increased from "1.2 times rated current" to "1.3 times rated current", and the voltage deviation threshold is relaxed from "±5%" to "±7%"; when the equipment operating years exceed 10 years, the rated temperature value is reduced by 10% (e.g., the original rated temperature of 60℃ is adjusted to 54℃), and the rated insulation value is reduced by 20% (e.g., the original rated insulation value of 1000MΩ is adjusted to 800MΩ); during humid seasons (relative humidity ≥80%), the humidity exceeding threshold is reduced from "85%RH" to "80%RH", while the insulation monitoring frequency is increased (from 0.1Hz to 0.2Hz). The risk threshold dynamic adjustment unit generates a threshold adjustment report every 24 hours, recording the basis for the adjustment, the threshold before and after the adjustment, and the expected impact. The report takes effect after being reviewed by the operation and maintenance experts (threshold rollback is triggered when the review pass rate is less than 80%), which is in line with the actual operating conditions.

[0022] Furthermore, the edge data preprocessing module also includes a data anomaly tracing unit. This unit analyzes the generation nodes and related data of abnormal data to pinpoint the cause of the anomaly, providing a basis for subsequent data quality optimization and equipment maintenance. The anomaly tracing process is as follows: After the data cleaning unit identifies abnormal data, the anomaly tracing unit first checks the status of the edge acquisition module (such as sensor power supply voltage and communication link signal strength). If the power supply voltage is <11V (standard 12V) or the signal strength is <-80dBm, it is determined to be "acquisition hardware abnormality". If the acquisition hardware is normal, it further compares the related data in the same time period (such as checking whether the voltage and power data are synchronously abnormal when the current is abnormal). If all related data are abnormal, it is determined to be "equipment operation abnormality". If only a single data is abnormal, it is determined to be "data transmission abnormality" (such as packet loss or bit error). The anomaly tracing unit generates an anomaly tracing report every hour, and statistically analyzes the distribution of anomaly types (percentage of hardware anomalies, percentage of equipment operation anomalies, percentage of transmission anomalies) and the IDs of high-frequency abnormal devices. The report is synchronized to the operation and maintenance module to formulate targeted operation and maintenance plans (such as arranging the replacement of sensors with hardware anomalies and arranging the maintenance of power distribution cabinets with equipment operation anomalies), so that the occurrence rate of abnormal data is reduced from 5% to below 1%.

[0023] Furthermore, the intelligent risk assessment module also includes a risk value correction unit. This unit combines historical risk management results with real-time operating conditions to correct the initially calculated risk value, improving the accuracy of the risk assessment. The risk value correction is based on historical data to construct a correction coefficient, calculated using the following formula:

[0024]

[0025] In the formula, The corrected risk value (dimensionless, range [0,1]). This is a preliminary calculated risk value (dimensionless, same as the initial risk value). Historical risk impact coefficient (dimensionless, value 0~0.3, 0.3 if a certain equipment has experienced a level IV risk in the past year, and 0 if there is no risk record). Historical risk frequency (times / year, e.g., if a certain device has experienced two Level III risks in the past year, H=2). The real-time operating condition optimization coefficient (dimensionless, value 0~0.2, 0.2 when the equipment is within 1 month after maintenance or when the operating environment meets the standards, otherwise 0). The compliance rate of operating conditions (% / 100, such as when the current temperature, humidity and insulation of the equipment meet the standards, G=1, and when one item fails to meet the standard, G=0.67).

[0026] Furthermore, the safety control execution module also includes a control effectiveness evaluation unit. This unit assesses the effectiveness of control measures by comparing changes in equipment parameters before and after control execution, providing a basis for subsequent control strategy optimization. The evaluation cycle is divided into "short-term evaluation (5 minutes after control execution)" and "long-term evaluation (24 hours after control execution)". Evaluation indicators include: parameter recovery rate = (parameter after control - parameter before control) / (rated parameter - parameter before control), where: a negative result indicates that the deviation of the parameter from the rated parameter after control has intensified; a positive result indicates that the parameter after control has recovered to the rated parameter; risk value reduction rate = (risk value before control - risk value after control) / risk value before control; secondary risk occurrence rate = number of new risks caused by control measures / total number of control measures. For example, regarding control measures for "current overload (350A before control, 300A rated)," a short-term assessment shows the current recovers to 280A, with a parameter recovery rate of (280-350) / (300-350) = (-70) / (-50) = 1.4 (over-rated recovery, deemed "effective"). The risk value decreases from 0.6 (Level III) to 0.35 (Level II), with a risk value reduction rate of (0.6-0.35) / 0.6 ≈ 0.417, and no secondary risks (occurrence rate 0%). A long-term assessment shows the current stabilizes at 280~290A within 24 hours, with the risk value stabilizing at 0.35~0.4, deemed "long-lasting." The control effectiveness assessment results are fed back to the intelligent risk assessment module in real time. "Ineffective" measures trigger control strategy optimization (such as adjusting load transfer or changing the model of heat dissipation equipment), while "long-lasting" measures are included in the best practice library for power distribution equipment control.

[0027] Furthermore, the cloud-edge collaborative scheduling module also includes an edge node load balancing unit. This unit dynamically allocates acquisition and preprocessing tasks by monitoring the CPU utilization, memory usage, and data processing volume of each edge node, avoiding data loss or delay caused by overload of a single node. The load balancing strategy is as follows: edge node load status is divided into "light load (CPU < 50%, memory < 40%)", "medium load (CPU 50%~70%, memory 40%~60%)", and "heavy load (CPU > 70%, memory > 60%)". When a node is under heavy load, the data acquisition tasks of some non-critical equipment under its jurisdiction (such as branch circuit breakers) are transferred to adjacent lightly loaded nodes (sampling frequency is reduced from 1kHz to 0.5kHz). When multiple nodes are under heavy load at the same time, cloud-assisted preprocessing is initiated (the cloud shares 30%~50% of the preprocessing tasks, such as data compression and anomaly identification). The edge node load balancing unit monitors the load status every 30 seconds, with task transfer time ≤1 second, maintaining the load of each node below medium load (heavy load node ratio ≤5%), data processing latency ≤100ms, and reducing data loss due to node overload (loss rate reduced from 3% to below 0.5%).

[0028] Furthermore, the multi-source data fusion module also includes a fusion task scheduling optimization unit. This unit optimizes task execution order and resource allocation by calculating the time complexity and resource requirements of fusion tasks, thereby improving fusion efficiency. The formula for calculating the priority of fusion task scheduling is as follows:

[0029]

[0030] In the formula, , where is the task scheduling priority (dimensionless, the larger the value, the higher the priority, and the priority is executed first), D is the task urgency weight (dimensionless, risk assessment related tasks take 8, equipment fault diagnosis takes 5, and routine data statistics take 2), and B is the cloud-edge transmission bandwidth (Mbps, real-time monitoring value). The task data size (MB). Reflects data transmission efficiency (the larger the value, the higher the efficiency), P is the computing power of the execution node (TOPS, 2 for edge nodes, 10 for the cloud), and M is the task data dimension (dimension). Reflects computational efficiency (the larger the value, the higher the efficiency).

[0031] Furthermore, the security protection module also includes a device identity blockchain authentication unit. This unit uses consortium blockchain technology to achieve trusted authentication of power distribution equipment, preventing unauthorized devices from forging identities to access the system. The blockchain network consists of power distribution management center nodes, edge nodes, and equipment manufacturer nodes (≥5 nodes, PBFT consensus mechanism, consensus latency ≤1s). Each power distribution device generates a unique identifier (device ID + national cryptographic SM2 public / private key pair) upon leaving the factory. The public key and equipment ledger information (model, production date, rated parameters) are uploaded to the blockchain for evidence storage (the stored data is tamper-proof, with a 100% tamper detection rate). When a device accesses the system, it must send an identity authentication request (including device ID, signature timestamp, and SM2 signature value) to the edge node. The edge node verifies the signature (verification time ≤50ms) and queries the blockchain evidence storage information. If the signature verification passes and the device ledger matches, access is allowed; if the signature verification fails or the ledger does not match, access is rejected and an alarm is triggered (pushed to the operations and maintenance manager).

