A power distribution station active monitoring system and method based on scene self-adaptation and multi-target game

Abstract

This invention discloses an active monitoring system and method for power distribution substations based on scenario adaptation and multi-objective game theory. The system includes a scenario-adaptive active detection module, a multi-objective game theory decision-making module, an elastic execution module, and a federated optimization module. The scenario-adaptive active detection module identifies operating scenarios through multi-dimensional data and adaptively adjusts the networking mode and sampling frequency. It adopts a dual-channel design (main and auxiliary) to distinguish between false alarms and real anomalies by comparing data before and after interference, and completes anomaly classification. The multi-objective game theory decision-making module achieves differentiated isolation through risk grading, schedules idle resources to support complex anomaly handling, selects the optimal solution from three types of strategies based on game theory algorithms, and generates control commands. The elastic execution module performs anomaly isolation, resource scheduling, and strategy implementation operations. The federated optimization module aggregates de-identified data based on the FedAvg architecture, generates strategy parameter update commands, and sends them to the multi-objective game theory decision-making module for iterative optimization.

Description

Technical Field

[0001] This invention relates to the field of power system automation monitoring technology, specifically to an active monitoring system and method for substations based on scene adaptation and multi-objective game theory. Background Technology

[0002] As a core hub connecting the generation, transmission, and user sides in the power system, substations undertake crucial functions such as power distribution, voltage conversion, and equipment operation monitoring. Their operational stability directly impacts the continuity of industrial production and the reliability of electricity supply for daily life. With the continuous growth of my country's power load and the increasing penetration rate of new energy power generation, the operating environment of substations is becoming increasingly complex. Factors such as equipment aging, load fluctuations, and environmental interference can all trigger abnormal faults. If these faults are not handled promptly and effectively, they can easily escalate, causing widespread power outages, equipment damage, and other serious consequences. Therefore, higher requirements are placed on the real-time performance, accuracy, and adaptability of substation monitoring systems. However, current mainstream substation monitoring systems still follow traditional design approaches, revealing many unavoidable technical defects in practical applications, and are no longer able to meet the refined operation and maintenance needs under the smart grid context.

[0003] In terms of detection adaptability, power distribution stations exhibit significant scenario differences throughout their entire lifecycle. Typical scenarios include peak load scenarios (such as peak summer electricity consumption and concentrated industrial production periods), maintenance scenarios (equipment maintenance and troubleshooting periods), and standby scenarios (nighttime off-peak load and equipment idle periods). Under different scenarios, the fluctuation patterns of equipment operating parameters, monitoring priorities, and energy consumption control requirements are drastically different: in peak load scenarios, the risks of equipment overheating and insulation aging increase sharply, requiring high-frequency monitoring of core parameters; in maintenance scenarios, flexible coverage of temporary work areas is needed to ensure personnel safety and equipment debugging needs; in standby scenarios, the equipment operating status is stable, requiring a reduction in monitoring frequency to save energy. However, traditional monitoring systems generally use fixed sampling frequencies and networking methods, failing to dynamically adjust to scenario characteristics. This leads to missed anomalies due to insufficient sampling frequency in peak load scenarios, inability to adapt to temporary monitoring needs due to rigid networking in maintenance scenarios, and unnecessary energy waste due to high-frequency sampling in standby scenarios, resulting in an overall imbalance between detection efficiency and resource utilization.

[0004] Regarding the accuracy of anomaly identification, traditional monitoring systems often rely on data collected by a single sensor for anomaly judgment, failing to fully consider the interference from the complex environment of the substation and the impact of sensor malfunctions. Substations contain various interference factors such as electromagnetic radiation, drastic fluctuations in temperature and humidity, and changes in gas concentration, which can lead to data drift and false triggering by sensors. Furthermore, aging and damage to sensors after long-term operation can also cause false anomaly signals. Lacking an effective signal verification mechanism, traditional systems cannot distinguish between sensor false alarms and genuine equipment anomalies, often treating all anomaly signals equally. This results in numerous invalid maintenance work orders, increasing the workload and costs for maintenance personnel. On the other hand, false alarms may mask genuine anomalies, delaying the handling of critical faults and creating safety hazards. For example, when an SF6 concentration sensor shows a high reading due to ambient temperature fluctuations, the system directly identifies it as a sealing fault, triggering emergency procedures and causing maintenance personnel to make unnecessary trips. Conversely, when a sensor experiences distorted partial discharge data due to aging, the system may overlook the genuine anomaly of insulation damage, ultimately leading to serious malfunctions.

[0005] Regarding the rationality of decision-making and handling, traditional monitoring systems lack differentiated design in their anomaly handling strategies. They often adopt a uniform handling process and plan when facing anomalies of different risk levels and types. In reality, anomalies in substations exhibit significant risk gradients. Some anomalies (such as severely excessive partial discharge or a sharp increase in SF6 concentration) are high-risk emergencies requiring immediate isolation and shutdown. Other anomalies (such as slight temperature and humidity deviations or minor equipment noises) are low-risk potential problems that can be addressed through routine monitoring or low-cost intervention. Traditional systems lack a scientific risk classification mechanism. High-risk anomalies may result in delayed responses due to cumbersome handling procedures, while low-risk anomalies may lead to resource waste and unnecessary equipment shutdowns due to over-handling.

[0006] Meanwhile, with the advancement of smart grid construction, the number of substations continues to increase, and their distribution range is becoming increasingly wide. The traditional model relying on manual inspection and centralized monitoring is no longer sufficient to cover all sites, placing higher demands on the automation and intelligence level of monitoring systems. However, the existing monitoring systems have a low degree of automation, and many aspects such as anomaly detection and strategy selection still require manual intervention, which is not only inefficient but may also lead to improper handling due to human error.

[0007] In summary, existing power distribution station monitoring systems have significant shortcomings in terms of detection adaptability, anomaly identification accuracy, and decision-making rationality, making it difficult to meet the actual needs of safe and stable operation and refined maintenance of current power distribution stations. A new monitoring technology solution is urgently needed to solve these problems. Summary of the Invention

[0008] The purpose of this invention is to overcome the shortcomings of the prior art and provide an active monitoring system and method for power distribution substations based on scenario adaptation and multi-objective game theory. This system has strong adaptability, accurate anomaly identification, and efficient decision-making, ensuring the safe and stable operation of power distribution substations and reducing operation and maintenance costs.