[0032] Furthermore, the system also includes an equipment health status prediction unit. This unit, based on multi-source fusion data and a machine learning model, predicts the health status of power distribution equipment for the next 1-3 months, enabling early detection of potential faults and achieving preventative maintenance. The health status prediction model employs an LSTM neural network (input layer dimension = fusion feature dimension, 2 hidden layers, 128 and 64 neurons respectively, learning rate 0.001, 300 iterations). Input features include electrical parameters (average / maximum current and voltage), status parameters (temperature and insulation change trends), environmental parameters (humidity and dust concentration statistics), and historical fault records for the past 3 months. The output is an equipment health index (range 0-1, 0 indicating fault and 1 indicating health) and potential fault types (such as "insulation aging", "contact overheating", and "coil overload"). The health status prediction unit generates a prediction report every month, marking equipment with a health index <0.6 as "key concern" and equipment with a health index <0.4 as "urgent maintenance", and pushes maintenance suggestions (such as "10kV switchgear health index 0.35, predicts that contact overheating may occur within 1 month, and it is recommended to replace the contacts").

[0033] Compared with existing technologies, the beneficial effects of this invention are:

[0034] This invention comprehensively overcomes the technical bottlenecks of traditional power distribution equipment management by constructing a full-process cloud-edge collaborative multi-source data fusion security management system, significantly improving the security management capabilities and operation and maintenance efficiency of medium-voltage power distribution networks. At the data processing level, the edge-end multi-source data acquisition module achieves comprehensive collection of electrical, environmental, and status parameters. Combined with a "rule + algorithm" cleaning strategy and standardized processing, various abnormal data are effectively eliminated. Then, through weighted averaging and federated learning fusion technology, the multi-source data fusion error is significantly reduced, providing high-quality, multi-dimensional data support for risk assessment and avoiding assessment bias caused by data quality issues.

[0035] At the risk assessment and control execution level, the dynamic risk threshold adjustment mechanism optimizes assessment standards by incorporating factors such as load, equipment age, and season, making risk level determination more closely aligned with actual operating conditions and reducing false alarms and missed alarms caused by fixed thresholds. The risk value correction unit introduces historical risk and real-time operating condition parameters to further improve assessment accuracy and ensure that the risk level accurately reflects the actual safety status of the equipment. The control execution module not only triggers automatic control and manual intervention measures for different risk levels but also assesses parameter recovery and risk reduction through short-term and long-term effect evaluations. Ineffective measures are optimized in a timely manner, and long-term measures are incorporated into the best practice library, forming a complete closed loop of "assessment-control-assessment-optimization," thereby improving the effectiveness and relevance of control measures.

[0036] At the cloud-edge collaboration and security protection level, edge node load balancing and bandwidth adaptation strategies ensure stable load on each node and efficient data transmission, avoiding data delays and loss due to node overload or insufficient bandwidth. Device identity blockchain authentication and multi-layered data encryption mechanisms effectively prevent unauthorized device access and data leakage, ensuring system operational security. Simultaneously, the device health status prediction unit can detect potential faults in advance, enabling preventative maintenance, reducing the incidence of sudden faults and power outages, and lowering maintenance costs. Overall, this invention realizes the transformation of power distribution equipment from "passive maintenance" to "proactive prevention and control," providing reliable technical support for the safe, stable, and efficient operation of medium-voltage power distribution networks, and offering significant safety and economic benefits. Attached Figure Description

[0037] Figure 1 This is a schematic block diagram of the intelligent power distribution equipment cloud-edge collaborative multi-source data fusion security management and control system proposed in this invention;

[0038] Figure 2 A comparison chart of power distribution data quality under different data processing schemes;

[0039] Figure 3 A comparison chart showing the assessment accuracy of different risk assessment schemes;

[0040] Figure 4A comparison chart of resource scheduling efficiency for different cloud-edge collaboration solutions;

[0041] Figure 5 A comparison chart of system security for different security protection schemes. 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] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0044] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified. Furthermore, the terms "installed," "connected," and "linked" should be interpreted broadly; for example, they may refer to a fixed connection, a detachable connection, or an integral connection; they may refer to a mechanical connection or an electrical connection; they may refer to a direct connection or an indirect connection through an intermediate medium; and they may refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances. The invention will now be described in further detail with reference to the accompanying drawings.

[0045] Reference Figures 1 to 5 A cloud-edge collaborative multi-source data fusion security management and control system for intelligent power distribution equipment includes the following modules:

[0046] Edge-end multi-source data acquisition module: Deployed at the edge nodes of power distribution rooms / substations, it includes multiple types of sensors and data acquisition terminals. Current sensors (model ACS712, measurement range 0~500A, accuracy ±2%, sampling frequency 1kHz) and voltage sensors (model LV28-P, measurement range 0~10kV, accuracy ±1%, sampling frequency 1kHz) acquire electrical parameters of the equipment; temperature sensors (model DS18B20, measurement range -20~85℃, accuracy ±0.5℃, sampling frequency 0.1Hz) and humidity sensors (model SHT30, accuracy ±3%RH, sampling frequency 0.1Hz) acquire environmental parameters; switch status sensors... The device uses a GY-31 sensor (with a response time of ≤10ms) and an insulation monitoring sensor (JCY-10, with a measurement range of 0~1000MΩ and an accuracy of ±5%) to collect device status parameters. The data acquisition terminal (equipped with an ARM Cortex-A9 processor, ≥2GB of memory, and ≥32GB of storage) supports Modbus RTU and IEC61850 protocols. It caches the collected data locally (with a cache validity of 24 hours), and uploads it to the edge preprocessing module after AES-256 encryption. The integrity of the collected data is ≥99.5%.

[0047] Edge data preprocessing module: includes a data cleaning unit, a data standardization unit, and a data compression unit. The data cleaning unit uses a "rule filtering + isolated forest algorithm". The rule layer removes obvious outliers such as "current < 0A" and "voltage > 12kV" (removal rate ≤ 1%), while the algorithm layer (100 trees, 256 sample subsets) identifies hidden anomalies such as "temperature surge > 5℃ / min" (identification accuracy ≥ 95%). The data standardization unit uses Z-Score standardization for electrical parameters (mapping current and voltage to the [-1,1] range) and one-hot encoding for status parameters (e.g., "switch open = 01, switch closed = 10"). The data compression unit uses the LZ4 compression algorithm (compression ratio 3:1, compression time ≤ 10ms / data) to reduce data transmission volume. The data error after compression is ≤ 0.5%. The preprocessed data is marked in the format of "device ID - acquisition time" and simultaneously uploaded to the cloud and local backup.

[0048] The cloud-based data storage and management module adopts a "distributed HDFS cluster + cloud object storage" architecture. The distributed cluster (≥5 nodes, single-node computing power ≥4 TOPS, storage capacity ≥100TB) stores high-frequency interaction data from the past year (read / write speed ≥1GB / s, IO latency ≤10ms), supporting tiered data storage by "region-device type-time" (e.g., "East China-10kV switchgear-202406"). The cloud object storage (compatible with S3 protocol, storage capacity ≥1PB) stores historical data (retention period ≥3 years, supports cold backup), employing a RAID5 redundancy strategy (fault tolerance of 1 node), with a data loss rate ≤10⁻.9 / year; The data management unit implements data indexing (index update frequency 1min / time), data retrieval (retrieval response time ≤1s) and data desensitization (desensitization processing of device IP and location information), and connects to the power distribution management system (PMS) and the power consumption information collection system to synchronize equipment ledgers (update cycle 1h / time) and power load data.