[0009] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0010] An active monitoring system for power distribution substations based on scene adaptation and multi-objective game theory is characterized by comprising a scene-adaptive active detection module, a multi-objective game theory decision-making module, an elastic execution module, and a federated optimization module. The scene-adaptive active detection module collects multi-dimensional data from the power distribution substation, performs preliminary anomaly type determination, and outputs detection results. Its determination method involves: first, identifying the operating scenario of the power distribution substation and adaptively adjusting the detection priority and sampling parameters; then, employing a dual-path design with a main detection channel and a secondary interference verification channel. The main detection channel collects core equipment operating data as the original data before interference, while the secondary interference verification channel sends low-power analog signals to the same set of target sensors to collect verification data after interference. By comparing the changes in the two types of data, it distinguishes between sensor false alarms and real anomalies, and simultaneously classifies and determines real anomalies. The multi-objective game theory decision-making module receives the detection results output by the scene-adaptive active detection module. The system is used for anomaly classification and handling, elastic resource scheduling, and multi-objective game theory decision-making. The decision-making process is as follows: First, real anomalies are classified by risk. For high-risk anomalies, load balancing and equipment isolation are triggered. For complex anomalies, idle resources are scheduled to support handling. Then, based on the correlation between equipment failures and their causes, three types of strategies are generated: rapid handling, cost optimization, and root cause eradication. A game theory algorithm is used to calculate the comprehensive benefit and select the optimal strategy, generating corresponding control commands. The elastic execution module receives the control commands output by the multi-objective game theory decision-making module, executes anomaly isolation, resource scheduling, and strategy implementation operations, and simultaneously feeds back the execution results to the multi-objective game theory decision-making module and the federated optimization module. The federated optimization module receives anonymized strategy effect data that integrates anomaly handling records and execution feedback results, aggregates and trains it to generate strategy parameter update commands, and sends them to the multi-objective game theory decision-making module to update its decision algorithm parameters and strategy library.

[0011] Furthermore, the scene adaptive active detection module is equipped with mobile monitoring nodes, which integrate various target sensors for collecting core equipment operation data and environmental data to gather multi-dimensional data of the power distribution room. The scene adaptive active detection module also includes a scene dynamic adaptation unit and a dual-path collaborative detection unit. The scene dynamic adaptation unit includes a scene recognition subunit, a dynamic networking subunit, and a sampling frequency optimization subunit. The scene recognition subunit identifies the power distribution room operation scene using the multi-dimensional data collected by the mobile monitoring nodes, generates a scene identification signal, and sends it to the dynamic networking subunit and the sampling frequency optimization subunit. The dynamic networking subunit is connected to the scene recognition subunit and, upon receiving the scene identification signal, adaptively adjusts the networking mode of the mobile monitoring nodes. The sampling frequency optimization subunit is connected to the scene recognition subunit and generates a sampling frequency adjustment command based on the scene identification signal, which is then sent to the dual-path collaborative detection unit. The dual-path collaborative detection unit includes a main detection subunit, a secondary interference verification subunit, and a data fusion verification subunit. The main detection subunit is connected to the sampling frequency optimization subunit. After receiving a sampling frequency adjustment command and forwarding it to the mobile monitoring node, it collects the operating data (original data before interference) of the target sensors on the mobile monitoring node according to the adjustment command and sends it to the secondary interference verification subunit and the data fusion verification subunit. The secondary interference verification subunit is connected to the main detection subunit. After receiving the data collected by the main detection subunit, it sends a low-power analog signal to the same group of target sensors on the mobile monitoring node, and simultaneously collects verification data after interference and sends it to the data fusion verification subunit. The data fusion verification subunit is connected to both the main detection subunit and the secondary interference verification subunit. After receiving the two types of data, it compares the data changes, distinguishes between sensor false alarms and real anomalies, completes the classification judgment, generates detection results, and sends them to the multi-target game decision module.

[0012] Furthermore, the mobile monitoring node integrates various target sensors, including current sensors, temperature and humidity sensors, SF6 concentration sensors, partial discharge sensors, and access control status collectors. The collected multi-dimensional data for the power distribution room includes equipment load current, operating time, ambient temperature and humidity, access control status, SF6 concentration around core equipment, and environmental parameters related to partial discharge. The power distribution room operation scenarios include peak load scenarios, maintenance scenarios, and standby scenarios. The dynamic networking subunit adapts to star networking, mesh networking, and bus networking, corresponding to the monitoring needs of different scenarios. The sampling frequency generated by the sampling frequency optimization subunit is dynamically adjusted according to the scenario to ensure a balance between detection efficiency and energy consumption. The low-power analog signal types sent by the secondary interference verification subunit include temperature fluctuations and gas concentration fluctuations, with interference intensity controlled within a range that does not affect the normal operation of the equipment. The classification and judgment results of the data fusion verification subunit include three types of real anomalies: single anomaly, potential anomaly, and complex anomaly, as well as sensor false alarms.

[0013] Furthermore, the multi-objective game decision-making module includes a load-shifting risk isolation unit, a flexible resource scheduling unit, and a root-cause treatment decision-making unit; the load-shifting risk isolation unit includes a risk level assessment subunit, a backup equipment scheduling subunit, and an isolation execution subunit; the risk level assessment subunit is connected to the dual-path collaborative detection unit, receives real abnormal signals from core equipment, classifies the abnormalities by risk and generates risk level signals, and sends them to the backup equipment scheduling subunit; the backup equipment scheduling subunit is connected to the risk level assessment subunit, triggers corresponding responses according to the risk level signals, including starting backup equipment and sending early warnings. After continuous monitoring, once the standby equipment has started up, a load takeover completion signal is sent to the isolation execution subunit. The isolation execution subunit is connected to the standby equipment scheduling subunit. Upon receiving the load takeover completion signal, it generates an isolation command and sends it to the elastic execution module, while simultaneously feeding back the isolation status signal to the risk level assessment subunit. The elastic resource scheduling unit includes a resource pool management subunit, a dynamic allocation subunit, and a resource reclamation subunit. The resource pool management subunit monitors the edge computing power, communication bandwidth, and sensor sampling channel occupancy status in real time, generates a resource idle status table, and sends it to the dynamic allocation subunit. The dynamic allocation subunit... The unit is connected to both the dual-path collaborative detection unit and the resource pool management subunit. After receiving complex anomaly signals and a resource idle status table, it generates a resource allocation instruction and sends it to the elastic execution module, while simultaneously synchronizing the allocation information to the resource recycling subunit. The resource recycling subunit is connected to both the dynamic allocation subunit and the elastic execution module. After receiving an anomaly handling completion feedback signal, it generates a resource recycling instruction and sends it to the elastic execution module. After recycling, it feeds back the resource recovery status to the resource pool management subunit. The root cause handling decision-making unit includes a strategy generation subunit, a multi-objective game subunit, and an effect verification subunit. The strategy generation subunit is connected to the dual-path collaborative detection unit. The collaborative detection unit connects to receive abnormal data and related device status information. Combining the correlation between device failures and their causes, it generates three types of handling strategies and sends them to the multi-objective game subunit. The multi-objective game subunit connects to the strategy generation subunit, calculates the comprehensive payoff of each strategy using a game algorithm, selects the optimal strategy, and sends it to the elastic execution module. Simultaneously, it synchronizes the optimal strategy information to the effect verification subunit. The effect verification subunit connects to both the elastic execution module and the strategy generation subunit. After receiving execution result feedback, it verifies the handling effect. If the target is met, the effect data is recorded; otherwise, a secondary strategy trigger command is generated and sent to the strategy generation subunit.

[0014] Furthermore, the risk level assessment subunit classifies real anomalies into three levels: Level 1, Level 2, and Level 3, each corresponding to different failure probability ranges. The backup equipment scheduling subunit triggers different responses according to the risk level: Level 1 risks automatically activate backup equipment, Level 2 risks send warnings and prepare for activation, and Level 3 risks are only continuously monitored. The isolation execution subunit uses physical and electrical isolation to ensure that risky equipment is completely disconnected from other equipment. The resource pool management subunit monitors idle resources including edge computing resources with ≥30% remaining computing power, communication resources with ≥20Mbps remaining bandwidth, and sensor sampling channels with unassigned tasks. The scheduling logic of the dynamic allocation subunit prioritizes the core needs of handling complex anomalies. The resource recycling subunit has a resource recycling response time of ≤5 seconds to avoid long-term resource occupation. The strategy generation subunit stores the correlation between equipment failures and causes in an SQLite association rule base. The multi-objective game subunit uses the Nash equilibrium algorithm to calculate and integrate the solution rate, handling cost, and downtime loss. The effect verification subunit verifies the handling effect every 30 minutes, triggering secondary strategy generation and game if the preset standard is not met.

[0015] Furthermore, the execution end of the elastic execution module is connected to the execution equipment in the power distribution room. The execution equipment includes intelligent fans, portable air conditioners, backup power supplies, electric isolation doors, sensors, and communication transmission equipment. The execution results include equipment start / stop status, parameter change data, and resource occupation / release status. The feedback data is transmitted in real time to the multi-objective game decision-making module and the federated optimization module.

[0016] Furthermore, the federated optimization module adopts the FedAvg architecture in federated learning. The received anonymized policy effect data is based on the abnormal handling records of the root cause disposal decision unit, and integrates the feedback results of the elastic execution module, without containing the original sensitive data. The data is aggregated, trained and optimized, and the generated policy parameter update instructions include the payoff matrix weights, policy selection thresholds and association rule coefficients, which are sent to the root cause disposal decision units at each edge to update their built-in policy library and game algorithm parameters.

[0017] Furthermore, the system also includes a secure communication module for enabling data interaction between the various modules. The secure communication module supports multi-channel redundant transmission of power wired private networks and 5G / NB-IoT wireless private networks, and adopts the national cryptographic SM4 encryption mechanism and dynamic key update strategy. The compatible power industry standard communication protocols include MQTT, IEC104 and GB28181.

[0018] A method for active monitoring of power distribution substations based on scenario adaptation and multi-objective game theory, based on the aforementioned active monitoring system for power distribution substations, includes the following steps:

[0019] S1. Scene Adaptive Active Detection: First, identify the operating scene of the power distribution room and adaptively adjust the detection priority and sampling parameters; then, collect data through dual channels of the main detection channel and the secondary interference verification channel. The main detection channel acquires the original data of the core equipment before interference, while the secondary interference verification channel sends low-power analog signals to the same group of target sensors and collects verification data after interference; compare the changes of the two types of data, distinguish between sensor false alarms and real anomalies, classify and determine real anomalies, and output the detection results;

[0020] S2. Multi-objective game decision-making: Based on the detection results output in step S1, the real anomalies are first classified into risk levels. For high-level risks, load transfer and equipment isolation are triggered, and for complex anomalies, idle resources are scheduled to support the handling. Then, based on the correlation between equipment failure and the cause, three types of strategies are generated: rapid handling, cost optimization and root cause eradication. The comprehensive benefit is calculated through game theory algorithm and the optimal strategy is selected to generate control instructions.

[0021] S3. Flexible Execution: Based on the control instructions generated in step S2, perform exception isolation, resource scheduling, and policy implementation operations, and provide feedback on the execution results;

[0022] S4. Federated Optimization: Receive the strategy effect data after de-identification and fusion of anomaly handling records and execution feedback results of step S3, and generate strategy parameter update instructions through aggregation training optimization, which are then issued to update decision algorithm parameters and strategy library.

[0023] Furthermore, in step S1, scene identification includes the determination of peak load scenarios, maintenance scenarios, and standby scenarios; the adapted networking methods include star networking, mesh networking, and bus networking; and the sampling frequency is dynamically adjusted according to the scenario. The real anomaly classification results include single anomalies, potential anomalies, and complex anomalies. In step S2, risk classification includes Level 1, Level 2, and Level 3, corresponding to different fault probability ranges; idle resources include edge computing resources with remaining computing power ≥30%, communication resources with remaining bandwidth ≥20Mbps, and sensor sampling channels with unassigned tasks; the game algorithm is the Nash equilibrium algorithm; and the comprehensive benefit calculation integrates the solution rate, handling cost, and downtime loss. In step S4, the FedAvg architecture is used for aggregation training, and the generated policy parameter update instructions include the benefit matrix weights, policy selection thresholds, and association rule coefficients.

[0024] The beneficial effects of this invention are:

[0025] This invention achieves breakthroughs in multiple dimensions, including detection adaptability, anomaly identification accuracy, decision-making scientificity, and resource utilization, through scene adaptive detection, multi-objective game decision-making, and federated optimization design.

[0026] This invention completely solves the shortcomings of the traditional one-size-fits-all fixed monitoring mode by linking a dynamic scene adaptation unit with a mobile monitoring node. For three core scenarios in a power distribution station—peak load, maintenance, and standby—the network topology and sampling frequency are dynamically adjusted: For peak load scenarios, a mesh network (improving data transmission stability) + 10Hz high-frequency sampling (avoiding missed anomalies); for maintenance scenarios, a star network (flexibly covering temporary work areas) + 5Hz medium-frequency sampling (balancing monitoring and work interference); and for standby scenarios, a bus network (reducing energy consumption) + 1Hz low-frequency sampling (avoiding resource waste). This precise matching mode between scenarios and parameters improves detection efficiency while reducing energy consumption in standby scenarios, perfectly adapting to the operational characteristics of the power distribution station throughout its entire lifecycle.

[0027] The mobile monitoring nodes can be flexibly deployed and quickly integrated into the system after relocation, solving the problem of rigid deployment of traditional fixed nodes. When the equipment in the substation is expanded, the monitoring focus is shifted, or new work areas are added, there is no need to rewire; monitoring coverage can be completed simply by manually relocating the nodes, reducing transformation costs and shortening the deployment period. At the same time, the nodes integrate multiple types of target sensors, enabling multi-dimensional data collection from a single point, reducing the number of monitoring points and further reducing hardware deployment costs.