[0049] The cloud-edge collaborative scheduling module includes a task allocation unit and a bandwidth adaptation unit. The task allocation unit, based on the principle of "edge priority, cloud supplementation," allocates tasks with high real-time requirements (such as overload warnings and switch control, with a response time ≤100ms) to the edge, and assigns complex computing tasks (such as monthly risk trend analysis and annual equipment health assessments) to the cloud. The bandwidth adaptation unit monitors the cloud-edge transmission bandwidth (monitoring frequency 1s / time). When the bandwidth is <50Mbps, it automatically reduces the transmission frequency of non-critical data (such as ambient humidity) (from 0.1Hz to 0.05Hz). When the bandwidth is ≥200Mbps, it improves the transmission quality of video surveillance data (such as power distribution room cameras, 1080P resolution, 25fps frame rate). It also supports local autonomy during edge disconnection (the edge independently executes control tasks after disconnection, and incremental data is synchronized to the cloud after connection recovery), maintaining system continuity.

[0050] Multi-source data fusion module: Includes a homogeneous data fusion unit and a heterogeneous data fusion unit. The homogeneous data fusion unit fuses multi-sensor data from the same device (e.g., the same phase current collected by three current sensors) using a weighted average method. The weights are dynamically allocated based on the sensor accuracy (sensors with ±1% accuracy have a weight of 0.6, and those with ±2% accuracy have a weight of 0.4), with a fusion error ≤1%. The heterogeneous data fusion unit uses a federated learning algorithm (≥10 clients, 50 iterations, learning rate 0.01) to fuse electrical parameters, environmental parameters, and state parameters, mining the correlation features of "current overload + temperature rise + insulation degradation". The feature dimension after fusion is controlled at 50~80 dimensions, providing comprehensive data support for risk assessment, with a fusion time ≤500ms / time.

[0051] The intelligent risk assessment module includes a risk indicator construction unit and a risk level determination unit. The risk indicator construction unit establishes a risk indicator system for power distribution equipment, including electrical risks (overload rate = actual current / rated current, voltage deviation rate = |actual voltage - rated voltage| / rated voltage), equipment condition risks (temperature exceedance rate = |actual temperature - rated temperature| / rated temperature, insulation degradation rate = (initial insulation value - current insulation value) / initial insulation value), and environmental risks (humidity exceedance rate = (actual humidity - rated humidity) / rated humidity, dust concentration exceedance rate). The risk level determination unit classifies risks into four levels: Level I (no risk, risk value < 0.3), Level II (low risk, 0.3 ≤ risk value < 0.5), Level III (medium risk, 0.5 ≤ risk value < 0.8), and Level IV (high risk, risk value ≥ 0.8). Risk value calculation incorporates equipment importance weights (e.g., main power transformer weight 1.2, branch circuit breaker weight 0.8), with an assessment cycle ≤ 1 minute and an assessment accuracy ≥ 92%.

[0052] Safety control execution module: Includes automatic control unit and manual intervention unit. The automatic control unit triggers preset measures for Level II and Level III risks: In case of electrical overload, it activates the load transfer device (response time ≤ 500ms) to transfer the overload circuit load to the backup circuit; in case of excessive temperature, it activates the cooling fan / air conditioner (control accuracy ±1℃); in case of insulation degradation, it disconnects the corresponding circuit breaker (opening time ≤ 200ms). The manual intervention unit generates intervention suggestions for Level IV risks (e.g., "Recommend maintenance of the 10kV main transformer, insulation value has dropped to 500MΩ"), providing a remote control entry point (supports Web / APP operation, operation log retention ≥ 1 year) and a local control interface (anti-misoperation lock, requires dual-user authorization); after control execution, it provides real-time feedback on the execution status (e.g., "Load transfer completed, current current 200A < rated 300A"), forming a control closed loop.

[0053] Security Protection Module: Includes a data security unit, a device security unit, and an access security unit. The data security unit employs "transmission encryption (AES-256) + storage encryption (SM4)" to prevent data leakage; the device security unit implements power distribution equipment identity authentication (based on the national cryptographic SM2 algorithm, with an authentication success rate ≥99.8%), rejecting unauthorized device access; the access security unit adopts the RBAC permission model, classifying users into administrators (full permissions), maintenance personnel (control + monitoring permissions), and viewers (monitoring permissions only). Operations require secondary verification (password + verification code), and abnormal access (such as login from a different location or multiple incorrect passwords) triggers account locking (lock duration 30 minutes); simultaneously, an intrusion detection system (IDS, detection rate ≥95%, false alarm rate ≤1%) is deployed to prevent network attacks.

[0054] Human-Computer Interaction Module: Includes a visualization unit and an alarm push unit. The visualization unit supports access via Web / APP and provides a power distribution network topology map (real-time updates of equipment status, red indicates faults, green indicates normal), data trend charts (such as the current change curve of the past 24 hours, the temperature change curve of the past 7 days), and a risk heat map (colored by region / equipment risk value, the darker the color, the higher the risk). The alarm push unit supports SMS (delivery rate ≥98%, response time ≤10s), WeChat / DingTalk (text and image alarms, including risk details and handling suggestions), and audible and visual alarms (deployed in the operation and maintenance center, sound pressure level ≥100dB, flashing frequency 1Hz). Level I risks are only logged, Level II risks are pushed to the operation and maintenance team, Level III risks are pushed to the operation and maintenance manager, and Level IV risks are pushed to senior management and the power regulatory department.

[0055] This invention also includes a data quality assessment unit. This unit calculates data credibility using multi-dimensional indicators, selects high-quality data for fusion and evaluation, and avoids risky misjudgments caused by low-quality data. Data credibility comprehensively considers data integrity, timeliness, and consistency, and the calculation formula is as follows:

[0056]

[0057] In the formula, The overall credibility of the data (dimensionless, range [0,1], R≥0.8 is used to determine high-quality data). The integrity weight is 0.4 (dimensionless, as integrity is critical to electrical data), and C is the number of fields actually included in the data (e.g., current data should include 3 fields: "sample value, timestamp, and sensor ID"). The total number of standard data fields (3 for current data). The timeliness weight is dimensionless and has a value of 0.35. The time decay coefficient (h) -1 The value is 0.2, and the higher the coefficient, the faster the data's timeliness decays over time. The data upload time (h, e.g., 15:00 on June 10, 2024). The data collection time (h, e.g., 14:30 on June 10, 2024). This is the consistency weight (dimensionless, value 0.25). The number of consistent data from multiple sensors on the same device (e.g., 2 out of 3 current sensors have consistent data). The total number of sensors on the same device (3 in this case). When R < 0.6, the data quality assessment unit automatically triggers the data re-acquisition process (sends a re-acquisition command to the edge acquisition module, with a re-acquisition timeout of 5 minutes). Data that still fails to meet the standards after re-acquisition is marked as "low quality" and excluded from the fused dataset.

[0058] This invention also includes a dynamic risk threshold adjustment unit. This unit adjusts the risk assessment index thresholds based on changes in power distribution load, equipment operating years, and seasonal factors to avoid false alarms or missed alarms caused by fixed thresholds. The adjustment criteria include: during peak power distribution load periods (e.g., summer peak electricity consumption, load rate ≥ 90%), the current overload threshold is increased from "1.2 times rated current" to "1.3 times rated current," and the voltage deviation threshold is relaxed from "±5%" to "±7%"; when equipment operating years exceed 10 years, the rated temperature is reduced by 10% (e.g., the original rated temperature of 60℃ is adjusted to 54℃), and the rated insulation value is reduced by 20% (e.g., the original rated insulation value of 1000MΩ is adjusted to 800MΩ); during humid seasons (relative humidity ≥ 80%), the humidity exceedance threshold is reduced from "85%RH" to "80%RH," while simultaneously increasing the insulation monitoring frequency (from 0.1Hz to 0.2Hz). The risk threshold dynamic adjustment unit generates a threshold adjustment report every 24 hours, recording the basis for the adjustment, the threshold before and after the adjustment, and the expected impact. The report takes effect after being reviewed by the operation and maintenance experts (threshold rollback is triggered when the review pass rate is less than 80%), which is in line with the actual operating conditions.