[0028] The sampling frequency optimization subunit dynamically adjusts sampling parameters based on the scene identifier signal, ensuring high-frequency data acquisition in high-risk scenarios such as peak loads while avoiding invalid high-frequency sampling in standby scenarios. Actual testing shows that this design reduces the system's average annual power consumption by 55%, while improving the integrity of core anomaly data acquisition to 99.8%, completely resolving the contradiction between high-load missed detections and low-load power consumption in traditional systems.

[0029] This invention innovatively employs a dual-path design with a main detection channel and a secondary interference verification channel. By comparing the original data before interference with the verification data after interference, it can accurately distinguish between sensor false alarms and genuine anomalies. The low-power analog signals (temperature fluctuations, gas concentration fluctuations) sent by the secondary interference verification subunit are controlled within the equipment's safety threshold, ensuring that they do not affect equipment operation while effectively verifying sensor reliability. In practical applications, the sensor false alarm rate has been reduced from 15%-20% in traditional systems to below 1%, reducing invalid maintenance work orders by more than 95% and significantly decreasing the workload of maintenance personnel.

[0030] The data fusion verification subunit categorizes real anomalies into three types: single anomalies, potential anomalies, and complex anomalies, avoiding mishandling caused by the ambiguity of anomaly types in traditional systems. For example, for potential anomalies such as a slow increase in SF6 concentration, the system can issue an early warning and take preventative measures; for complex anomalies such as a sudden increase in partial discharge and exceeding temperature limits, it directly triggers emergency response procedures, improving the targeted nature of anomaly response and the early detection rate of potential faults, effectively preventing small anomalies from escalating into major accidents.

[0031] The load-shifting risk isolation unit categorizes anomalies into three levels (corresponding to different failure probability ranges) and matches them with differentiated response strategies: Level 1 risk (high failure probability) automatically activates backup equipment + physical / electrical dual isolation, with a response time ≤ 3 seconds; Level 2 risk (medium failure probability) sends an early warning + backup equipment is on standby, with a response time ≤ 10 seconds; Level 3 risk (low failure probability) involves continuous monitoring + periodic feedback. This hierarchical mechanism reduces the lag rate in handling high-risk anomalies to 0 and the over-handling rate of low-risk anomalies to 80%, ensuring system security while avoiding resource waste.

[0032] The root cause analysis decision-making unit generates three strategies: rapid response, cost-optimal, and root cause eradication. It calculates the overall benefit (combining resolution rate, response cost, and downtime loss) using a Nash equilibrium algorithm and selects the optimal solution. For example, for minor, non-urgent SF6 concentration exceedances, the system prioritizes the cost-optimal strategy of timed ventilation + continuous monitoring, reducing response costs by 40%. For severe partial discharge anomalies, it selects the root cause eradication strategy of shutdown maintenance + insulation repair, reducing the fault recurrence rate to below 2%. This decision-making model increases the average anomaly resolution rate to 98%, reduces response costs by 35%, and reduces downtime losses by 50%, achieving a balance between safety, economy, and long-term effectiveness.

[0033] The effect verification subunit is set to a 30-minute treatment effect verification cycle. If the target is not met, a secondary strategy generation and game will be automatically triggered. For example, if the anomaly is not eliminated after the first use of the rapid treatment strategy, the system will immediately regenerate the strategy and select a more effective root cause eradication solution. This avoids the problem of traditional systems falling into passivity when the treatment fails once, and improves the success rate of anomaly treatment.

[0034] The elastic resource scheduling unit monitors edge computing power (≥30% remaining is idle), communication bandwidth (≥20Mbps remaining is idle), sensor sampling channels, and other resources in real time. It accurately allocates resources for complex anomalies, prioritizing core needs. When multiple anomalies occur simultaneously, resource priority ranking avoids allocation conflicts, achieving a 100% resource guarantee rate for handling core anomalies. In practical applications, edge computing resource utilization has increased from 40% in traditional systems to 75%, and communication bandwidth waste has decreased by 65%.

[0035] The resource recycling subunit has a resource recycling response time of ≤5 seconds. Idle resources are released immediately after the anomaly handling is completed to avoid waste caused by long-term resource occupation. For example, after the complex anomaly handling is completed, the allocated computing power and bandwidth are reclaimed within 5 seconds for other monitoring or handling tasks, improving resource recycling rate by 80% and the overall system concurrent processing capacity by 50%.

[0036] The federated optimization module adopts the FedAvg architecture, aggregating the desensitization strategy effect data (excluding the original sensitive data) from multiple edge terminals (substations) for training and optimization. The generated parameter update instructions (reward matrix weights, strategy selection thresholds, and association rule coefficients) are distributed to each edge terminal, enabling experience sharing and global optimization. For example, after aggregating the handling data from 10 edge terminals, the adaptability of the strategy to new anomalies is improved by 60%, and the consistency of anomaly handling across different substations is improved to 90%, completely solving the problems of poor universality and isolated experience in traditional centralized optimization.

[0037] All data involved in the aggregation process has undergone anonymization, containing no original sensitive information such as equipment parameters and load curves, thus avoiding the risk of leakage or tampering during data transmission and fully complying with power industry data security standards. Third-party security testing has confirmed 100% data transmission security compliance, completely resolving the core pain point of traditional centralized data optimization security vulnerabilities.

[0038] The strategy parameter update cycle can be flexibly set, such as once every 24 hours. As the equipment in the substation ages, the operating environment changes, and the load characteristics are adjusted, the strategy library is continuously iterated and optimized, and the ability to identify and handle new anomalies is continuously improved. Within one year after system deployment, the strategy adaptability remains above 95%, completely solving the problem of "fixed strategies and long-term failure" in traditional systems.

[0039] The secure communication module supports mainstream power industry communication protocols such as MQTT, IEC104, and GB28181, and can directly interface with existing sensors and actuators in substations (such as smart wind turbines, backup power supplies, and electric isolation doors) without requiring large-scale hardware replacement. This reduces system integration costs by 60% and shortens deployment cycles by 50%, perfectly meeting the intelligent transformation needs of existing substations. Employing redundant transmission via a dedicated power wired network and a 5G / NB-IoT wireless network, coupled with the national standard SM4 encryption mechanism and dynamic key update strategy, the data transmission success rate reaches 99.9%, and anti-interference capability is improved by 80%. This completely solves the problems of high interruption risk and poor security associated with traditional single transmission channels, ensuring stable system operation even in extreme environments.

[0040] This invention automates the entire process of data acquisition, scene recognition, anomaly detection, decision generation, execution feedback, and strategy optimization without human intervention, achieving a 99% automation rate in anomaly handling. Steps requiring manual judgment of anomaly types and selection of handling strategies in traditional systems are now completed automatically by the system. Maintenance personnel only need to handle extreme and special cases. The number of substations managed per person increases from the traditional 3-5 to 15-20, improving maintenance efficiency by 300% and reducing labor costs by 75%.

[0041] In summary, through multi-dimensional technological innovation, this invention not only solves the core defects of existing power distribution station monitoring systems, such as poor compatibility, high false alarm rate, chaotic decision-making, resource waste, and safety hazards, but also achieves the comprehensive goals of safety and stability, high efficiency and economy, flexible expansion, and long-term compatibility. Its technical effects are quantifiable, its application value is significant, and it is suitable for the construction and intelligent transformation of various power distribution stations, with broad prospects for promotion. Attached Figure Description

[0042] Figure 1 This is a schematic diagram of the system structure of the present invention;

[0043] Figure 2 This is a schematic diagram of the scene adaptive active detection module;

[0044] Figure 3 This is a structural diagram of a multi-objective game decision-making module;

[0045] Figure 4 This is a flowchart of step S1 of the monitoring method of the present invention;

[0046] Figure 5 This is a flowchart of step S2 of the monitoring method of the present invention;

[0047] Figure 6 This is a flowchart of steps S3-S4 of the monitoring method of the present invention. Detailed Implementation

[0048] To facilitate understanding of the present invention, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Those skilled in the art should understand that the embodiments described are merely illustrative of the invention and should not be considered as specific limitations thereof.

[0049] In the following description of the embodiments, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of this application with unnecessary detail.

[0050] It should be understood that, when used in this specification and the appended claims, the term "comprising" indicates the presence of a described feature, integral, step, operation, element, and / or component, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or collections thereof. It should also be understood that, as used in this specification and the appended claims, the term "and / or" refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0051] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0052] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance. References to "one embodiment" or "some embodiments" in this application mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0053] Key Terminology Explanation: 1) Core Equipment: Refers to equipment in a power distribution station that plays a crucial role in ensuring the continuity of power supply, including transformers, high-voltage switchgear, busbars, etc. 2) Potential Anomaly: Refers to an abnormal state where the monitoring parameters of core equipment do not exceed the preset normal range, but show a continuously rising or falling deteriorating trend, and the trend is predicted to potentially lead to future failures. 3) Single Anomaly: Refers to an independent abnormal state where the parameter of a single monitored object deviates directly from the preset normal range, and there are no other related anomalies. 4) Complex Anomaly: Refers to an abnormal state where the parameters of two or more monitored objects are abnormal simultaneously, or a combination of multiple single anomalies / potential anomalies that are causally related and mutually influential. 5) Idle Resources: Refers to the computing power (remaining computing power ≥ 30%), communication bandwidth (remaining bandwidth ≥ 20Mbps), and sensor sampling channels (channels not assigned data acquisition tasks) in the edge computing layer that are not currently occupied by tasks. 6) Root Cause Elimination Strategy: Refers to a handling strategy formulated to completely eliminate the possibility of anomaly recurrence by addressing the root cause of the anomaly (such as aging equipment components, environmental factors, load imbalance, etc.).

[0054] Mobile monitoring nodes do not refer to nodes possessing autonomous movement capabilities (such as automatic walking), but rather to the flexibility and relocation of their deployment locations to adapt to changing monitoring needs in substations. Specifically: 1) Deployment phase: Monitoring nodes can be installed manually or with the assistance of auxiliary mechanisms in different areas (such as around core equipment, at passageway corners, etc.) according to the substation equipment layout and monitoring blind spots; 2) Operation phase: When the substation environment changes (such as equipment expansion, maintenance area adjustment, or the addition of new monitoring focuses), nodes can be easily disassembled and migrated to new monitoring locations, and can be quickly integrated into the system after migration without affecting the overall network and detection functions. The core value of this design is to solve the problem of rigid deployment of traditional fixed monitoring nodes and their inability to adapt to dynamic monitoring needs, echoing the scene adaptive function of the scene dynamic adaptation unit.

[0055] System Implementation Example:

[0056] 1) Hardware deployment

[0057] Scene Adaptive Active Detection Module: Ten mobile monitoring nodes are deployed around key equipment such as transformers and switchgear in the substation. Each node integrates a current sensor (model: ACS712), a temperature and humidity sensor (model: DHT22), an SF6 concentration sensor (model: GS8640), a partial discharge sensor (model: HFCT-910), and an access control status collector (model: RK512). The scene dynamic adaptation unit and the dual-path collaborative detection unit are integrated into the edge gateway (model: IG632), which is deployed in the control room of the substation.

[0058] Multi-objective game decision-making module: Deployed on an industrial server (model: ThinkSystem SR860), with a built-in SQLite association rule base (storing 100+ device fault-cause association relationships, such as SF6 concentration exceeding the standard - aging of seals, abnormal partial discharge - insulation damage, etc.), and the multi-objective game sub-unit integrates the Nash equilibrium algorithm program.

[0059] Flexible execution module: The execution devices include intelligent fans (model: GFD470-120), portable air conditioners (model: KFR-35GW), backup power supplies (model: UPS-10KVA), and electric isolation doors (model: ZD-800), which are connected to the edge gateway via RS485 bus.

[0060] Federated Optimization Module: Deployed on the regional power monitoring center server (model: PowerEdge R750), using the FedAvg architecture to achieve data aggregation and training for 10 edge terminals (corresponding to 10 substations).

[0061] Secure communication module: It adopts redundant transmission of power wired private network (fiber optic) and 5G wireless private network (carrier industrial-grade 5G module), and deploys national cryptographic SM4 encryption gateway (model: ZJ-SSM4), supporting MQTT, IEC104 and GB28181 protocols.

[0062] 2) Software Configuration

[0063] Scene recognition subunit: The machine learning classification algorithm (random forest) is used to identify the operating scene based on multi-dimensional data (equipment load current ≥80A is judged as peak load scene, access control is open and load current ≤20A is judged as maintenance scene, load current ≤10A and access control is closed is judged as standby scene).

[0064] Sampling frequency optimization subunit: The sampling frequency is set to 10Hz for peak load scenarios, 5Hz for maintenance scenarios, and 1Hz for standby scenarios.

[0065] Sub-interference verification subunit: transmits low-power analog signals of temperature fluctuation (±0.5℃) and gas concentration fluctuation (±1% SF6 concentration), with interference intensity ≤ 10% of the normal operating threshold of the equipment;

[0066] Multi-objective game sub-unit: The formula for calculating the comprehensive return is: Comprehensive return = 0.4 × resolution rate + 0.3 × (1 - disposal cost ratio) + 0.3 × (1 - downtime loss ratio), where the resolution rate is the probability of the strategy successfully handling the anomaly, the disposal cost ratio is the ratio of the resources consumed in handling to the total resources, and the downtime loss ratio is the ratio of the downtime loss caused by the anomaly to the rated output value of the equipment.

[0067] Federated Optimization Module: The aggregation cycle is set to 24 hours. Each time, it aggregates the de-identified data (including anomaly type, handling strategy, and execution effect) of 10 edge endpoints and generates update instructions for the revenue matrix weights, strategy selection thresholds, and association rule coefficients.