[0059] In this invention, the edge data preprocessing module also includes a data anomaly tracing unit. This unit analyzes the generation nodes and related data of abnormal data to locate the cause of the anomaly, providing a basis for subsequent data quality optimization and equipment maintenance. The anomaly tracing process is as follows: After the data cleaning unit identifies abnormal data, the anomaly tracing unit first checks the status of the edge acquisition module (such as sensor power supply voltage and communication link signal strength). If the power supply voltage is <11V (standard 12V) or the signal strength is <-80dBm, it is determined to be "acquisition hardware abnormality". If the acquisition hardware is normal, it further compares the related data in the same time period (such as checking whether the voltage and power data are synchronously abnormal when the current is abnormal). If all related data are abnormal, it is determined to be "equipment operation abnormality". If only a single data is abnormal, it is determined to be "data transmission abnormality" (such as packet loss or bit error). The anomaly tracing unit generates an anomaly tracing report every hour, and statistically analyzes the distribution of anomaly types (percentage of hardware anomalies, percentage of equipment operation anomalies, percentage of transmission anomalies) and the IDs of high-frequency abnormal devices. The report is synchronized to the operation and maintenance module to formulate targeted operation and maintenance plans (such as arranging the replacement of sensors with hardware anomalies and arranging the maintenance of power distribution cabinets with equipment operation anomalies), so that the occurrence rate of abnormal data is reduced from 5% to below 1%.

[0060] In this invention, the intelligent risk assessment module also includes a risk value correction unit. This unit combines historical risk management results with real-time operating conditions to correct the initially calculated risk value, thereby improving the accuracy of the risk assessment. The risk value correction is based on historical data to construct a correction coefficient, and the calculation formula is as follows:

[0061]

[0062] In the formula, The corrected risk value (dimensionless, range [0,1]). This is a preliminary calculated risk value (dimensionless, the same as the risk value mentioned above). Historical risk impact coefficient (dimensionless, value 0~0.3, 0.3 if a certain equipment has experienced a level IV risk in the past year, and 0 if there is no risk record). Historical risk frequency (times / year, e.g., if a certain device has experienced two Level III risks in the past year, H=2). The real-time operating condition optimization coefficient (dimensionless, value 0~0.2, 0.2 when the equipment is within 1 month after maintenance or when the operating environment meets the standards, otherwise 0). The compliance rate of operating conditions (% / 100, such as when the current temperature, humidity and insulation of the equipment meet the standards, G=1, and when one item fails to meet the standard, G=0.67). For example, if a main transformer has an initial risk value of F=0.45 (Level II), no risk record in the past year (α=0, H=0), and all current operating conditions are met (β=0.2, G=1), then F'=0.45×(1+0-0.2×1)=0.36 (still Level II); if the main transformer experiences one Level IV risk in the past year (α=0.3, H=1), and one operating condition is not met (β=0.2, G=0.67), then F'=0.45×(1+0.3×1-0.2×0.67)=0.45×(1+0.3-0.134)=0.45×1.166≈0.52 (upgraded to Level III). The corrected risk level better reflects the actual risk status of the equipment.

[0063] In this invention, the safety control execution module also includes a control effect evaluation unit. This unit evaluates the effectiveness of control measures by comparing changes in equipment parameters before and after control execution, providing a basis for subsequent control strategy optimization. The evaluation cycle is divided into "short-term evaluation (5 minutes after control execution)" and "long-term evaluation (24 hours after control execution)". Evaluation indicators include: parameter recovery rate = (parameter after control - parameter before control) / (rated parameter - parameter before control), where: a negative result indicates that the deviation of the parameter from the rated parameter after control has intensified; a positive result indicates that the parameter after control has recovered to the rated parameter; risk value reduction rate = (risk value before control - risk value after control) / risk value before control; secondary risk occurrence rate = number of new risks caused by control measures / total number of control measures. For example, regarding control measures for "current overload (350A before control, 300A rated)," a short-term assessment shows the current recovers to 280A, with a parameter recovery rate of (280-350) / (300-350) = (-70) / (-50) = 1.4 (over-rated recovery, deemed "effective"). The risk value decreases from 0.6 (Level III) to 0.35 (Level II), with a risk value reduction rate of (0.6-0.35) / 0.6 ≈ 0.417, and no secondary risks (occurrence rate 0%). A long-term assessment shows the current stabilizes at 280~290A within 24 hours, with the risk value stabilizing at 0.35~0.4, deemed "long-lasting." The control effectiveness assessment results are fed back to the intelligent risk assessment module in real time. "Ineffective" measures trigger control strategy optimization (such as adjusting load transfer or changing the model of heat dissipation equipment), while "long-lasting" measures are included in the best practice library for power distribution equipment control.

[0064] In this invention, the cloud-edge collaborative scheduling module also includes an edge node load balancing unit. This unit dynamically allocates acquisition and preprocessing tasks by monitoring the CPU utilization, memory usage, and data processing volume of each edge node, avoiding data loss or delay caused by overload of a single node. The load balancing strategy is as follows: the edge node load status is divided into "light load (CPU < 50%, memory < 40%)", "medium load (CPU 50%~70%, memory 40%~60%)", and "heavy load (CPU > 70%, memory > 60%)". When a node is under heavy load, the data acquisition tasks of some non-critical equipment under its jurisdiction (such as branch circuit breakers) are transferred to adjacent lightly loaded nodes (the sampling frequency is reduced from 1kHz to 0.5kHz). When multiple nodes are under heavy load at the same time, cloud-assisted preprocessing is initiated (the cloud shares 30%~50% of the preprocessing tasks, such as data compression and anomaly identification). The edge node load balancing unit monitors the load status every 30 seconds, with task transfer time ≤1 second, maintaining the load of each node below medium load (heavy load node ratio ≤5%), data processing latency ≤100ms, and reducing data loss due to node overload (loss rate reduced from 3% to below 0.5%).

[0065] In this invention, the multi-source data fusion module further includes a fusion task scheduling optimization unit. This unit optimizes the task execution order and resource allocation by calculating the time complexity and resource requirements of the fusion tasks, thereby improving fusion efficiency. The formula for calculating the fusion task scheduling priority is as follows:

[0066]

[0067] In the formula, , where is the task scheduling priority (dimensionless, the larger the value, the higher the priority, and the priority is executed first), D is the task urgency weight (dimensionless, risk assessment related tasks take 8, equipment fault diagnosis takes 5, and routine data statistics take 2), and B is the cloud-edge transmission bandwidth (Mbps, real-time monitoring value). The task data size (MB). Reflects data transmission efficiency (the larger the value, the higher the efficiency), P is the computing power of the execution node (TOPS, 2 for edge nodes, 10 for the cloud), and M is the task data dimension (dimension). This reflects computational efficiency (the higher the value, the higher the efficiency). For example, in a risk assessment fusion task (D=8, B=100Mbps, L=50MB, P=2TOPS, M=50 dimensions), then... For a typical data statistics task (D=2, B=100Mbps, L=100MB, P=10TOPS, M=10 dimensions), then... Risk assessment tasks have a higher priority and are executed first to ensure the real-time nature of risk assessments and avoid risk delays caused by task queuing.