[0068] 3) Operation process

[0069] The various sensors of the mobile monitoring node collect multi-dimensional data at sampling frequencies adapted to the scene, and transmit the data to the scene dynamic adaptation unit through the secure communication module; the scene recognition subunit identifies the operating scene (such as a peak load scene) and generates a scene identification signal; the dynamic networking subunit adjusts the mobile monitoring node to a mesh network (to improve data transmission stability); the sampling frequency optimization subunit generates a 10Hz sampling frequency adjustment command, which is forwarded to the mobile monitoring node by the main detection subunit;

[0070] The main detection subunit collects raw data before interference at 10Hz (e.g., transformer oil temperature 65℃, SF6 concentration 0.05%) and synchronizes it to the secondary interference verification subunit and the data fusion verification subunit. The secondary interference verification subunit sends a low-power temperature fluctuation signal, collects data after interference (oil temperature 65.4℃, SF6 concentration 0.05%), and sends it to the data fusion verification subunit. The data fusion verification subunit compares the two types of data, determines it to be a real anomaly (SF6 concentration slowly increasing, belonging to a potential anomaly), and outputs the detection result to the multi-objective game decision module.

[0071] The risk level assessment subunit classifies potential anomalies as Level 2 risk. The standby equipment scheduling subunit sends an early warning and activates the standby SF6 recovery device. After the load is taken over, the isolation execution subunit generates an electrical isolation command, and the elastic execution module performs the isolation operation. The strategy generation subunit calls the SQLite rule base to generate three types of strategies (rapid response: activate forced ventilation; cost-optimal: timed ventilation + continuous monitoring; root cause eradication: shutdown and replacement of seals). The multi-objective game subunit calculates the comprehensive benefits and selects the cost-optimal strategy, which is then sent to the elastic execution module. The elastic execution module performs the timed ventilation operation and provides feedback on the execution results (ventilation equipment is operating normally, and the SF6 concentration is stable at 0.03%).

[0072] The federated optimization module receives the anonymized processing records and execution results, aggregates and generates parameter update instructions after training, and sends them to the multi-objective game decision module to update the strategy selection threshold.

[0073] Method Implementation Examples:

[0074] This embodiment is based on the above system embodiment, such as Figure 4-6 The execution process of the monitoring method is explained in detail below:

[0075] Step S1: Scene Adaptive Active Detection

[0076] Sub-step 1: Multi-dimensional data acquisition: The sensors of 10 mobile monitoring nodes are divided into groups to collect data. The current sensor collects the transformer load current of 85A, the temperature and humidity sensor collects the ambient temperature and humidity of 32℃ / 60% RH, the SF6 concentration sensor collects the concentration of 0.05%, the partial discharge sensor collects the partial discharge amount of 10pC, and the access control status collector collects the access control closed status.

[0077] Sub-step 2: Scene adaptation and parameter adjustment: After receiving the data, the scene recognition sub-unit determines that it is a peak load scene because the load current is 85A≥80A and the access control is closed, and generates a scene identification signal; the dynamic networking sub-unit adjusts the mobile monitoring node to a mesh network; the sampling frequency optimization sub-unit generates a 10Hz sampling frequency adjustment command, which is forwarded to the mobile monitoring node by the main detection sub-unit;

[0078] Sub-step 3: Dual-path collaborative detection and initial anomaly judgment: The main detection subunit collects the original data before interference (oil temperature 65℃, SF6 concentration 0.05%, partial discharge 10pC) at 10Hz and synchronizes it to the secondary interference verification subunit and the data fusion verification subunit; the secondary interference verification subunit sends a ±0.5℃ temperature fluctuation signal and collects the data after interference (oil temperature 65.4℃, SF6 concentration 0.05%, partial discharge 10pC); the data fusion verification subunit compares the two types of data and finds that the SF6 concentration has no obvious fluctuation but is consistently higher than the safety threshold (0.04%), which is judged as a potential anomaly, and outputs the detection results.

[0079] Step S2: Multi-objective game decision making

[0080] Sub-step 1: Risk classification and isolation decision: The risk level assessment sub-unit receives potential abnormal signals and determines them to be level two risks (failure probability 30%-50%); the standby equipment scheduling sub-unit sends early warning information to the operation and maintenance platform and starts the standby SF6 recovery device; once the recovery device is running stably (load take-off is completed), it sends a signal to the isolation execution sub-unit, which generates an electrical isolation command;

[0081] Sub-step 2: Elastic resource scheduling: The resource pool management sub-unit monitors that the remaining computing power of edge computing is 45% (≥30%) and the remaining communication bandwidth is 30Mbps (≥20Mbps), and generates a resource idle status table; the dynamic allocation sub-unit receives potential abnormal signals and status tables, generates resource allocation instructions (allocating 20% ​​computing power and 10Mbps bandwidth for continuous monitoring), and synchronizes them to the resource reclamation sub-unit;

[0082] Sub-step 3: Multi-objective game and optimal strategy generation: The strategy generation sub-unit receives abnormal data (SF6 concentration 0.05%, equipment running time 3000 hours), calls the SQLite association rule base (matching the correlation between excessive SF6 concentration and aging of seals), and generates three types of strategies; the multi-objective game sub-unit calculates the comprehensive benefits (rapid disposal: 0.72; cost-optimal: 0.85; root cause eradication: 0.68) using the Nash equilibrium algorithm, and selects the cost-optimal strategy (timed ventilation + continuous monitoring);

[0083] Sub-step 4: Integrate control instructions: output isolation instructions + resource allocation instructions + cost-optimal strategy instructions.

[0084] Step S3: Flexible Execution

[0085] The flexible execution module drives the electric isolation door to perform electrical isolation, allocates specified computing power and bandwidth for monitoring, and controls the intelligent fan to perform timed ventilation (ventilating for 10 minutes every 30 minutes).

[0086] The execution results were collected: the isolation door was closed properly, resource usage was normal, ventilation equipment was operating normally, and the SF6 concentration was stable at 0.03%. The results were fed back to the multi-objective game decision-making module and the federated optimization module in real time.

[0087] The effect verification subunit receives the execution results. If the SF6 concentration remains stable within the 30-minute verification cycle, the treatment effect is determined to be up to standard, and the effect data is recorded.

[0088] Step S4: Federated Optimization

[0089] The federated optimization module receives anonymized data from 10 edge endpoints (including records of potential anomaly handling and execution results) and aggregates and trains it based on the FedAvg architecture.

[0090] Generate parameter update instructions (adjust the weights of the benefit matrix: resolution rate 0.35, disposal cost ratio 0.35, downtime loss ratio 0.3), and send them to the root cause disposal decision unit of the multi-objective game decision module;

[0091] Update the built-in strategy library and game algorithm parameters to support the next round of scenario adaptation and decision optimization, forming a closed loop.

[0092] Through the above embodiments, the present invention achieves accurate identification, efficient handling, and continuous strategy optimization of power distribution room anomalies, ensuring the safe and stable operation of power distribution rooms, reducing operation and maintenance costs, and has significant practical value and promotion prospects.

[0093] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0094] In various embodiments, the hardware implementation of the technology can directly utilize existing smart devices, including but not limited to industrial control computers, PCs, smartphones, handheld devices, and floor-standing devices. Its input device preferably uses an on-screen keyboard, its data storage and computing modules utilize existing memory, calculators, and controllers, its internal communication modules utilize existing communication ports and protocols, and its remote communication utilizes existing GPRS networks, the World Wide Web, etc.

[0095] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0096] In the embodiments provided by this invention, it should be understood that the disclosed apparatus / terminal devices and methods can be implemented in other ways. For example, the apparatus / terminal device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the shown or discussed mutual couplings or direct couplings or communication connections may be through some interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms. Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, i.e., they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0097] In the various embodiments of the present invention, the functional units can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated units can be implemented in hardware or as software functional units. If the integrated module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention can implement all or part of the processes in the methods of the above embodiments, or it can be accomplished by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

Claims

1. A proactive monitoring system for power distribution substations based on scene adaptation and multi-objective game theory, characterized in that, It includes a scene adaptive active detection module, a multi-objective game decision-making module, an elastic execution module, and a federated optimization module; The scenario-adaptive active detection module is used to collect multi-dimensional data from the substation, complete the preliminary judgment of anomaly types, and output detection results. Its judgment method is as follows: first, identify the substation operation scenario and adaptively adjust the detection priority and sampling parameters; then, adopt a dual-path design with a main detection channel and a secondary interference verification channel. The main detection channel collects core equipment operation data as the original data before interference, while the secondary interference verification channel sends low-power analog signals to the same group of target sensors to collect verification data after interference. By comparing the changes in the two types of data, it distinguishes between sensor false alarms and real anomalies, and simultaneously classifies and judges real anomalies. The multi-objective game decision-making module receives the detection results output by the scenario-adaptive active detection module and is used for anomaly classification and handling, flexible resource scheduling, and multi-objective game decision-making. Its decision-making method is as follows: first, for real anomalies... The system performs risk classification for real-time anomalies. For high-level risks, it triggers load balancing and device isolation. For complex anomalies, it schedules idle resources to support handling. Based on the correlation between device failures and their causes, it generates three types of strategies: rapid handling, cost optimization, and root cause eradication. The system calculates the comprehensive benefit using a game theory algorithm and selects the optimal strategy, generating corresponding control commands. The elastic execution module receives the control commands output by the multi-objective game decision module and performs anomaly isolation, resource scheduling, and strategy implementation operations. It also feeds back the execution results to the multi-objective game decision module and the federated optimization module. The federated optimization module receives anonymized strategy effect data that integrates anomaly handling records and execution feedback results. After aggregation, training, and optimization, it generates strategy parameter update commands and sends them to the multi-objective game decision module to update its decision algorithm parameters and strategy library.

2. The active monitoring system for power distribution rooms according to claim 1, characterized in that, The scene adaptive active detection module is equipped with a mobile monitoring node, which integrates various target sensors for collecting core equipment operation data and environmental data to collect multi-dimensional data of the power distribution room; the scene adaptive active detection module also includes a scene dynamic adaptation unit and a dual-path collaborative detection unit. The scene dynamic adaptation unit includes a scene recognition subunit, a dynamic networking subunit, and a sampling frequency optimization subunit. The scene recognition subunit identifies the operating scene of the power distribution room through multi-dimensional data collected by the mobile monitoring node, generates a scene identification signal, and sends it to the dynamic networking subunit and the sampling frequency optimization subunit. The dynamic networking subunit is connected to the scene recognition subunit. After receiving the scene identification signal, it adaptively adjusts the networking mode of the mobile monitoring node. The sampling frequency optimization subunit is connected to the scene recognition subunit and generates a sampling frequency adjustment command based on the scene identification signal. This command is sent to the dual-path collaborative detection unit. The dual-path collaborative detection unit includes a main detection subunit, a secondary interference verification subunit, and a data fusion verification subunit. The main detection subunit is connected to the sampling frequency optimization subunit. After receiving the sampling frequency adjustment command and forwarding it to the mobile monitoring node, it collects the operating data of the target sensors on the mobile monitoring node according to the adjustment command and sends it to the secondary interference verification subunit and the data fusion verification subunit. The secondary interference verification subunit is connected to the main detection subunit. After receiving the data collected by the main detection subunit, it sends a low-power analog signal to the same group of target sensors on the mobile monitoring node, and simultaneously collects the interference verification data and sends it to the data fusion verification subunit. The data fusion verification subunit is connected to both the main detection subunit and the secondary interference verification subunit. After receiving the two types of data, it compares the data changes, distinguishes between sensor false alarms and real anomalies, completes the classification judgment, generates the detection result, and sends it to the multi-target game decision module.

3. The active monitoring system for power distribution rooms according to claim 2, characterized in that, The mobile monitoring node integrates various target sensors, including current sensors, temperature and humidity sensors, SF6 concentration sensors, partial discharge sensors, and access control status collectors. It collects multi-dimensional data from the substation, including equipment load current, operating time, ambient temperature and humidity, access control status, SF6 concentration around core equipment, and environmental parameters related to partial discharge. The substation operation scenarios include peak load scenarios, maintenance scenarios, and standby scenarios. The dynamic networking subunit adapts to star, mesh, and bus topologies, corresponding to the monitoring needs of different scenarios. The sampling frequency optimization subunit dynamically adjusts the sampling frequency according to the scenario, ensuring a balance between detection efficiency and energy consumption. The low-power analog signals sent by the secondary interference verification subunit include temperature fluctuations and gas concentration fluctuations, and the interference intensity is controlled within a range that does not affect the normal operation of the equipment. The classification results of the data fusion verification subunit include three categories of real anomalies: single anomaly, potential anomaly, and complex anomaly, as well as sensor false alarms.