[0068] In this invention, the security protection module also includes a device identity blockchain authentication unit. This unit uses consortium blockchain technology to achieve trusted authentication of power distribution equipment, preventing unauthorized devices from forging identities to access the system. The blockchain network consists of power distribution management center nodes, edge nodes, and equipment manufacturer nodes (number of nodes ≥ 5, consensus mechanism adopts PBFT, consensus latency ≤ 1s). Each power distribution device generates a unique identity identifier (device ID + national cryptographic SM2 public / private key pair) upon leaving the factory. The public key and equipment ledger information (model, production date, rated parameters) are uploaded to the blockchain for evidence storage (the evidence storage data is tamper-proof, with a tamper detection rate of 100%). When a device accesses the system, it needs to send an identity authentication request (including device ID, signature timestamp, and SM2 signature value) to the edge node. The edge node verifies the signature (verification time ≤ 50ms) and queries the blockchain evidence storage information. If the signature verification is successful and the device ledger matches, access is allowed; if the signature verification fails or the ledger does not match, access is rejected and an alarm is triggered (pushed to the operation and maintenance manager). The device identity blockchain authentication unit updates the device identity information quarterly (e.g., the status is updated to "normal" after device maintenance, and "cancelled" for scrapped devices) to ensure the timeliness of identity authentication. The interception rate of unauthorized device access is ≥99.9%, avoiding control command tampering or data leakage caused by device forgery.

[0069] In this invention, the system also includes an equipment health status prediction unit. This unit, based on multi-source fusion data and a machine learning model, predicts the health status of power distribution equipment for the next 1-3 months, detects potential faults in advance, and enables preventative maintenance. The health status prediction model uses an LSTM neural network (input layer dimension = fusion feature dimension, 2 hidden layers, 128 and 64 neurons respectively, learning rate 0.001, 300 iterations). Input features include electrical parameters (average / maximum current and voltage), status parameters (temperature, insulation change trend), environmental parameters (humidity, dust concentration statistics), and historical fault records for the past 3 months. The output is an equipment health index (range 0-1, 0 indicates fault, 1 indicates health) and potential fault types (such as "insulation aging", "contact overheating", "coil overload"). The health status prediction unit generates a monthly forecast report, marking equipment with a health index <0.6 as "high-priority concern" and equipment with a health index <0.4 as "urgent maintenance," and pushes maintenance suggestions (e.g., "10kV switchgear health index 0.35, predicted contact overheating within 1 month, contact replacement recommended"). Through this unit, the early detection rate of power distribution equipment faults has increased to over 85%, the incidence of sudden faults has decreased from 10% to below 3%, maintenance costs have decreased by 20%~30%, and the power outage time caused by sudden faults has been reduced (from an average of 8 hours / time to 2 hours / time).

[0070] Example 1

[0071] Safety management and control of 10kV distribution network in urban core area (high load fluctuation scenario)

[0072] This embodiment addresses the challenges of managing a 10kV power distribution network (including 20 distribution rooms, 120 switchgear, 30 10kV / 0.4kV transformers, and 80 circuit breakers) in a city's CBD by applying a cloud-edge collaborative multi-source data fusion security management and control system for intelligent power distribution equipment. This system solves the management and control problems in the area, which include "large fluctuations in power load (92% load rate during the morning peak and 95% during the evening peak), dense equipment (an average of 6 switchgear per distribution room), and high requirements for operation and maintenance response."

[0073] I. Hardware Deployment and Basic Parameters

[0074] Edge deployment: One edge data acquisition terminal is deployed in each power distribution room. An ACS712 current sensor, an LV28-P voltage sensor, and a GY-31 switch status sensor are installed inside the switch cabinet. A DS18B20 temperature sensor (deployed on top of the switch cabinet and near the transformer, a total of 3 locations) and an SHT30 humidity sensor (1 location) are also installed in the power distribution room. A JCY-10 insulation monitoring sensor is installed at the transformer neutral point. The edge terminal supports Modbus RTU and IEC61850 protocols, with a 24-hour buffering time. Collected data is encrypted with AES-256 and uploaded via a 5G private network (transmission rate 150Mbps, latency 35ms).

[0075] Cloud Deployment: The cloud uses a 5-node HDFS cluster (8 TOPS computing power per node, 200TB storage) to store nearly one year's worth of data (read / write speed 1.2GB / s); cloud object storage (2PB capacity, S3 compatible) stores historical data with RAID5 redundancy; it connects to the city's power distribution management system (PMS) and electricity information collection system, synchronizing equipment ledgers (such as switch cabinet model KYN28-12, transformer rated capacity 1000kVA) and load data (morning peak 9:00-11:00, evening peak 18:00-20:00) every hour.

[0076] Security protection configuration: AES-256 encryption is used for transmission between edge terminals and the cloud, and SM4 encryption is used for cloud storage; IDS intrusion detection system is deployed (96% detection rate, 0.8% false alarm rate); device identity authentication uses SM2 algorithm, and the blockchain nodes include the power distribution management center, 5 edge nodes, and equipment manufacturers (a total of 7 nodes, PBFT consensus latency of 0.8s).

[0077] II. Full-process technology implementation

[0078] Multi-source data acquisition and preprocessing: During the morning rush hour at 9:30, the current sensor of phase A of a switchgear collected data of 320A (rated 300A), the voltage sensor collected line voltage of 10.5kV (rated 10kV), the temperature sensor collected 48℃ (rated 60℃), and the humidity was 65%RH. After the edge terminal cached the data, the data cleaning unit removed one obvious outlier value of "current -5A", and the isolated forest algorithm identified one latent outlier of "temperature surge of 6℃ / min" (corresponding to transformer overload). The current data was standardized using Z-Score: x'=(320-280) / 50=0.8 (280A is the average value of the past hour, and 50A is the standard deviation). The switch status "closed" was coded as 10. The LZ4 compression algorithm compressed 100 data points from 100KB to 32KB (compression ratio 3.1:1) with an error of 0.3%.

[0079] Data quality assessment and fusion: The reliability of current data is calculated according to the formula: C=3 (including sampled value, timestamp, sensor ID), C_t=3, w_1=0.4; t-t0=0.5h, k=0.2, w_2=0.35; two of the three current sensors have consistent data (S=2, S_r=3, w_3=0.25), then R=0.4×(3 / 3)+0.35×e^{-0.2×0.5}+0.25×(2 / 3)=0.4+0.35×0.9048+0.25×0.6667≈0.4+0.3167+0.1667≈0.883 (≥0.8, high quality). When the same source is fused, the voltage sensor with an accuracy of ±1% has a weight of 0.6 and the current sensor with an accuracy of ±2% has a weight of 0.4. After fusion, the current of phase A is 318A (error 0.6%). When the different source is fused, federated learning (10 edge clients, 50 iterations) is used to mine the correlation features of "current 318A (overload) + temperature 48℃ (rising) + insulation 900MΩ (normal)" and generate a 60-dimensional feature vector.

[0080] Risk assessment and threshold adjustment: Risk index calculation: Overload rate = 318 / 300≈1.06, Voltage deviation rate = |10.5-10| / 10=5%, Temperature exceedance rate = |48-60| / 60=20%; Due to the morning peak load rate of 92% (≥90%), the risk threshold dynamic adjustment unit increases the current overload threshold from 1.2 times to 1.3 times and relaxes the voltage deviation threshold from ±5% to ±7%. The initial risk value F = (1.06 × 0.3 + 5% × 0.2 + 20% × 0.3) × 1.2 (switch cabinet weight) = (0.318 + 0.01 + 0.06) × 1.2 ≈ 0.388 × 1.2 ≈ 0.466 (Level II); the switch cabinet has no risk in the past year (α = 0, H = 0), and the current humidity meets the standard (β = 0.2, G = 1), then F' = 0.466 × (1 + 0 - 0.2 × 1) = 0.466 × 0.8 ≈ 0.373 (still Level II).

[0081] Control Execution and Effectiveness Evaluation: The safety control execution module initiates automatic load transfer (response time 450ms), transferring 10% of the load (approximately 30A) of the switchgear to the standby circuit; short-term evaluation (after 5 minutes) shows the current drops to 285A, parameter recovery rate = (285-318) / (300-318) = (-33) / (-18)≈1.83 (over-rated recovery), risk value drops to 0.32; long-term evaluation (after 24 hours) shows the current stabilizes at 280-290A, risk value is 0.3-0.33, it is judged as "long-term effective", and the measures are included in the best practice library. Cloud-edge collaboration and security authentication: Cloud-edge transmission bandwidth monitoring is 80Mbps (≥50Mbps), and the bandwidth adaptation unit maintains the environmental data transmission frequency at 0.1Hz; when the CPU utilization of a certain edge node is 75% (heavy load), the load balancing unit reduces the sampling frequency of the two branch circuit breakers under its jurisdiction from 1kHz to 0.5kHz, and after 30s, the CPU utilization drops to 60% (medium load); when a new switch cabinet is connected, the blockchain authentication unit verifies its SM2 signature (takes 40ms), allows access after matching the ledger information, and intercepts one forged device access request.

[0082] III. Performance Verification Form for Distribution Network in Urban Core Area

[0083]

[0084] The table data is based on one month of continuous operation statistics (averaging 1.2 million data entries per day). Traditional systems suffer from coarse data fusion (simple averaging only), resulting in errors exceeding 5%; risk assessment uses fixed thresholds, leading to a false alarm rate exceeding 8.5%; and there is no effect evaluation after control, resulting in an effectiveness rate of less than 78%. Edge nodes lack load balancing, resulting in an overload ratio exceeding 25% and a data loss rate of 3.2%. Device authentication is simple, resulting in an illegal interception rate of less than 70%. This invention, through multi-dimensional fusion, dynamic thresholds, effect evaluation, and load balancing, reduces the fusion error to below 0.8%, controls the false alarm rate to within 1.8%, achieves a control effectiveness rate exceeding 94%, reduces the edge overload ratio to ≤5%, and achieves an illegal interception rate of 99.9%. It is fully adaptable to high-load fluctuation scenarios in urban core areas, ensuring the reliability of power supply in CBDs (reducing power outage time from an average of 4 hours / month to 0.5 hours / month).

[0085] Example 2

[0086] Safety management of 35kV power distribution network in industrial park (equipment aging scenario)

[0087] This embodiment focuses on a 35kV distribution network in a chemical industrial park (including 8 box-type substations, 40 35kV switchgear, 15 35kV / 10kV main transformers, and 60 circuit breakers). The average service life of the equipment in this park is 12 years (60% of the equipment is over 10 years old), and there are problems such as "rapid equipment aging, high risk of insulation deterioration, and large impact of motor starting load".

[0088] I. Hardware Deployment and Basic Parameters

[0089] Edge Terminal Deployment: Each transformer substation is equipped with one edge terminal (ARM Cortex-A9 processor, 4GB RAM, 64GB storage). The 35kV main transformer high-voltage side is equipped with an ACS712 current sensor (1kHz sampling frequency) and an LV28-P voltage sensor (1kHz sampling frequency). The main transformer itself is equipped with three DS18B20 temperature sensors (top, middle, and bottom layers) and one JCY-10 insulation sensor (0.2Hz sampling frequency, increased to 0.2Hz during humid seasons). A GY-31 switch status sensor is installed in the motor control cabinet, and an SHT30 humidity sensor (monitoring humidity during humid seasons) is installed in the park's power distribution room. The edge terminal cache has a 24-hour validity period, and data is encrypted with AES-256 and uploaded via industrial Ethernet (120Mbps transmission rate, 40ms latency).

[0090] Cloud deployment: 6-node HDFS cluster in the cloud (6 TOPS computing power per node, 150TB storage), with a data read / write speed of 1GB / s for the past year; 1.5PB of object storage in the cloud, retaining 3 years of historical data; connected to the park's energy management system, synchronizing motor start / stop plans (e.g., starting 5 10kV motors at 8:00) and equipment maintenance records every hour.

[0091] Health prediction configuration: The equipment health status prediction unit adopts an LSTM model (60-dimensional input layer, 128 / 64 neurons in the hidden layer, and a learning rate of 0.001). The input features include the average main transformer current (200A), the maximum temperature (55℃), the insulation value change (from 1000MΩ to 850MΩ), and historical faults (1 insulation drop).

[0092] II. Full-process technology implementation

[0093] Data Acquisition and Preprocessing: During the humid season (82%RH), a 35kV main transformer current sensor collected 220A (rated 250A), at a temperature of 52℃ (originally rated 65℃, after 12 years of operation, the threshold was dynamically adjusted to 58℃), and an insulation value of 820MΩ (originally rated 1000MΩ, adjusted to 800MΩ). The edge terminal data cleaning unit removed two outliers of "voltage 38kV" (over 35kV +10%), and the isolated forest algorithm identified one latent anomaly of "insulation value suddenly dropped by 100MΩ / day". The current data was Z-Score standardized: x'=(220-200) / 30≈0.67, and the insulation status "normal" was coded as 10. The LZ4 compressed data was reduced from 80KB to 26KB (compression ratio 3.1:1), with an error of 0.4%.

[0094] Data fusion and health prediction: For homogeneous fusion, the weights of the three temperature sensors (accuracy ±0.5℃) are 0.6, and the weight for ±1℃ is 0.4, resulting in a fused temperature of 51.8℃ (error 0.3%). For heterogeneous fusion, federated learning (8 edge clients, 50 iterations) is used to fuse the features "current 220A + temperature 51.8℃ + insulation 820MΩ + humidity 82%RH" to generate a 70-dimensional vector. The health prediction unit inputs this vector and outputs a health index of 0.38 (<0.4, emergency maintenance), identifying a potential fault as "insulation aging," and recommending "replace the main transformer insulating oil; the insulation value is expected to drop below 800MΩ within one month."

[0095] Risk Assessment and Control: Risk Indicator Calculation: Overload Rate = 220 / 250 = 0.88, Temperature Exceedance Rate = |51.8-58| / 58 ≈ 10.7%, Insulation Deterioration Rate = (1000-820) / 1000 = 18%; Preliminary Risk Value F = (0.88×0.4+10.7%×0.3+18%×0.3)×1.5 (Main Transformer Weight) = (0.352+0.0321+0. 054) × 1.5 ≈ 0.438 × 1.5 ≈ 0.657 (Level III); This main transformer has experienced one Level III risk in the past year (α=0.2, H=1), with humidity exceeding the standard (β=0.2, G=0.67), then F'=0.657×(1+0.2×1-0.2×0.67)=0.657×(1+0.2-0.134)=0.657×1.066≈0.700 (still Level III). The control unit initiates an increase in the insulation monitoring frequency (from 0.2Hz to 0.5Hz) and generates a maintenance work order (maintenance personnel must arrive within 48 hours).

[0096] Control effectiveness and safety protection: After maintenance (replacing insulating oil), short-term assessment (5 min) shows the insulation value rises to 950 MΩ, parameter recovery rate = (950-820) / (1000-820) = 130 / 180≈0.72, and the risk value drops to 0.45; long-term assessment (24 h) shows the insulation stabilizes at 940-950 MΩ, and the health index rises to 0.75 (key focus). The blockchain authentication unit updates the equipment status quarterly, marking the main transformer as "normal after maintenance"; the access security unit intercepts one attempt by non-maintenance personnel to remotely control the circuit breaker (triggers account lockout for 30 min).