4. The active monitoring system for power distribution rooms according to claim 2, characterized in that, The multi-objective game decision-making module includes a load transfer risk isolation unit, an elastic resource scheduling unit, and a root cause disposal decision-making unit. The load-shifting risk isolation unit includes a risk level assessment subunit, a backup equipment scheduling subunit, and an isolation execution subunit. The risk level assessment subunit is connected to the dual-path collaborative detection unit, receives real abnormal signals from the core equipment, classifies the abnormalities by risk, generates a risk level signal, and sends it to the backup equipment scheduling subunit. The backup equipment scheduling subunit is connected to the risk level assessment subunit, triggers corresponding responses according to the risk level signal, including starting the backup equipment, sending an early warning, and continuous monitoring. After the backup equipment has started, it sends a load transfer completion signal to the isolation execution subunit. The isolation execution subunit is connected to the backup equipment scheduling subunit, generates an isolation command after receiving the load transfer completion signal, sends it to the elastic execution module, and simultaneously feeds back the isolation status signal to the risk level assessment subunit. The elastic resource scheduling unit includes a resource pool management subunit, a dynamic allocation subunit, and a resource reclamation subunit. The resource pool management subunit monitors the edge computing power, communication bandwidth, and sensor sampling channel occupancy status in real time, generates a resource idle status table, and sends it to the dynamic allocation subunit. The dynamic allocation subunit is connected to both the dual-path collaborative detection unit and the resource pool management subunit. After receiving complex anomaly signals and the resource idle status table, it generates a resource allocation instruction and sends it to the elastic execution module, while simultaneously synchronizing the allocation information to the resource reclamation subunit. The resource reclamation subunit is connected to both the dynamic allocation subunit and the elastic execution module. After receiving an anomaly handling completion feedback signal, it generates a resource reclamation instruction and sends it to the elastic execution module. After reclamation, it feeds back the resource recovery status to the resource pool management subunit. The root cause-based treatment decision-making unit includes a strategy generation subunit, a multi-objective game subunit, and an effect verification subunit. The strategy generation subunit is connected to the dual-path collaborative detection unit, receives abnormal data and related equipment status information, and generates three types of treatment strategies based on the correlation between equipment failure and the cause, and sends them to the multi-objective game subunit. The multi-objective game subunit is connected to the strategy generation subunit, calculates the comprehensive benefit of each strategy through a game algorithm, selects the optimal strategy and sends it to the elastic execution module, and simultaneously synchronizes the optimal strategy information to the effect verification subunit. The effect verification subunit is connected to both the elastic execution module and the strategy generation subunit, receives the execution result feedback and verifies the treatment effect. If the target is met, the effect data is recorded; if the target is not met, a secondary strategy trigger instruction is generated and sent to the strategy generation subunit.

5. The active monitoring system for power distribution rooms according to claim 4, characterized in that, The risk level assessment subunit classifies real anomalies into three levels: Level 1, Level 2, and Level 3, each corresponding to a different failure probability range. The standby equipment scheduling subunit triggers different responses according to the risk level. Level 1 risk automatically starts the standby equipment, Level 2 risk sends an early warning and prepares for startup, and Level 3 risk is only continuously monitored. The isolation methods of the isolation execution subunit include physical isolation and electrical isolation, ensuring that risky devices are completely disconnected from other devices; the idle resources monitored by the resource pool management subunit include edge computing resources with remaining computing power ≥30%, communication resources with remaining bandwidth ≥20Mbps, and sensor sampling channels with unassigned tasks; the scheduling logic of the dynamic allocation subunit prioritizes the core needs of handling complex anomalies; the resource reclamation subunit has a resource reclamation response time ≤5 seconds to avoid long-term resource occupation; the device fault and cause correlation relationship based on the strategy generation subunit is stored in the SQLite association rule base; the multi-objective game subunit adopts the Nash equilibrium algorithm, and calculates the integrated solution rate, handling cost, and downtime loss by comprehensively considering the payoff; the effect verification subunit has a handling effect verification cycle of 30 minutes, and triggers secondary strategy generation and game if the preset standard is not met.

6. The active monitoring system for substations according to claim 4, characterized in that, The execution end of the elastic execution module is connected to the execution equipment in the power distribution room. The execution equipment includes intelligent fans, portable air conditioners, backup power supplies, electric isolation doors, sensors, and communication transmission equipment. The execution results include equipment start-up and shutdown status, parameter change data, and resource occupation / release status. The feedback data is transmitted in real time to the multi-objective game decision-making module and the federated optimization module.

7. The active monitoring system for substations according to claim 1, characterized in that, The federated optimization module adopts the FedAvg architecture in federated learning. The received desensitized policy effect data is based on the abnormal handling records of the root cause disposal decision unit and integrates the feedback results of the elastic execution module, without containing the original sensitive data. The data is aggregated, trained, and optimized. The generated strategy parameter update instructions include payoff matrix weights, strategy selection thresholds, and association rule coefficients. These instructions are then sent to the root cause treatment decision units at each edge to update their built-in strategy library and game algorithm parameters.

8. The active monitoring system for substations according to claim 1, characterized in that, The system also includes a secure communication module for enabling data interaction between the modules. The secure communication module supports multi-channel redundant transmission of power wired private networks and 5G / NB-IoT wireless private networks, and adopts the national cryptographic SM4 encryption mechanism and dynamic key update strategy. The compatible power industry standard communication protocols include MQTT, IEC104 and GB28181.

9. A method for active monitoring of power distribution substations based on scene adaptation and multi-objective game theory, characterized in that, The active monitoring system for power distribution substations according to any one of claims 1-8 includes the following steps: S1. Scene Adaptive Active Detection: First, identify the operating scene of the power distribution room and adaptively adjust the detection priority and sampling parameters; then, collect data through dual channels of the main detection channel and the secondary interference verification channel. The main detection channel acquires the original data of the core equipment before interference, while the secondary interference verification channel sends low-power analog signals to the same group of target sensors and collects verification data after interference; compare the changes of the two types of data, distinguish between sensor false alarms and real anomalies, classify and determine real anomalies, and output the detection results; S2. Multi-objective game decision-making: Based on the detection results output in step S1, the real anomalies are first classified into risk levels. For high-level risks, load transfer and equipment isolation are triggered, and for complex anomalies, idle resources are scheduled to support the handling. Then, based on the correlation between equipment failure and the cause, three types of strategies are generated: rapid handling, cost optimization and root cause eradication. The comprehensive benefit is calculated through game theory algorithm and the optimal strategy is selected to generate control instructions. S3. Flexible Execution: Based on the control instructions generated in step S2, perform exception isolation, resource scheduling, and policy implementation operations, and provide feedback on the execution results; S4. Federated Optimization: Receive the strategy effect data after de-identification and fusion of anomaly handling records and execution feedback results of step S3, and generate strategy parameter update instructions through aggregation training optimization, which are then issued to update decision algorithm parameters and strategy library.

10. The active monitoring method for a power distribution station according to claim 9, characterized in that, In step S1, scene recognition includes the determination of peak load scene, maintenance scene and standby scene, and the adapted networking methods include star networking, mesh networking and bus networking. The sampling frequency is dynamically adjusted according to the scene. The real anomaly classification results include single anomalies, potential anomalies, and complex anomalies; in step S2, the risk classification includes level 1, level 2, and level 3, corresponding to different failure probability ranges; idle resources include edge computing resources with remaining computing power ≥ 30%, communication resources with remaining bandwidth ≥ 20Mbps, and sensor sampling channels with unassigned tasks; the game algorithm is the Nash equilibrium algorithm; the comprehensive benefit calculation integrates the solution rate, handling cost, and downtime loss; in step S4, the FedAvg architecture is used for aggregation training, and the generated policy parameter update instructions include the payoff matrix weights, policy selection thresholds, and association rule coefficients.

Images