[0097] III. Performance Verification Form for Industrial Park Power Distribution Network

[0098]

[0099] The data in the table is based on statistics from two months of continuous operation. Traditional systems lack health prediction, have a fault early detection rate of less than 50%, an insulation degradation identification rate of over 60%, a sudden fault occurrence rate of over 9%, and each fault results in an 8-hour power outage. The lack of preventative maintenance leads to high maintenance costs. This invention, through LSTM health prediction, achieves a fault early detection rate of over 85%, an insulation degradation identification rate of over 96%, reduces sudden faults to below 3%, shortens power outage time by 80%, and reduces maintenance costs by 25%. It is fully adaptable to the aging equipment scenarios in industrial parks, avoiding chemical production interruptions caused by sudden equipment failures and ensuring safe production in the park.

[0100] Figure 2This diagram focuses on the core foundational data quality for power distribution management, validating the comprehensive advantages of the data processing solution of this invention. Traditional manual processing relies on manual screening, resulting in an anomaly detection rate of less than 45%, a fusion error exceeding 6%, and lacks compression capabilities, making it unable to handle the daily processing demands of millions of data entries. This invention, through "rule + isolated forest" cleaning, Z-Score standardization, and weighted fusion, combined with a data quality assessment formula (calculated reliability R≥0.8), increases the anomaly detection rate to 95%, reduces the fusion error to below 1.5%, and achieves a reliability compliance rate exceeding 98%. For example, in urban core area switchgear current data, traditional processing leads to a 15% deviation in risk assessment due to large fusion errors. The solution of this invention has an error of only 1%, and the compression time is reduced to 10ms / entry, reducing cloud-edge transmission volume by 3 times. This provides high-quality data support for subsequent risk assessment and control execution, solving the pain point of "poor data quality leading to management failure" in traditional systems.

[0101] Figure 3 This diagram verifies the accuracy and scenario adaptability of the risk assessment scheme of this invention, solving the problems of "high false alarm rate and lack of dynamic adjustment" in traditional schemes. Traditional fixed threshold schemes do not consider load and equipment age variations, resulting in a false alarm rate exceeding 15%. For example, in an industrial park, an aging main transformer might be mistakenly judged as safe at 52℃ (actually close to fault) because the threshold was not lowered. This invention, through dynamic threshold adjustment (raising the overload threshold during peak load periods and lowering the temperature threshold for aging equipment) and a risk value correction formula, reduces the false alarm rate to below 1.8%, with a dynamic threshold adaptability exceeding 90% and a historical risk correction effect exceeding 85%. For example, in a city's core area, during the morning peak, a switchgear current of 318A (traditional fixed threshold of 1.2 times is considered overload, while this invention raises it to 1.3 times for normal operation) is assessed in just 2 seconds, fully adapting to complex scenarios and ensuring that the risk level accurately reflects the equipment status, avoiding ineffective maintenance due to false alarms and safety hazards caused by missed alarms.

[0102] Figure 4This diagram highlights the resource optimization capabilities of the cloud-edge collaboration solution of this invention, solving the problems of "node overload, high latency, and bandwidth waste" in traditional solutions. Traditional solutions without collaboration suffer from edge node overload exceeding 30% and data loss rate of 7%, making multi-node collaboration impossible. This invention, through edge load balancing (transferring tasks from heavily loaded nodes to lightly loaded nodes), bandwidth adaptation (reducing the frequency of non-critical data transmission during low bandwidth periods), and a task scheduling formula (Tp=D+B / L+P / M, prioritizing risk assessment tasks), reduces the overload percentage to below 5%, reduces transmission latency to 40ms, achieves a data loss rate of ≤1%, bandwidth utilization exceeding 85%, and task scheduling accuracy exceeding 92%. For example, in an industrial park where multiple edge nodes simultaneously process motor start-up load data, the traditional solution has two heavily loaded nodes (80% CPU usage) with a transmission latency of 90ms; the solution of this invention has no heavily loaded nodes, a latency of 35ms, and prioritizes risk assessment tasks, ensuring timely identification of load impact risks and guaranteeing efficient and stable operation of cloud-edge collaboration.

[0103] Figure 5 This diagram verifies the comprehensiveness and reliability of the security protection system of this invention, solving the problems of "weak protection and vulnerability to attack" in traditional solutions. Traditional cryptographic authentication schemes have an illegal device interception rate of less than 70% and a data leakage rate of over 8%, failing to cope with device forgery and data theft. This invention, through AES-256 / SM4 encryption, RBAC permission model, and blockchain authentication of device identity (PBFT consensus, SM2 signature), increases the illegal device interception rate to 99.9%, reduces the data leakage rate to below 0.1%, achieves an abnormal access interception rate of over 95%, and has a 100% blockchain evidence tampering detection rate. For example, in urban power distribution systems, unauthorized personnel attempt to forge switchgear access to the system. Traditional solutions fail to intercept this, leading to the tampering of control commands. This invention's solution verifies device identity through blockchain (signature mismatch), achieving 100% interception of illegal access. Simultaneously, encrypted transmission prevents leakage of current and voltage data, ensuring the safety of the power distribution system and preventing power accidents caused by security vulnerabilities.

[0104] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A cloud-edge collaborative multi-source data fusion security management and control system for intelligent power distribution equipment, characterized in that, Includes the following modules: Edge-end multi-source data acquisition module: Deploys current and voltage sensors on variable edge nodes to acquire electrical parameters of the equipment; temperature and humidity sensors to acquire environmental parameters; switch status sensors and insulation monitoring sensors to acquire equipment status parameters; after data acquisition, it is cached locally, encrypted, and then uploaded to the edge preprocessing module; Edge data preprocessing module: includes data cleaning unit, data standardization unit and data compression unit; The data cleaning unit uses rule filtering and the isolated forest algorithm. The rule layer removes obvious outliers, and the algorithm layer identifies hidden anomalies. The data standardization unit uses Z-Score standardization for electrical parameters and one-hot encoding for status parameters. The data compression unit uses the LZ4 compression algorithm to preprocess the data and marks it in the device ID-acquisition time format. The data is then synchronously uploaded to the cloud and backed up locally. Cloud data storage and management module: adopts a distributed HDFS cluster + cloud object storage rack; the distributed cluster stores high-frequency interaction data from the past year, and supports data tiered storage by region, device type, and time; Historical data is stored in the cloud object storage system, which adopts a RAID5 redundancy strategy. The data management unit realizes data indexing, data retrieval and data anonymization, and also connects to the power distribution management system and the power consumption information collection system to synchronize equipment ledgers and power load data. The cloud-edge collaborative scheduling module includes a task allocation unit and a bandwidth adaptation unit. The task allocation unit is based on the principle of edge priority and cloud supplementation, allocating tasks with high real-time requirements to the edge and complex computing tasks to the cloud. The bandwidth adaptation unit monitors the cloud-edge transmission bandwidth and supports local autonomy when the edge is disconnected.

2. The intelligent power distribution equipment cloud-edge collaborative multi-source data fusion security management and control system according to claim 1, characterized in that, Also includes: Multi-source data fusion module: includes a same-source data fusion unit and a different-source data fusion unit; the same-source data fusion unit uses a weighted average method to fuse data from multiple sensors of the same device, with the weights dynamically allocated according to the sensor accuracy; the different-source data fusion unit uses a federated learning algorithm to fuse electrical parameters, environmental parameters and state parameters, and to mine the correlation features of current overload + temperature rise + insulation decline, with the feature dimension after fusion controlled at 50~80 dimensions; The intelligent risk assessment module includes a risk indicator construction unit and a risk level determination unit. The risk indicator construction unit establishes a risk indicator system for power distribution equipment, including electrical risk, equipment condition risk, and environmental risk. The risk level determination unit classifies risks into Level I, Level II, Level III, and Level IV, and the risk value calculation combines the equipment importance weight. Safety control execution module: includes an automatic control unit and a manual intervention unit; the automatic control unit triggers preset measures for Level II and Level III risks: in case of electrical overload, it activates the load transfer device to transfer the overload circuit load to the backup circuit; in case of excessive temperature, it activates the cooling fan / air conditioner; in case of insulation degradation, it disconnects the circuit breaker of the corresponding circuit; the manual intervention unit generates intervention suggestions for Level IV risks, and provides remote control entry and local control interface; after control execution, the execution status is fed back in real time. Security protection module: includes data security unit, device security unit and access security unit; the data security unit adopts AES-256 transmission encryption + SM4 storage encryption to prevent data leakage; the device security unit realizes the identity authentication of power distribution equipment and refuses unauthorized device access; the access security unit adopts the RBAC permission model, classifying users into administrators, maintenance personnel and viewers, requiring secondary verification for operations, and triggering account lockout for abnormal access; an intrusion detection system is also deployed. Human-computer interaction module: includes a visualization display unit and an alarm push unit; The visualization unit supports access via Web / APP and provides power distribution network topology diagrams, data trend charts, and risk heat maps; The alarm push unit supports SMS, WeChat / DingTalk, and sound and light alarms, and pushes alarms to different levels of personnel according to the risk level; The data quality assessment unit calculates data credibility using multi-dimensional indicators, selects high-quality data for fusion and evaluation, and reduces the risk of misjudgment caused by low-quality data; the calculation formula is as follows: In the formula, To ensure the overall credibility of the data, For integrity weighting, C is the actual number of fields included in the data. The total number of standard data fields. As a weight for timeliness, The time decay coefficient, For data upload time, For data collection time, For consistency weight, The consistency of data from multiple sensors on the same device. The total number of sensors on the same device; when R < 0.6, the data quality assessment unit automatically triggers the data re-collection process. Data that still does not meet the standards after re-collection is marked as "low quality" and excluded from the fused dataset.

3. The intelligent power distribution equipment cloud-edge collaborative multi-source data fusion security management and control system according to claim 1, characterized in that, It also includes a risk threshold dynamic adjustment unit, which adjusts the risk assessment index thresholds based on changes in power distribution load, equipment operating years, and seasonal factors, thereby reducing false alarms or missed alarms caused by fixed thresholds. The adjustment criteria include: during peak power distribution load periods, the current overload threshold will be raised and the voltage deviation threshold will be relaxed; When the equipment has been in operation for more than 10 years, the temperature rating and insulation rating will be lowered. During the humid season, the humidity exceeding threshold will be lowered, and the insulation monitoring frequency will be increased. The risk threshold dynamic adjustment unit will generate a threshold adjustment report every 24 hours, recording the basis for the adjustment, the threshold before and after the adjustment, and the expected impact. The report will take effect after being reviewed by the operation and maintenance experts.

4. The intelligent power distribution equipment cloud-edge collaborative multi-source data fusion security management and control system according to claim 1, characterized in that, The edge data preprocessing module also includes a data anomaly tracing unit. This unit analyzes the generation nodes and related data of abnormal data to locate the cause of the anomaly, providing a basis for subsequent data quality optimization and equipment maintenance. The anomaly tracing process is as follows: After the data cleaning unit identifies abnormal data, the anomaly tracing unit first checks the status of the edge acquisition module. If the power supply voltage is <11V or the signal strength is <-80dBm, it is determined to be an acquisition hardware anomaly. If the acquisition hardware is normal, it further compares the related data in the same time period. If all the related data are abnormal, it is determined to be an equipment operation anomaly. If only a single data is abnormal, it is determined to be a data transmission anomaly. The anomaly tracing unit generates an anomaly tracing report every hour, statistically analyzes the distribution of anomaly types and the IDs of high-frequency abnormal devices, and synchronizes the report to the maintenance module.

5. The intelligent power distribution equipment cloud-edge collaborative multi-source data fusion security management and control system according to claim 1, characterized in that, The intelligent risk assessment module also includes a risk value correction unit. This unit combines historical risk management results with real-time operating conditions to correct the initially calculated risk value. The risk value correction is based on historical data to construct a correction coefficient, and the calculation formula is as follows: In the formula, This is the corrected risk value. This is a preliminary calculation of the risk value. This represents the historical risk impact coefficient. For historical risk frequency, To optimize coefficients for real-time operating conditions, The rate of compliance with operating conditions.

6. The intelligent power distribution equipment cloud-edge collaborative multi-source data fusion security management and control system according to claim 1, characterized in that, The safety control execution module also includes a control effectiveness evaluation unit. By comparing the changes in equipment parameters before and after control execution, the effectiveness of control measures is evaluated. The evaluation cycle is divided into short-term evaluation and long-term evaluation. Evaluation indicators include: parameter recovery rate = (parameter after control - parameter before control) / (rated parameter - parameter before control), where: a negative result indicates that the deviation of the parameter from the rated parameter after control has increased; a positive result indicates that the parameter after control has recovered to the rated parameter; risk value reduction rate = (risk value before control - risk value after control) / risk value before control; secondary risk occurrence rate = number of new risks caused by control measures / total number of control measures.

7. The intelligent power distribution equipment cloud-edge collaborative multi-source data fusion security management and control system according to claim 1, characterized in that, The cloud-edge collaborative scheduling module also includes an edge node load balancing unit, which dynamically allocates collection and preprocessing tasks by monitoring the CPU utilization, memory usage and data processing volume of each edge node. The load balancing strategy is to classify the load status of edge nodes into light load, medium load, and heavy load. When a node is under heavy load, the data acquisition tasks of some non-critical devices under its jurisdiction are transferred to adjacent lightly loaded nodes; when multiple nodes are under heavy load at the same time, cloud-assisted preprocessing is started; the edge node load balancing unit monitors the load status once every 30 seconds, the task transfer time is ≤1 second, and the load of each node is maintained below medium load.

8. The intelligent power distribution equipment cloud-edge collaborative multi-source data fusion security management and control system according to claim 1, characterized in that, The multi-source data fusion module also includes a fusion task scheduling optimization unit. By calculating the time complexity and resource requirements of the fusion task, it optimizes the task execution order and resource allocation. The formula for calculating the fusion task scheduling priority is as follows: In the formula, Here, D represents the task scheduling priority, B represents the task urgency weight, and D represents the cloud-edge transmission bandwidth. For the amount of task data, Reflecting data transmission efficiency, P represents the computing power of the execution node, and M represents the task data dimension. It reflects computational efficiency.

9. The intelligent power distribution equipment cloud-edge collaborative multi-source data fusion security management and control system according to claim 1, characterized in that, The security protection module also includes a device identity blockchain authentication unit, which uses consortium blockchain technology to achieve trusted authentication of power distribution equipment and prevent unauthorized devices from forging identities to access the system. The blockchain network consists of power distribution management center nodes, edge nodes, and equipment manufacturer nodes. Each power distribution device generates a unique identity identifier when it leaves the factory, and the public key and device ledger information are uploaded to the blockchain for storage. When a device accesses the system, it needs to send an identity authentication request to the edge node. The edge node verifies the signature and queries the blockchain evidence information. If the signature verification is successful and the device ledger matches, access is allowed. If the signature verification fails or the ledger does not match, access is rejected and an alarm is triggered.

10. The intelligent power distribution equipment cloud-edge collaborative multi-source data fusion security management and control system according to claim 1, characterized in that, The system also includes an equipment health status prediction unit, which predicts the health status of power distribution equipment for the next 1 to 3 months based on multi-source fusion data and machine learning models, thereby identifying potential faults in advance and enabling preventive operation and maintenance. The health status prediction model uses an LSTM neural network, and the input features include electrical parameters, status parameters, environmental parameters and historical fault records for the past 3 months. The output is the equipment health index and potential fault types. The health status prediction unit generates a prediction report every month.

Images